Rights / License: Research Collection In Copyright - Non ...5927/eth-5927-02.pdf · 2 budget of the lawn ecosystem had to be adjusted by ˇ135 g C m 2 s 1 to determine the natural

Research Collection

Doctoral Thesis

Greenhouse gasesFrom the ecosystem scale to a regional integration

Author(s): Hiller, Rebecca V.

Publication Date: 2012

Permanent Link: https://doi.org/10.3929/ethz-a-7355117

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For moreinformation please consult the Terms of use.

ETH Library

https://doi.org/10.3929/ethz-a-7355117

http://rightsstatements.org/page/InC-NC/1.0/

https://www.research-collection.ethz.ch

https://www.research-collection.ethz.ch/terms-of-use

GREENHOUSE GASES

FROM THE ECOSYSTEM SCALETO A REGIONAL INTEGRATION

REBECCA V. HILLER

Dissertation for the degreeof Doctor of Science

DISS. ETH NO. 20183

GREENHOUSE GASES

FROM THE ECOSYSTEM SCALETO A REGIONAL INTEGRATION

A dissertation submitted to

ETH Zurich

for the degree of

Doctor of Sciences

presented by

Rebecca Veronika HillerMaster of Science in Geography, University of Bern

22.07.1982

citizen ofZurich (ZH)

Prof. Dr. Nina BuchmannPD Dr. Werner Eugster

Prof. Dr. Bruno G. Neininger

2012

To my grand parents Gertrud and Arnold Ernst-Eugster.

Abstract

Today’s greenhouse gas inventories are typically based on bottom-up estimates, but com-parisons to atmospheric measurements are rare and can disagree by a factor of two or more.Therefore, top-down assessments and direct flux measurements of individual ecosystems areneeded to verify the commonly used emission factors.

In Switzerland, 83% of the anthropogenic methane emissions are attributed to the agricul-tural sector. To validate the methane emission inventory estimates from this most relevantsource in Switzerland, we performed aircraft measurements over the agriculturally domi-nated landscape of the Reuss Valley in Central Switzerland on 16 days in 2009 and 2010.Aircraft eddy covariance flux estimates (0.30 µg CH4 m�2 s�1) were lower than the meaninventory flux for this area (0.41 µg CH4 m�2 s�1), while fluxes estimated by a simplifiedbox model were higher (0.89 µg CH4 m�2 s�1).

A similar study was performed at Lake Wohlen, a run-off-river hydropower reservoir closeto Bern, Switzerland, on 30 September 2009. Previous studies showed considerable methaneemissions from this lake. A full-featured box model and a simplified version yielded sim-ilar values for the noon (19 µg CH4 m�2 s�1 for both approaches) and afternoon overflight(12 µg CH4 m�2 s�1 and 8 µg CH4 m�2 s�1). These values were higher than fluxes from thethree chambers deployed on the same day that averaged 5.5 µg CH4 m�2 s�1. Relativelyhigh uncertainties are associated with the aircraft measurements questioning the represen-tativeness of the performed measurements for the total box. Contrastingly, the chamberflux measurements rely only on three chambers that were sampled four times and hencethe spatial representativeness can be questioned. Given these limitations, a good agreementbetween the bottom-up and top-down approach was found.

Urban and suburban areas are known for their heterogenic land use and are generally con-sidered a CO2 source. However, they also include a considerable vegetated area includingturf-grass lawns that take-up a considerable amount of CO2. Since dimensions of lawns aretypically small, flux measurements by the eddy covariance method may be influenced bynearby CO2 sources such as passing traffic. This was also the case for eddy covariance fluxmeasurements performed over a turf-grass lawn in the first-ring suburb of Minneapolis–SaintPaul, MN, USA. Traffic contribution was estimated from winter flux data when vegetation

v

was dormant, the footprint size calculated by a parameterized version of a three-dimensionalLagrangian footprint model, and continuous traffic counts. The CO2 budget of the lawnecosystem had to be adjusted by � 135 g C m�2 s�1 to determine the natural flux in 2007and 2008, even though the road crossed the footprint only at its far edge.

Fast response optical analyzers based on laser absorption spectroscopy are the preferredtools to measure field-scale mixing ratios and fluxes of a range of trace gases. Severalstate-of-the-art instruments have become commercially available and are gaining popularity.Two compact, cryogen-free, and fast response instruments that measure CH4 were criticallyevaluated in a field intercomparison: a quantum cascade laser based absorption spectrometerfrom Aerodyne Research, Inc., and an off-axis integrated cavity output spectrometer fromLos Gatos Research Inc. Both flux setups compared well with a prescribed flux and theinstruments showed good overall performance. However, both instruments showed cross-sensitivities to H2O, which influence especially measurements of small fluxes.

Cross-sensitivities caused by H2O were assessed in respect to the influences on fluxmeasurements for a Fast Methane Analyzer (FMA) and a Fast Greenhouse Gas Analyzer(FGGA, both Los Gatos Research Inc.). These cross-sensitivities are also observed for otherlaser spectroscopic instruments and generally enlarge the dilution effect of water. H2O notonly influences the measured power spectrum, sorption processes at the intake tube wallsand filters delay and dampen the observed H2O signal more strongly than the sampled tracegas. Thereby, a shift between the trace gas and the H2O fluctuations is introduced. Theknowledge of this behavior of H2O in the instrument is important because the density fluxcorrection for H2O introduced density fluctuations depends on the water vapor flux in the cellat the time the trace gas of interest is sampled. With our specific setup for eddy covarianceflux measurements, the cross-sensitivity counteracts the damping effects, which compensateeach other. Hence, the new correction only deviates very slightly from the traditional Webb,Pearman, and Leuning density correction, which can be calculated from separate measure-ments of atmospheric water vapor.

vi

Zusammenfassung

Die heutigen Treibhausgasinventare basieren typischerweise auf extrapolierten Abschatz-ungen der einzelnen Quellen. Direkte Vergleiche mit atmospharischen Messungen sind rarund unterscheiden sich um Faktor zwei oder mehr von diesen. Um die gangigen Emis-sionsfaktoren zu prufen, werden deshalb integrale Abschatzungen wie auch Messungen vonindividuellen Flussen der einzelnen Okosysteme benotigt.

In der Schweiz stammen 83% der vom Mensch verursachten Methanemissionen aus derLandwirtschaft. Um die inventarbasierten, landwirtschaftlichen Abschatzungen mit regio-nalen Flussmessungen zu vergleichen, fuhrten wir an 16 Tagen in den Jahren 2009 und 2010Flugzeugmessungen im landwirtschaftlich gepragten Reusstal, in der Zentralschweiz, durch.Flugzeugbasierte Eddy Kovarianz Flussmessungen liegen etwas tiefer (0.30 µg CH4 m�2 s�1)als der gemittelte Fluss aus dem raumlich hochaufgelosten Inventar fur diese Region (0.41µg CH4 m�2 s�1). Abschatzungen mit einem vereinfachten Boxmodel ergaben im Vergleichhohere Werte (0.89 µg CH4 m�2 s�1), wobei auch die Streuung grosser ausfiel.

Eine ahnliche Studie wurde am Stausee eines Flusswasserkraftwerks, dem Wohlensee,in der Nahe von Bern (Schweiz), am 30. September 2009 durchgefuhrt. VorhergehendeStudien zeigten fur diesen See deutliche Methanemissionen. Ein vollstandiges und einvereinfachtes Boxmodell ergaben fur den Mittags- und Nachmittaguberflug ahnliche Werte.Am Mittag schatzten beide Modelle einen Fluss von 19 µg CH4 m�2 s�1. Der Nachmit-tagswert fur das vollstandige Modell fiel mit 12 µg CH4 m�2 s�1 etwas hoher aus als die 8µg CH4 m�2 s�1 fur das vereinfachten Modells. Kammermessungen, welche am selben Tagdurchgefuhrt wurden, lagen mit durchschnittlich 5.5 µg CH4 m�2 s�1 tiefer als die flugzeug-basierten Abschatzungen. Die Flugzeugmessungen unterliegen relativ hohen Unsicherheiten,welche die Reprasentativitat der Flugmessungen fur die ganze Box betreffen. Die Kammer-messungen ihrerseits beruhen nur auf den Messungen von drei einzelnen Kammern, welchean diesem Tag vier Mal beprobt wurden, und sind somit raumlich beschrankt reprasentativ.In Anbetracht dieser Einschrankungen stimmten die extrapolierten und integrativ gemesse-nen Flusse gut miteinander uberein.

Stadtische und vorstadtische Gebiete sind fur ihre heterogene Struktur bekannt und wer-den generell als CO2 Quellen betrachtet. Dabei werden Grunflachen, welche eine beachtliche

vii

Menge an CO2 aufnehmen, nicht beachtet. Zu diesen Grunflachen gehoren auch Rasenflach-en, die haufig klein sind. Flussmessungen von solchen Rasenflachen mittels der Eddy Ko-varianz Methode konnen deshalb von naheliegenden CO2 Quellen, wie Durchgangsverkehr,beeinflusst werden. Dies trifft auch fur Eddy Kovarianz Flussmessungen zu, die uber einemRasen in der Agglomeration von Minneapolis–Saint Paul durchgefuhrt wurden. Die Auswir-kungen der Verkehrsemissionen wurden anhand von Winterflussen abgeschatzt. Bei ruhen-der Vegetation bewirken nur die vom Verkehr verursachten Emissionen einen beobachtba-ren Tagesgang. Der Verkehrseinfluss wurde mit Hilfe der Footprintgrosse der Flussmessungsowie kontinuierlichen Verkehrszahlungen evaluiert. Das CO2 Budget des Rasenokosystemsmusste in den Jahren 2007 und 2008 um je�135 g C m�2 s�1 korrigiert werden, obwohl dieStrasse den Footprint nur, wenn uberhaupt, am aussersten Ende querte.

Optische Sensoren mit einer kurzen Ansprechzeit, welche auf der Theorie der Laserab-sorptionsspektroskopie beruhen, werden bevorzugt zum Messen von Mischungsverhaltnis-sen und Flussen von Spurengasen verwendet. Diverse Gerate – dem letzten Stand der For-schung entsprechend – wurden kommerziell verfugbar und erlangen immer grossere Be-liebtheit. Zwei kompakte und kuhlmittelfreie Instrumente mit kurzer Ansprechzeit wurdeneinem kritischen Vergleich im Feld unterzogen. Dabei handelte es sich um einen auf Ab-sorptionsspektroskopie basierenden Quantenkaskadenlaser von Aerodyne Research Inc. undeinen Cavity Ring-down Laserabsorptionsspektroskop von Los Gatos Research Inc. BeideMesssysteme massen den kunstlich vorgegebenen Methanfluss prazise und zeigten eine guteGesamtleistung. Allerdings wiesen beide Instrumente Querempfindlichkeiten zu Wasser auf,welche speziell die Messungen von kleinen Flussen beeinflussen.

Der Einfluss von H2O Querempfindlichkeiten auf die Flussmessungen wurde fur einenFast Methane Analyzer (FMA) und einen Fast Greenhouse Gas Analyzer (FGGA, beide LosGatos Research Inc.) evaluiert. Solche Querempfindlichkeiten wurden auch fur andere laser-spektroskopische Instrumente beobachtet und vergrossern in der Regel den Verdunnungsef-fekt durch Wasser. H2O beeinflusst aber nicht nur das gemessene Leistungsspektrum. Sorp-tionsprozesse an den Wanden der Ansaugschlauche und in Filtern verzogern und dampfendas H2O Signal starker als das gemessene Spurengas. Dadurch wird eine Verschiebungzwischen den Fluktuationen des Spurgases und H2O herbeigefuhrt. Dieser Sachverhaltist wichtig, da fur die Dichteflusskorrekur vom Wasserdampffluss in der Messzelle zumMesszeitpunkt des Spurengases ausgegangen wird. Bei unserer Installation fur Eddy Kovar-ianz Flussmessungen wirkt die Querempfindlichkeit den Dampfungseffekten entgegen unddie beiden Effekte kompensieren sich gegenseitig. Daher weicht die neue Korrektur nur ger-ingfugig von der traditionellen Webb, Pearman und Leuning Dichtekorrektur ab, welche ausseparaten Messungen des atmospharischen Wasserdampfs berechnet werden kann.

viii

Contents

1. General introduction 1

1.1. Inventories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.2. Footprints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3. MAIOLICA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.4. Methane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.5. Flux measurements of methane . . . . . . . . . . . . . . . . . . . . . . . . 61.6. Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.7. Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.8. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2. Aircraft based CH4 flux estimates for an agriculturally dominated valley in

Switzerland 19

2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.2.1. Site and flight pattern . . . . . . . . . . . . . . . . . . . . . . . . 212.2.2. Aircraft measurements . . . . . . . . . . . . . . . . . . . . . . . . 222.2.3. Flux calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.2.4. Spatially explicit CH4 inventory . . . . . . . . . . . . . . . . . . . 252.2.5. Footprint calculations . . . . . . . . . . . . . . . . . . . . . . . . . 26

2.3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.3.1. Performance of the methane analyzer . . . . . . . . . . . . . . . . 272.3.2. Methane fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272.3.3. Footprint location sensitivity . . . . . . . . . . . . . . . . . . . . . 29

2.4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292.4.1. Uncertainties of the methane inventory . . . . . . . . . . . . . . . 292.4.2. Variations of the CH4 flux measurements . . . . . . . . . . . . . . 31

2.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.7. Figures and tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

ix

Contents

3. Aircraft based CH4 emissions estimates for a Swiss lowland hydropower

reservoir 45

3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

3.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.2.1. Aircraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

3.2.2. Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.2.3. Weather conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3.2.4. Flight pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.2.5. Box model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.2.6. Simplified box model . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.2.7. Chamber measurements . . . . . . . . . . . . . . . . . . . . . . . 52

3.3. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

3.3.1. Atmospheric CH4 distribution over Lake Wohlen . . . . . . . . . . 52

3.3.2. Atmospheric concentration gradients along the lake . . . . . . . . . 53

3.3.3. Conservation of air mass and mass balance closure . . . . . . . . . 53

3.3.4. Flux estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

3.4. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.4.1. Regional CH4 distribution . . . . . . . . . . . . . . . . . . . . . . 55

3.4.2. Model performance . . . . . . . . . . . . . . . . . . . . . . . . . . 55

3.4.3. Flux estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

3.5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

3.6. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

3.7. Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4. Interpreting CO2 Fluxes Over a Suburban Lawn 67

4.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

4.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.2.1. Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.2.2. Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4.2.3. Data processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

4.2.4. Gap filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

4.2.5. Traffic counts and CO2 emission factors . . . . . . . . . . . . . . . 73

4.2.6. Estimating traffic effects on measured CO2 fluxes – development ofan empirical model . . . . . . . . . . . . . . . . . . . . . . . . . . 74

4.3. Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4.3.1. CO2 flux measurements . . . . . . . . . . . . . . . . . . . . . . . 75

x

Contents

4.3.2. Empirical model of traffic CO2 fluxes . . . . . . . . . . . . . . . . 764.3.3. Empirical model compared to traffic CO2 flux estimates based on

emission factors . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.3.4. Impact of traffic emissions on annual CO2 fluxes at this site . . . . 78

4.4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.6. Figures and tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

5. Field intercomparison of two optical analyzers for CH4 eddy covariance flux

measurements 91

5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 925.2. Material and methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93

5.2.1. Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . 935.2.2. Measurement site . . . . . . . . . . . . . . . . . . . . . . . . . . . 955.2.3. Trace gas flux generation . . . . . . . . . . . . . . . . . . . . . . . 955.2.4. Flux measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 96

5.3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 975.3.1. Instrument characterisation . . . . . . . . . . . . . . . . . . . . . . 985.3.2. CH4 mixing ratio and fluxes . . . . . . . . . . . . . . . . . . . . . 105

5.4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1075.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1085.6. Figures and tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

6. Flux correction for closed-path laser spectrometers without internal water

vapor measurements 121

6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1226.2. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1236.3. Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126

6.3.1. Instruments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1266.3.2. Dilution and cross-sensitivity experiment . . . . . . . . . . . . . . 1266.3.3. Flux measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 126

6.4. Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1306.4.1. Dilution and cross-sensitivity experiment . . . . . . . . . . . . . . 1306.4.2. Damping loss correction . . . . . . . . . . . . . . . . . . . . . . . 1316.4.3. Signal desynchronisation . . . . . . . . . . . . . . . . . . . . . . . 1326.4.4. Flux comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 1336.4.5. Adapted WPL correction . . . . . . . . . . . . . . . . . . . . . . . 133

xi

Contents

6.5. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1346.6. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1366.7. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

7. General discussion 151

7.1. Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1517.2. Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547.3. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

Appendix 157

A. Seasonal contributions of vegetation types to suburban evapotranspiration 157

Acknowledgements 175

xii

1General introduction

The atmosphere enables life on Earth by its protecting properties and provides many ecosys-tem services (Thornes, 2010). Harmful radiation is absorbed and never reaches the Earth’ssurface. Additionally, part of the outgoing long-wave radiation is scattered back, increasingthe surface temperature (Le Treut et al., 2007). Without atmosphere, average surface tem-perature would be only –18 ıC providing unfavorable conditions for life. Today’s averagesurface temperature is 15 ıC with an increasing trend. The rescattering of the outgoing long-wave radiation is caused by certain gases in the atmosphere, the so-called greenhouse gases(GHG). The most important one is H2O, followed by CO2, CH4, and N2O. Concentrationsof CO2, CH4, and N2O have been increasing substantially since the industrial revolution.Measurements from ice cores dating back 800’000 years never showed atmospheric GHGconcentrations as high as currently observed (Luthi et al., 2008; Spahni et al., 2005). CO2

is released to the atmosphere by the combustion of fossil fuels whereas the CH4 increaserelates to intensified agriculture of rice and ruminants to maintain the food supply for thegrowing population (Crutzen, 2002). Knowing the warming potential of the GHGs, concernarose due to the increasing trend of GHG concentrations. Global climate model scenariosproject up to 4 ıC higher global average temperatures from 2000 to 2100, leading to a liftof the sea level and changing weather pattern (IPCC, 2007). Additionally, extreme weatherevents tend to occur more often. These consequences directly affect human well-being allaround the globe. Hence, the United Nations (UN) took a leading position by composingthe UN General Assembly Resolution 43/53 on 6 December 1988. The IntergovernmentalPanel on Climate Change (IPCC) was established by the World Meteorological Organiza-tion (WMO) and the United Nations Environment Program (UNEP), both UN agencies. Theobjective of the IPCC is to report on the most recent scientific findings on a regular base andprovide therewith a basis for decisions and measures taken by policy makers and the in-

1

Chapter 1: General introduction

dustry. International pacts, such as the Kyoto Protocol that was adopted in 1997, specifythe emission reductions of the signing nations. However, direct measures still apply to theindividual countries and thus cannot be forced from outside without the commitment of thecountry itself. Additionally, no penalties are currently foreseen for failing the commitmentsof the Kyoto Protocol. Nevertheless, to control the performance of the individual countries,national emission inventories are needed. These inventories also provide fundamental inputto climate models used to generate future climate scenarios.

1.1. Inventories

Inventories are used for different purposes and thus have to meet different requirementsdepending on their application, whether they are used for regulation or scientific purposes.

For regulation application, country or region based estimates are sufficient. The invento-ries are used e.g. to control whether a country fulfills the meetings of a ratified agreementsuch as the Kyoto Protocol. Therefore, all countries have to follow a standard approach.In case of the Kyoto Protocol, these are the IPCC guidelines (IPCC, 2006). Depending onthe data availability, countries can choose between different approaches. The annual reportshowever have to follow the same procedure every year to maintain consistency. The targetof the Kyoto Protocol is to reduce the GHG emissions on average by 5% compared to 1990from 2008–2012. Since the target is defined relative to the emissions in 1990, consistencyof the reporting is far more important than choosing the most accurate method for the indi-vidual year. An under- or overestimation of all emissions still results in a reduction of thedefined percentage of the true estimations if the same reporting approach is used. Validationin respect to reporting inventories means the control of the reporting procedure and not theverification of methods to determine the source strength (IPCC, 2000). Even though weak-nesses in the reporting procedure cannot be adjusted immediately, verification of the existinginventories is important. Better knowledge of the individual emissions helps to specify themost effective way to reduce the individual sources. New methods can be introduced intothe reporting framework during revisions to achieve more accurate results in the future.

Many inventories, including the national inventory reports following the IPCC guidelines,base on a bottom-up approach. A quantity, e.g. the number of vehicles, is multiplied with anemission factor, e.g. the amount of CO2 emitted per vehicle and year. The emission factormay be different for different vehicle types and the fleet age. Large cars emit more CO2

than small cars while new vehicles are more efficient than old ones. The fleet compositionand age distribution changes from country to country as well as gasoline mixture and theaverage number of driven kilometers per vehicle. Hence, many factors are involved in the

2

1.2. Footprints

emission estimate. Other anthropogenic sources are estimated by similar approaches. Eventhough most emission factors relate to direct measurements at the respective source, thesum of the total emissions might be questionable. Many uncertainties are introduced inthe calculation of the total emission, e.g. the annually driven kilometer per vehicle is notknown very well. Thus, the question arises whether the bottom-up estimates correspondto the true emission. Direct comparisons are scarce and sometimes inventories disagree byfactor of two or three (Nisbet and Weiss, 2010) compared to high precision atmosphericmeasurements (Levin et al., 2010; Muhle et al., 2010; Stohl et al., 2009; Weiss et al., 2008;Bergamaschi et al., 2005) or satellite retrievals (Bergamaschi et al., 2009). Verifications ofthe existing inventories are therefore needed.

Scientific requirements on inventories are different. Spatially explicit inventories areused to prescribe GHG emissions in models and therefore, accurate inventories are required(Lamarque et al., 2010). The better the source strength and location is known, the better themodel performs. Inventory requirements largely depend on the application. While globalmodels operate on a coarse grid of a few hundreds of kilometers, regional models requestmore detailed information with a resolution of hundreds of meters or even less. Hence,the spatial distribution and strength of the individual sources is crucial for spatially highlyresolved inventories. Bottom-up estimates are one approach to create such an inventory(Olivier et al., 1999). Therefore, the sources are in addition to the national inventory reportspatially distributed. Another possibility to derive or improve spatially resolved inventoriesis inverse modeling (Weiss and Prinn, 2011).

Especially for scientific applications of inventories, verification of the available bottom-up inventories against top-down emission estimated from direct measurements helps to testtheir reliability. Therefore, the field of view of the measurement instrument needs to bespecified. Footprint models are a potential tool to determine the required information.

1.2. Footprints

The footprint describes the area seen by a sensor and the related regional contribution differ-ences. Mathematically, the footprint is specified by a source weight function (Schuepp et al.,1990). The source area or fetch is the total area viewed by a sensor and can be interpretedas the integral of the source weight function (Schmid, 1994). For an outgoing shortwave ra-diometer, the fetch is for example equal to a circle conically projected on the ground whichsize depends on the measurement height (Schmid et al., 1991). The determination of thefootprint for a turbulently transported property requires more sophisticated approaches. Theflux footprint itself is mainly located up-wind of the sensor with an approximately ellipti-

3


cal shape. The source area strongly depends on the sensor height, the surface roughness,the atmospheric stability, and the wind speed (Leclerc and Thurtell, 1990). The higher themeasurement height, the more mixing occurs between the source and the instrument thatintegrates air with different origin. Therefore, larger footprints are associated with increas-ing measurement height. A very rough rule of thump estimates the fetch by 100 times themeasurement height (Horst and Weil, 1994). Rougher surface produces more turbulencethat promotes the mixing and hence shorter footprint are observed compared to over smoothsurfaces. The same concept is associated with the stability of the atmosphere. During stableconditions mixing is attenuated and hence larger footprints are observed. The combinationof many footprints over a specified time period indicates the most flux relevant areas for thisperiod, the footprint climatology (Rebmann et al., 2005).

To assess the flux footprint of the measured trace gas, various models have been developed(Vesala et al., 2010). These models either follow the Eulerian or Lagrangian framework todescribe motion (Finnigan, 2004): (1) In the Eulerian framework, the mass conservationequation is solved for all points in a predefined grid and the dispersion of a trace gas is de-scribed by concentration changes at each individual grid point (e.g. deployed in the modelsby Schuepp et al., 1990; Schmid, 1994; Kormann et al., 2001); (2) the Lagrangian frame-work follows a single particle and describes its motion and gas molecules are tracked in theair (Kljun et al., 2002).

Lagrangian particle dispersion models (LPDM) are widely used to calculate footprints forfluxes (Kljun et al., 2002), but also for trace gas concentrations (Lin et al., 2011). To derivethe footprint, thousands of particles are released at the location of the instrument. The modelitself is operated backward and the trajectories of the individual particles are calculated. Thetime and location where the trajectories touch the surface are converted to a footprint. Thesecalculations are very time consuming while the application of flux footprints often operateswith 30-minute footprints over longer periods e.g. one year. Therefore, parametrizationsof the Lagrangian models were derived to improve the computing time (Kljun et al., 2004).However, these models are restricted to certain conditions. The application of footprintanalyses include: (1) the determination of the main flux contributing area for a flux tower(Zeeman et al., 2010; Hiller et al., 2008); (2) tracer studies where an artificial source isissued to either test the performance of the footprint model (Gockede et al., 2005; Foken andLeclerc, 2004) or to test and compare flux setups (Tuzson et al., 2010); (3) flux partitioningof the measured flux for known sources and sinks (Hiller et al., 2010); (4) scaling studiesthat measured fluxes at different spatial scales (Peters et al., 2011); (5) inverse modelingwhere measured concentrations are distributed to create or verify inventories (Bergamaschiet al., 2005).

4

1.3. MAIOLICA

Hence, footprint calculations are a powerful tool to compare local with regional measure-ments, one of many objectives in the project MAIOLICA.

1.3. MAIOLICA

MAIOLICA (Modelling And experIments On Land-surface Interactions with atmosphericChemistry and climAte) was founded by the Competence Center Environment and Sus-tainability (CCES) of the ETH domain and is executed by ten different groups within fiveinstitutions of the ETH domain. The project aims to combine different pieces from individ-ual approaches to an entire mosaic, in analogy to the Italian tin-glazed pottery dating fromthe Renaissance called Maiolica. The mosaic will improve our understanding of GHGs atvarious spatial scales under different environmental conditions. Pieces are represented bydirect measurements at the local and regional scales, compilation of emission inventories,process, regional, and global model runs incorporating different climate sensitivities. On thelocal scale, point source measurements contribute to a better understanding of the drivingprocesses and are representative for a single ecosystem. In order to scale up and connectthe individual pieces, a regional integration is needed. Locally measured GHG fluxes arequantitatively compared against regional fluxes from aircraft measurements. This link willcontribute to close the gap between the local fluxes and the regional to global models thatwill be developed and run within the MAIOLICA project. The research mainly focuses onthe GHG methane. The project itself was split into three activities: (1) Experiments at localand regional scales, (2) soil/land/vegetation modeling at local, regional and global scales,and (3) atmospheric modeling at regional and global scales. Within the experimental ac-tivity of the project, measurements of GHG fluxes from the local to the regional scale areperformed on the Swiss Plateau.

1.4. Methane

Methane is the second most important anthropogenic GHG. Even though the residence timeof CH4 in the atmosphere is only about one tenth compared to CO2, its warming potential is25 times larger (Forster et al., 2007). Since the industrial revolution, the global atmosphericCH4 concentration more than doubled from 0.7 ppm to 1.7 ppm (BUWAL, 1996), got ratherstable during the past few years (Forster et al., 2007), but started to rise again since early2007 (Rigby et al., 2008). Methane attributes 18% to the global warming potential comparedto 7% in Switzerland. While at global scale natural wetlands are the largest CH4 source(Wofsy et al., 2007), agriculture dominates the Swiss emissions (FOEN, 2010). The Swiss

5


national Inventory only includes anthropogenic emissions. However, wetland contributionsare assumed to contribute marginally to the total emissions. The Swiss river courses werecorrected and the associated flood plains drained in the last 300 years. The major globalCH4 sink is chemical oxidation in the troposphere. A small amount of CH4 is lost to thestratosphere or taken up by soils (Wofsy et al., 2007).

1.5. Flux measurements of methane

Many different techniques exist to measure the CH4 fluxes. Their applicability stronglydepends on the desired spatial resolution, but also on the availability of resources.

Chambers

Static chamber measurements are the cheapest method and can be deployed on land or wa-ter. The chambers are closed at the beginning of the measurement. The concentration in thechamber is measured for different closure times, either by taking gas samples or by directlymeasuring with a gas specific sensor. A concentration increase with advancing closure timeindicates a net source whereas a decrease is observed in case of a net sink. The gas samplesare analyzed in the laboratory by a gas chromatograph or a mass spectrometer for the gasesor isotopes of interest. The flux is calculated from the concentration change over time, thechamber volume and its surface area. More sophisticated chambers close, sample, and openautomatically. Steady state through flow chambers operate on a slightly different approach.Fresh air is continuously pumped through the chamber and sampled at the chamber inletand the outlet. The change in concentration between inlet and outlet, the flow rate of the airthrough the chamber and the chamber surface area determine the measured flux. A detaileddescription of the chamber method is provided in Livingston and Hutchinson (1995). Cham-bers can be deployed on land (Pumpanen et al., 2004), over water (DelSontro et al., 2010),or enclose the medium of interest, e.g. a tree (Barthel et al., 2011) or a cow (Klevenhusenet al., 2010). However one chamber only represents the fluxes of one spot or object andhence, several chambers need to be deployed to receive a representative result (Baldocchiet al., 1988).

Eddy covariance

The eddy covariance method integrates over a larger spatial scale than individual chambermeasurements. In general, the concentration of the trace gas of interest and vertical wind

6

1.5. Flux measurements of methane

speed of the air is measured at very high temporal resolution, e.g. 10 Hz. Turbulent move-ments transport air parcels up and down an imaginary surface parallel plane at the height ofthe sensors. Budgeting the individual air parcels over a given time period of e.g. 30 minallows to asses the transport of the trace gas of interest and hence the calculation of the net-flux. Therefore, the covariance between the vertical wind and the concentration of the tracegas of interest is calculated (Baldocchi, 2003). In the post-processing, different correctionsare applied depending on the measurement setup. Discrete sampling and other set-up spe-cific factors such as sorption and desorption in the tubes lead to a damping of the measuredsignal (Mammarella et al., 2009). This damping can be corrected either by an integratedmethod accounting for various damping sources (e.g. Eugster and Senn, 1995) or by ap-plying individual transfer functions that account for the individual damping processes (e.g.Moore, 1986; De Ligne et al., 2010). Density effects arise as warm and humid air is lighterand hence is more likely transported upwards (Webb et al., 1980). Temperature effects aremost important for open-path instruments, where the air is directly measured and the val-ues are recorded in density units (Leuning, 2007). The dry mole fraction does not changewith temperature fluctuations and thus no correction is needed. Setups with closed-path in-struments pull air through tubes into the measurement cell for sampling and hence densityeffects due to temperature fluctuations are less pronounced (Leuning, 2005). The tubes andthe waste heat of the instrument equilibrate the temperature of the air sample. Correctionsfor density effects due to water vapor apply if the air is not dried before sampling and humidmole fractions or density units are recorded. In order to apply an appropriate correction,water vapor flux measurements are needed. Best results are achieved if the same instrumentmeasures water vapor since the water vapor flux in the measurement cell determines thedilution effect (Ibrom et al., 2007). Otherwise, the water vapor flux might need post calibra-tion to fit the one that would be observed in the cell at the time of the measurement of thetrace gas of interest.

Eddy covariance measurements of CH4 have been deployed over wetlands, pastures, lakesand landfills (Detto et al., 2011; Eugster et al., 2011; Eugster and Pluss, 2010; Tuzson et al.,2010; Hendriks et al., 2010; Kroon et al., 2010, and many others). Even though moderninstruments allow measurements of various GHG species at high temporal resolution, theirprecision is not sufficient yet to measure very small fluxes as for example CH4 uptake bysoils. To assess this process, chamber measurements are still the most reliable approach.

Aircraft based flux measurements

As chambers do not necessary represent ecosystem scale fluxes, tower based eddy covari-ance fluxes may not be representative for the entire region. While fluxes measured from

7


a tall towers already integrate over a larger area, aircraft flux measurements and boundarylayer budgeting definitively do. The principles of airborne eddy covariance measurementsare similar to the tower approach. Instead of averaging data over time, the analyses are per-formed on spatial transects (Desjardins et al., 1989). Many aircraft based flux studies wereaccomplished for sensible heat, H2O, CO2, and O3 fluxes (Graber et al., 1998; Lehning et al.,1998; Barr et al., 1997; Desjardins et al., 1997, 1995; Mahrt et al., 1994). Only two publishedstudies (Ritter et al., 1994, 1992) exist to our knowledge about CH4 eddy covariance mea-surements whereas more studies deployed conditional or relaxed eddy accumulation (e.g.Pattey et al., 2006; Beswick et al., 1998). For this method, air is conditionally sampled in anup and down bin depending on the vertical motion of the air. These air samples are thereafteranalyzed in the laboratory for the property of interest (Denmead, 2008). Recent instrumentdevelopment however resulted in stable, robust, and relatively light instruments without ac-tive cooling of the detector with liquid nitrogen that are suitable for airborne application andmeasure at high temporal frequencies that allow eddy covariance measurements.

Boundary layer budgets are deployed in the cumulative mixing layer during the day.Therefore, the area of interest is enclosed by an imaginary box and the important flowsin and out of this box as well as concentration changes within the box are assessed. The im-balance between the fluxes into and out of the box is attributed to the source or sink strengthwithin the box (Denmead et al., 1999). Many applications of boundary layer budgets arefound in the literature (Desai et al., 2011; Alfieri et al., 2010; Laubach and Fritsch, 2002;Lehning et al., 1996; Betts, 1992), including studies about CH4 (Choularton et al., 1995). AllCH4 applications base on air samples analyzed in the lab after the flight, except one study(Mays et al., 2009) where CH4 was measured continuously.

1.6. Objectives

This thesis aims to contribute to a better understanding of greenhouse gas fluxes at theecosystem and regional scale. Special focus is paid on the scalability of fluxes betweenthe different spatial scales. The objectives can be split into the following main aspects:

1. Quantify regional scale CH4 fluxes and assess their reliability.

2. Compare the regional scale fluxes with spatially highly resolved inventory and ecosys-tem scale measurements.

3. Quantify ecosystem scale fluxes and relate those with the help of footprints to knownsources.

8

1.7. Thesis outline

4. Assess the performance of new instruments to measure CH4 fluxes.

1.7. Thesis outline

In the first manuscript (Chapter 2), regional scale CH4 fluxes are estimated by two indi-vidual methods and contrasted against each other. Further, these results are compared toinventory based emissions whose representativeness is discussed in more detail. The secondmanuscript (Chapter 3) investigates the CH4 fluxes of an aquatic system by comparing air-borne flux estimates with chamber and eddy covariance based fluxes. In the third manuscript(Chapter 4) the focus is changed to CO2 fluxes. Side effects of an adjacent road on the netecosystem exchange of a suburban lawn are assessed with the help of a footprint model.This study demonstrates the power of footprint models. The forth manuscripts (Chapter 5)switches back to CH4 fluxes. Footprints serve to scale artificial CH4 emissions released ina tracer experiment. The flux measurements of two individual eddy covariance systems arecompared against the issued reference flux. The last manuscript (Chapter 6) carefully ex-amines instrument specific side effects of water vapor on the CH4 flux measurements. Thefindings are used to derive an appropriate correction procedure that can be applied to themeasured fluxes.

1.8. References

Alfieri, S., U. Amato, M. Carfora, M. Esposito, and V. Magliulo (2010). Quan-tifying trace gas emissions from composite landscapes: A mass-budget ap-proach with aircraft measurements. Atmospheric Environment 44(15), 1866–1876.doi:10.1016/j.atmosenv.2010.02.026.

Baldocchi, D. D. (2003). Assessing the eddy covariance technique for evaluating carbondioxide exchange rates of ecosystems: past, present and future. Global Change Biol-ogy 9(4), 479–492. doi:10.1046/j.1365-2486.2003.00629.x.

Baldocchi, D. D., B. B. Hincks, and T. P. Meyers (1988). Measuring biosphere-atmosphereexchanges of biologically related gases with micrometeorological methods. Ecol-ogy 69(5), 1331–1340.

Barr, A. G., A. K. Betts, R. L. Desjardins, and J. I. MacPherson (1997). Compar-ison of regional surface fluxes from boundary-layer budgets and aircraft measure-ments above boreal forest. Journal of Geophysical Research. 102(D24), 29213–29218.doi:10.1029/97JD01104.

9


Barthel, M., A. Hammerle, P. Sturm, T. Baur, L. Gentsch, and A. Knohl (2011). The dielimprint of leaf metabolism on the ı13C signal of soil respiration under control and droughtconditions. New Phytologist 192(4), 925–938. doi:10.1111/j.1469-8137.2011.03848.x.

Bergamaschi, P., C. Frankenberg, J. F. Meirink, M. Krol, M. G. Villani, S. Houweling,F. Dentener, E. J. Dlugokencky, J. B. Miller, L. V. Gatti, A. Engel, and I. Levin (2009).Inverse modeling of global and regional CH4 emissions using SCIAMACHY satelliteretrievals. Journal of Geophysical Research. 114, D22301. doi:10.1029/2009JD012287.

Bergamaschi, P., M. Krol, F. Dentener, A. Vermeulen, F. Meinhardt, R. Graul, M. Ramonet,W. Peters, and E. J. Dlugokencky (2005). Inverse modelling of national and europeanCH4 emissions using the atmospheric zoom model TM5. Atmospheric Chemistry andPhysics 5(9), 2431–2460. doi:10.5194/acp-5-2431-2005.

Beswick, K. M., T. W. Simpson, D. Fowler, T. W. Choularton, M. W. Gallagher, K. J. Har-greaves, M. A. Sutton, and A. Kaye (1998). Methane emissions on large scales. Atmo-spheric Environment 32(19), 3283–3291.

Betts, A. K. (1992). FIFE atmospheric boundary layer budget methods. Journal of Geo-physical Research. 97, 18,523–18,531. doi:10.1029/91JD03172.

BUWAL (1996). Luftschadstoff-Emissionen aus natrlichen Quellen in der Schweiz.Schriftenreihe Umwelt – Luft 257, pp. 31.

Choularton, T. W., M. W. Gallagher, K. N. Bower, D. Fowler, M. Zahniser, A. Kaye, J. L.Monteith, and R. J. Harding (1995). Trace gas flux measurements at the landscape scaleusing boundary-layer budgets. Philosophical Transactions of the Royal Society: Physicaland Engineering Sciences 351(1696), 357–369. doi:10.1098/rsta.1995.0039.

Crutzen, P. J. (2002). Geology of mankind. Nature 415(6867), 23–23. doi:10.1038/415023a.

De Ligne, A., B. Heinesch, and M. Aubinet (2010). New transfer functions for correcting tur-bulent water vapour fluxes. Boundary-Layer Meteorology 137, 205–221. 10.1007/s10546-010-9525-9.

DelSontro, T., D. F. McGinnis, S. Sobek, I. Ostrovsky, and B. Wehrli (2010). Extrememethane emissions from a Swiss hydropower reservoir: Contribution from bubbling sedi-ments. Environmental Science & Technology 44(7), 2419–2425. doi:10.1021/es9031369.

Denmead, O. (2008). Approaches to measuring fluxes of methane and nitrous oxide betweenlandscapes and the atmosphere. Plant and Soil 309(1), 5–24. doi:10.1007/s11104-008-9599-z.

Denmead, O., R. Leuning, D. Griffith, and C. Meyer (1999). Some recent developments intrace gas flux measurement techniques. In A. Bouwman (Ed.), Approaches to Scaling ofTrace Gas Fluxes in Ecosystems, Volume 24, pp. 69–84. Elsevier.

10

1.8. References

Desai, A. R., D. J. P. Moore, W. K. M. Ahue, P. T. V. Wilkes, S. F. J. De Wekker, B. G.Brooks, T. L. Campos, B. B. Stephens, R. K. Monson, S. P. Burns, T. Quaife, S. M.Aulenbach, and D. S. Schimel (2011). Seasonal pattern of regional carbon balance in thecentral Rocky Mountains from surface and airborne measurements. Journal of Geophys-ical Research. 116(G4), G04009. doi:10.1029/2011JG001655.

Desjardins, R. L., J. I. MacPherson, L. Mahrt, P. Schuepp, E. Pattey, H. Neumann, D. Bal-docchi, S. Wofsy, D. Fitzjarrald, H. McCaughey, and D. W. Joiner (1997). Scaling upflux measurements for the boreal forest using aircraft-tower combinations. Journal ofGeophysical Research. 102(D24), 29125–29133. doi:10.1029/97JD00278.

Desjardins, R. L., J. I. Macpherson, H. Neumann, G. D. Hartog, and P. H. Schuepp (1995).Flux estimates of latent and sensible heat, carbon dioxide, and ozone using an aircraft-tower combination. Atmospheric Environment 29(21), 3147–3158. doi:10.1016/1352-2310(95)00007-L.

Desjardins, R. L., J. I. MacPherson, P. H. Schuepp, and F. Karanja (1989). An evaluationof aircraft flux measurements of CO2, water vapor and sensible heat. Boundary-LayerMeteorology 47, 55–69. doi:10.1007/BF00122322.

Detto, M., J. Verfaillie, F. Anderson, L. Xu, and D. Baldocchi (2011). Comparinglaser-based open- and closed-path gas analyzers to measure methane fluxes using theeddy covariance method. Agricultural and Forest Meteorology 151(10), 1312–1324.doi:10.1016/j.agrformet.2011.05.014.

Eugster, W., T. DelSontro, and S. Sobek (2011). Eddy covariance flux measurements confirmextreme CH4 emissions from a Swiss hydropower reservoir and resolve their short-termvariability. Biogeosciences 8(9), 2815–2831. doi:10.5194/bg-8-2815-2011.

Eugster, W. and P. Pluss (2010). A fault-tolerant eddy covariance system formeasuring CH4 fluxes. Agricultural and Forest Meteorology 150(6), 841–851.doi:10.1016/j.agrformet.2009.12.008.

Eugster, W. and W. Senn (1995). A cospectral correction model for measurement of turbulentNO2 flux. Boundary-Layer Meteorology 74(4), 321–340. doi:10.1007/BF00712375.

Finnigan, J. (2004). The footprint concept in complex terrain. Agricultural and ForestMeteorology 127(3-4), 117–129. doi:10.1016/j.agrformet.2004.07.008.

FOEN (2010). Switzerland’s greenhouse gas inventory 1990–2008. pp. 417.

Foken, T. and M. Leclerc (2004). Methods and limitations in validation offootprint models. Agricultural and Forest Meteorology 127(3-4), 223–234.doi:10.1016/j.agrformet.2004.07.015.

Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D. Fahey, J. Haywood, J. Lean,D. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz, and V. D. R. (2007).Changes in atmospheric constituents and in radiative forcing. In S. Solomon, D. Qin,

11


M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, H. L. Miller, and Eds (Eds.),Climate Change 2007: The Physical Science Basis. Contribution of Working Group I tothe Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp.129–234. Cambridge, United Kingdom and New York, NY, USA: Cambridge UniversityPress.

Gockede, M., T. Markkanen, M. Mauder, K. Arnold, J.-P. Leps, and T. Foken(2005). Validation of footprint models using natural tracer measurements froma field experiment. Agricultural and Forest Meteorology 135(1-4), 314–325.doi:10.1016/j.agrformet.2005.12.008.

Graber, W. K., B. Neininger, and M. Furger (1998). CO2 and water vapour exchangebetween an alpine ecosystem and the atmosphere. Environmental Modelling and Soft-ware 13(3-4), 353–360.

Hendriks, D., J. van Huissteden, and A. Dolman (2010). Multi-technique assessment ofspatial and temporal variability of methane fluxes in a peat meadow. Agricultural andForest Meteorology 150(6), 757–774. doi:10.1016/j.agrformet.2009.06.017.

Hiller, R., J. McFadden, and N. Kljun (2010). Interpreting CO2 fluxes over a suburbanlawn: The influence of traffic emissions. Boundary-Layer Meteorology 138, 215–230.doi:10.1007/s10546-010-9558-0.

Hiller, R., M. J. Zeeman, and W. Eugster (2008). Eddy-covariance flux measurements in thecomplex terrain of an Alpine valley in Switzerland. Boundary-Layer Meteorology 127(3),449–467. doi:10.1007/s10546-008-9267-0.

Horst, T. W. and J. C. Weil (1994). How far is far enough?: The fetch requirements formicrometeorological measurement of surface fluxes. JOURNAL OF ATMOSPHERICAND OCEANIC TECHNOLOGY 11(4), 1018–1025.

Ibrom, A., E. Dellwik, S. E. Larsen, and K. Pilegaard (2007). On the use ofthe Webb–Pearman–Leuning theory for closed-path eddy correlation measurements.TELLUS SERIES B-CHEMICAL AND PHYSICAL METEOROLOGY 59(5), 937–946.doi:10.1111/j.1600-0889.2007.00311.x.

IPCC (2000). Good practice guidance and uncertainty management in national greenhousegas inventories (IPCC GPG). Intergovernmental Panel on Climate Change.

IPCC (2006). 2006 IPCC Guidelines for national greenhouse gas inventories. Prepared bythe National Greenhouse Gas Inventories Programme, Eggleston H.S., Buendia L., MiwaK., Ngara T., and Tanabe K. (eds). IGES, Japan.

IPCC (2007). Summary for policymakers. In M. L. Parry, O. F. Canziani, J. P. Palutikof, P. J.van der Linden, and C. E. Hanson (Eds.), Climate Change 2007: The Physical ScienceBasis. Contribution of Working Group I to the Fourth Assessment Report of the Intergov-ernmental Panel on Climate Change, pp. 7–22. Cambridge, UK: Cambridge UniversityPress.

12

1.8. References

Klevenhusen, F., S. Bernasconi, M. Kreuzer, and C. Soliva (2010). Experimental vali-dation of the intergovernmental panel on climate change default values for ruminant-derived methane and its carbon-isotope signature. Anim. Prod. Sci. 50(3), 159–167.doi:10.1071/AN09112.

Kljun, N., P. Calanca, M. Rotach, and H. Schmid (2004). A simple parameterisa-tion for flux footprint predictions. Boundary-Layer Meteorology 112(3), 503–523.doi:10.1023/B:BOUN.0000030653.71031.96.

Kljun, N., M. Rotach, and H. Schmid (2002). A three-dimensional backward lagrangianfootprint model for a wide range of boundary-layer stratifications. Boundary-Layer Me-teorology 103(2), 205–226. doi:10.1023/A:1014556300021.

Kormann, R., H. Muller, and P. Werle (2001). Eddy flux measurements of methane overthe fen Murnauer Moos, 11ı11’E, 47ı39’N, using a fast tunable diode laser spectrometer.Atmospheric Environment 35(14), 2533–2544. doi:10.1016/S1352-2310(00)00424-6.

Kroon, P., A. Hensen, H. Jonker, H. Ouwersloot, A. Vermeulen, and F. Bosveld (2010).Uncertainties in eddy covariance flux measurements assessed from CH4 and N2O obser-vations. Agricultural and Forest Meteorology 150(6), 806–816.

Lamarque, J.-F., T. C. Bond, V. Eyring, C. Granier, A. Heil, Z. Klimont, D. Lee, C. Liousse,A. Mieville, B. Owen, M. G. Schultz, D. Shindell, S. J. Smith, E. Stehfest, J. Van Aar-denne, O. R. Cooper, M. Kainuma, N. Mahowald, J. R. McConnell, V. Naik, K. Riahi,and D. P. van Vuuren (2010). Historical (1850–2000) gridded anthropogenic and biomassburning emissions of reactive gases and aerosols: methodology and application. Atmo-spheric Chemistry and Physics 10(15), 7017–7039. doi:10.5194/acp-10-7017-2010.

Laubach, J. and H. Fritsch (2002). Convective boundary layer budgets derived from aircraftdata. Agricultural and Forest Meteorology 111(4), 237–263.

Le Treut, H., R. Somerville, U. Cubasch, Y. Ding, C. Mauritzen, A. Mokssit, T. Peterson,and M. Prather (2007). Historical overview of climate change. In S. Solomon, D. Qin,M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, H. L. Miller, and Eds (Eds.),Climate Change 2007: The Physical Science Basis. Contribution of Working Group I tothe Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp.93–127. Cambridge, United Kingdom and New York, NY, USA: Cambridge UniversityPress.

Leclerc, M. Y. and G. W. Thurtell (1990). Footprint prediction of scalar fluxes using a marko-vian analysis. Boundary-Layer Meteorology 52, 247–258. doi:10.1007/BF00122089.

Lehning, M., H. Richner, and G. L. Kok (1996). Pollutant transport over complex terrain:Flux and budget calculations for the POLLUMET field campaign. Atmospheric Environ-ment 30(17), 3027–3044. 10.1016/1352-2310(96)00007-6.

Lehning, M., H. Richner, G. L. Kok, and B. Neininger (1998). Vertical exchange andregional budgets of air pollutants over densely populated areas. Atmospheric Environ-ment 32(8), 1353–1363. doi:10.1016/S1352-2310(97)00249-5.

13


Leuning, R. (2005). Measurements of trace gas fluxes in the atmosphere using eddy co-variance: WPL corrections revisited. In X. Lee, W. Massman, B. Law, L. A. Mysak,and K. Hamilton (Eds.), Handbook of Micrometeorology, Volume 29 of Atmospheric andOceanographic Sciences Library, pp. 119–132. Springer Netherlands. doi:10.1007/1-4020-2265-4 6.

Leuning, R. (2007). The correct form of the Webb, Pearman and Leuning equation for eddyfluxes of trace gases in steady and non-steady state, horizontally homogeneous flows.Boundary-Layer Meteorology 123(2), 263–267.

Levin, I., T. Naegler, R. Heinz, D. Osusko, E. Cuevas, A. Engel, J. Ilmberger, R. L. Langen-felds, B. Neininger, C. v. Rohden, L. P. Steele, R. Weller, D. E. Worthy, and S. A. Zimov(2010). The global SF6 source inferred from long-term high precision atmospheric mea-surements and its comparison with emission inventories. Atmospheric Chemistry andPhysics 10(6), 2655–2662. doi:10.5194/acp-10-2655-2010.

Lin, J., D. Brunner, and C. Gerbig (2011). Studying atmospheric transport through la-grangian models. EOS Transactions 92, 177–178.

Livingston, G. and G. Hutchinson (1995). Enclosure-based measurement of trace gas ex-change: applications and sources of error. In P. Matson and H. R. C. (Eds.), Biogenictrace gases: measuring emissions from soil and water, pp. 14–51. Blackwell ScienceCambridge.

Luthi, D., M. Le Floch, B. Bereiter, T. Blunier, J.-M. Barnola, U. Siegenthaler, D. Raynaud,J. Jouzel, H. Fischer, K. Kawamura, and T. F. Stocker (2008). High-resolution carbondioxide concentration record 650,000–800,000 years before present. Nature 453(7193),379–382. doi:10.1038/nature06949.

Mahrt, L., J. I. Macpherson, and R. Desjardins (1994). Observations of fluxes over heteroge-neous surfaces. Boundary-Layer Meteorology 67(4), 345–367. doi:10.1007/BF00705438.

Mammarella, I., S. Launiainen, T. Gronholm, P. Keronen, J. Pumpanen, U. Rannik, andT. Vesala (2009). Relative humidity effect on the high-frequency attenuation of watervapor flux measured by a closed-path eddy covariance system. Journal of Atmosphericand Oceanic Technology 26(9), 1856–1866. doi:10.1175/2009JTECHA1179.1.

Mays, K. L., P. B. Shepson, B. H. Stirm, A. Karion, C. Sweeney, and K. R. Gurney (2009).Aircraft-based measurements of the carbon footprint of Indianapolis. Environmental Sci-ence & Technology 43(20), 7816–7823. doi:10.1021/es901326b.

Moore, C. J. (1986). Frequency response corrections for eddy correlation systems.Boundary-Layer Meteorology 37(1), 17–35. doi:10.1007/BF00122754.

Muhle, J., A. L. Ganesan, B. R. Miller, P. K. Salameh, C. M. Harth, B. R. Greally, M. Rigby,L. W. Porter, L. P. Steele, C. M. Trudinger, P. B. Krummel, S. O’Doherty, P. J. Fraser,P. G. Simmonds, R. G. Prinn, and R. F. Weiss (2010). Perfluorocarbons in the globalatmosphere: tetrafluoromethane, hexafluoroethane, and octafluoropropane. AtmosphericChemistry and Physics 10(11), 5145–5164. doi:10.5194/acp-10-5145-2010.

14

1.8. References

Nisbet, E. and R. Weiss (2010). Top-down versus bottom-up. Science 328(5983), 1241–1243. doi:10.1126/science.1189936.

Olivier, J., A. Bouwman, J. Berdowski, C. Veldt, J. Bloos, A. Visschedijk, C. van der Maas,and P. Zandveld (1999). Sectoral emission inventories of greenhouse gases for 1990 ona per country basis as well as on 1ı�1ı. Environmental Science & Policy 2(3), 241–263.doi:10.1016/S1462-9011(99)00027-1.

Pattey, E., I. B. Strachan, R. L. Desjardins, G. C. Edwards, D. Dow, and J. I. MacPherson(2006). Application of a tunable diode laser to the measurement of CH4 and N2O fluxesfrom field to landscape scale using several micrometeorological techniques. Agriculturaland Forest Meteorology 136(3-4), 222–236.

Peters, E. B., R. V. Hiller, and J. P. McFadden (2011). Seasonal contributions of vegeta-tion types to suburban evapotranspiration. Journal of Geophysical Research. 116(G1),G01003. doi:10.1029/2010JG001463.

Pumpanen, J., P. Kolari, H. Ilvesniemi, K. Minkkinen, T. Vesala, S. Niinisto, A. Lo-hila, T. Larmola, M. Morero, M. Pihlatie, I. Janssens, J. C. Yuste, J. M. Grunzweig,S. Reth, J.-A. Subke, K. Savage, W. Kutsch, G. Ostreng, W. Ziegler, P. Anthoni,A. Lindroth, and P. Hari (2004). Comparison of different chamber techniques formeasuring soil CO2 efflux. Agricultural and Forest Meteorology 123(3-4), 159–176.doi:10.1016/j.agrformet.2003.12.001.

Rebmann, C., M. Gockede, T. Foken, M. Aubinet, M. Aurela, P. Berbigier, C. Bern-hofer, N. Buchmann, A. Carrara, A. Cescatti, R. Ceulemans, R. Clement, J. A. Elbers,A. Granier, T. Grunwald, D. Guyon, K. Havrankova, B. Heinesch, A. Knohl, T. Lau-rila, B. Longdoz, B. Marcolla, T. Markkanen, F. Miglietta, J. Moncrieff, L. Montagnani,E. Moors, M. Nardino, J.-M. Ourcival, S. Rambal, U. Rannik, E. Rotenberg, P. Sedlak,G. Unterhuber, T. Vesala, and D. Yakir (2005). Quality analysis applied on eddy covari-ance measurements at complex forest sites using footprint modelling. Theoretical andApplied Climatology 80(2), 121–141. doi:10.1007/s00704-004-0095-y.

Rigby, M., R. G. Prinn, P. J. Fraser, P. G. Simmonds, R. L. Langenfelds, J. Huang,D. M. Cunnold, L. P. Steele, P. B. Krummel, R. F. Weiss, S. O’Doherty, P. K.Salameh, H. J. Wang, C. M. Harth, J. Muhle, and L. W. Porter (2008). Renewedgrowth of atmospheric methane. Geophysical Research Letters 35(L22805), L22805.doi:10.1029/2008GL036037.

Ritter, J. A., J. D. W. Barrick, G. W. Sachse, G. L. Gregory, M. A. Woerner, C. E. Watson,G. F. Hill, and J. E. Collins Jr (1992). Airborne flux measurements of trace species inan Arctic boundary layer. Journal of Geophysical Research. 97(D15), 16601–16625.doi:10.1029/92JD01812.

Ritter, J. A., J. D. W. Barrick, C. E. Watson, G. W. Sachse, G. L. Gregory, B. E. Anderson,M. A. Woerner, and J. E. Collins Jr (1994). Airborne boundary layer flux measurementsof trace species over Canadian boreal forest and northern wetland regions. Journal ofGeophysical Research. 99(D1), 1671–1685. doi:10.1029/93JD01859.

15


Schmid, H. P. (1994). Source areas for scalars and scalar fluxes. Boundary-Layer Meteorol-ogy 67(3), 293–318. doi:10.1007/BF00713146.

Schmid, H. P., H. A. Cleugh, C. S. B. Grimmond, and T. R. Oke (1991). Spatial vari-ability of energy fluxes in suburban terrain. Boundary-Layer Meteorology 54, 249–276.doi:10.1007/BF00183956.

Schuepp, P. H., M. Y. Leclerc, J. I. MacPherson, and R. L. Desjardins (1990). Footprintprediction of scalar fluxes from analytical solutions of the diffusion equation. Boundary-Layer Meteorology 50, 355–373. doi:10.1007/BF00120530.

Spahni, R., J. Chappellaz, T. F. Stocker, L. Loulergue, G. Hausammann, K. Kawamura,J. Fluckiger, J. Schwander, D. Raynaud, V. Masson-Delmotte, and J. Jouzel (2005). At-mospheric methane and nitrous oxide of the late pleistocene from antarctic ice cores.Science 310(5752), 1317–1321. doi:10.1126/science.1120132.

Stohl, A., P. Seibert, J. Arduini, S. Eckhardt, P. Fraser, B. R. Greally, C. Lunder, M. Maione,J. Muhle, S. O’Doherty, R. G. Prinn, S. Reimann, T. Saito, N. Schmidbauer, P. G. Sim-monds, M. K. Vollmer, R. F. Weiss, and Y. Yokouchi (2009). An analytical inversionmethod for determining regional and global emissions of greenhouse gases: Sensitivitystudies and application to halocarbons. Atmospheric Chemistry and Physics 9(5), 1597–1620. doi:10.1126/science.1120132.

Thornes, J. (2010). Atmospheric services. In Ecosystem Services, Volume 30, pp. 70–104.The Royal Society of Chemistry. doi:10.1039/9781849731058-00070.

Tuzson, B., R. V. Hiller, K. Zeyer, W. Eugster, A. Neftel, C. Ammann, and L. Emmenegger(2010). Field intercomparison of two optical analyzers for CH4 eddy covariance flux mea-surements. Atmospheric Measurement Techniques 3(6), 1519–1531. doi:10.5194/amt-3-1519-2010.

Vesala, T., N. Kljun, U. Rannik, J. Rinne, A. Sogachev, T. Markkanen, K. Sabelfeld, T. Fo-ken, and M. Leclerc (2010). Modelling of pollutants in complex environmental systems.In G. Hanrahan (Ed.), Flux and concentration footprint modeling, pp. 339–355. Advancedtopics in environmental sciences series, ILM Publications.

Webb, E., G. Pearman, and R. Leuning (1980). Correction of flux measurements for densityeffects due to heat and water vapour transfer. Quarterly Journal of the Royal Meteorolog-ical Society 106(447), 85–100.

Weiss, R. F., J. Muhle, P. K. Salameh, and C. M. Harth (2008). Nitrogen trifluoride in theglobal atmosphere. Geophys. Res. Lett. 35(20), L20821. doi:10.1029/2008GL035913.

Weiss, R. F. and R. G. Prinn (2011). Quantifying greenhouse-gas emissions from at-mospheric measurements: a critical reality check for climate legislation. Philosophi-cal Transactions of the Royal Society A: Mathematical,Physical and Engineering Sci-ences 369(1943), 1925–1942.

16

1.8. References

Wofsy, S. C., X. Zhang, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, andC. Heinze (2007). Couplings between changes in the climate system and biogeochemistry.In S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, andH. L. Miller (Eds.), Climate Change 2007: The Physical Science Basis. Contributionof Working Group I to the Fourth Assessment Report of the Intergovernmental Panel onClimate Change, pp. 499–587. Cambridge, United Kingdom and New York, NY, USA:Cambridge University Press.

Zeeman, M. J., R. Hiller, A. K. Gilgen, P. Michna, P. Pluss, N. Buchmann, and W. Eugster(2010). Management and climate impacts on net co2 fluxes and carbon budgets of threegrasslands along an elevational gradient in switzerland. Agricultural and Forest Meteo-rology 150(4), 519–530. doi:10.1016/j.agrformet.2010.01.011.

17

This manuscript is intended for submission to Atmospheric Environ-

ment

2Aircraft based CH4 flux estimates for an

agriculturally dominated valley inSwitzerland

Rebecca V. Hiller1, Bruno G. Neininger2, Dominik Brunner3,

Daniel Bretscher4, Thomas Kunzle5, Christoph Gerbig6,

Nina Buchmann1,Werner Eugster1

1Institute of Agricultural Sciences, ETH Zurich, Zurich, Switzerland2MetAir AG, airborne observations, airfield LSZN, Hausen am Albis, Switzerland

3Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Air

Pollution and Environmental Technology, Dubendorf, Switzerland4Agroscope Reckenholz-Tanikon Research Station ART, Zurich, Switzerland

5Meteotest, Bern, Switzerland6Max Planck Institute for Biogeochemistry, Jena, Germany

Abstract

Today’s greenhouse gas inventories are typically based on bottom-up estimates,but comparisons to atmospheric measurements are rare and can disagree by afactor of two or more. Therefore, top-down assessments to verify the com-monly used emission factors are needed. In Switzerland, almost 80% of thetotal methane emissions are attributed to the agricultural sector. To validate

19

Chapter 2: Aircraft based CH4 flux estimates

emission inventory estimates from this most relevant source in Switzerland, weperformed aircraft measurements in the Reuss Valley, which is situated in a widepre-alpine valley that is dominated by agriculture.

Flight legs were chosen to follow the terrain at constant heights of 50–400 mabove ground level. In general, a concentration gradient along the flight legs wasobserved. Fluxes were calculated by a simplified model approach and the eddycovariance method. Compared to the inventory that includes anthropogenic pluswetland emissions (0.41 µg CH4 m�2 s�1), the eddy covariance method yieldedslightly lower fluxes (0.30 µg CH4 m�2 s�1) whereas the box model fluxes werehigher (0.89 µg CH4 m�2 s�1).

2.1. Introduction

Inventories are a widely used tool to assess greenhouse gas (GHG) budgets of a specifiedarea. In policy, inventories serve to track the development of GHG emissions and to checkthe progress of the undertaken measures to reduce GHG emissions. All Annex I KyotoProtocol Parties have to report their current emissions to the United Nations FrameworkConvention on Climate Change (UNFCCC). In science, spatially explicit inventories are ofmost interest. Climate models that are used to produce climate scenarios rely on the inputof GHG emissions (Lamarque et al., 2010). Recently, the creditability of inventories wasquestioned by Nisbet and Weiss (2010). They stated that direct comparisons of the inventoryagainst measurements are rare and sometimes disagree by factor of two or three (Levin et al.,2010; Muhle et al., 2010; Bergamaschi et al., 2009; Stohl et al., 2009; Weiss et al., 2008;Bergamaschi et al., 2005).

While many investigations were accomplished for CO2, e.g. within the framework ofthe international programs CarboEurope (http://www.carboeurope.org/) and Fluxnet(http://www.fluxnet.ornl.gov/fluxnet/index.cfm), the focus lately shifted to non-CO2 GHGs. Methane is the second most important anthropogenic GHG after CO2. Eventhough the residence time of CH4 in the atmosphere is only about one tenth compared toCO2, its warming potential is 25 times larger at a 100-yr time horizon (Forster et al., 2007).On the global scale, wetlands are the most important source (30%), followed by agriculture(rice cultivation 9% and ruminants 15%). This distribution is shifted towards the agriculturalsector in Switzerland (79% of Swiss emissions) where ruminants emit the most CH4.

Recent developments in laser spectroscopy resulted in stable, robust, and relatively lightinstruments without liquid nitrogen cooling of the detector. These instrument are also suit-able for aircraft deployment. While many studies on airborne in situ measurements of CO2

20

2.2. Methods

and ozone in the planetary boundary layer were published (Graber et al., 1998; Lehninget al., 1998; Barr et al., 1997; Desjardins et al., 1997, 1995; Mahrt et al., 1994), to ourknowledge only three flux studies exist for CH4. Ritter et al. (1992) presented CH4 eddy co-variance measurements for the Arctic in Alaska and over Canadian boreal forest and northernwetland regions (Ritter et al., 1994). Mays et al. (2009) estimated the carbon footprint ofIndianapolis, USA. The fluxes were calculated from the urban plume intensively sampled inthe lee of the city. Other airborne CH4 studies involved flask sampling (Choularton et al.,1995) or conditional eddy correlation, where the air is sampled into different containers de-pending on up-draft or down-draft of the air (Pattey et al., 2006; Beswick et al., 1998). Allapproaches demonstrate the applicability of aircraft measurements to derive regional scalefluxes.

In this study, we present the first airborne CH4 flux estimates for an agriculturally dom-inated valley calculated with a boundary layer budget approach and the eddy covariancemethod. Flights took place on 16 days in 2009 and 2010. The measured fluxes are comparedwith a spatially explicit CH4 emission inventory accounting for 90% of the anthropogenicemissions as well as wetland emissions, including lakes. The limitations of the inventorybased approach as well as of the measured fluxes are discussed.

2.2. Methods

2.2.1. Site and flight pattern

The Reuss Valley is situated in central Switzerland at the southern border of the SwissPlateau (see Figure 2.1). Before the straightening of the Reuss river in the middle of the19th century, meanders winded through the wide valley and the plain was flooded on a regu-lar basis. The river correction prevents from flooding and combined with drainage made theland suitable for agriculture (Eugster, 1994). The national soil aptitude map classifies thearea into suitable to very suitable for fodder and suitable for crop production (soil aptitudemap FSO GEOSTAT, 1980). Today, 74% of the land in our study area is used for agricultureand 18% is covered by forests (CORINE FSO GEOSTAT, 1990).

During fair weather conditions, the valley wind system of the Alpine valleys penetratesinto the Alpine foreland and controls the local wind system. During the day, air movestowards the Alps (NNW winds) whereas during the night, cold air drainage flow prevails(SSE winds). Approximately 14 km long flight legs were flown along the valley axis atconstant flight levels between 50 m and 500 m a.g.l. to sample the air as it travels along thevalley. On 16 flight days, the flight pattern was repeated two to three times per day, namely

21


in the late morning, around noon, and in the afternoon.

The aircraft measurements were complemented with ground based energy flux as wellas micrometeorological observations at the ETH research station Chamau (8ı 240 3800 E,47ı 120 3700N, 400 m a.s.l.), situated at the southern end of the flight leg. Air temperatureand relative humidity (Tair and rh, 2 m, HygroClip S3, Rotronic AG, Basserdorf, Switzer-land), soil temperature (Tsoil, –15 cm, TB107, Markasub AG, Olten, Switzerland), soilwater content (swc, –15 cm, ML2x, Delta-T Devices Ltd., Cambridge, United Kingdom),wind direction (wdir 2 m, W200P, Vector Instruments, Rhyl, UK), and wind speed (u, 2 m,A100R, Vector Instruments, Rhyl, UK) were measured. A clear sky factor was introducedthat represents the ratio between the measured incoming shortwave radiation (2 m, CNR1,Kipp & Zonen B.V., Delft, The Netherlands) and the maximal expected incoming shortwaveradiation (Allen, 1996) in percent. The higher the value, the less clouds retarded the shortwave radiation to reach the surface. A more detailed description of the measurement systemcan be found in Zeeman et al. (2010). The planetary boundary layer height hPBL was deter-mined visually from the aircraft measurements of a flown profile for each day. The growthrate was estimated from the sensible heat flux measured on the ground (Lyra et al., 1992) andadded or subtracted from the profile determined height to derive the diurnal cycle of hPBL

growth throughout the day. Table 2.1 summarizes the weather conditions for the individualflight days.

2.2.2. Aircraft measurements

Aircraft measurements were performed on fair weather days from March to September in2009 and 2010. This selection of the flight days assured a well-mixed boundary layer. Thesmall research aircraft (METAIR-DIMO) of the type DIMO HK36 TTC-ECO (Diamondaircrafts, Wiener-Neustadt, Austria) was operated by MetAir AG (meteorological airborneobservations measurements and expertise, airfield LSZN, Hausen am Albis, Switzerland).The measurement instruments were situated in underwingpods and in the cockpit. Generalmeteorological variables as air temperature, atmospheric pressure, 3D turbulence and tracegas concentrations of CO2 and CO were measured continuously during the flight as part ofthe standard measurements performed with the METAIR-DIMO (see Neininger et al., 2001,for a detailed description of the aircraft and the instrumentation).

Methane

Additionally, CH4 concentrations were measured with a fast methane analyzer (FMA, LosGatos Research Inc., Mountain View, CA, USA) at 5 Hz. A commercially available instru-

22

2.2. Methods

ment was modified to weigh less and fit into one of the underwingpods. The original casewas replaced by an isolated aluminum box to minimize temperature fluctuations in the in-strument during the flight. The internal pump of the instrument was replaced to increasethe flow rate. A membrane pump (Vacuubrand MZ2C Vario SP, Vacuubrand gmbh + cokg, Wertheim, Germany) instead drew the air from the inlet through the instrument. Thepump was regulated to keep the cell pressure of the FMA in the automatically regulatedrange, independent of the varying atmospheric pressure between the different flight (Schnei-der, 2009). A 1/2” PETE tube through wing connected the instrument in the pod with thepump in the cockpit. The inlet outside the pod, a 33 cm long and 6 mm inner diametertube (Synflex-1300, Eaton Performance Plastics, Cleveland, OH, USA) was followed by aparticle filter (SMC, Japan, model AF20-F03 with 0.3 µm filter) including droplet separatorto prevent liquid water to enter the measurement system. The original instrument itself con-tained an internal 2 µm filter (SS-4FW4-2, Swagelok, Solon, OH, USA), which was kept inour modified version.

Flasks

Complementary to the continuous measurements, 4–17 grab samples were taken in 1 literglass flasks throughout each flight day. The air was analyzed for concentrations of CH4, CO,CO2, N2O and SF6 in the GasLab at the Max Planck Institute for Biogeochemistry in Jena,Germany. The CH4 flask concentration was compared against the continuous measurementsby the FMA. Since the ambient air was not dried before sampling, dry mole fraction of CH4

were calculated including the correction for spectroscopic water interferences (Hiller et al.,2012) based on the water measurements of the LI-7200 (Li-Cor Inc., Lincoln, NB, USA).The 1 Hz data were thereafter averaged over the flask flushing and filling period using theweighting function proposed by Chen et al. (2012). Only flasks that did not show unusualpressure fluctuations during the filling and flushing period were included in the comparison.Additionally, flasks, where the averaged continuous measurements deviated by more than˙25 ppb from the flask concentration, were dismissed. The outliers were not consecutiveand therefore we assumed that some problems occurred in the process of flask sampling,storage, and analysis and the in-situ measurements are of good quality.

2.2.3. Flux calculations

The high frequency measurements of CH4 and the wind vector allowed the determination ofthe persistent fluxes in the Reuss Valley by two different methods, (1) a simplified boundarylayer budget approach that relies on the advective component of the CH4 signal and (2) the

23


eddy covariance method that incorporates the turbulent fluctuations of the signal.

Boundary layer budget approach

Simple one-box models were used in many urban air pollution studies to estimate atmo-spheric trace gas or pollutant emissions from a known source as a function of time (Arya,1999; Oke, 1987; Hanna et al., 1982). We employed a similar model to estimate the CH4

emissions from the Reuss Valley. The area of interest was therefore enclosed by an imag-inary box and all relevant flows into and out of the box were quantified (see Figure 2.2).The box height was defined equal to the boundary layer and hence the box includes the wellmixed part of the atmosphere. Exchange with the air above the boundary layer was consid-ered to be small compared to the major fluxes and was thus neglected. The long axis of thebox was placed along the main valley axis and hence followed the main flow in the valleyduring days with a pronounced valley wind system. The flow through sidewalls was assumedto be small and thus negligible in our considerations. The only considered faces of the boxwere the front and end walls as well as the surface area. The fluxes through the front andend wall (Fin and Fout) were defined by the area (b�hPBL Œm�2�) times the mean wind speed( Nu Œms�1�) and the mean concentration ( N�CH4,in, N�CH4,out Œµgm�3�) at the respective faces. Themean wind speed did not show a clear height dependency and hence the average wind speedin direction of the valley axis was used for the flux calculations. The flux through the surfacearea Fsource was defined by the area (a � b) times the source strength (S ).

Assuming steady state (no accumulation of methane inside the box), mass conservationrequires that the outflow equals the sum of the inflow plus the total flux through the surface.Hence,

Fin C Fsource D Fout (2.1)

b hPBL Nu N�in C a b S D b hPBL Nu N�out: (2.2)

Solved for S , the equation yields:

S D hPBL NuN�out � N�in

a: (2.3)

For homogeneous sources, a linear concentration increase along the flight leg is expected.Hence, Ncout�Ncin

ais equal to the observed linear concentration increase along the transect ( dc

dx)

and is replaced therewith in the flux calculations.

24

2.2. Methods

Eddy covariance fluxes

Fluxes for the individual 14 km flight legs were calculated with the eddy covariance methodfrom the 1 Hz aircraft data. Before the flux calculation, trends in the /methane concentrationwere removed by removing the linear trend along the flown leg. Thereafter, the flux wascalculated as FCH4

D< w0c0 > where w0 are the deviations from the mean vertical windspeed and c0 deviations from the detrended mean CH4 concentration. The brackets indicateaveraging over one flight leg.

Quality control

Fluxes were quality flagged according the following criteria. Values were rejected when thestandard deviation of the wind direction was larger than 50ı. The linearity of the concen-tration gradient is for both methods of concern. In the simplified box model, the observedgradient is used to derive the flux whereas the time series are linearly detrended for the eddycovariance flux calculations. Hence, all records with standard errors of the slope estimatingthe concentration gradient greater than 0.11 ng CH4 m�1 were dismissed.

2.2.4. Spatially explicit CH4 inventory

To compare the emissions reported in the National Inventory Report (NIR, FOEN, 2010)with measurements, CH4 sources were spatially distributed over Switzerland to a 500 m grid.In 2008, total anthropogenic CH4 emissions in Switzerland were estimated to be 165’000 tCH4. The report lists about 620 different CH4 sources. However, 98% of the total emissionswere caused by only 25 processes. The 8 most dominant processes that contribute more than2.5% to the total include 90% of all emissions and are included in the spatially disaggregatedinventory. These processes are in the order of their contribution: enteric fermentation ofdairy cattle (43.5%), of young cattle (16%), manure of dairy cattle (9.5%), landfills (6%),grid losses in gas distribution (5%), enteric fermentation of non-dairy cattle (3.5%), manureof swine (3.5%), and enteric fermentation of sheep (2.5%).

Emissions from agriculture were estimated with a bottom-up approach. The livestockquantities from the agricultural establishment census (BFS 2009) were multiplied with thecorresponding emission factor from the Swiss national air pollution data base (EMIS) ofthe Federal Office of the Environment (FOEN). The emissions were spatially distributedaccording to the national land use statistics (BFS 1992/97) with the implications that 80%of the emissions for cattle, 20% for sheep and 100% for swine were produced in the stablewhereas the rest was distributed on the meadows. The agricultural establishment census also

25


reports the number of cattle that was moved to the Alps for summer grazing. Assuming asummer grazing period of 90 days, 4% of the total agricultural CH4 emissions were allocatedon Alpine meadows.

Grid losses from gas distribution were also calculated according to the factors reportedin the emission database FOEN. No CH4 escapes from the welded high pressure pipes.Leakage however occurs in the low pressure distribution network. The national emissionswere proportionally distributed to those 1 ha grids with more than two houses that heatwith gas (National building and apartment survey, BFS 2010). The result was thereafteraggregated to a 500 m raster.

Landfill emissions were estimated by a bottom-up approach with national factors. Thenational emissions were proportionally distributed according the land use statistics (BFS1992/97) since no additional information about the characteristics of the individual landfillswas available.

The wetland emissions CH4 inventory by Hobi (2011) were added to the anthropogenicemissions to account for the most important natural CH4 source in Switzerland. The wetlandinventory includes emissions from mires, marshes, lakes, and rivers following the approachby (Saarnio et al., 2009).

2.2.5. Footprint calculations

The Lagrangian Particle Dispersion Model COSMO-FLEXPART (Henne et al., 2010; Stohlet al., 2005) was used to calculate backward trajectories of the last 60 hours. 50’000 particleswere released every 3 minutes along the flight track. The residence time within the lowest500 m above ground were summed for each grid point in the 2.2�2.2 km raster. Thesespatial distribution maps are called footprints in the following.

To get a better idea about the source area responsible for the observed changes in CH4

concentration along the transects in the Reuss Valley, a footprint at the end of the transectwas subtracted from one at the beginning of the transect. The difference in residence timeresults in a positive and negative contribution area that indicates the area influencing thechanges in measured concentrations. The coordinates of the two areas were averaged andused as an indication for the footprint location.

26

2.3. Results

2.3. Results

2.3.1. Performance of the methane analyzer

The FMA performed excellent in the aircraft (see Fig. 2.3). A linear regression betweenthe flask concentrations and the continuous measurements resulted in ŒCH4�f lask=.�39 ˙10/ C .0:978 ˙ 0:005/ � ŒCH4�FMA with an R2 D 0:9949. On average, the continuousmeasurements are 4.8 ppb lower than the flask reference with a standard deviation of 5.3ppb.

In a next step, we compared the difference between the continuous and flask CH4 con-centration against the environmental variables relative humidity, water content of the air,atmospheric pressure, and air temperature, as well as the time to account for time trendsand the date for the inter day performance of the FMA. The differences can be explainedby inter diurnal variations (15.5%), time trend (6.7%), and water vapor (3.4%). The contri-bution of pressure, relative humidity, and air temperature was insignificant (p < 0:1) andcontributed less than 2% to the explained variance of total 27.2%. However, these numbershave to be contrasted against the very good agreement between continuous measurementsand flask CH4 concentrations. Hence, the listed variables influence the measured continuousCH4 concentration only marginally.

For the following analysis, only relative CH4 concentrations are used and therefore theobserved offset is irrelevant for the calculation of the fluxes. The error introduced by theslope is relatively small compared to the other measurement uncertainties and therefore nocorrection was applied.

2.3.2. Methane fluxes

Flux estimates from 108 transects which were classified as good quality data represent 13measurement days in 2009 and 2010. Since the data were not normally distributed, the non-parametric Wilcoxon test was applied. Eddy covariance fluxes (median=0.38 µg CH4 m�2 s�1,mean=0.49 µg CH4 m�2 s�1) were not significantly different from the inventory bases fluxestimate (mean=0.41 µg CH4 m�2 s�1) for the Reuss Valley (p=0.3), while the box modelyields significantly higher fluxes (median=1.01 µg CH4 m�2 s�1, mean=1.44 µg CH4 m�2 s�1)than the inventory and the eddy covariance method (both p<0.0001). Restricting the datato conditions when the mean wind followed the main box axis did not shift the distributiontowards higher or lower values. For the eddy covariance (p=0.6) and for the box modelderived fluxes (p=0.8), no significant difference to the previous selection was found. The

27


shape of the probability density function for the eddy covariance fluxes changed marginally,whereas the distribution was less well defined and got wider for the box model fluxes (seeFigure 2.4). Eddy covariance fluxes scattered less (sd=0.65 µg CH4 m�2 s�1) compared tothe box model estimates (sd=1.86 µg CH4 m�2 s�1).

The scatterplot summarizing the fluxes for the individual flight days supports the highervariation in the box model based fluxes (see Figure 2.5). While the eddy covariance valuesclosely fluctuated around the inventory based estimate, being often slightly lower than theinventory, more scattering was observed for the box model fluxes. Three measurement daysclearly separated from the general cloud with substantially higher values resulting from thebox model (see Figure 2.5). For these days, eddy covariance fluxes were also slightly higher,but could not compete with box model fluxes. On the other hand, the eddy covariancemethod also resulted in higher than average values on two days. On one of those days, thebox model yielded an estimate that is very close to the inventory, but shows large variationsto higher fluxes. Box model fluxes for the second day were even higher than the eddycovariance fluxes.

The eddy covariance fluxes of the individual flight days were not significantly differentfrom each other (ANOVA, p=0.15), whereas for the box model, fluxes tended to vary amongthe days (p<0.015). Hence, a correlation analyses including the following environmentalvariables measured at the ETH research station Chamau was performed: Cloudiness in-dex, incoming shortwave radiation (SWin), air temperature (Tair), soil temperature (Tsoil),soil water content (swc), relative humidity (rh), water vapor pressure (e), wind speed (u),footprint location in west-east direction (CHx(footprint)), footprint location in south-northdirection (CHy(footprint)), hour of day (hour), and day in year (doy). To also detect nonlin-ear dependencies, the Spearman’s rank correlation coefficient was chosen. Since the winddirection changed between the different overflights on some flight days, the analysis wasimplemented on the values for the individual overflights. For the eddy covariance fluxes,highest correlation was found for the shortwave incoming radiation, followed by the loca-tion of the footprint (see Table 2.2). The box model fluxes yielded highest correlation withthe wind speed, followed by shortwave incoming radiation. While the footprint locationseemed to be less important for the box model fluxes than for the eddy covariance fluxes,the sign of the correlation coefficient for changes in the footprint location in the west-eastdirection was opposite.

28

2.4. Discussion

2.3.3. Footprint location sensitivity

Swiss methane emissions source mostly from agriculture and hence highest emissions arefound in agriculturally dominated regions with livestock farming. This regions are situatedin the southern part of the Swiss Plateau that is located north of the Alps and south of the Juramountains (see Figure 2.1). The northern part of the Swiss Plateau is more densely populatedand consequently less land is available for agriculture, which transfers into the methaneemission distribution. The southern part of the Swiss Plateau also receives more orographicprecipitation due to its proximity to the Alps. Due to the relatively wet conditions, it isless suited for vegetable and crop production, which is mainly cultivated in the northernand western part of the Swiss Plateau. Within the Alps, agriculture is found in the valleyfloors. During the summer month, part of the livestock is moved to the Alpine meadowsfor grazing to save the resources in the valley for the winter. This practice is part of theSwiss traditional three-stage farming system and therefore methane emissions can also befound in very loosely populated areas of the Alps. In order to produce the spatially explicitinventory, geospatial information about the source locations were used. However, someassumptions were made as e.g. livestock numbers were only available at community level.To assess the spatial variability of the inventory, the following experiment was performed:The mean flux within a rectangle covering the Reuss Valley was calculated. This rectanglewas moved 60 km in 1 km incremental steps from south to north and 100 km from west toeast (see Figure 2.6). In both directions, the Reuss Valley corresponds to the location withhighest methane emissions. Only south-west of the Reuss Valley, even higher emissionsare expected. Most emissions source from agriculture. In urbanized areas, contributionsby methane escaping from natural gas distribution networks increase, but never exceed theagricultural emissions. Lakes also add up a considerable amount of methane if most of therectangle falls on water bodies. Variations in north-south directions are higher than in west-east, which can be explained by more pronounced land use changes with higher populatedareas in the northern part of the represented area.

2.4. Discussion

2.4.1. Uncertainties of the methane inventory

Since the inventory based fluxes serve as reference for evaluating the aircraft based esti-mates, it is useful to quantify the uncertainty of the inventory values. We limit the discussionto emissions from the agricultural sector, which largely dominate in the Reuss Valley.

29


The uncertainty for the agricultural sector has been estimated based on the approach usedin the Swiss national greenhouse gas inventory (ART, 2008). Applying a simple Gaussianerror propagation approach the uncertainty of the combined emission factor for enteric fer-mentation and manure management lies in the range of ˙17% (95% confidence interval).This corresponds very well to the default value of 20% suggested in the IPCC Guidelines(IPCC, 1997, 2000, 2006). The methane conversion rate for enteric fermentation (Ym) andthe methane conversion factor for manure management (MCF) are the most significant con-tributions to the overall uncertainty. The current Swiss methane emission inventory relieson the IPCC (2000) default conversion rates for Ym and MCF. Recent studies indicated thatthese values are not fully appropriate for Switzerland (Zeitz et al., 2012). They showed thatthe methane conversion rate depends on the diet. Swiss diet contains less feed concentratecompared to the IPCC propagated mitigation and hence Ym should be somewhat higherthan the default value but would still lie within the given uncertainty range. Only for forage-based fattening of bulls a value of Ym lower than proposed by IPCC was found (Staerfl et al.,2012). Overall, methane emissions from enteric fermentation are supposed to be higher thancurrently reported in the national greenhouse gas inventory that is based on the IPCC (2000)default values (Ym = 6%) (Zeitz et al., 2012). Notably, also the new IPCC (2006) guidelinessuggest a higher Ym of 6.5% for all cattle categories. Hence, adopting the new guidelineswould lead to almost 10% higher emissions than at present.

For the manure management methane conversion factor in contrast, much lower valueswere found by Zeitz et al. (2012). One main reason is the considerably lower storage tem-peratures than those currently used in the inventory, especially in winter. A first analysissuggests that applying country specific Ym and MCF would result in a compensation ofthe higher emissions from enteric fermentation by lower emissions form manure manage-ment (Zeitz et al., 2012). However, further evaluations are needed to determine the exactimplications of the country specific values.

Uncertainty estimates of livestock statistics used in the bottom-up approach lie in the or-der of 6% (ART, 2008). However, uncertainties in the spatial distribution of livestock andmanure management facilities also need to be considered. The confidence interval of theagricultural CH4 emissions along the north south and west east transects (see Section 2.3.3)is approximately mean˙ 5%. The uncertainty in the spatial distribution is even lower sincethe inventory variation also includes real regional emission differences. Overall uncertaintyof the bottom-up approach has been estimated as 18.8% accounting for uncertainties of theemission factor (17%), livestock statistics (6.4%) and spatial allocation (5%). In accor-dance with good practice for national greenhouse gas inventories (IPCC, 2000) a normalprobability distribution is assumed since no evidence could be found that would suggest an

30

2.4. Discussion

alternative distribution.

2.4.2. Variations of the CH4 flux measurements

Box model fluxes from September 2009 were much higher than for all other measurementdays whereas eddy covariance fluxes were only slightly enhanced. On 8 and 30 September2009, the synoptic wind blew from northeast. This implied that the air arriving over theReuss Valley traveled over the region between the plain (rectangle in Figure 2.6) and LakeZurich. Here, CH4 emissions show a north-south gradient with higher emissions in the south.Such a gradient was also observed when moving the rectangle in Figure 2.6 northwards intothe less agriculturally dominated area. If the methane sources are not homogeneously dis-tributed and air travels perpendicular to the observed source strength gradient, this gradientis translated into the concentration of the air. In our case, we would expect higher CH4 con-centrations in the south than in the north, even before the air enters the imaginary box overthe Reuss Valley. Hence, the observed concentration gradient does not only represent theemissions within the box, it also includes the concentration gradient that developed beforethe air enters the box. This results in a steeper than expected gradient and thereby a over-estimation of the effective CH4 emissions. This might also explain the opposite correlationcoefficient for the footprint location in west-east direction and the CH4 flux estimated by thetwo methods. On 7 September 2009, only one of 19 flown transects was qualified as goodquality data. CO2, CO, and aerosols showed a similar pattern as CH4 along the box mainaxis. Hence, it is very likely that the observed gradient was not caused by emissions, but bymoving air masses. In the southern part of the box, the wind field was inhomogeneous andsupported this hypothesis. Therefore, the value should better be treated as an outlier. Eddycovariance fluxes were not affected by sidewards affected gradients as the CH4 concentrationtime series was linearly detrended before calculating the covariances. The enhanced eddyfluxes might hence represent actual fluxes.

Higher than usual eddy covariance CH4 fluxes were observed on 24 June and 2 July 2009as well as on 1 June 2010. On those days, the cloudiness index (Table 2.1) was lower thanon most other days. However, cloud cover alone should not affect the measurements. On24 June 2009, a strong synoptical northeasterly wind even dominated the normally observedvalley wind system over the Reuss Valley. The wind direction also explained the relativelyhigh CH4 fluxes estimated with the box model. On 2 July 2009, a strong and narrow CH4

peak (� 0.15 ppm enhancement) was observed along the flown transect. This peak is likelyincorporated in the calculated covariance and negatively affects the flux. Therefore, thefluxes were considered as outliers. The peak itself was visible in all flown transects for this

31


day and thus likely represented a local source. Since CO and CO2 did not peak along withCH4, combustion could be excluded as source. CH4 peaks were also observed along thetransects flown on 1 June 2010.

Removing the outliers resulted in significantly lower fluxes than observed by the inventoryfor the eddy covariance method (Wilcoxon test, p<0.02). Other studies also showed lowereddy covariance measurements from aircraft compared to tower fluxes (Desjardins et al.,1997, 1995). Desjardins et al. (1989) argues that short-term observations cannot resolvelong-lived convection eddies. These effects are less important in the surface boundary layer.The reason for comparable higher fluxes by the box model approach might result from theirsensitivity to sidewards advected concentration gradients. Hence, we expect the true flux torange somewhere between the fluxes from the two approaches.

2.5. Conclusion

To our knowledge, this is the first attempt to directly compare a spatially explicit CH4 inven-tory with regional scale flux measurements. We were able to show that aircraft based fluxestimates provide a useful tool to determine CH4 emissions rates from a relatively homoge-neous source. While eddy covariance based fluxes tended to underestimate the emissions,the variability and the mean in the box model derived fluxes were considerably higher. How-ever, sampling all sidewalls of the box would help to apply a more complete boundary layerbudget and certainly would improve the flux estimates by this method.

Acknowledgements We thank the MetAir crew (Moritz Isler, Lorenz Muller, Dave Oldani,and Boris Schneider, Yvonne Schwarz) for their great effort during the measurementcampaign, Silas Hobi, and Elke Hodson for their contribution to the spatially explicitCH4 inventory of Switzerland, Hans-Ruedi Wettstein and his team for their supportat the ETH Research station Chamau, and Susanne Burri her helpful comments onthe manuscript. This project was funded by the Competence Center Environment andSustainability of the ETH Domain (CCES).

2.6. References

Allen, R. (1996). Assessing integrity of weather data for reference evapotranspiration esti-mation. Journal of irrigation and drainage engineering 122(2), 97–106.

ART (2008). Uncertainty in agricultural CH4 and N2O emissions of Switzerland. Internaldocumentation by Bretscher, D. and Leifeld, J.,. Agroscope Reckenholz-Tanikon Re-search Station, Zurich.

32

2.6. References

Arya, S. (1999). Air pollution meteorology and dispersion. Oxford University Press, USA.pp. 310.

Barr, A. G., A. K. Betts, R. L. Desjardins, and J. I. MacPherson (1997). Compar-ison of regional surface fluxes from boundary-layer budgets and aircraft measure-ments above boreal forest. Journal of Geophysical Research. 102(D24), 29213–29218.doi:10.1029/97JD01104.

Bergamaschi, P., C. Frankenberg, J. F. Meirink, M. Krol, M. G. Villani, S. Houweling,F. Dentener, E. J. Dlugokencky, J. B. Miller, L. V. Gatti, A. Engel, and I. Levin (2009).Inverse modeling of global and regional CH4 emissions using SCIAMACHY satelliteretrievals. Journal of Geophysical Research. 114, D22301. doi:10.1029/2009JD012287.

Bergamaschi, P., M. Krol, F. Dentener, A. Vermeulen, F. Meinhardt, R. Graul, M. Ramonet,W. Peters, and E. J. Dlugokencky (2005). Inverse modelling of national and EuropeanCH4 emissions using the atmospheric zoom model TM5. Atmospheric Chemistry andPhysics 5(9), 2431–2460. doi:10.5194/acp-5-2431-2005.

Beswick, K. M., T. W. Simpson, D. Fowler, T. W. Choularton, M. W. Gallagher, K. J. Har-greaves, M. A. Sutton, and A. Kaye (1998). Methane emissions on large scales. Atmo-spheric Environment 32(19), 3283–3291.

Chen, H., J. Winderlich, C. Gerbig, K. Katrynski, A. Jordan, and M. Heimann (2012). Val-idation of routine continuous airborne CO2 observations near the Bialystok tall towers.Atmospheric Measurement Techniques 5(4), 873–889.

Choularton, T. W., M. W. Gallagher, K. N. Bower, D. Fowler, M. Zahniser, A. Kaye, J. L.Monteith, and R. J. Harding (1995). Trace gas flux measurements at the landscape scaleusing boundary-layer budgets. Philosophical Transactions of the Royal Society: Physicaland Engineering Sciences 351(1696), 357–369. doi:10.1098/rsta.1995.0039.

Desjardins, R. L., J. I. MacPherson, L. Mahrt, P. Schuepp, E. Pattey, H. Neumann, D. Bal-docchi, S. Wofsy, D. Fitzjarrald, H. McCaughey, and D. W. Joiner (1997). Scaling upflux measurements for the boreal forest using aircraft-tower combinations. Journal ofGeophysical Research. 102(D24), 29125–29133. doi:10.1029/97JD00278.

Desjardins, R. L., J. I. Macpherson, H. Neumann, G. D. Hartog, and P. H. Schuepp (1995).Flux estimates of latent and sensible heat, carbon dioxide, and ozone using an aircraft-tower combination. Atmospheric Environment 29(21), 3147–3158. doi:10.1016/1352-2310(95)00007-L.


Eugster, W. (1994). Mikrometeorologische Bestimmung des NO2-Flusses an der GrenzflacheBoden/Luft. Ph. D. thesis, Bern : Geographisches Institut der Universitat Bern.

33


FOEN (2010). Switzerland’s greenhouse gas inventory 1990–2008. pp. 417.

Forster, P., V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D. Fahey, J. Haywood, J. Lean,D. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga, M. Schulz, and V. D. R. (2007).Changes in atmospheric constituents and in radiative forcing. In S. Solomon, D. Qin,M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor, H. L. Miller, and Eds (Eds.),Climate Change 2007: The Physical Science Basis. Contribution of Working Group I tothe Fourth Assessment Report of the Intergovernmental Panel on Climate Change, pp.129–234. Cambridge, United Kingdom and New York, NY, USA: Cambridge UniversityPress.

FSO GEOSTAT (1980). Soil aptitude map.

FSO GEOSTAT (1990). Corine landuse map.

Graber, W. K., B. Neininger, and M. Furger (1998). CO2 and water vapour exchangebetween an alpine ecosystem and the atmosphere. Environmental Modelling and Soft-ware 13(3-4), 353–360.

Hanna, S. R., G. A. Briggs, and R. P. Hosker (1982). Handbook on Atmospheric Diffusion.Technical Informaiton Center U.S. Department of Energy. pp. 102.

Henne, S., D. Brunner, D. Folini, S. Solberg, J. Klausen, and B. Buchmann (2010). Assess-ment of parameters describing representativeness of air quality in-situ measurement sites.Atmospheric Chemistry and Physics 10(8), 3561–3581. doi:10.5194/acp-10-3561-2010.

Hiller, R. V., C. Zellweger, A. Knohl, and W. Eugster (2012). Flux correction for closed-pathlaser spectrometers without internal water vapor measurements. Atmospheric Measure-ment Techniques Discussions 5(1), 351–384.

Hobi, S. (2011). Spatial allocation of methane sources and sinks in Switzerland. Master’sthesis, ETH Zurich. pp. 76.

IPCC (1997). Greenhouse gas inventory: Reference manual – Revised 1996 IPCC guidelinesfor national greenhouse gas inventories. Intergovernmental Panel on Climate Change.

IPCC (2000). Good practice guidance and uncertainty management in national greenhousegas inventories (IPCC GPG). Intergovernmental Panel on Climate Change.

IPCC (2006). 2006 IPCC Guidelines for national greenhouse gas inventories. Prepared bythe National Greenhouse Gas Inventories Programme, Eggleston H.S., Buendia L., MiwaK., Ngara T., and Tanabe K. (eds). IGES, Japan.

Lamarque, J.-F., T. C. Bond, V. Eyring, C. Granier, A. Heil, Z. Klimont, D. Lee, C. Liousse,A. Mieville, B. Owen, M. G. Schultz, D. Shindell, S. J. Smith, E. Stehfest, J. Van Aar-denne, O. R. Cooper, M. Kainuma, N. Mahowald, J. R. McConnell, V. Naik, K. Riahi,and D. P. van Vuuren (2010). Historical (1850–2000) gridded anthropogenic and biomassburning emissions of reactive gases and aerosols: methodology and application. Atmo-spheric Chemistry and Physics 10(15), 7017–7039. doi:10.5194/acp-10-7017-2010.

34

2.6. References

Lehning, M., H. Richner, G. L. Kok, and B. Neininger (1998). Vertical exchange andregional budgets of air pollutants over densely populated areas. Atmospheric Environ-ment 32(8), 1353–1363. doi:10.1016/S1352-2310(97)00249-5.

Levin, I., T. Naegler, R. Heinz, D. Osusko, E. Cuevas, A. Engel, J. Ilmberger, R. L. Langen-felds, B. Neininger, C. v. Rohden, L. P. Steele, R. Weller, D. E. Worthy, and S. A. Zimov(2010). The global SF6 source inferred from long-term high precision atmospheric mea-surements and its comparison with emission inventories. Atmospheric Chemistry andPhysics 10(6), 2655–2662. doi:10.5194/acp-10-2655-2010.

Lyra, R., A. Druilhet, B. Benech, and C. B. Biona (1992). Dynamics above a dense equa-torial rain forest from the surface boundary layer to the free atmosphere. Journal ofGeophysical Research. 97(D12), 12953–12965.

Mahrt, L., J. I. Macpherson, and R. Desjardins (1994). Observations of fluxes over heteroge-neous surfaces. Boundary-Layer Meteorology 67(4), 345–367. doi:10.1007/BF00705438.

Mays, K. L., P. B. Shepson, B. H. Stirm, A. Karion, C. Sweeney, and K. R. Gurney (2009).Aircraft-based measurements of the carbon footprint of Indianapolis. Environmental Sci-ence & Technology 43(20), 7816–7823. doi:10.1021/es901326b.

MeteoSchweiz (2009). 2009 Annalen Annales Annali. pp. 160.

MeteoSchweiz (2010). 2010 Annalen Annales Annali. pp. 160.

Muhle, J., A. L. Ganesan, B. R. Miller, P. K. Salameh, C. M. Harth, B. R. Greally, M. Rigby,L. W. Porter, L. P. Steele, C. M. Trudinger, P. B. Krummel, S. O’Doherty, P. J. Fraser,P. G. Simmonds, R. G. Prinn, and R. F. Weiss (2010). Perfluorocarbons in the globalatmosphere: tetrafluoromethane, hexafluoroethane, and octafluoropropane. AtmosphericChemistry and Physics 10(11), 5145–5164. doi:10.5194/acp-10-5145-2010.

Neininger, B., W. Fuchs, M. Baeumle, A. Volz-Thomas, A. Prevot, and J. Dommen (2001).A small aircraft for more than just ozone: Metair’s ’DIMONA’ after ten years of evolvingdevelopments. In 11th Symposium on Meteorological Observations and Instrumentation.

Nisbet, E. and R. Weiss (2010). Top-down versus bottom-up. Science 328(5983), 1241–1243. doi:10.1126/science.1189936.

Oke, T. (1987). Boundary layer climates. London : Routledge. pp. 420.

Pattey, E., I. B. Strachan, R. L. Desjardins, G. C. Edwards, D. Dow, and J. I. MacPherson(2006). Application of a tunable diode laser to the measurement of CH4 and N2O fluxesfrom field to landscape scale using several micrometeorological techniques. Agriculturaland Forest Meteorology 136(3-4), 222–236.

Ritter, J. A., J. D. W. Barrick, G. W. Sachse, G. L. Gregory, M. A. Woerner, C. E. Watson,G. F. Hill, and J. E. Collins Jr (1992). Airborne flux measurements of trace species inan Arctic boundary layer. Journal of Geophysical Research. 97(D15), 16601–16625.doi:10.1029/92JD01812.

35


Ritter, J. A., J. D. W. Barrick, C. E. Watson, G. W. Sachse, G. L. Gregory, B. E. Anderson,M. A. Woerner, and J. E. Collins Jr (1994). Airborne boundary layer flux measurementsof trace species over Canadian boreal forest and northern wetland regions. Journal ofGeophysical Research. 99(D1), 1671–1685. doi:10.1029/93JD01859.

Saarnio, S., W. Winiwarter, and J. L. ao (2009). Methane release from wet-lands and watercourses in Europe. Atmospheric Environment 43(7), 1421–1429.doi:10.1016/j.atmosenv.2008.04.007.

Schneider, B. (2009). Regelung einer vakuumpumpe. Studienarbeit. Hochschule Karlsruhe.pp. 25.

Staerfl, S. M., J. O. Zeitz, M. Kreuzer, and C. R. Soliva (2012). Methane conversion rate ofbulls fattened on grass or maize silage as compared with the ipcc default values, and thelong-term methane mitigation efficiency of adding acacia tannin, garlic, maca and lupine.Agriculture, Ecosystems & Environment 148(0), 111 – 120.

Stohl, A., C. Forster, A. Frank, P. Seibert, and G. Wotawa (2005). Technical note: Thelagrangian particle dispersion model flexpart version 6.2. Atmospheric Chemistry andPhysics 5(9), 2461–2474. doi:10.5194/acp-5-2461-2005.

Stohl, A., P. Seibert, J. Arduini, S. Eckhardt, P. Fraser, B. R. Greally, C. Lunder, M. Maione,J. Muhle, S. O’Doherty, R. G. Prinn, S. Reimann, T. Saito, N. Schmidbauer, P. G. Sim-monds, M. K. Vollmer, R. F. Weiss, and Y. Yokouchi (2009). An analytical inversionmethod for determining regional and global emissions of greenhouse gases: Sensitivitystudies and application to halocarbons. Atmospheric Chemistry and Physics 9(5), 1597–1620. doi:10.1126/science.1120132.

Weiss, R. F., J. Muhle, P. K. Salameh, and C. M. Harth (2008). Nitrogen trifluoride in theglobal atmosphere. Geophys. Res. Lett. 35(20), L20821. doi:10.1029/2008GL035913.

Zeeman, M. J., R. Hiller, A. K. Gilgen, P. Michna, P. Pluss, N. Buchmann, and W. Eugster(2010). Management and climate impacts on net co2 fluxes and carbon budgets of threegrasslands along an elevational gradient in switzerland. Agricultural and Forest Meteo-rology 150(4), 519–530. doi:10.1016/j.agrformet.2010.01.011.

Zeitz, J., C. Soliva, and K. M. (2012). Swiss diet types for cattle: How accurately are theyreflected by the IPCC default values? Journal of Integrative Environmental Sciences,1–18. doi:10.1080/1943815X.2012.709253.

36

2.7. Figures and tables


Figure 2.1.: Spatially highly resolved CH4 inventory for the eight most important anthro-pogenic sources accounting for 90% of the anthropogenic emissions. Theseinclude agriculture, landfills, and gas distribution in Switzerland. The blackrectangle indicates the Reuss Valley where this inventory is cross-validated withdirect regional scale flux measurements from a small research aircraft. In thebottom-left corner, the box used for the flux calculation of is indicated in yellow(Map source: Google Earth, 2009). The red lines show the flight track from 7April 2010 and indicate the transects flown along the box main axis.

37


[CH4]

x

hPBL

a

b

Cin

Cout

u

Figure 2.2.: Scheme of the simplified box model. The imaginary box with a length a, widthb, and height hPBL encloses the air volume of interest. The large black arrowsindicate the methane flow in and out of the box that is forced by the wind speed u(black arrow) along the main box axis. The small arrows at the surface of the boxindicate a homogeneous CH4 source. As the air travels over the source throughthe box, the CH4 concentration increases gradually. This observed linear CH4

gradient along the box main axis x is indicated by the red line.

38


1800 1900 2000 2100 2200

1800

1900

2000

2100

2200

[CH4]FMA [ppb]

[CH

4]fla

sks

[ppb

]

[CH4]flasks = (39 ± 10) + (0.978 ± 0.005)[CH4]FMA

R2 = 0.994995% confidence interval1:1 relationship

●

●● ●

●

●●●

●●

●

●

●●●●

●

●●

●

●

●

●

●

●

● ●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●●

●●

●

●

●

●●

●

●

●

●●●●

●

●

●

●

●

●

●

●●

●

●

●

●

●

●●

●

●●

●●

●

●

●

●

●

●

●●

●

●●

●

●●

●

●

●●●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●

●●

●●

●●●●●●

●●

●

●

●

[CH4]flasks − [CH4]FMA [ppb]

Fre

quen

cy

−10 0 10 20

0

10

20

30

Figure 2.3.: Scatterplot of flask CH4 concentration and weight averaged continuous mea-surements of the FMA. The grey area represents the 95% confidence intervaland the solid line the found relationship of the linear model. The dashed lineindicates the 1:1 relationship. In the lower right corner, the histogram of thedifferences between continuous and flask CH4 measurements is shown.

39


−5

0

5

10

F CH

4 [μ

gC

H4

m−2

s−1]

EC Simplified box model

0 0.25 0.5 0.75 1f(x)

0 0.1 0.2 0.3f(x)

good qualitygood quality & valleyinventory

#108#37

13 days7 days

Figure 2.4.: Probability density function for the good quality fluxes (solid line) calculatedby the eddy covariance method (left) and the simplified box model (right). Inan additional filtering step, only records when the wind direction followed thevalley axis ˙15ı were selected (dashed line). The counts in the legend indicatethe number of records represented by the probability density function as well asthe number of days represented by these records.

40


0 1 2 3 4 5 6 7 8

0

1

2

FCH4,box [μg CH4 m−2 s−1]

F CH

4,EC [μ

gC

H4

m−2

s−1]

●●

●●

●●

●●

●

●

●●●

Figure 2.5.: Scatterplot between the CH4 fluxes obtained from the eddy covariance and thebox model method. The dots represent the median values of the individual flightdays whereas the crosses span the interquartile range of the measured fluxes.Points located in the grey areas show cases where both methods either under oroverestimate the inventory flux. The dashed line indicates the 1:1 relationship.

41


Figure 2.6.: Sensitivity of footprint location. The map is overprinted by the inventory foranthropogenic methane emissions, including sources from agriculture, landfills,and gas distribution. Map units are Swiss coordinates in meters. The rectanglein the middle of the map indicates the location of the box used for the flux cal-culations. This box was moved south-north and west-east along the indicatedlines to simulate the influence of the footprint location on the expected flux. Theleft bar shows the changes in methane emissions for the south-north direction.The total emissions are split into the categories agriculture, landfills, gas dis-tribution, and wetlands/lakes. The latter is not included in the anthropogenicemissions presented on the map. The boxplot next to the graph summarizesthe total emissions by indicating the median (solid line), the interquartile range(box), and the lowest and highest observed value (whiskers). The bottom barshows the emission variation in west-east direction and the boxplot next to itsummarizes the corresponding values. Basemap reproduced with the authoriza-tion of swisstopo (JA100120)

42


T abl

e2.

1.:W

eath

ersi

tuat

ion

(Met

eoSc

hwei

z,20

09,2

010)

and

obse

rved

met

eoro

logi

cal

vari

able

sat

the

field

stat

ion

Cha

mau

for

the

indi

vidu

alfli

ghtd

ays.

Val

ues

are

aver

aged

fort

hepe

riod

10:0

0–17

:00

CE

T,w

hen

the

fligh

tmea

sure

men

tsw

ere

perf

orm

ed.

The

boun

dary

laye

rhei

ghth

PBL

isth

ees

timat

eca

lcul

ated

acco

rdin

gSe

ctio

n2.

2.1

for1

5:00

CE

T.

Dat

eW

eate

rsitu

atio

nC

lear

sky

fact

orT

air

Tso

ilsw

crh

wdi

ru

hPB

Lat

15:0

0C

ET

[%]

[ıC

][ı

C]

[%]

[%]

[ı]

[ms�1]

[ma.

g.l.]

24Ju

ne20

09Fl

atpr

essu

repa

ttern

79

19

15:6

44

69

21

2:4

1130

02Ju

ly20

09Fl

atpr

essu

repa

ttern

91

29

18:6

42

50

320

1:8

900

06Ju

ly20

09L

owpr

essu

rem

oves

toSc

an-

dina

via

and

dire

cts

hum

id-

war

mai

rtow

ards

the

alps

58

23

19:4

44

67

216

1:9

510

13Ju

ly20

09H

otai

rfro

mN

W64

26

17:7

43

60

344

1:3

680

07Se

ptem

ber2

009

Hig

hpr

essu

re94

21

16:3

43

64

299

1:0

680

08Se

ptem

ber2

009

Hig

hpr

essu

re93

21

16:6

43

64

340

1:8

720

29Se

ptem

ber2

009

Rid

geof

high

pres

sure

over

Irel

and

95

21

15:7

35

68

294

1:0

720

30Se

ptem

ber2

009

Rid

geof

high

pres

sure

over

Irel

and

98

20

15:3

35

64

314

1:5

840

18M

arch

2010

Hig

hpr

essu

re90

13

4:3

44

37

273

1:0

910

06A

pril

2010

Hig

hpr

essu

re100

12

8:2

45

41

312

1:8

890

07A

pril

2010

Hig

hpr

essu

reov

erSc

andi

-na

vie,

wea

kfr

onto

verF

ranc

e100

16

8:2

45

45

281

1:1

770

01Ju

ne20

10H

umid

airf

rom

N51

15

14:6

45

58

61:4

680

04Ju

ne20

10H

igh

pres

sre

over

the

Nor

thSe

a96

21

14:4

45

55

41:8

970

19A

ugus

t201

0H

igh

pres

sure

deve

lopm

ent

62

20

17:8

44

72

356

1:0

650

20A

ugus

t201

0Fl

atpr

essu

repa

ttern

79

24

18:2

44

70

217

1:1

710

26A

ugus

t201

0R

idge

ofhi

ghpr

essu

refr

omSp

ain

toE

aste

rnE

urop

e95

27

20:0

44

44

149

1:2

710

43


Table 2.2.: Correlation matrix (Spearman’s rank correlation coefficiant) between the CH4

flux and selected environmental variables. Correlation coefficitions of 0.2 andhigher are printed in bold face.

FCH4;EC FCH4;box

cloudiness –0.08 0.16SWin 0.25 0.27Tair –0.04 –0.04Tsoil –0.10 –0.13swc –0.01 0.03rh 0.18 0.16e 0.06 0.04u 0.13 0.48CHx(footprint) –0.13 0.10CHy(footprint) 0.28 0.01hour –0.19 –0.04doy –0.10 –0.03

44

This manuscript will be submitted to a Scientific Journal

3Aircraft based CH4 emissions estimates

for a Swiss lowland hydropower reservoir

Rebecca V. Hiller1, Bruno G. Neininger2, Torsten Diem3,

Tonya DelSontro4,5, Werner Eugster1

1Institute of Agricultural Sciences, ETH Zurich, Zurich, Switzerland2MetAir AG, airborne observations, airfield LSZN, Hausen am Albis, Switzerland

3School of Geography and Geosciences, University of St Andrews, St Andrews, UK4Eawag, Swiss Federal Institute of Aquatic Science and Technology, Seestrasse 79, 6047

Kastanienbaum, Switzerland5Institute for Biogeochemistry and Pollutant Dynamics, ETH Zurich, Zurich, Switzerland

Abstract

Methane emissions from reservoirs blur the climate creditability of the renew-able hydroelectricity. Especially for large tropical reservoirs where biomassrich surfaces were flooded, the emitted methane is of concern. So far, emissionassessments were performed with chambers or the eddy covariance technique.Both methods cover only a part of the lake surface whereas aircraft based CH4

flux estimates integrate over larger areas. In this study, we conducted airbornewind speed and CH4 concentration measurements over a mid-latitudes run-off-river reservoir to quantify the CH4 emissions from the reservoir on September

45

Chapter 3: Aircraft based CH4 emissions estimates for a Swiss lowland hydropower reservoir

30 2009. For comparison reasons with earlier chamber based studies, threechambers were deployed on the lake on that day and sampled four times each.The airborne fluxes were determined with two different models: (1) The boxmodel approach assesses the advective flow through the individual sidewalls ofthe box and the imbalance is attributed to the CH4 emissions of the lake (19µg CH4 m�2 s�1 and 12 µg CH4 m�2 s�1 for the noon and afternoon overflight,respectively). (2) The simplified approach estimates the flux from mixing layerheight, the observed concentration gradient and the mean wind speed along val-ley axis (19 µg CH4 m�2 s�1 and 8 µg CH4 m�2 s�1 for the noon and afternoonoverflight, respectively). These fluxes compare well with eddy covariance andchamber measurements from earlier studies. However, the chamber fluxes at theflight day were about factor two lower than the airborne measurements.

3.1. Introduction

Climate creditability of hydroelectricity is currently a highly debated topic (Giles, 2006).Hydroelectricity is sold as renewable green energy (Hoffert et al., 1998; Victor, 1998). Sev-eral studies however show that reservoirs can be a considerable source of greenhouse gases,especially CH4 (Gueerin et al., 2006; DelSontro et al., 2010; Eugster et al., 2011; Teodoruet al., 2011). The emission strength depends on reservoir age and latitude (Barros et al.,2011). Even though Barros et al. (2011) estimated one order of magnitude lower emissionsthan had been previously reported by St. Louis et al. (2000), Wehrli (2011) concluded thathydropower is renewable, but not free of carbon emissions. Reservoirs emit about 48 Tg Cas CO2 and 3 Tg C as CH4 per year (Barros et al., 2011). Converted to CO2-equivalents,this yields 276 Tg CO2-eq per year. In 2008, 3.3 PWh electricity were produced by hy-dropower (International Energy Agency, 2008). This results in an emission factor of 0.08g CO2-eq Wh�1. However, the emission factor for hydroelectricity largely depends on var-ious variables describing the reservoir, such as temperature, the type of flooded ecosystem,depth, and age of reservoir (St. Louis et al., 2000).

Emissions from Swiss reservoirs are considerably lower than those observed in the tropicsdue to lower water temperatures and less available organic material for decomposition. LakeWohlen belongs to one of the most productive Swiss reservoirs in terms of greenhouse gasemissions (Diem et al., 2008). The lake is relatively shallow and situated on the SwissPlateau (481 m a.s.l.) whereas most Swiss reservoirs are located in an alpine environment.

Direct chamber flux measurements of Lake Wohlen indicate average CH4 emissions >150 mg CH4 m�2 d�1, the highest ever reported rates for a midlatitude reservoir (DelSon-

46

3.2. Methods

tro et al., 2010). Measurements by the eddy covariance method confirm the extremely highemissions (Eugster et al., 2011). Both methods however spatially cover only part of the lake.Up-scaling the observed flux to the total lake surface is possible by assuming that the mea-surements are representative for the total lake area. This introduces a considerable amountof uncertainty, especially for spatially variable fluxes such as CH4 (DelSontro, 2011). Thisshortcoming can be overcome by aircraft-based flux measurements that provide a relativelystable estimate at the regional scale (Ritter et al., 1994).

In the following, we assess the potential of aircraft-based flux estimates to evaluate thegreenhouse gas emissions of a reservoir. Several transects over the lake were flown threetimes on September 30 2009. The observed concentration patterns and concentration gradi-ents of CH4, CO, and CO2 are discussed. Two different box model approaches are appliedto derive CH4 fluxes. The results are compared with CH4 fluxes from chamber and eddycovariance measurement conducted at the same lake.

3.2. Methods

3.2.1. Aircraft

A small research aircraft (METAIR-DIMO) of the type DIMO HK36 TTC-ECO (Diamondaircrafts, Wiener-Neustadt, Austria) was equipped and operated by MetAir AG (meteoro-logical airborne observations measurements and expertise, airfield LSZN, Hausen am Albis,Switzerland). Environmental measurement instruments were operated in underwingpodsas well as in the cockpit. The standard instrumentation included various sensors to mea-sure meteorological variables such as air temperature, atmospheric pressure as well as tracegas concentration measurements of CO2 and CO (see Neininger et al. (2001) for a detaileddescription of the aircraft and the instrumentation). CO2 was measured by two redundantinfrared gas analyzers (modified LI-6262 and LI-7500, both Li-Cor, Lincoln, NB, USA). Amodified Aerolaser AL-5003 (Aero-Laser GmbH, Garmisch-Partenkirchen, Germany) re-ported CO concentrations. Atmospheric turbulence was measured by a five-hole pressuresonde (Crawford and Dobosy, 1992) mounted on a boom in front of the right pod. Theground based wind vector was derived from these measurements in combination with datafrom an INS/IRS (Inertial Navigation/Reference System, RT3102, Oxford Technical Solu-tions, Oxfordshire, UK).

CH4 was measured with a fast methane analyzer (FMA, Los Gatos Research Inc., Moun-tain View, CA, USA). A commercially available instrument was modified to weigh less andfit into one of the pods below the wings. The original case was replaced by an isolated alu-

47


minum box to minimize temperature fluctuations in the instrument during the flight. Theinternal pump of the instrument was removed. A membrane pump (Vacuubrand MZ2CVario SP, Vacuubrand gmbh + co kg, Wertheim, Germany) instead drew the air from theinlet through the instrument. A 1/2” PETE tube in the wing connected the instrument inthe pod with the pump in the cockpit. The inlet outside the pod, a 33 cm long and 6 mminner diameter tube (Synflex-1300, Eaton Performance Plastics, Cleveland, OH, USA) wasfollowed by a particle filter (SMC, Japan, model AF20-F03 with 0.3 µm filter) includingdroplet separator to prevent liquid water to enter the measurement system. The original in-strument itself contains an internal 2 µm filter (SS-4FW4-2, Swagelok, Solon, OH, USA),which was kept also in our modified version.

3.2.2. Site description

Lake Wohlen (46ı58’N/7ı19’E, 481 m a.s.l.) is a run-off-river reservoir that was artificiallycreated between 1917 and 1922 to use the water of the river Aare for hydroelectricity produc-tion. The lake is situated in a relatively narrow valley eroded by the river into the glaciallystructured Molasse Basin (Durst Stucki et al., 2010), close to the city of Bern. A 24 m talldam retains the water. The lake extends 12 km along the river with a maximal width of 650m and a maximum depth of 18 m. On average, the water remains 40 h in the lake (DelSontroet al., 2010). The lake is known for its comparably high CH4 emissions (Diem et al., 2008;DelSontro et al., 2010; Eugster et al., 2011).

3.2.3. Weather conditions

The ridge of a high-pressure area over Ireland dominated the weather in Switzerland from27 to 30 September 2009. Conditions were sunny and warm with local fog patches alongwater bodies in the morning (MeteoSchweiz, 2009). Temperatures in Bern, the closest me-teorological station to Lake Wohlen at 10 km to the northeast, ranged between 7.0 and 21.3ıC on 30 September (Source: MeteoSwiss). The region of Bern was fog free and no cloudsdeveloped during the course of the day. Over the Swiss Plateau, the synoptic wind prevailedfrom northeast on 30 September. The wind system in the research area was however dom-inated by the local valley wind system: down-valley winds from east were observed in themorning until up-valley winds from west took over.

Water temperature at the Schonau monitoring site, about 25 km upstream of the lake, were17 ıC and the flow rate reached 65 m3 s�1 (Swiss Federal Office of Environment). The watertemperature measured at the dam at 10 m depth was reported with 16 ıC by the hydroelectriccompany BKW.

48

3.2. Methods

3.2.4. Flight pattern

The wind system within the valley is dominated by the local valley wind system and thusthe main wind direction followed the valley axes. CH4 concentrations in an air parcel areexpected to increase as the parcel travels over a CH4 source, in our case the lake. Therefore,several transects from approx. 100 to 300 m a.g.l. were flown over the lake along the valleyaxis (see Figure 3.1). We assumed that the air in the valley was well mixed and the observedconcentration changes should hence represent the CH4 emissions in the valley. The sameflight pattern was repeated three times at 10:00, 13:00, and 14:30 CEST on 30 September2009.

3.2.5. Box model

The continuity equation for mass, momentum, moisture, heat, and a scalar quantity describesthe change of the property in space and time based on the conservation principle. For mass,the continuity equation can be written as (Arya, 1999, p. 34):

@�

@t„ƒ‚…Storage

[email protected]�/

@[email protected]�/

@[email protected]�/

@z„ ƒ‚ …Advection

D 0; (3.1)

where u, v, and w are the wind velocity components in the x, y, and z directions. Ifstationarity is assumed, the density of air � does not change over time t and hence @�

@tD 0.

In the continuity equation for scalar quantities c, an additional source/sink term S ac-counts for emissions and removal of the scalar within the considered air volume. Moleculardiffusion is also involved in the transportation process. Compared to the turbulent and advec-tive transport, this process is very ineffective and is thus neglected in further considerations,the continuity equation then yields (Stull, 1988, p. 80):

@c

@t„ƒ‚…Storage

[email protected]/

@[email protected]/

@[email protected]/

@z„ ƒ‚ …Advection

D S„ƒ‚…Source/Sink

(3.2)

As for the mass conservation (Equation 3.1), stationarity is assumed for the time periodof one overflight and hence the storage term can be neglected. The transport can be dividedinto the mean flow and the turbulent portion by applying Reynold’s decomposition (Arya,1999, p. 137):

S D@uc

@xC@u0c0

@xC@vc

@yC@v0c0

@yC@wc

@zC@w0c0

@z(3.3)

49


The continuity equation considers the air within an arbitrary small volume around a point.At the scale of a box volume, all points within the box however need consideration. To bephysically consistent, the overall integral over the box must also fulfill the mass conservation(Arya, 1999, p. 130). The integration of Equation 3.2 yields:R a

xD0

R byD0

R hPBL

zD0

�S�dxdydz D

R axD0

R byD0

R hPBL

zD0

�@uc@xC

@u0c0

@xC

@vc@yC

@v0c0

@yC

@wc@zC

@w 0c0

@z

�dxdydz ;

(3.4)

where a defines the box length, b the box width, and hPBL the box height (see Figure 3.2).

To estimate the average source/sink term of the total box, it is sufficient to only considerthe points on the faces of the box, assessing the quantities that go in and out of the box. Thesource is restricted to the box surface and is thus equal to zero at all other faces:

�R axD0

R byD0

�SzD0

�dxdy D

Z b

yD0

Z hPBL

zD0

�ucxDa C u0c0xDa

�dydz„ ƒ‚ …

Feast

�

Z b

yD0

Z hPBL

zD0

�ucxD0 C u0c0xD0

�dydz„ ƒ‚ …

Fwest

C

Z a

xD0

Z hPBL

zD0

�vcyDb C v0c0yDb

�dxdz„ ƒ‚ …

Fnorth

�

Z a

xD0

Z hPBL

zD0

�vcyD0 C v0c0yD0

�dxdz„ ƒ‚ …

Fsouth

C

Z a

xD0

Z b

yD0

�wczDhPBL C w

0c0zDhPBL

�dxdy„ ƒ‚ …

Ftop

�

Z a

xD0

Z b

yD0

�wczD0 C w0c0zD0

�dxdy„ ƒ‚ …

Fbottom

(3.5)

The aircraft measurements only specify the wind vector and CH4 concentration at certainpoints on individual faces of the box. The wind concentration product was calculated for thethree wind vectors for the individual measurement points. To get a best guess, this quantitywas averaged for the individual faces of the box. At the east and west face, all measurementswithin a 200 m wide slide parallel to the face were averaged. The north and south faces were

50

3.2. Methods

defined by the flown transects on the corresponding side of the box.

Further simplification of the equation apply for certain box faces during special conditions(Arya, 1999, p. 138f): (1) The turbulent flux in direction of the main flow u0c0 is much lowerthan the transportation by the mean flow uc and is therefore negligible. (2) The observedsource is assumed to be constant over the total surface area and hence no concentrationgradient exists across the box. v0c0 D 0 can also expressed in terms of mean concentrationgradients: v0c0 D �Ky @c@y . As @c

@yD 0, v0c0 D 0 also yields zero. (3) At the surface, the

mean vertical wind speed is assumed to be zero and hence wczD0 D 0. (4) The same appliesfor the top of the well-mixed boundary layer (wczDhPBL D 0). (5) No turbulent flow existsdirectly at the surface and therefore w0c0zD0 D 0. (6) On the lid of the box, the inversionat the top of the boundary layer is relatively impermeable and hence the w0c0hPBL D 0. Thesimplified equation yields:

a � b � �S D b � hPBL � ucxDa„ ƒ‚ …Feast

� b � hPBL � ucxD0„ ƒ‚ …Fwest

C

a � hPBL � vcyDb„ ƒ‚ …Fnorth

� a � hPBL � vcyD0„ ƒ‚ …Fsouth

:(3.6)

The resulting sign convention counts fluxes into the box positive, and out of the box neg-ative. In micrometeorology, fluxes from the ground to the atmosphere are defined positive.The source term needs therefore to be multiplied with –1 to follow the micrometeorologicalsign convention. Solved for S , the equation thereafter writes:

S D hPBL�.ucxDa � ucxD0/=aC .vcyDb � vcyD0/=b

�: (3.7)

hPBL was estimated from the dust layer height in pictures taken during the flight.

The same decomposition is valid for the mass conservation equation (Equation 3.1) withthe difference that the source term is zero. Hence, the flow through the individual facesof the box should add to zero if the measurements are representative. To force the massbalance to close, the crosswinds were adjusted by multiplying with a scaling factor. Thescaled crosswinds were thereafter used to calculate the CH4 net exchange.

3.2.6. Simplified box model

A simple one-box model is widely used in urban air pollution studies to estimate atmospherictrace gas or pollutant concentrations from a known source as a function of time (Arya, 1999;Oke, 1987; Hanna et al., 1982). An imaginary box following the mean wind therefore en-closes the area of interest and all relevant flows into and out of the box are quantified (see

51


Figure 3.2), as in the box model presented above. In addition, transport through the sidewallsis assumed to be small compared to the transport in the mean wind direction. Further, themain wind is set constant throughout the entire box. Hence, Equation (3.7) further simplifiesto:

S D hPBL ucout � cin

a: (3.8)

For homogeneous sources, a linear concentration increase is expected. Hence, cout�cina

isequal to the observed slope of the measured c along the x-axis and is replaced therewith inthe flux calculations.

3.2.7. Chamber measurements

Lake surface CH4 emissions were quantified with the same floating chambers as on 23, 2429, and 30 July 2008 (DelSontro et al., 2010) to allow comparison with those results. Thechambers consisted of a cylindrical bucket attached to buoys for floatability and were keptupright by weights. Diffusion from the water body to the atmosphere and emerging bubbleswere captured in the chamber. Fluxes were calculated from the CH4 increase in the chamberover time. A more detailed description of the chambers can be found in (Eugster et al., 2011;DelSontro et al., 2010).

3.3. Results

3.3.1. Atmospheric CH4 distribution over Lake Wohlen

Methane concentrations measured by aircraft over Lake Wohlen showed a distinct patternon 30 September 2009 (see Figure 3.3). Especially for the noon and afternoon overflights(Figure 3.3bc), CH4 concentrations increased from east to west. In the morning flight (Fig-ure 3.3a), a concentration gradient was only visible in the northern transects whereas theconcentrations along the southern transects were rather homogeneous. The well-mixed partof the planetary boundary layer was very shallow and thus only the lowest transects weredirectly influenced by the local processes. Transects above the lowest inversion layer werethus excluded from further analyses. Additionally to the gradient, a methane hot-spot aroundpoint (502 km/201.5 km) was observed during the noon and afternoon overflight.

52

3.3. Results

3.3.2. Atmospheric concentration gradients along the lake

CH4 concentrations showed a significant increase on the first 2 km from west to east (p<0.0001)during the noon and afternoon overflight (see Figure 3.4a). From 2.5 to 3.5 km, no CH4 in-creases were observed and the slopes were not significantly different from 0 (p=0.35). Thepattern was opposite in the morning. In the western part, the concentration decreased andthen slightly increased in the east.

CO2 concentration increases were observed along the western part of the transects, whichwere even enhanced on the eastern part during the morning overflight (see Figure 3.4b). Atnoon and in the afternoon, the concentrations stabilized in the eastern part of the transectand the slopes were not significantly different form zero any more (p=0.25).

CO concentrations strongly increased along the transect in the morning (see Figure 3.4c).Until noon, this strong CO gradient disappeared and concentrations gently increased fromwest to east. In the afternoon, the CO concentrations were relatively stable in the westernpart of the transect, but significantly increased in the eastern part (p<0.0001). While thesign of the slopes differed between CH4 and CO, they follow a common pattern for CO2 andCO.

3.3.3. Conservation of air mass and mass balance closure

The mass balance of the specified box did not close for all three overflights (see Figure 3.5).The model resulted in a accumulation of air (�) in all three cases. In the morning, 37%of the outflow was missing, and 53% and 59% at noon and in the afternoon, respectively.Total mass flow was lowest in the morning, peaked around noon and got smaller again inthe afternoon. In the morning, flow through the sidewalls was higher than through the windfacing sides. At noon and in the afternoon, the flow in direction of the mean wind dominatedthe total exchange. The flow through the side walls differed between the south and north sideof the box. At noon, the flow through the south face was higher than the flow through thenorth face. This pattern was reversed in the afternoon.

The lack of closure was overcome by scaling the crosswind such that the fluxes throughthe individual faces of the box added to zero. The scaling factor was variable betweenthe individual overflights and yielded –1.17 in the morning, 0.25 at noon, and 0.32 in theafternoon. This measure significantly reduced the exchange through the side walls for thenoon and afternoon overflight.

53


3.3.4. Flux estimates

Airborne measurements of vertical fluxes

The pattern of the CH4 flown through the individual box faces was similar to the massflow pattern (compare Figure 3.5 and 3.6). A huge CH4 sink of –20, -630, and –330µg CH4 m�2 s�1 resulted from the application of the box model with unscaled winds. Atnoon and in the afternoon, the scaled box model resulted in very reasonable fluxes of 19and 12 µg CH4 m�2 s�1. In the morning, the model result still calculated a CH4 sink of -115 µg CH4 m�2 s�1. The calculated CH4 sinks or sources were very small compared to theflows through the box faces.

The simplified box model fluxes depend on the variables wind speed, mixing layer height,and observed CH4 concentration gradient along the valley axis. The calculated fluxes variedbetween close to 0 up to about 20 µg CH4 m�2 s�1 (see Figure 3.7a). The flux in the morningwas lower compared to the afternoon since the mixing layer height was lower. However,absolute values of wind speed and the concentration gradient were also slightly lower. Themixing layer was highest in the afternoon, but the flux was lower compared to the flux atnoon. Lower wind speed and a slightly shallower concentration gradient resulted in a fluxreduction. The fluxes calculated with simplified model compared well with the box modelapproach at noon and in the afternoon. For both methods, fluxes were lower in the afternooncompared to at noon.

The full featured box model with scaled crosswinds was also applied to the trace gasesCO2 and CO. The CO2 flux yielded 2.7 mg CO2 m�2 s�1 and 2.1 mg CO2 m�2 s�1 for thenoon and afternoon overflight respectively while the CO flux yielded 10.0 µg CO m�2 s�1

and 8.8 µg CO m�2 s�1.

Chamber fluxes

No distinct diurnal pattern was found in the chamber-based CH4 fluxes (see Figure 3.7b).The measurements fluctuated around the mean of 5.5 µg CH4 m�2 s�1 which was in the sameorder of magnitude as the airborne flux estimates.

54

3.4. Discussion

3.4. Discussion

3.4.1. Regional CH4 distribution

The observed CH4 hot spot in Figure 3.3 originated from a point source. Depending on thewind speed, the plume was shifted along the main wind. Since neither CO2 nor CO showed apeak in the same region, combustion can be excluded from list of possible processes leadingto the CH4 peak. Landfills are a known source of CH4. The nearest landfills (Illiswil andTeuftal) were located outside of the area of interest. The CH4 plume of those was howevervisible on transects flown outside of the box where CH4 peaked above the landfills (notshown). The remaining potential sources were emissions from agriculture or bio energyproduction. An aquatic hot spot was rather unlikely as none of the previous publicationsabout methane emissions from Lake Wohlen have found a very local and strong CH4 hotspot in this region (Eugster et al., 2011; DelSontro, 2011; DelSontro et al., 2010).

The opposite CH4 concentration gradients between the morning and the later overflightsoriginated from reverse wind directions. In the morning, down-valley wind prevailed whereaslater in the day, the wind blew up-valley. DelSontro et al. (2010) observed highest CH4 emis-sions in the area in between 1 to 2 km from the dam. We also observed steeper slopes in thisregion of the lake (up to approx. 2 km in Figure 3.4). For the morning transect, the positiveslope at the eastern part resulted from a very poor model fit (R2 = 0.01).

The missing similarities between the CH4 gradients and trace gases CO2 and CO indicatedthat different processes were responsible for the observed concentration patterns. CO2 andCO however did show a common pattern and hence one of the dominant emission processeswas most likely combustion. The ratio between the slopes of CO2 and CO was higher thanthe typical ratio for traffic exhaust (Wieser and Kurzweil, 2004). Thus, traffic was onlypartly responsible for the CO2 and CO emissions that resolved in the observed concentrationgradient. CH4 is also a byproduct of combustion, but the proportions in the exhaust arevery small compared to CO2 and CO in case of proper combustion. Hence, the combustioncontribution to observed CH4 gradients was minor and can be neglected.

3.4.2. Model performance

The consistency of the box model approach was tested on the air mass balance closure. Thebalance showed a mass gain for all three overflights, but most was dominant for the noonand afternoon overflights. During these periods of the day, a well-developed valley-windsystem dominated the flow in the valley. The mixing layer grew to a height of 820 and 970

55


m a.g.l., respectively. Sampling however only took place in the lower 300 m of the box.Following the theory of the valley wind system (Whiteman, 2000), the crosswinds at thevalley bottom of the mixing layer follow the slope terrain (see Figure 3.8). Thermally drivenup-slope winds are observed during the day and cold air drainage down-slope winds duringthe night. At higher elevation, the bottom flow is compensated. The air is redirected to thevalley center during the day. The strength of the up-slope winds depends on the heating ofthe surface and thus the slope exposition. The symmetric cells (see Figure 3.8) may shifttowards the more shaded slope or even collapse in a single cell where up-slope winds areobserved on the sun exposed slope and down-slope winds on the other.

Our crosswind measurements were not representative for the whole sidewall, because thebox walls intersected with the slope-wind circulation. We assumed the mixing layer heightto be constant during one overflight and attributed the mass gain in the box model to thecrosswinds. Their strength should nevertheless represent the activity of the circulation at therespective slope. Thus, we scaled the crosswinds and forced the mass balance to close. Thisscaling resulted in very weak crosswinds for the noon and afternoon overflight. Thereafter,the contribution of side flow to the total flow turned considerably lower compared to the flowin the main wind direction (see Figure 3.5).

During the morning overflight, the valley wind system was still in the night phase andwe observed down-valley winds. The winds were already weak and the transition to theday phase was close. The local wind system was strongly influenced by local differencesin irradiation and hence topography. Additionally, only a few transects were located in themixed boundary layer. The flux estimate bases on less measurement points than during theother periods. Averaged values might not be representative and thus the box model not thebest method of choice to derive regional scale fluxes under these conditions. The resultsshould be herein interpreted with care.

3.4.3. Flux estimates

CH4 fluxes from the full featured and the simplified box model were similar for the noonand afternoon overflight. The additional simplification of neglecting crosswind transportdid not considerably change the model outcome. The constant wind-speed over the wholebox might even have compensated for this simplification. Additionally, the simplified modelmight have compensated some shortcomings in the accuracy of the wind measurements.While the full-featured box models relied on accurate wind measurements, the mean windspeed in the simplified model incorporated measurements from transects flown with andagainst the wind. Hence, a possible sensor bias was compensated in the simplified model.

56

3.4. Discussion

The morning flux estimate was considerably lower than the other values. In the full-featured box model, a strong negative flux was estimated. As discussed for the mass balance,this time period was associated with a relatively high uncertainties and therefore will be leftout in the further discussion of the CH4 fluxes. For CO and CO2, fluxes were very highand not explainable with known CO2 and CO producing processes in the valley where LakeWohlen is situated. Therefore, also the CH4 fluxes should be interpreted with care andregarded as a first best estimate. For a representative assessment, further investigations areneeded.

Chamber measurements performed on the lake surface at the same day as the flights re-vealed about factor two lower fluxes than measured by the aircraft. Chamber measurementswere restricted to a small part of the lake, where highest methane emissions were expected(DelSontro et al., 2010). Lake methane emissions occur very locally (DelSontro, 2011) andhence three chambers might not be representative for the lake emissions. Nevertheless, theygave a reference for the order of magnitude of the expected lake emissions.

Airborne CH4 fluxes also involved uncertainties. The uncertainty related to the windmeasurements was already discussed above. Additionally, the mixing layer height wasonly estimated and not directly measured. Since the mass balance was performed onlyon the vertical faces of the box, the mixing layer height did not influence the result. Thedifference between inflow and outflow of the box was scaled on the box surface to esti-mate the CH4 flux. Hence, the mixing layer height was essential in this case. The low-est temperature inversion in the sounding from Payerne (46.81ıN/6.95ıE, 491 m a.s.l., lo-cated approx 35 km from our site) was observed between 1350 and 1525 m a.s.l. at 13CEST (http://weather.uwyo.edu/cgi-bin/sounding?region=europe&TYPE=TEXT%3ALIST&YEAR=2009&MONTH=09&FROM=3000&TO=3012&STNM=06610&REPLOT=1, accessed:20 November 2011). This corresponded to a mixing layer height of 880 to 1030 m, whichwas close to our estimate. Hence, the uncertainty introduced by the estimation of the mixinglayer height was likely less than 20% of the total height.

The estimated CH4 fluxes from aircraft measurements from 2009 compared well withobserved CH4 fluxes measured by eddy covariance and with chambers in 2008 (Eugsteret al., 2011; DelSontro, 2011; DelSontro et al., 2010). The magnitude of the eddy covariance(EC) fluxes in 2008 was in the same range for periods with wind across the lake (Eugsteret al., 2011). This corresponded to the area where the chambers were deployed during theaircraft campaign in 2009 and also during the EC measurements in 2008. The EC CH4 fluxmeasurements were on average lower as the chamber measurements reported for the sametime period (Eugster et al., 2011), but agree well if the wind blew from the area were thechamber measurement were located. Hydroacoustic estimations of the CH4 ebullition results

57


in slightly lower fluxes compared to estimates with floating chambers (DelSontro, 2011).

3.5. Conclusions

Airborne CH4 flux measurements agreed with ground base estimates, but the presented re-sults were still associated with considerable uncertainties. The topography of the valley onone hand simplified the examinations due to the canalling of the main flow. On the otherhand, the valley-wind system also introduced additional uncertainty to the flow through thesidewalls. For similar topography, we would suggest to sample the individual sidewalls ofthe box throughout the total height of the mixing layer. This would reduce the uncertaintyrelated to the crosswinds observed in a well-developed valley system. The same applies to amore open terrain. Nevertheless, the study showed the potential of airborne CH4 flux mea-surements to resolve landscape scale fluxes. Despite all assumptions and simplifications, themodels estimated reasonable CH4 fluxes for the assessed lake ecosystem.

Acknowledgements We thank the MetAir crew (Lorenz Muller, Dave Oldani, and BorisSchneider) for their great effort during the measurement campaign, Dorte Carstens andGijs Nobbe for their contributions to the chamber measurements, and Susanne Burriand Jacqueline Stieger for their helpful comments on the manuscript. This project wasfunded by the Competence Center Environment an Sustainability at the ETH Domain(CCES).

3.6. References

Arya, S. (1999). Air pollution meteorology and dispersion. Oxford University Press, USA.pp. 310.

Barros, N., J. J. Cole, L. J. Tranvik, Y. T. Prairie, D. Bastviken, V. L. M. Huszar,P. del Giorgio, and F. Roland (2011, September). Carbon emission from hydroelec-tric reservoirs linked to reservoir age and latitude. Nature Geoscience 4(9), 593–596.doi:10.1038/ngeo1211.

Crawford, T. L. and R. J. Dobosy (1992). A sensitive fast-response probe to measure tur-bulence and heat flux from any airplane. Boundary-Layer Meteorology 59, 257–278.doi:10.1007/BF00119816.

DelSontro, T., D. F. McGinnis, S. Sobek, I. Ostrovsky, and B. Wehrli (2010). Extrememethane emissions from a Swiss hydropower reservoir: Contribution from bubbling sedi-ments. Environmental Science & Technology 44(7), 2419–2425. doi:10.1021/es9031369.

58

3.6. References

DelSontro, T. S. (2011). Quantifying methane emissions from reservoirs: From basin-scaleto discrete analyses with a focus on ebullition dynamics. Ph. D. thesis, ETH Zurich. pp.155.

Diem, T., S. Koch, S. Schwarzenbach, B. Wehrli, and C. J. Schubert (2008). Greenhousegas emissions (CO2, CH4 and N2O) from perialpine and alpine hydropower reservoirs.Biogeosciences Discussions 5(5), 3699–3736.

Durst Stucki, M., R. Reber, and F. Schlunegger (2010). Subglacial tunnel valleys in theAlpine foreland: An example from Bern, Switzerland. Swiss Journal of Geosciences 103,363–374. doi:10.1007/s00015-010-0042-0.

Eugster, W., T. DelSontro, and S. Sobek (2011). Eddy covariance flux measurements confirmextreme CH4 emissions from a Swiss hydropower reservoir and resolve their short-termvariability. Biogeosciences 8(9), 2815–2831. doi:10.5194/bg-8-2815-2011.

Giles, J. (2006). Methane quashes green credentials of hydropower. Nature 444(7119),524–525. doi:10.1038/444524a.

Gueerin, F., G. Abril, S. Richard, B. Burban, C. Reynouard, P. Seyler, and R. Delmas (2006).Methane and carbon dioxide emissions from tropical reservoirs: Significance of down-stream rivers. Geophys. Res. Lett. 33(21), L21407. doi:10.1029/2006GL027929.

Hanna, S. R., G. A. Briggs, and R. P. Hosker (1982). Handbook on Atmospheric Diffusion.Technical Informaiton Center U.S. Department of Energy. pp. 102.

Hoffert, M. I., K. Caldeira, A. K. Jain, E. F. Haites, L. D. D. Harvey, S. D. Potter,M. E. Schlesinger, S. H. Schneider, R. G. Watts, T. M. L. Wigley, and D. J. Wuebbles(1998). Energy implications of future stabilization of atmospheric CO2 content. Na-ture 395(6705), 881–884. doi:10.1038/27638.

International Energy Agency (2008). Electricity/heat in world in 2008. available from http:

//www.iea.org/stats/electricitydata.asp?COUNTRY_CODE=29 (accessed: 2011-08-17).

MeteoSchweiz (2009). Witterungsbericht September 2009. Bundesamt fur Meteorologieund Klimatologie MeteoSchweiz. pp. 6.

Neininger, B., W. Fuchs, M. Baeumle, A. Volz-Thomas, A. Prevot, and J. Dommen (2001).A small aircraft for more than just ozone: Metair’s ’DIMONA’ after ten years of evolvingdevelopments. In 11th Symposium on Meteorological Observations and Instrumentation.

Oke, T. (1987). Boundary layer climates. London : Routledge. pp. 420.

Ritter, J., J. Barrick, G. Sachse, J. C. Jr, B. Anderson, G. Hill, M. Woerner, and J. H. Jr(1994). The role of airborne eddy correlation measurements in global change studies.Advances in Space Research 14(1), 183 – 186. doi:10.1016/0273-1177(94)90368-9.

59


St. Louis, V. L., C. A. Kelly, E. Duchemin, J. W. M. Rudd, and D. M. Rosen-berg (2000, September). Reservoir surfaces as sources of greenhouse gases to theatmosphere: a global estimate. BioScience 50(9), 766–775. doi:10.1641/0006-3568(2000)050[0766:RSASOG]2.0.CO;2.

Stull, R. (1988). An introduction to boundary layer meteorology. Kluwer Academic Pub.pp. 666.

Teodoru, C., Y. Prairie, and P. del Giorgio (2011). Spatial heterogeneity of surface CO2fluxes in a newly created Eastmain-1 reservoir in northern Quebec, Canada. Ecosys-tems 14, 28–46. doi:10.1007/s10021-010-9393-7.

Victor, D. G. (1998). Strategies for cutting carbon. Nature 395(6705), 837–838.doi:10.1038/27532.

Wehrli, B. (2011). Climate science: Renewable but not carbon-free. Nature Geoscience 4(9),585–586. doi:10.1038/ngeo1226.

Whiteman, C. D. (2000). Mountain meteorology : fundamentals and applications. NewYork : Oxford University Press. pp. 355.

Wieser, M. and A. Kurzweil (2004). Emissionsfaktoren als Grundlage fur die osterreichischeLuftschadstoff-Inventur. In Energieberichts 2003. Osterreichischen Bundesregierung.

60

3.7. Figures

3.7. Figures

0.00.5

1.01.5

2.02.5

3.0 km

NDHM25 and SWISSIMAGE © 2010 swisstopo (JA082267)

Figure 3.1.: Flight pattern above Lake Wohlen. The red line indicates the flight track flownin the afternoon on 30 September 2009. The orange line indicates the imagi-nary box that was arranged along the valley axis to calculate the CH4 flux fromthe aircraft measurements. (Orthophoto: Reproduced with the authorization ofswisstopo (JA082267))

[CH4]

x

hPBL

a

b

Cin

Cout

u

Figure 3.2.: Scheme of the modified box model (Reproduced from Chapter 2, Figure 2.2).

61


CH4 [ppm](a)

200

201

202

203

© 2011 swisstopo (JA100120)

30 Sept 09:50 CEST

589 590 591 592 593 594

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●●

●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●

2.10 2.15 2.20 2.25 2.30

CH4 [ppm](b)


30 Sept 12:50 CEST

589 590 591 592 593 594

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

●●

●●

●●

●●

●●

●●

1.95 2.00 2.05 2.10 2.15

CH4 [ppm](c)

200

201

202

203


30 Sept 14:30 CEST

589 590 591 592 593 594

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●●●●●●●●●●●●●●●●●●●●●●

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

1.90 1.95 2.00 2.05

Figure 3.3.: Flight patterns and CH4 concentration distribution for the individual overflights.The black line indicates the flight pattern whereas the colored points show theCH4 concentration. The coordinates are given in units of km in Swiss coordinate(CH1903). (Background map PK100: Reproduced with the authorization ofswisstopo (JA100120))

62

3.7. Figures

0.0 1.0 2.0 3.0

1.9

2.0

2.1

2.2

a [km]

CH

4 [pp

m]

(a)

W E0.0 1.0 2.0 3.0

390

400

410

420

430

440

450

a [km]

CO

2 [pp

m]

(b)

W E

Sept 30 09:50Sept 30 12:50Sept 30 14:30

0.0 1.0 2.0 3.0

150

200

250

a [km]

CO

[ppb

]

(c)

W E

Figure 3.4.: Observed concentration gradients of CH4, CO2, and CO along the valley axis.The relative distance a represents the projected distance between the individualmeasurement points along the x-axis of the box. The fine dark gray lines showthe concentration gradients for the individually flight transects whereas the boldblack lines represent the linear regression through the individual points for oneoverflight consisting of multiple transects. The relationship was once fit fora < 2 km and once for a > 2:5 km.

W E S N ∆

10:00 CEST

Mas

s flo

w [G

g s−1

]

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5unscaledscaled

W E S N ∆

13:00 CEST

Mas

s flo

w [G

g s−1

]

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5

W E S N ∆

14:30 CEST

Mas

s flo

w [G

g s−1

]

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5

−1.5

−1.0

−0.5

0.0

0.5

1.0

1.5

Figure 3.5.: Mass flux through the individual faces of the box for the three overflights.

63


W E S N ∆

10:00 CEST

FC

H4 [

kg s

−1]

−2

−1

0

1

2

−2

−1

0

1

2unscaledscaled

W E S N ∆

13:00 CEST

FC

H4 [

kg s

−1]

−2

−1

0

1

2

−2

−1

0

1

2

W E S N ∆

14:30 CEST

FC

H4 [

kg s

−1]

−2

−1

0

1

2

−2

−1

0

1

2

Figure 3.6.: CH4 flux through the individual faces of the box for the three overflights.

●

●

●

Aircraft(a)

−10

−5

0

5

10

15

20

FC

H4 [

µg C

H4

m−2

s−1

], s

[ng

CH

4 m

−1]

30 Sep10:00

30 Sep13:00

30 Sep14:30

shu

FCH4(box)

FCH4(simple)

shpblu

●

h

h

h

−500

0

500

1000

h PB

L [m

a.g

.l.]

u

u

u

−1.0

−0.5

0.0

0.5

1.0

1.5

2.0

u [m

s−1

]

s

s

s

−10

−5

0

5

10

15

20

FC

H4 [

µg C

H4

m−2

s−1

]

30 Sep12:00

30 Sep13:00

30 Sep14:00

30 Sep15:00

Chambers(b)

Figure 3.7.: Aircraft (a) and chamber based (b) CH4 fluxes. (a) Flux estimates including thedriving variables mixing layer height (hPBL), wind speed (u), and concentrationgradient along the valley axis (s). (b) Boxplot summarizing the measurementsof 3 chambers.

64

3.7. Figures

Figure 3.8.: Sketch of the cross-wind profile. The arrows indicate the expected air circula-tion and the dashed lines show the outline of the imaginary box used for the fluxcalculations.

65

This chapter was published in Boundary Layer Meteorology,

2011, 138, 215–230, doi:10.1007/s10546-010-9558-0

4Interpreting CO2 fluxes over a suburbanlawn: The influence of traffic emissions

Rebecca V. Hiller1, Joseph P. McFadden2, Natascha Kljun3

1Institute of Plant, Animal and Agroecosystem Sciences, ETH Zurich, Zurich, Switzerland2Department of Geography, University of California, Santa Barbara, Santa Barbara, CA, USA

3Department of Geography, Swansea University, Swansea, UK

Abstract

Turf-grass lawns are ubiquitous in the United States, however direct measure-ments of land–atmosphere fluxes using the eddy covariance method above lawnecosystems are challenging due to the typically small dimensions of lawns andthe heterogeneity of land use in an urbanized landscape. Given their typicallysmall patch sizes, there is the potential that CO2 fluxes measured above turf-grass lawns may be influenced by nearby CO2 sources such as passing traffic.In this study, we report on two years of eddy covariance flux measurementsabove a 1.5 ha turf-grass lawn in which we assess the contribution of nearbytraffic emissions to the measured CO2 flux. We use winter data when the vege-tation was dormant to develop an empirical estimate of the traffic effect on themeasured CO2 fluxes, based on a parameterized version of a three-dimensional,Lagrangian footprint model and continuous traffic count data. The CO2 budgetof the ecosystem had to be adjusted by 135 g C m�2 in 2007 and by 134 g C m�2

67

Chapter 4: Interpreting CO2 Fluxes Over a Suburban Lawn

in 2008 to determine the natural flux, even though the road crossed the footprintonly at its far edge. We show that bottom-up flux estimates based on CO2 emis-sions factors of the passing vehicles combined with the crosswind-integratedfootprint at the distance of the road agreed very well with the empirical estimateof the traffic contribution that we derived from the eddy covariance measure-ments. The approach we developed may be useful for other sites as investigatorsseek to make eddy covariance measurements on small patches within heteroge-neous landscapes where there are significant contrasts in flux rates. However, wecaution that the modelling approach is empirical and it will need to be adaptedindividually to each site.

4.1. Introduction

The recent decades have seen large increases in the extent of urban and built-up land use.From 1982 to 1997, built-up areas increased by 37% in the United States and they are pro-jected to increase by another 79% by 2025 (Alig et al., 2004), with much of the growth insuburban areas (Kahn, 2000). Although European cities traditionally have been much morecompact, their built-up area has also increased, by 20% from 1980 to 2000 (European Com-mission, 2006). Urbanized areas represent a major source of CO2 to the atmosphere dueto the concentration of human activities that depend on energy from fossil fuel combustion.At the same time, many built-up areas, especially suburban land use-types, contain signif-icant amounts of vegetation that takes up CO2 by photosynthesis and releases it throughmetabolic activity. Against this background, there has been increasing interest in measuringland–atmosphere fluxes of CO2 in different types of urban settings (e.g. Grimmond et al.,2002; Nemitz et al., 2002; Soegaard and Moller-Jensen, 2003; Moriwaki and Kanda, 2004;Vogt et al., 2006; Coutts et al., 2007; Schmidt et al., 2008; Vesala et al., 2008).

Suburban areas in the United States are often dominated by single-family detached housessurrounded by turf-grass lawns, trees, and other green spaces. Milesi et al. (2005) estimatedthat cultivated turf-grasses covered 163,800 km2 of the continental United States, whichwould represent a large potential area for CO2 exchange. However, direct measurementsof land–atmosphere fluxes using the eddy covariance method above lawn ecosystems aredifficult due to the typically small dimensions of lawns and the heterogeneity of land usein an urbanized landscape. For example, in suburban Minneapolis–Saint Paul, Minnesota,our study site surrounding the KUOM tall tower had 34% cover of turf-grass lawns withina residential area of 4 km2. However, there was only one lawn of sufficient size for eddycovariance measurements that was not a sports field or golf course, neither of which would be

68

4.1. Introduction

suitable for a long-term installation of micrometeorological instruments. Given the typicallysmall patch size of urban turf-grass lawns, there is the potential that CO2 fluxes measuredabove them may be influenced by nearby CO2 sources such as passing traffic.

Many neighbourhood-scale studies show that traffic is an important source of short-timescalevariations in CO2. Grimmond et al. (2002) implicitly state that the CO2 flux largely dependson emissions of fixed (industrial, commercial, institutional) and mobile (traffic) sources, inaddition to the variations in vegetation cover. To separate the fluxes originating from differ-ent land surfaces and mobile sources requires highly detailed information about the spatialdistribution of the sources and sinks, as well as the spatial extent of the flux source area,or footprint. For this reason, previous studies have used approaches such as wind sectoranalysis of surfaces such as park versus main traffic route (Nemitz et al., 2002; Vesala et al.,2008; Burri, 2009) or have installed multiple towers in different areas of the city (Couttset al., 2007). In order to verify such top-down approaches to separating the sources andsinks of CO2, smaller, ecosystem-scale flux measurements can be of potential benefit. Atthe scale of a single turf-grass lawn, we can assume horizontal homogeneity and often a flatsurface. However, there is the potential that CO2 from traffic sources located beyond thefetch of the turf-grass field could affect the measured fluxes. In this respect, the problemthat fluxes from outside of an ecosystem of interest may influence the measured fluxes isalso common to natural or managed ecosystems that normally occur in small patches, e.g.in arctic tundra (McFadden et al., 1998), at mountainous sites (Hammerle et al., 2007; Hilleret al., 2008), or a single agricultural field (Ammann et al., 2007).

In this study, we show that even when the flux footprint was mainly within a turf-grassfield and the CO2 fluxes followed expected ecological patterns, a very small contributionfrom traffic CO2 emissions introduced a bias in the annual carbon budget. We analyzedwinter data when the vegetation was dormant to develop an empirical estimate of the trafficeffect on the measured CO2 fluxes, based on the footprint model of Kljun et al. (2004) andcontinuous traffic count data. As a check on the magnitude of the traffic flux predicted byour empirical model, we used published emissions factors (INFRAS, 2004) to independentlyestimate the amount of CO2 released by passing traffic. These analyses also provided uswith a unique way of assessing the performance of the footprint model that was roughlyanalogous to a tracer experiment, but using a long-term site where the well-defined trafficsource represented a natural tracer (e.g. Foken and Leclerc, 2004; Gockede et al., 2005).

69


4.2. Methods

4.2.1. Site description

The study site was located in a first-ring suburb of Minneapolis–Saint Paul, Minnesota, USA(44ı5904200N 93ı1101100W, 300 m a.s.l.). We made eddy covariance measurements of CO2,water vapour, energy, and momentum exchanges over a 1.5 ha turf-grass field (Figure 4.1)from November 2005 to May 2009. For the present study, we analyzed data from 2007 and2008 since continuous traffic counts were available for those years. The dominant speciesat the site were the C3 cool-season turf-grasses Kentucky bluegrass (Poa pratensis L.), tallfescue (Festuca arundinacea Schreb.), and perennial ryegrass (Lolium perenne L.). The sitewas representative of low-maintenance lawns in the area, such as those found in residentialneighbourhoods or a city park. The site was not irrigated and it received one applicationof fertilizer per year. During the growing season (mid-April to mid-October), the grass wasmowed weekly to a height of 70 mm, and the clippings were left in place to decompose.A two-lane county road carrying primarily commuting traffic (�10,000 vehicles day�1)was located on the western edge of the turf-grass field, at a distance of 60 m from thetower (Figure 4.1). The surrounding landscape was a residential neighbourhood consistingof single-family detached houses with a golf course located across the road to the west ofthe turf-grass study site.

Minnesota exhibits a continental climate, characterized seasonally by different air masses.Cold polar air can intrude in any season, most frequently in the winter. Conversely, bothwet and dry subtropical air masses move in from the south mainly in summer (Seeley andJensen, 2006). The mean air temperature (based on the period 1971–2000) is –10.5ıC inJanuary and 22.9ıC in July (NCDC, 2003). The mean annual precipitation is approximately750 mm, peaking in the summer months when thunderstorms are common (NCDC, 2003).The average snow cover in January and February is about 0.10–0.15 m (Minnesota StateClimatology Office), but varies widely within the season and among years. Depending onthe air temperature and the thickness and composition of the snow pack, the soil freezes to adepth >0.05 m for at least part of the winter.

4.2.2. Instrumentation

CO2 fluxes were measured at 1.35 m above the ground surface using an eddy covariance sys-tem consisting of a CR5000 data-logger, a CSAT3 sonic anemometer (both Campbell Sci-entific, Inc., Logan, Utah, USA) and an open-path infrared gas analyzer (IRGA) (LI-7500,

70

4.2. Methods

LI-COR, Inc., Lincoln, Nebraska, USA). The measurement height was not adjusted duringwinter for changes in snow depth because the open site was often wind blown, and snowaccumulation was generally low. Soil temperature was recorded at two locations at 0.05m below ground (STP1, Radiation and Energy Balance Systems, Inc., Seattle, Washington,USA). The incident photosynthetically active photon flux density (PPFD) was measured at2 m above ground (LI-190SA, LI-COR, Inc., Lincoln, Nebraska, USA).

4.2.3. Data processing

Fluxes were calculated using a slightly modified version of the eth-flux program developedby Werner Eugster (http://www.ipw.agrl.ethz.ch/�eugsterw/eth-flux/index.html). This pro-gram was part of the CarboEurope software inter-comparison (Mauder et al., 2008). Weadded a new spike filter to the software, following Vickers and Mahrt (1997). Using a 300s point-to-point moving window, values exceeding the mean by ˙6 standard deviations (� )were removed. The maximum number of consecutive spikes was set to 5, and the proce-dure was repeated up to 3 times for each 30-minute block of data. Prior to running eth-flux,raw data records were screened and replaced by missing values if the instrument flags werehigh or if the measurements exceeded the instrument range, indicating measurement prob-lems. Fluxes were computed over 30-minute periods by applying a time lag if needed, andcalculating the covariance between the vertical wind speed and the scalar, e.g. the CO2concentration, using the block averaging method.

High-frequency losses in the fluxes were corrected using Moore’s [1986] transfer func-tions for line averaging (sensible heat flux) and additionally sensor separation for the watervapour and CO2 fluxes, with the resulting correction factor trimmed to maximum of 1.5.The sonic temperature was corrected following Schotanus et al. (1983) and corrections fordensity effects were made following Webb et al. (1980). No adjustments for self-heating ofthe LI-7500 (e.g. Burba et al., 2008) were applied because we did not observe systematicapparent net CO2 uptake during winter and because of the uncertainty of the empirical cor-rections due to their dependence on the tilt angle of the LI-7500, the wind direction, and thewind speed. The CO2 flux was corrected for changes in storage in the air column below thesensor by calculating the change in the CO2 concentration measured by the LI-7500 overeach 30-minute period.

Before further analysis, we screened out periods with wind directions from 16ı–135ı,a wind sector which included experimental and irrigated lawn patches that were locatedbehind the eddy covariance tower (see Figure 4.1). We screened out precipitation eventsusing measurements from a rain gauge at a weather station located 750 m away from our

71


study site. We removed records when the path of the IRGA was obscured (automatic gaincontrol >69). Outliers were detected using a day-to-day running window of 13 days if thepoint occurred outside the mean ˙3 � . This screening was done separately for daytimeand nighttime data. Using the algorithm of Gu et al. (2005), a low u� threshold of 0.06m s�1 was determined. Data were screened out if this threshold was not exceeded or themean wind speed was <0.6 m s�1. We followed the sign convention that positive CO2 fluxdensities represent net efflux, and negative flux densities represent net uptake, of CO2 bythe ecosystem. We assessed the footprint climatology over the study period by calculatingthe footprint for each valid 30-minute period using the model of Kljun et al. (2004) and thenplotting contours of the average relative contributions of different land areas to the measuredfluxes as in Rebmann et al. (2005) (see Figure 4.1). We note that the measured fluxes didnot originate only from within the area enclosed by the isopleths in Figure 4.1. Rather, theisopleths represent the average relative contributions of different land areas over the two-year period; however, individual 30-minute flux footprints did intersect the road when windswere blowing from that direction. This, combined with a relatively high source magnitude,meant that the traffic emissions contributed significantly to the measured fluxes, and thiswas especially apparent in winter when biological activity was low. Data are reported withrespect to local standard time (UTC–6h).

The overall data availability is shown by the gray shaded lines above the CO2 flux finger-print in Figure 4.2. The longer data gaps were due to failures of the buried power systemduring snow melt in spring or after lightning events in summer, and the shorter gaps weredue to instrument maintenance and technical problems (13% in 2007 and 2008, white areas).Apart from the longer gaps, the overall data availability was reasonable, and only 23% of themeasured fluxes (grey areas) were rejected due to our screening criteria, as described above.

4.2.4. Gap filling

We used a gap-filling approach similar to the one described by Falge et al. (2001). Day-time gaps (PPFD >10 µmol m�2 s�1) were filled using light-response curves (Ruimy et al.,1995) and nighttime gaps were filled using a temperature-driven model of ecosystem respi-ration (Lloyd and Taylor, 1994). For the ecosystem respiration model, we implemented theparameterization following Moureaux et al. (2006). The gap-filling process was performediteratively, and the parameters for both models were determined using an overlapping, cen-tred, running window that started with a length of 7 days and increased by 2 days in eachloop. The parameters were accepted only if at least 100 valid measurements were availablein the window and the model parameters were significant (Student’s t-test, p<0.1) (Desai

72

4.2. Methods

et al., 2005). We determined the start and end of the winter period when there was no plantuptake of CO2 by examining the diurnal cycle of fluxes and additional meteorological data.During the winter period, the light-response model was not used and, thus, only the ecosys-tem respiration model was used for gap filling over the full diurnal cycle. The light-responsemodel was again used immediately following spring snowmelt because the turf-grass beganto green up within one week after it was free of snow cover. Because the growing seasonof turf-grasses normally ends abruptly following a hard frost, we added break points to thegap-filling routine to indicate sudden changes in ecosystem productivity. At these points,we forced the moving window to break and thus prevented the mixing of data from beforeand after the event. Additional break points were added to indicate the onset and end of themidsummer drought period. All break points were determined examining the diurnal cyclefor significant, abrupt changes in CO2 exchange, along with additional meteorological data.

Missing meteorological data were filled by linear regression using data from a NationalWeather Service cooperative station operated by the University of Minnesota that was lo-cated 750 m from our measurement site.

4.2.5. Traffic counts and CO2 emission factors

Traffic was counted continuously by the Ramsey County Department of Public Works usingan inductive loop detector located at the nearest intersection along the road that passed ourstudy site. We obtained counts of the number of vehicles that crossed the intersection ineach direction of travel for every 15-minute period. In order to assess the vehicle fleetcomposition, we recorded traffic using a web cam during a weekday in April 2008. Usingthe video recording, we manually counted traffic and assigned each vehicle to the followingclasses: motorcycles, passenger cars, pickup trucks and sport utility vehicles (SUVs), vans,buses, single-unit trucks, and trailer trucks.

We obtained CO2 emission factors for each vehicle class from the Handbook EmissionFactors for Road Transport (Version 2.1, INFRAS, 2004). We retrieved emission factorsthat were appropriate for the fuel type distribution and vehicle age distribution typical ofMinneapolis–Saint Paul suburban areas in 2007 (Adam Boies, pers. com. 2008), as wellas the road characteristics (i.e., suburban, flat terrain, and 64 km h�1 (nominally 40 milesh�1) speed limit). Then, we weighted the emission factors by the frequency of each vehicleclass and multiplied by the measured traffic counts to calculate the total CO2 emissions fromtraffic for 30-minute intervals matching our eddy covariance measurements.

73


4.2.6. Estimating traffic effects on measured CO2 fluxes –

development of an empirical model

We assessed the influence of traffic on the observed CO2 fluxes by using a flux footprintmodel in combination with the continuous traffic count data. We used the parameterization(Kljun et al., 2004) of the footprint model from Kljun et al. (2002). In this model, thecrosswind-integrated footprint prediction f y represents the influence of the area located ata distance x from the tower and integrates the flux contribution over the entire width of thefootprint. In order to assess the contribution of the road to the total CO2 flux, we computedf y for each 30-minute measuring interval, setting x equal to the distance between towerand the road along the direction of the mean wind (i.e., the fetch length). The boundary-layer height was defined at a constant value of 1000 m. As the traffic contribution decreaseswith increasing fetch whereas the uncertainty of the footprint grows, we excluded from theempirical model any 30-minute intervals when the fetch length was >350 m (correspondingto 190ı < wdir < 350ı and the footprint being parallel to or pointing away from the road)and, instead, we set the estimated traffic CO2 flux to zero.

In the following, the influence of the road within the footprint is denoted by f y road. Thisterm varied with turbulence and surface characteristics, depending on the following vari-ables: the distance between the tower and the road in the prevailing wind direction, thefriction velocity, the instrument height, the standard deviation of the vertical wind, and theroughness length. For a given wind direction (i.e., constant fetch length), and instrumentheight, f y road varied with atmospheric stability and wind speed. In this case, a larger foot-print corresponded to a greater influence from the road. On the other hand, for a given setof atmospheric conditions, f y road varied with wind direction because the angle of the windaltered the fetch length between the tower and the road.

To estimate traffic CO2 fluxes, we developed a multiple regression model that accountedfor variations in both traffic and meteorological conditions, the latter through the informa-tion on the footprint and the fetch length that was represented by f y road. To exclude otherenvironmental drivers, we fitted the model using winter data when the daily maximum soiltemperature was below the freezing point and the snow cover was continuous. Under thoseconditions, there was no photosynthetic uptake by turf-grass, and we could assume that theonly ecosystem contribution to the measured CO2 flux was a very low and nearly constantefflux due to soil respiration. During this period, the Pearson’s product-moment correlation(r) between the CO2 flux and the traffic volume was 0.47, while the correlation between theCO2 flux and f y road was 0.31. We fitted the following multiple regression model, includingall interaction terms:

74

4.3. Results and Discussion

FCO2;meas D intercept C af y road„ ƒ‚ …FCO2;eco

C bnvehicles C cf y roadnvehicles„ ƒ‚ …FCO2;traf

; (4.1)

where FCO2;meas was the measured flux at the tower and nvehicles denoted the number ofpassing vehicles during the 30-minute interval. The terms intercept and af y road did notdepend on nvehicles and, together, they represented the ecosystem CO2 flux, FCO2;eco. ThequantityFCO2;eco is equivalent to the net ecosystem exchange (NEE) of CO2 that is commonlyreported for flux sites in natural ecosystems. The sum of the other two terms, bnvehicles C

cf y roadnvehicles, provided an estimate of the traffic related flux, FCO2;traf.


4.3.1. CO2 flux measurements

At first sight, the flux measurements over the turf-grass field did not appear to show signs thattraffic emissions had contaminated the normal temporal patterns of ecosystem CO2 fluxes.If traffic emissions were important, we would have expected to find positively biased fluxesduring rush hours in the morning and early evening throughout the annual cycle. However,the flux fingerprint for the years 2007 and 2008 (Figure 4.2) showed CO2 fluxes that hadgenerally similar diurnal, weekly, and annual patterns as compared to natural grasslands andpastures (Flanagan et al., 2002; Byrne et al., 2005). The annual cycle of CO2 fluxes followeda similar pattern in both years. When soils were frozen and the ground was snow covered, thefluxes had small positive magnitudes and the ecosystem was a net source of CO2. Followingsnowmelt in April, turf-grass photosynthesis began rapidly and the first negative fluxes wereobserved, increasing throughout the spring. This period of strong CO2 uptake was followedby warmer and drier weather in midsummer, and the flux fingerprint showed a mid-summerdecline in growth rates that is characteristic of cool-season turf-grasses (Fry and Huang,2004). In 2007, an unusually dry spring was followed by a warmer than average summer,and the flux fingerprint showed a more severe mid-summer decline in CO2 uptake duringthat year. In both years, when cooler conditions returned in late summer and fall, turf-grassgrowth increased and a second period net CO2 uptake was observed in the flux fingerprint.

In addition to the flux fingerprint analyses, we examined the data set in the followingways for evidence that traffic emissions had affected the measured CO2 fluxes. First, wereasoned that CO2 concentrations would be elevated when the wind direction was from thewest, which represented the shortest distance between the tower and the road. However, wedid not find a relationship between CO2 concentration and wind direction. On the contrary,

75


during the summer we found that CO2 concentrations were commonly lower when the windwas from the west because this direction was also affected by strong plant uptake of CO2 dueto a golf course on the other side of the road. Second, if traffic emissions were important, weexpected that light-response curves of the CO2 flux would differ between weekday versusweekend periods. However, the scatter in the light-response curves was large relative to themagnitude of the traffic contribution, and thus no such pattern was found in the data. Third,we binned the data into periods when the wind was blowing over the road versus other winddirections, and then compared light-response curves of the CO2 flux. However, this com-parison was confounded by the characteristically different weather conditions that occurredwith the different wind sectors. Whereas north-westerly winds occurred with cooler weather,southerlies brought warm and humid air from the Gulf of Mexico. The systematic tempera-ture differences that were associated with the wind patterns, and their consequent effects onecosystem respiration, likely explain why we did not observe a signal of traffic emissions inthese analyses.

4.3.2. Empirical model of traffic CO2 fluxes

A final set of analyses was limited to mid-winter periods when the soil was frozen to adepth at least 0.05 m and the ground was covered by snow. During this period, there was nophotosynthetic uptake by turf-grass, and we could assume that the ecosystem contributionto the measured CO2 flux was very low and varied within a narrow range depending on soiltemperature. Despite this, the variations in the measured winter fluxes were only weaklycorrelated with soil temperature (r = 0.05). Instead, we found that the measured CO2 fluxeswere strongly related to the number of vehicles passing on the nearby road (upper panels inFigure 4.3). The correlation between the measured CO2 flux and the traffic counts was 0.46on weekdays and 0.27 on weekends. With this information, we fitted the model described inSection 4.2.6 to separate traffic CO2 emissions from the ecosystem CO2 flux.

We fitted Equation 4.1, including all interaction terms, and obtained a parameterizationwith R2 = 0.35 (Table 4.1). That the model explained only 35% of the variation in the mea-sured CO2 flux was not unexpected given that ecosystem fluxes (i.e., soil respiration drivenby temperature) also contributed to the measured CO2 fluxes, even in winter. In addition, thefootprint climatology (Figure 4.1) showed that the relative influence of locations as far awayfrom the tower as the road would have been small for the entire years 2007 and 2008. How-ever, despite the fact that the main source area of the measured fluxes was located withinthe turf-grass field, the source strength of motor vehicle emissions was large and there was aclear signal of the traffic in the CO2 flux when the wind was blowing from the road towards

76


the tower (upper panels of Figure 4.3). In order to remove the traffic CO2 emissions fromour measurements, we subtracted the estimated traffic related flux F CO2;traf (i.e., boldfaceparameters in Table 4.1). After subtracting F CO2;traf using this approach, there was no longera diurnal cycle in the observed fluxes, which was the pattern we would have expected to findover a field that was covered with snow (lower panels of Figure 4.3).

Figure 4.4 illustrates the characteristics of our empirical model of traffic CO2 fluxes. Plot-ting the estimated F CO2;traf by the wind direction resulted in a bell shaped curve. The winddirection is directly related to the length of the fetch over the turf-grass field, and the highestfluxes from traffic occurred when winds blew directly from the road to the tower and thefetch length was shortest. The scatter within each wind sector may be attributed to bothvariations in the dimensions of the footprint with atmospheric conditions and variations intraffic counts with time. Variations in roughness length with season or wind direction wereaccounted for in the empirical model because it was based on the footprint information, andmeasured values of z0 were used to compute the footprint for each 30-minute period.

4.3.3. Empirical model compared to traffic CO2 flux estimates based

on emission factors

As a check on the magnitude of F CO2;traf predicted by our empirical model, we used pub-lished emissions factors to independently estimate the amount of CO2 released by passingtraffic. First, we assessed whether there was a linear relationship between the CO2 emissionsby traffic, as estimated in Section 4.2.5, and the number of passing vehicles (Figure 4.5). Alinear relationship between the number of passing vehicles and the observed flux wouldbe expected if the fleet composition did not vary significantly over time. This relationshipyielded an R2 of 0.999 and the variations in fleet composition over the course of the daywere minor (lower panel in Figure 4.5); therefore, we determined that a linear relationshipwas appropriate for our site.

For the following analysis we used the same winter data set that was used to derive themodel, but we included only data with wind directions from 240–300ı, so that traffic con-tributions would be at their maximum. For the same reason, we also excluded data whenmore than 90% of the footprint was within the turf-grass area and did not overlap the road.These restrictions reduced the data set to 74 valid 30-minute records. The linear relationshipbetween F CO2;traf and the number of passing vehicles yielded an average of 53.32 mg C m�1

vehicle�1 (upper panel in Figure 4.5). We multiplied this emission factor by the number ofpassing vehicles to produce an estimate of the traffic related CO2 emissions for each 30-minflux measurement interval. In order to compare this estimated traffic related flux to the one

77


obtained from the eddy covariance system, we multiplied traffic related emission rate withf y road to account for the size of the footprint at that time period and devided it by 1800 toobtain an emission rate per second instead of per 30-minutes.

Figure 4.6 shows a scatterplot of FCO2;meas against the CO2 fluxes that were estimated us-ing emissions factors, as well as the linear regression between these variables. The interceptof 0.004 mg C m�2 s�1 was similar to the residual flux after applying the empirical trafficcorrection model (see Figure 4.3) and it can be attributed to the average winter soil respi-ration. At the same time, we note that the empirical traffic correction model described inSection 4.2.6 had the advantages that potential biases in estimating per-vehicle traffic emis-sions were not translated into the traffic correction and we did not need to know the footprintwidth.

4.3.4. Impact of traffic emissions on annual CO2 fluxes at this site

We assessed the overall influence of the traffic related CO2 flux on an annual basis by runningthe gap-filling procedure described in Section 4.2.4, once using a data set in which theinfluence of traffic was removed and once where it was not. Removing the CO2 flux causedby traffic changed the annual carbon budget by 135 g C m�2 in 2007, and by 134 g C m�2

in 2008 (see Figure 4.7). We note that these values do not represent an exact quantificationof the CO2 flux contribution from traffic. This is because our gap-filling procedure, likemost such approaches, did not account for differences in CO2 flux with wind direction. Thismeans that, when we ran the gap-filling procedure without first having removed traffic effectsfrom the data, gross primary productivity could have been underestimated and ecosystemrespiration overestimated on either side of the gap, thereby propagating effects of a giventraffic flux into gap-filled values. After subtracting FCO2;traf, the pattern of FCO2;eco betweenthe two years showed the impact of the observed weather conditions: the dry period in spring2007 reduced photosynthetic uptake, leading to a higher annual NEE, while the relativelylong and cold winter of 2008 reduced winter respiration, leading to a lower annual NEE.The similar magnitude of the traffic effect between 2007 and 2008 is consistent with thesimilar amounts of traffic in the two years (total volume of 3:51 � 106 vehicles in 2007 and3:40 � 106 vehicles in 2008). In addition, it suggests that variations of the flux footprintover time were similar between the two years on average. The fact that traffic related CO2

fluxes were essentially equal in 2007 and 2008 means that the relative differences in NEEbetween years were robust. Even though the road was outside of the main footprint area andthe effect of traffic emissions on each 30-minute flux observation was small, these valuesshow that traffic had a significant impact on the annual carbon budget because it represented

78

4.4. Conclusions

a small, continuous positive bias. For our study site, the correction converted the ecosystemfrom a net source to a net sink in 2008.

4.4. Conclusions

We measured CO2 fluxes above a lawn that was relatively large for a patch of homogeneousturf-grass cover in a suburban neighbourhood. We found that, even with a low measurementheight of 1.35 m, the limited fetch (from 60 to a few hundred meters) was insufficient tocompletely exclude fluxes from adjacent areas. These results are consistent with the estab-lished rule of thumb for micrometeorological measurements that the uniform fetch shouldequal at least 100 times the measurement height (Horst and Weil, 1994). However, we alsofound that a footprint model could successfully quantify the effects of traffic emissions fromoutside the measurement area, given that the location of the road relative to the flux towerwas known and the CO2 source from traffic could be estimated using traffic counts. Ourempirical model of traffic CO2 fluxes agreed closely with a bottom-up approach that usedfootprint-weighted emission factors. This further suggests that the footprint model producedreasonable estimates even at the edge of its domain. While we believe the empirical modelaccurately quantified the main effects of traffic emissions on the measured fluxes, we wouldrecommend nonetheless that the corrected, ecosystem CO2 flux data be interpreted withcaution. The approach we developed may be useful for other sites as investigators seekto make eddy covariance measurements on small patches within heterogeneous landscapeswhere there are significant contrasts in flux rates. However, we caution that the modellingapproach is empirical and it will need to be adapted individually to each site.

Acknowledgements We thank Duane Schilder of Ramsey County for providing trafficcount data, Dave Ruschy for providing data from the University of Minnesota climatestation, and Brian Horgan for advice and access to the field site. Special thanks goes toAdam Boies, Institute of Technology, University of Minnesota, for his data set on thefleet age, Katrina Hill, David Rittenhouse, and Christopher Buyarski for counting traf-fic, Emily Peters and Yana Sorkin for help in the field, Nicole Ifill for statistical input,David Levinson for advice on traffic counts, Mario Keller of INFRAS for providingus with emission factor data, and Werner Eugster for comments on the manuscript Wealso thank the two anonymous reviewers for their valuable comments and critical feed-back. This research was funded by grants to J.P.M. from the University of MinnesotaGraduate School and the NASA Earth Science Program (NNG04GN80G).

79


4.5. References

Alig RJ, Kline JD, Lichtenstein M (2004) Urbanization on the us landscape: looking aheadin the 21st century. Landscape Urban Plan 69(2-3):219–234

Ammann C, Flechard C, Leifeld J, Neftel A, Fuhrer J (2007) The carbon budget of newly es-tablished temperate grassland depends on management intensity. Agricult Ecosys Environ121(1-2):5 – 20

Burba GG, Mcdermitt DK, Grelle A, Anderson DJ, Xu L (2008) Addressing the influence ofinstrument surface heat exchange on the measurements of CO2 flux from open-path gasanalyzers. Global Change Biol 14(8):1854–1876

Burri S (2009) CO2 Fluxes and Concentrations overv an Urban Surface in Cairo, Egypt.Master’s thesis, University of Basel, 66 pp.

Byrne KA, Kiely G, Leahy P (2005) CO2 fluxes in adjacent new and permanent temperategrasslands. Agric Forest Meteorol 135(1-4):82 – 92

Coutts AM, Beringer J, Tapper NJ (2007) Characteristics influencing the variability of urbanCO2 fluxes in Melbourne, Australia. Atmos Environ 41(1):51–62

Desai AR, Bolstad PV, Cook BD, Davis KJ, Carey EV (2005) Comparing net ecosystem ex-change of carbon dioxide between an old-growth and mature forest in the upper Midwest,USA. Agric Forest Meteorol 128(1-2):33–55

European Commission (2006) Urban sprawl in Europe. Office for Official Publications ofthe European Communities, 56 pp.

Falge E, Baldocchi D, Olson R, Anthoni P, Aubinet M, Bernhofer C, Burba G, Ceulemans R,Clement R, Dolman H, Granier A, Gross P, Grunwald T, Hollinger D, Jensen NO, KatulG, Keronen P, Kowalski A, Lai CT, Law BE, Meyers T, Moncrieff J, Moors E, MungerJW, Pilegaard K, Rannik U, Rebmann C, Suyker A, Tenhunen J, Tu K, Verma S, VesalaT, Wilson K, Wofsy S (2001) Gap filling strategies for defensible annual sums of netecosystem exchange. Agr Forest Meteorol 107(1):43–69

Flanagan LB, Wever LA, Carlson PJ (2002) Seasonal and interannual variation in carbondioxide exchange and carbon balance in a northern temperate grassland. Global ChangeBiol 8(7):599–615

Foken T, Leclerc M (2004) Methods and limitations in validation of footprint models. AgricForest Meteorol 127(3-4):223 – 234

Fry J, Huang B (2004) Applied turfgrass science and physiology. John Wiley & Sons, Hobo-ken, New Jersey, 320 pp.

Gockede M, Markkanen T, Mauder M, Arnold K, Leps JP, Foken T (2005) Validation offootprint models using natural tracer measurements from a field experiment. Agric ForestMeteorol 135(1-4):314 – 325

80

4.5. References

Grimmond CSB, King TS, Cropley FD, Nowak DJ, Souch C (2002) Local-scale fluxesof carbon dioxide in urban environments: methodological challenges and results fromChicago. Environ Pollut 116(Supplement 1):S243–S254

Gu L, Falge EM, Boden T, Baldocchi DD, Black T, Saleska SR, Suni T, Verma SB, VesalaT, Wofsy SC, Xu L (2005) Objective threshold determination for nighttime eddy fluxfiltering. Agric Forest Meteorol 128(3-4):179–197

Hammerle A, Haslwanter A, Schmitt M, Bahn M, Tappeiner U, Cernusca A, WohlfahrtG (2007) Eddy covariance measurements of carbon dioxide, latent and sensible energyfluxes above a meadow on a mountain slope. Boundary-Layer Meteorol 122(2):397–416

Hiller R, Zeeman MJ, Eugster W (2008) Eddy-covariance flux measurements in the complexterrain of an Alpine valley in Switzerland. Boundary-Layer Meteorol 127(3):449–467

Horst TW, Weil JC (1994) How far is far enough?: The fetch requirements for micrometeo-rological measurement of surface fluxes. J Atmos Oceanic Tech 11(4):1018–1025

INFRAS (2004) HBEFA 2.1, Handbuch Emissionsfaktoren des Strassenverkehrs 2.1,Grundlagenbericht, im Auftrag von UBA Berlin / BUWAL Bern / UBA Wien, 127 pp.

Kahn ME (2000) The environmental impact of suburbanization. J Policy Anal Manag19(4):569–586

Kljun N, Rotach M, Schmid H (2002) A three-dimensional backward lagrangian foot-print model for a wide range of boundary-layer stratifications. Boundary-Layer Meteorol103(2):205–226

Kljun N, Calanca P, Rotach M, Schmid H (2004) A simple parameterisation for flux footprintpredictions. Boundary-Layer Meteorol 112(3):503–523

Lloyd J, Taylor JA (1994) On the temperature dependence of soil respiration. Funct Ecol8(3):315–323

Mauder M, Foken T, Clement R, Elbers JA, Eugster W, Grunwald T, Heusinkveld B, KolleO (2008) Quality control of carboeurope flux data - part 2: Inter-comparison of eddy-covariance software. Biogeosci 5(2):451–462

McFadden JP, Chapin I F Stuart, Hollinger DY (1998) Subgrid-scale variability in the surfaceenergy balance of arctic tundra. J Geophys Res 103(D22):28,947–28,961

Milesi C, Running S, Elvidge C, Dietz J, Tuttle B, Nemani R (2005) Mapping and mod-eling the biogeochemical cycling of turf grasses in the United States. Environ Manage36(3):426–438

Moore CJ (1986) Frequency response corrections for eddy correlation systems. Boundary-Layer Meteorol 37(1):17–35

Moriwaki R, Kanda M (2004) Seasonal and diurnal fluxes of radiation, heat, water vapor,and carbon dioxide over a suburban area. J Appl Meteorol 43(11):1700–1710

81


Moureaux C, Debacq A, Bodson B, Heinesch B, Aubinet M (2006) Annual net ecosystemcarbon exchange by a sugar beet crop. Agric Forest Meteorol 139(1-2):25–39

NCDC (2003) Climatography of the united states no. 20 1971-2000. http://cdo.ncdc.noaa.gov/climatenormals/clim20/mn/215435.pdf. Accessed 22 April 2010

Nemitz E, Hargreaves KJ, McDonald AG, Dorsey JR, Fowler D (2002) Micrometeorologicalmeasurements of the urban heat budget and CO2 emissions on a city scale. Environ SciTechnol 36(14):3139–3146

Rebmann C, Gockede M, Foken T, Aubinet M, Aurela M, Berbigier P, Bernhofer C, Buch-mann N, Carrara A, Cescatti A, Ceulemans R, Clement R, Elbers JA, Granier A, GrunwaldT, Guyon D, Havrankova K, Heinesch B, Knohl A, Laurila T, Longdoz B, Marcolla B,Markkanen T, Miglietta F, Moncrieff J, Montagnani L, Moors E, Nardino M, OurcivalJM, Rambal S, Rannik U, Rotenberg E, Sedlak P, Unterhuber G, Vesala T, Yakir D (2005)Quality analysis applied on eddy covariance measurements at complex forest sites usingfootprint modelling. Theor Appl Climatol 80(2):121–141

Ruimy A, Jarvis P, Baldocchi D, Saugier B (1995) CO2 fluxes over plant canopies and solarradiation: A review. Academic Press Volume 26:1–68

Schmidt A, Wrzesinsky T, Klemm O (2008) Gap filling and quality assessment of co2and water vapour fluxes above an urban area with radial basis function neural networks.Boundary-Layer Meteorol 126(3):389–413

Schotanus P, Nieuwstadt F, Bruin H (1983) Temperature measurement with a sonicanemometer and its application to heat and moisture fluxes. Boundary-Layer Meteorol26(1):81–93

Seeley M, Jensen B (2006) Minnesota weather almanac. Minnesota Historical Society Press,294 pp.

Soegaard H, Moller-Jensen L (2003) Towards a spatial CO2 budget of a metropolitan regionbased on textural image classification and flux measurements. Remote Sens Environ 87(2-3):283–294

Vesala T, Jarvi L, Launiainen S, Sogachev A, Rannik U, Mammarella I, Siivola E, Keronen P,Rinne J, Riikonen A, Nikinmaa E (2008) Surface—atmosphere interactions over complexurban terrain in Helsinki, Finland. Tellus B 60(2):188–199

Vickers D, Mahrt L (1997) Quality control and flux sampling problems for tower and aircraftdata. J Atmos Oceanic Tech 14(3):512–526

Vogt R, Christen A, Rotach MW, Roth M, Satyanarayana ANV (2006) Temporal dynamicsof CO2 fluxes and profiles over a central european city. Theor Appl Climatol 84(1):117–126

Webb E, Pearman G, Leuning R (1980) Correction of flux measurements for density effectsdue to heat and water vapour transfer. Q J Roy Meteorol Soc 106(447):85–100

82



10

3050

30

10

50

5

+

North

–100 100–50 500

–50

50

0

m

m

Figure 4.1.: Aerial photograph of the turf-grass flux site (U.S. Geological Survey, 2004)with the footprint climatology. The white lines are isopleths indicating landareas having the same relative contribution to the measured fluxes, averagedcumulatively over all valid measurements during 2007 and 2008. Solid linesrepresent 10% contour intervals and the dashed line indicates the 5% isopleth.Even though the footprint climatology does not intersect with the road, individ-ual 30-min footprints (not shown) will. The white cross indicates the positionof the flux tower. Areas to the east of the flux tower were behind the instrumentarray and included irrigated experimental plots of turf-grass and some smallmaple tree saplings; these wind sectors were screened out of our analyses.

83


0000

1200

2400

Jan Mar May Jul Sep Nov Jan Mar May Jul Sep Nov

CO2 F

lux [m

g C

m!2 s!

1 ]

2007 2008

Hour

of D

ay [U

TC-6

]

!0.20!0.15!0.10!0.05

0.000.050.100.15

Figure 4.2.: Fingerprint of the CO2 flux using gap-filled data for the years 2007 and 2008.Positive fluxes represent net efflux, whereas negative values represent net up-take, of CO2 by the ecosystem. The gray-shaded lines above the flux fingerprintindicate data availability. Black points indicate high-quality data, gray pointsrepresent data removed due to low quality, and white areas indicate gaps due toinstrument failure, tower maintenance, or power outages.

84


040

080

012

00

weekdays

!

!!

!

!!

!

!

!!

!

! !!

! !

!!

!

!

!!

!

!

!

!!

! !!

!

!

!

!

!!

!

!

!!

!

!

!

! !!

!

!

n veh

icles

[veh

icles

h!1 ]

!

!

high road impactlow road impact

no road impact

weekends and holidays

!!

!

!!

! !

!

!!

! !

!! !

!! !

!!

!!

!!

0.00

0.01

0.02

0.03

F CO 2

,mea

s [mg

C m

!2s!

1 ]

measured

0000 0600 1200 1800

040

080

012

00

!!

!

!

! !

!!

!

!

!!

!!

!

!

!

!!

!

!!

!

!!

!!

! !

! !

! !

!!

!

!

!

!

!

!

!!

!

!! !

!

hour of day [UTC-6]

n veh

icles

[veh

icles

h!1

]

trafficFCO2

0000 0600 1200 1800 2400

!!

!

!!

! !

!

! ! !

!

!

!!

!!

!! !

! !! !

0.00

0.01

0.02

0.03

F CO 2

,eco

[mg

C m!

2 s!1 ]

traffic flux removed

hour of day [UTC-6]

Figure 4.3.: Diurnal cycle of the traffic volume and the CO2 flux for weekdays (left panels)and for weekends and holidays (right panels) binned into 1-hour classes. Theselected data were from all winter days in 2007 and 2008 when the soil wasfrozen to a depth of at least 0.05 m and there was a continuous snow cover. Thegray band shows the 95% confidence interval for the traffic volume. The CO2fluxes are means for each 1-hour class and the corresponding error bars indicate˙1 standard error. For the weekdays, the fluxes were split into three classes,one with high road impact (f y road >0.001, round symbols and lined error bars),one with low road impact (0.001 > f y road >0.00005, open symbols and linederror bars), and no traffic impact (f y road D0 and wind direction <180ı, trianglesymbols and arrow error bars).

85


!

!!!!!!!!!!!! !

!!!!!

!!

!

!

!!

!!!

!!

!!

!!!!

!! !

!!!!

!

!!!!

!!

!!

!

!

!!!!

!

!!!

!

!

!

!!!!!

!!!

!

!!!

!!

!

!!

!

!!!!!!

!

!!

!!!!!!!!!!!

!!!!!

!!!!!!

!

!!!!!

!!

!!!!!!!!!!!!!!!!!!

!!

!!!!!

!!!!

!!!!!!!!!

!!!

!!

190 230 270 310 350

0.00

0.02

0.04

wind direction [deg]

F CO 2

,traf [m

g C

m!2 s!

1 ]

fetch [m]

346 78 60 78 346

Figure 4.4.: The figure shows how the traffic related flux FCO2;traf varied with the fetch size,i.e. the distance x from the tower to the road (upper horizontal axis), which inturn varied with the prevailing wind direction (lower horizontal axis). Boxesshow the quartiles and horizontal bars indicate the median for each wind direc-tion class. Whiskers show the lesser of the full data range or ˙ 1.5 times theinter-quartile range and open circles indicate data points >1.5 times the inter-quartile range. Within each box, the calculated CO2 contribution per vehicle (FCO2;traf=nvehicles) is linearly dependent on the footprint size (f y road).

86


!

!

!

!

!

!

!!!

!

!

!!!

!

!

!

!!

!

!

!

!

!

!

0 100 200 300 400 500 600

1000

015

000

2000

025

000

3000

0

nvehicles

emiss

ions [

mg C

m! 1

]emissions = (53.32 ± 0.32) " nvehiclesR2 = 0.999

0 100 200 300 400 500 600

nvehicles

hour

of d

ay [U

TC!6

]

0800

1000

1200

1400

1600

1800

62%52%

59%53%

55%59%

52%58%

49%48%50%

50%48%56%52%

52%53%

54%55%

55%58%

61%62%

59%70%

36%45%

37%42%

41%35%

41%35%

44%46%

45%43%

45%39%42%

42%43%

41%42%

42%40%

36%36%

38%28%

MotorcyclesTrucks, Busses, and VansPickups and SUVsPassenger Cars

Figure 4.5.: Upper panel: Relationship between the modelled CO2 emissions from trafficand the number of passing vehicles. Lower panel: Half hourly traffic composi-tion based on manual counts. The numbers in the bars show the proportion ofeach bar.

87


!

!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!!

!

!

!

!

!

!

!

!

!

!!

!

!

!

!!

!

!

!

!

!

!

!

!!

!

!!

!

!

!

!

!!

!

!

!

!

!!

!

!!

!!

!

!

!

!

!!

!

!

!

0.00 0.02 0.04 0.06 0.08 0.10

0.00

0.01

0.02

0.03

0.04

0.05

0.06

Fmodel [mg C m!2s!1]

F CO 2

,mea

s [mg

C m

!2s!

1 ]

FCO2,meas = (0.004 ± 0.001) + (0.38 ± 0.03) " Fmodel

R2=0.70

Figure 4.6.: Measured CO2 flux compared to modelled CO2 emissions by vehicles, scaled bythe crosswind-integrated footprint for the distance of the road. The line showsthe linear model between the plotted variables. The intercept of the linear re-gression can be interpreted as the biogenic wintertime flux.

2007 2008

[g C

m!2

year

!1]

!100

010

020

030

0

FCO2, measFCO2, trafFCO2, eco = NEE

Figure 4.7.: Annual CO2 budget separated into FCO2;meas (black), FCO2;traf (gray) and the dif-ference between these two fluxes, FCO2;eco (white).

88


Table 4.1.: Parameterization of the empirical model presented in Equation 4.1. The parame-ters in boldface were used to calculate the traffic related flux F CO2;traf.

Estimate Std. Estimate p-value R2

intercept Œmg Cm2 s � 5.6e-03 4.4-04 <0.001

a Œmg Cm s � 1.2e-01 2.3e-01 0.61

b Œ mg Cm2 s vehicle � 3.2e-05 1.7e-06 <0.001

c Œ mg Cm s vehicle � 5.8e-03 8.7e-04 <0.001

0.35

89

This chapter was published in Atmospheric Measure-

ment Techniques, 2010, 3, 1519–1531, doi:10.5194/amt-

3-1519-2010.

5Field intercomparison of two optical

analyzers for CH4 eddy covariance fluxmeasurements

Bela Tuzson1, Rebecca V. Hiller2, Kerstin Zeyer1, Werner Eugster2,

Albrecht Neftel3, Christof Ammann3, and Lukas Emmenegger1

1Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Air

Pollution and Environmental Technology, Dubendorf, Switzerland2Institute of Plant, Animal and Agroecosystem Sciences, ETH Zurich, Zurich, Switzerland

3Agroscope Reckenholz-Tanikon Research Station ART, Zurich, Switzerland

Abstract

Fast response optical analyzers based on laser absorption spectroscopy are thepreferred tools to measure field-scale mixing ratios and fluxes of a range of tracegases. Several state-of-the-art instruments have become commercially availableand are gaining in popularity. This paper aims for a critical field evaluation andintercomparison of two compact, cryogen-free and fast response instruments: aquantum cascade laser based absorption spectrometer from Aerodyne Research,Inc., and an off-axis integrated cavity output spectrometer from Los Gatos Re-search, Inc. In this paper, both analyzers are characterized with respect to pre-cision, accuracy, response time and also their sensitivity to water vapour. The

91

Chapter 5: Field intercomparison

instruments were tested in a field campaign to assess their behaviour under vari-ous meteorological conditions. The instrument’s suitability for eddy covarianceflux measurements was evaluated by applying an artificial flux of CH4 gener-ated above a managed grassland with otherwise very low methane exchange.This allowed an independent verification of the flux measurements accuracy,including the overall eddy covariance setup and data treatment. The retrievedfluxes were in good agreement with the known artificial emission flux, which ismore than satisfactory, given that the analyzers were attached to separate sonicanemometers placed on individual eddy towers with different data acquisitionsystems but similar data treatment that are specific to the best practice used bythe involved research teams.

5.1. Introduction

Understanding the temporal dynamics of methane emission at the global scale requires con-tinuous and long-term field measurements of CH4 fluxes at representative sites, where the re-lationships between landscape-scale flux measurements and their environmental drivers canbe investigated (Bartlett and Harriss, 1993; Bubier and Moore, 1994). The well establishedchamber technique, however, poorly captures the spatial and temporal variability of gas ex-change despite its undebated usefulness for small scale (plot-level) studies. The generalproblems associated with chamber flux measurements are leaks, inhibition of fluxes throughconcentration build-up and pressure effects, which are well known limitations of this method(Matthias et al., 1978; Bain et al., 2005; Camarda et al., 2009). Alternatively, micrometeoro-logical methods like eddy covariance (EC) integrate trace gas exchange over extended areas(typically hundreds of m2) – appropriate for landscape scale studies. However, these tech-niques rely on fast response, field-deployable and high-sensitivity instruments, which canrapidly resolve small (preferably better than 0.1%) concentration changes in CH4 at ambientlevel. Until recently, EC flux measurements of CH4 required frequent re-calibration and/orliquid nitrogen for the analyzers (Verma et al., 1992; Fowler et al., 1995; Friborg et al., 1997;Werle and Kormann, 2001) leading to logistic limitations for unattended field deployment.

Recently, a number of new instruments appeared on the market, which may overcomethese shortcomings. However, just a very few studies have addressed their application in thefield for EC flux measurements and they focus on only one single type of analyzer (Eug-ster et al., 2007; Eugster and Pluss, 2010; Neftel et al., 2007; Kroon et al., 2007; Hendrikset al., 2008; Smeets et al., 2009). For novel instruments, it is rather challenging to validatethe retrieved EC flux data, because one cannot rely on a reference method, and CH4 emis-

92

5.2. Material and methods

sions of natural sources are often very small. Although the analyzers may give the correctmixing ratio under laboratory conditions, these readings do not necessarily translate to arepresentative surface fluxes during a field campaign.

Here we characterize and compare two cryogen-free optical analyzers that allow for fastand high precision measurement of methane mixing ratios in ambient air. For the validationof flux data during field deployment, an artificial methane flux was generated at the surface ofan intensively managed grassland with otherwise very low methane flux. To our knowledge,this is the first example of an intercomparison for methane flux measurements where eddycovariance CH4 flux data were validated against a known, artificial source.


5.2.1. Instrumentation

In this study, we investigated two commercially available optical analyzers that are cryogen-free and allow for CH4 measurements at high (>10Hz) temporal resolution: 1) an off-axisintegrated-cavity output spectrometer (FMA, Fast Methane Analyzer, Model 908-0001, LosGatos Research Inc., Mountain View, CA) and 2) a dual continuous-wave (cw) quantumcascade laser based direct absorption spectrometer (QCLAS, Model QCL-76-D, AerodyneResearch Inc., Billerica, MA). Even though the optical implementation and the signal pro-cessing techniques used in the two instruments are quite different, the underlying principlesof infrared (IR) absorption spectroscopy apply to both analyzers. The trace gas mixing ratiosare determined based on the Beer-Lambert law:

ln.I0=I / D k�nL D Si��nL (5.1)

which relates the attenuation of light (I and I0 being the transmitted and incident light in-tensity, respectively) to the number density n [molecule/cm3] of absorbing molecules alongthe light path of length L [cm]. In this equation k� [1/(molecule cm�2)] denotes the spec-tral absorption coefficient for an isolated transition i , and it is defined by the line strengthof the transition Si [cm�1/(molecule cm�2)] and by the normalized line shape function ��[1/cm�1]. The number density, n, can easily be expressed in mole fraction of the absorbingspecies using the ideal gas law. Given that the minimum detectable mixing ratio is deter-mined by the noise-equivalent-power, absorption path length, absorption cross-section andthe incident laser power, it is evident that the detection sensitivity can be optimized usingseveral approaches.

93


The QCLAS employs a mid-IR quantum cascade laser as light source which is particularlyattractive due to its stable single mode spectral output, high power and near room tempera-ture operating condition. The advantage of selecting the mid-IR spectral region is that themajority of trace gas molecules of interest have their fundamental ro-vibrational absorptionbands located at these energies and therefore the absorption strengths are also the highesthere. This explains the ability of the analyzer, equipped with two lasers, to measure not onlyCH4 and H2O, but also the mixing ratios of N2O and NO2, respectively (see Table 5.1). Forthis study, however, only the CH4 and H2O data were considered. The signal-to-noise ratiois significantly improved by employing a multipass absorption cell which increases the ef-fective light path through the absorber by more than two orders of magnitude. Nevertheless,the main challenge in the direct absorption approach is to accurately measure small changes(�IDI0�I ) of two large quantities. This limitation can be overcome by the rapid sweepintegration technique, where the absorption spectrum is typically acquired at 5–10 kHz andthe co-addition of many individual scans results in rapid gains in signal-to-noise ratio. Op-tical fringes mainly resulting from unwanted etalons formed by reflections and scattering inthe multipass cell are effectively suppressed by a proper mechanical modulation of the cell’srear mirror using a piezo-crystal (McManus et al., 2006).

The FMA, on the other hand, operates in the near-IR region where low-cost distributedfeedback diode laser sources as well as photodetectors with high detectivity values are read-ily available, and fibre optics can be efficiently used. This results in very compact, robustand easy-to-build systems. However, the absorption strength (overtone transitions) of themolecules is 10 to 103 times weaker than in the mid-IR region. To circumvent this con-straint, a high-finesse optical cavity with high reflectivity mirrors (R>0:9999) is used toachieve an effective path length of light on the order of kilometers (see Table 5.1). Thus,the laser beam is trapped for tens of microseconds. The absorption signal is obtained in thiscase by the time integrated intensity of the light passing through the cavity with and withoutthe absorbing medium and expressed as:

�I=I0 D GA=.1CGA/ (5.2)

where A is the single-pass absorption [AD1� exp.�k�d/] and GDR=.1�R/. R is themirror reflectivity, and d is the interaction length of the laser beam with the absorbingmedium. Once the mirror reflectivity is determined from a known concentration standard, italso serves for the absolute calibration of the FMA.

For both analyzers, the absorption features are fitted to a Voigt profile, and the mixingratios are retrieved based on measured pressure, temperature, path length and molecular pa-

94


rameters, which are listed in spectral data-bases such as HITRAN (Rothman et al., 2009).While the FMA is stated to be calibrated by the manufacturer, the QCLAS requires calibra-tion by the user. More detailed descriptions of the instrumental design and signal processinghave been published by Nelson et al. (2002) for QCLAS and by Baer et al. (2002) for FMA,respectively.

5.2.2. Measurement site

The field measurements were conducted at the Oensingen experimental site, which is locatedon the Swiss Plateau (7ı440 E, 47ı170N) at an altitude of 450 m a.s.l. The managed grass-land field is part of the global FLUXNET observation network and the two large integratedEuropean projects CarboEurope and NitroEurope. The field is divided into two parts andthey differ in the management intensity (Ammann et al., 2007). However, during the presentexperiment in spring, both fields were covered by similar short grass vegetation and did notshow a significant methane exchange.

All experiments assessing the analyzers performance were performed in this field site,which was motivated by our effort to test the instruments under the same conditions as theyare employed in the EC measurements. The only exception was the water dilution experi-ment, which required, as it will be discussed in Sect. 5.3.1, a temperature-controlled labo-ratory environment for an accurate water vapour generation and consecutive measurementsby the analyzers.

5.2.3. Trace gas flux generation

An objective evaluation of the suitability and accuracy of the instruments for measuringmethane fluxes by the EC method was achieved by an artificial flux generation system,which mimics an amplified ecosystem-atmosphere gas exchange. For this purpose, stainlesssteel tubing of 1/400 outer diameter was used to construct a rectangular grid containing 30flow orifices (30 µm, Lenox Laser, USA) as gas outlets. As illustrated in Fig. 5.1, the gridconsisted of three parallel lines of 30 m length, 5 m apart to approximate a surface area of300 m2. Each line embodied 10 flow orifices placed every 3 m along the tubing to obtain aneven release of trace gas across the fumigation area.

The gas release experiment took place in two periods between 25–29 March (south-westerly wind directions, see Fig. 5.1) and 30 March–6 April 2009 (north-easterly winddirections). For the first period, the trace gas was obtained through continuous dilution ofpure CH4 with N2 to a concentration of 5%, while during the second period, a pressurizedcylinder (40 L, Messer, Switzerland) containing 1% mole fraction of methane in pure N2

95


was coupled to the manifold through a pressure regulator. The trace gas mixture was contin-uously released at a rate of 0.8 standard liters per minute (slpm) set by a mass flow controller(Vogtlin Instruments, Switzerland) at an overpressure of 182 kPa.

To relate the area of the artificial flux to the area seen by the measuring system, thefootprint of the flux measurements was calculated according to Neftel et al. (2008). Thecalculations are based on the analytical footprint model by Kormann et al. (2001). Thefootprint density function of a flux sensor is determined using readily available data from thestandard eddy covariance measurements. This footprint density function is then integratedover the fumigation source area given as quadrangular polygon (Fig. 5.1). Its fraction givesthe dilution factor of the measured flux in relation to average emission flux in the source area.In some cases, the footprint was also influenced by the neighbouring extensive grasslandfield. This was, however, not problematic since both fields had similar short grass vegetationduring the experiment in spring.

5.2.4. Flux measurements

During the field measurements, the FMA was placed in a weatherproof housing close to theeddy flux tower, whereas the QCLAS was kept in an air-conditioned (23˙2 ıCC) trailer tominimize temperature fluctuations. Additionally, the entire optical module of the QCLASwas insulated and maintained at 35˙0:1 ıCC. A constant flow of dry air purged the optics toprevent condensation, which may occur due to the low temperature (16 ıCC) of the coolingwater used for the QCL (Alpes Lasers, Switzerland ) and the IR-detectors (VIGO System,Poland). The trailer was located about 25 m away from the eddy flux tower in crosswinddirection. The instrument automatically switched between ambient air (four hours) andcalibration gas (two minutes). The FMA ran without any calibration procedure.

For the flux measurements, each trace gas analyzer was connected to a separate sonicanemometer (with the air inlet 0.25 m below the center of the sonic head), thus forming twoindividual eddy flux systems (see Table 5.1). The two EC systems were mounted close toeach other at a lateral distance of 0.7 m. In addition, CO2 and H2O fluxes were routinelymeasured with another EC system on the same field (Ammann et al., 2007) positioned at adistance of about 7 m. In order to better match the overlap between the size of the flux foot-print and the CH4 fumigation area, the measurement height of the EC systems was chosenrelatively low at 1.2 m above ground.

For the EC flux measurements, the FMA was connected to an external vacuum scrollpump (XDS-35i BOC, Edwards, UK) with a maximum pumping capacity of 580 lpm. Thesample was drawn through 6.7 m long tubing (8 mm i.d.) followed by two serially mounted

96

5.3. Results and discussion

filters (5 µ m and 0.3 µ m, AFM Series, SMC Inc., Japan) with droplet separator. The FMAhas a pressure controller that throttles the flow to maintain the cell at its target pressure of�19 kPa. Due to the pressure drop within the sampling system and analyzer, the pumpingspeed was reduced to an effective gas flow of about 30.5 slpm.

The QCLAS was connected to a scroll pump (TriScroll 600, Varian Inc., Italy) with amaximum pumping capacity of 420 lpm. As the QCLAS has no built-in pressure or flowcontrol system, it is left to the user to deal with these issues. In the present study, a mass flowcontroller (Red-y SmartSeries, Vogtlin Instruments, Switzerland) regulated the gas flow intothe absorption cell, while a high-flow throttle valve (SS-63TS8, Swagelok, USA) was usedto provide manual control of the downstream flow to keep the cell pressure at about 8 kPa.Since the analyzer was placed in a trailer 25 m away from the eddy flux tower, a custommade PTFE sampling tube (6 mm i.d., Hot Tube, Clayborn Lab, USA) was used to conductthe air stream to the instrument. The air sample was first filtered at the inlet by a slightlyheated Quartz filter (MK 360, Munktell Filter, Sweden) and secondly before entering theabsorption cell using a 7 µ m sintered metal filter (SS-4F-VCR-7, Swagelok, USA). Thesampling tubing was maintained at about 10 ıCC above ambient temperature by applying aconstant current through its heating elements. The air was dried using a Nafion drier (PD-100T, Perma Pure, USA) just before the analyzer. This setup resulted in a turbulent flow of13.5 slpm.

More details about the EC setups and data processing methods are given by Eugster andPluss (2010) and Neftel et al. (2010) for the FMA and QCLAS setup, respectively. In con-trast to Eugster and Pluss (2010), the methane signal of the FMA system was linearly de-trended prior to the flux calculations to be consistent with the data processing strategy of theQCLAS EC system.


In this section, we first present the results of the analyzers’ performance tests and discuss thevarious effects which may influence the retrieved mixing ratio values. Then, in situ methanemixing ratio measurements and CH4 flux determination of the artificially fumigated area willbe discussed. Finally, we compare the experimentally determined fluxes with the knownemissions of the fumigation area.

97


5.3.1. Instrument characterisation

Precision

The instrument performance in terms of detection limit and stability over time was charac-terized in the field by the Allan variance technique (Werle et al., 1993). For this purpose, theanalyzers were connected through a T-split to a pressurized air cylinder and they measuredthe same dry gas over 16 h. The gas flow rate was 2 slpm and the analyzers were config-ured for low flow mode, but maintaining the same cell pressure as in the high-flow modeused for EC measurements (see Table 5.1). The total gas consumption of the analyzers was1.28 slpm, while the rest was released through an overflow. The methane mixing ratio wasmeasured with a time resolution of one second. The precision was 1 and 3 ppb Hz�1=2 forFMA and QCLAS, respectively (Fig. 5.2). From the associated Allan variance plots an op-timum averaging time on the order of 500 s can be derived. This corresponds to a detectionprecision of 0.23 and 0.58 ppb, respectively. For integration times beyond 500 s, the Allanvariance levels off and even increases, which indicates the presence of 1=f -type and “drift”noise. It is noteworthy that the Allan variance may have significant variations in time. Thisbehaviour is indicated in Fig. 5.2 where the range defined by many individual 30 min Allanplots is shown (shaded area). Such representation illustrates that it is always possible toobtain “nice” Allan plots with low drift, and hence good precision, by selecting the best datasegment from an extended time series. Figure 5.2 also shows that for half hour averaging(commonly used in EC techniques) the advantage of the FMA’s higher short-time precisionis somewhat diminished by low frequency instrumental drifts.

The Allan variance plots in Fig. 5.2 are typical in the sense that they contain frequency in-dependent white noise and frequency dependent 1=f ˛ (˛>1) noise (Werle et al., 1993). Thelatter encompasses a noise at low frequencies which can be considered as drift. Accordingly,the Allan variance has a characteristic “V-shape” defined by the white-noise dominated re-gion at short integration times and by the domain of various drift effects at longer integra-tion times. Our results are consistent with other recent work (Kroon et al., 2007; Smeetset al., 2009; Eugster and Pluss, 2010; Bowling et al., 2009), suggesting that the achiev-able detection limit for methane with the existing analyzers is about 1 to 5 ppb Hz�1=2,which implies that variations of 0.05% in ambient methane mixing ratio are readily cap-tured. The only strikingly different Allan variance plot has recently been published byHendriks et al. (2008). This plot has no clear domains nor local minima, and the reportedprecision (6:1� 10�3 ppb Hz�1=2) is at least two orders of magnitude better than what is ex-pected from the stated precision (4.74 ppb) at a sampling rate of 10 Hz. Assuming a purewhite-noise behaviour of the analyzer, the estimated precision should rather be 4:74=

p10

98


or 1.5 ppb Hz�1=2.

Accuracy

In principle, direct absorption spectroscopy is an absolute method that should allow for astraightforward calculation of the mixing ratio from the measured signals. However, theaccuracy of this calculation depends on the knowledge of the molecular line parameters in-cluding line strength, pressure and temperature dependence (listed in the spectral databases),as well as instrumental laser line width and shape. Additionally, non-unity gain factors andpotential pressure broadening effects also bias the retrieved mixing ratios. Whereas theFMA is delivered with factory calibration, the QCLAS is shipped without calibration, andall corrections need to be performed by the user during a post processing step.

Three reference gases were used to determine the accuracy and the linearity of the ana-lyzers and to calibrate the QCLAS. One cylinder (1838:5˙0:3 ppb CH4) was prepared andcalibrated at the World Meteorological Organization (WMO) Central Calibration Labora-tory (CCL), while the other two (2273:1˙0:5 ppb and 2478:8˙0:5 ppb, respectively) weresecondary standards linked to the WMO mole fraction scale. A four-point calibration (in-cluding zero) was performed, but given the stability and linearity of the analyzers a two pointcalibration would be sufficient. The calibration factor was 1.0003 and 1.0516 for FMA andQCLAS, respectively. The reproducibility determined by replicate measurements of the tar-get tank was 0.9 ppb for FMA and 1.7 ppb for the QCLAS. It is noteworthy that the excellentaccuracy of the FMA has been maintained even after transport and under field deployment.

Response to water vapour

It is well known that density fluctuations arising from heat and water vapour fluxes affectthe measured flux densities of trace gases according to the Webb-Pearman-Leuning (WPL)theory (Webb et al., 1980). The following paragraphs are dealing with considerations relatedto water vapour that are not directly covered by the WPL theory, such as the effect of watervapour on the spectral line shape. Drying the air sample or measuring its absolute moisturecontent is required for an accurate methane mole fraction determination. This is, becausethe laser spectrometers measure the mixing ratio of CH4 to total pressure (Pt), including alsothe vapour partial pressure (PwDŒH2O�Pt), whereas the methane mixing ratio in calibrationgas is defined as dry air mole fraction. A dilution correction must thus be applied to thereported raw mixing ratio values

ŒCH4�dry=ŒCH4�wetPt=.Pt�Pw/ (5.3)

99


Furthermore, changing the gas composition by varying its water content will influencethe spectral line shape of the absorbing molecule (in this case CH4) due to the collisionalbroadening effect and thus, the derived mixing ratio value. As the line shape function (seeEq. 5.1) is determined by the physical mechanisms that perturb the energy levels of the tran-sition (Varghese and Hanson, 1984), it is obvious that perturbations caused by molecularcollisions may differ as a function of the colliding molecules type. Within the binary colli-sion assumption the collisional broadening effect is introduced as a Lorentz function with afull-width at half maximum ( c) that is proportional to the sample pressure

c DXj

jPj (5.4)

where j (cm�1 atm�1) is the transition dependent broadening coefficient quantifying theability of a molecular species j (e.g. N2, O2, and H2O) from the air sample to cause pressurebroadening due to the collisions with the absorbing molecules. The role of this collision-induced interference of spectral lines in ro-vibrational spectroscopy is in many cases un-derestimated. It is often stated that the instruments are interference free, which, however,only means that there is no spectral overlap between the absorption line of the trace gas ofinterest and the absorption features of other molecular species. Therefore, one might mis-leadingly assume that there is no cross-sensitivity to other ambient air species. Nevertheless,this assumption should be investigated carefully in the case of trace gas flux measurements,because water vapour concentrations can vary rapidly and are often correlated with verticalwind speed. Thus, any cross-sensitivity of trace gases to water may lead to an apparent flux.In principle, it would be possible to implement in the retrieval software an algorithm to ac-count for this collision-induced effect based on the theory of pressure broadening. However,the commonly used spectral database, HITRAN, contains only the dry air broadened half-width ( air) for most of the absorption lines (Brown et al., 2003). Investigations performedon simulated spectra which included the pressure broadening effect showed that quantifi-cation using least-square fit of a Voigt profile will systematically underestimate the mixingratio values. For ambient water vapour levels (<4%v), the magnitude of this underestima-tion can be well approximated using a linear dependence on the vapour partial pressure, andthus a simple cross-sensitivity correction can be applied to the measured CH4 mixing ratio:

ŒCH4�=ŒCH4�dry C bctPw=Pt (5.5)

where bct is the cross-sensitivity coefficient of the particular device (Neftel et al., 2010),which can also be determined empirically as shown below. Before doing so, we should

100


mention an important aspect regarding the above corrections. While the dilution effect de-pends only on the absolute amount of water vapour in the air sample, i.e. it is always presentin humid air sample, the water pressure broadening causing the cross-sensitivity effect willdepend linearly on the mixing ratio of the absorbing trace gas. This has an impact on thedetermination of the bct value. For example, the slope will be twice as steep when the mix-ing ratio of the absorbing trace gas is doubled, but it vanishes as the mixing ratio approacheszero.

To investigate the effect of increasing humidity on the retrieved CH4 mixing ratio, theanalyzers were installed in an air conditioned laboratory and measured the gas from a pres-surized dry air cylinder coupled to a water vapour generation system (LI-610, LI-COR Inc.,USA). This dew-point generator gradually increased the humidity level of the gas every20 min. The QCLAS simultaneously measured the methane mixing ratio and the watervapour in the humidified samples. Thus, every single CH4 mixing ratio measurement can benormalized to dry conditions (see Eq. 5.3) given that the humidity data are calibrated. TheLI-610 has been considered as reference point for absolute water concentration. The resultsof the water dilution experiment are shown in Fig. 5.3. It clearly indicates that correcting fordilution by water vapour only is not sufficient and, because no direct spectral interference towater vapour is expected in this spectral region, the observed behaviour is attributed to theline broadening effect of the increasing water content in the gas-matrix (Neftel et al., 2010).Moreover the effect shows, in accordance with the theoretical prediction, a clear dependenceon the water vapour, which although is non-linear, it still can be approximated by a linearfunction at usual ambient water vapour concentrations (<4%v). In our case its magnitudeis about 16% of the density correction, but as mentioned earlier, the exact value depends onthe CH4 mixing ratio as well as on the absorption lines which are used to retrieve the mixingratios. In this context, it is rather a coincidence that both analyzers showed exactly the samecross-sensitivity effect, because the absorption lines analyzed by the instruments are locatedat different (near and mid-IR, respectively) spectral regions. In contrast to the QCLAS, theFMA measures a spectral feature of CH4 that is a combination of four individual, but closelyspaced, and thus not resolved, absorption lines (see Table 5.1). Nevertheless, this empiri-cal approach allows for the determination of the cross-sensitivity coefficient bct without theknowledge of the water pressure broadening coefficient value.

We also tried to repeat the dilution experiment in the field by gradually adding watervapour to the gas stream from a pressurized air cylinder as described by Neftel et al. (2010),but in this case the FMA showed only a random response to the increased humidity. Onecould only speculate about the nature of such behaviour. One possible source could bethe presence of two serially mounted (5 µ m and 0.3 µ m) filters with droplet separators

101


upstream of the FMA. These elements were used to further protect the cell mirrors duringthe campaign, but they may also have dampened the water vapor levels considerably. Sincewe had no means to determine the actual water content inside the cell of the FMA, the datainterpretation became difficult.

To summarize, the dilution experiment gives us solid evidence that water vapour flux mayintroduce an apparent methane flux similar to the water vapour density effect. This is mostsignificant for small CH4 fluxes and large water vapour fluxes. In the density flux correctionequation proposed by Webb et al. (1980) there are three terms:

F D w0�0c„ƒ‚…I

C� .�c=�a/ w0�0v„ ƒ‚ …II

C .1C ��/��c=T

�w0T 0„ ƒ‚ …

III

; (5.6)

where w is vertical wind speed in m s�1, �a, �v, and �c are the densities of air, watervapour, and a trace gas c, respectively, in kg m�3, T is air temperature in K, and �Dma=mv

and �D�v=�a. ma and mv are the molar masses (“weights”) of dry air and water vapour,respectively, in kg mol�1. Term (I) is the measured flux, term (II) is the moisture flux cor-rection, and term (III) is the sensible heat flux correction. Since temperature fluctuations arelargely under control in a closed-path CH4 analyzer such as the ones used here, this termbecomes negligibly small and hence we will focus on term (II). This term is non-negligibleif water vapour is not removed from the air stream prior to analysis or the measured mix-ing ratios are not individually corrected for the dilution effect. As shown by Neftel et al.(2010), the water cross-sensitivity on CH4 can formally be treated similar to the density cor-rection. Its influence increases linearly with the water flux and will consequently be mostpronounced in summer conditions with high transpiration of the plant canopy. This effect isillustrated by a series of measurements over the same grassland field taken in August 2008,using the same QCLAS setup, but without drying the air samples. Consequently both, thedensity (WPL) and the cross-sensitivity corrections had to be applied. This was performednumerically on the individual 20 Hz raw data values. This procedure avoids the potentialoverestimation of the correction due to the fact that the residence time of water moleculesin the sampling line is slightly longer than that of non polar molecules such as CH4 (Ibromet al., 2007). Figure 5.4 shows the scatter plot of the evaluated CH4 EC fluxes at differentstages of corrections. The corrected fluxes have been evaluated with a prescribed lag thatwas derived from the maximum of the covariance function of the non corrected fluxes. Thislag corresponds to the expected time delay based on the tube length, diameter and pumpspeed. The strong negative apparent CH4 fluxes up to �15 nmol m�2 s�1 are evidently in-duced by water vapour fluxes, because the CH4 exchange fluxes of the grassland system inOensingen are expected to be very small. Applying the density correction reduces these

102


negative fluxes by about a factor of 5. However, this still exceeds plausible limits of CH4

uptake fluxes (Neftel et al., 2010; Smith et al., 2000). The second correction brings the CH4

fluxes to slightly positive values that are individually not significantly different from zero,but do show a tendency to emission fluxes and positive correlation with the water vapourflux. These positive fluxes are on average higher than potential emissions from grasslandvegetation reported by Keppler et al. (2006). Yet the results are strongly dominated by theuncertainties in the applied correction factors (water calibration of the QCLAS, cross inter-ferences of the water vapour on the CH4 mixing ratio) and are sensitive to any slight changein these factors.

Although not considered in this study, it should still be mentioned that an additional effectmay also influence the final CH4 flux values. This effect has its origin in the well knownphysical process of adsorption of gas molecules on solid surfaces predominantly due toattractive Van der Waals forces. In the classical model of gas adsorption (Langmuir’s model)it is assumed that gas molecules striking the surface have a given probability of becomingadsorbed, while molecules already adsorbed similarly have a given probability of desorbing,which leads to a dynamic equilibrium where the average number of molecules per unit areaof surface per time interval are constant. The accumulation of a specific molecule on a solidsurface is influenced by many factors involved in the interactions between the solid surface,water and the species concerned. This also means that there is a possibility that trace gaseslike CH4 could be influenced by changes in the air humidity due to the competitive sorptionof water vapour on the surface. Given that water molecules have a high potential to replaceother molecules from surfaces and variation in the water vapour concentrations in high fluxconditions are on the order of 10%, i.e. about 5 orders of magnitude higher than the expectedCH4 variations, it may be worthwhile to consider this effect in future investigations.

Influence of temperature

The QCLAS is rather sensitive to ambient temperature fluctuations. This is due to the relativelarge number of optical elements and the thermoelectrically cooled (TEC) infrared detectors(Vigo Systems, PL). As mentioned in Sect. 5.2.1, the signal-to-noise ratio can be stronglyinfluenced by optical fringes whose magnitude mainly depends on the alignment accuracyof the optical system. Furthermore, the employed detectors are particularly sensitive to theaiming angle of incident light. This is due to the small size of the photodiodes (0.5 mm2

active area) and to their optical immersion into high refractive index GaAs hyperhemispher-ical lenses. The influence of the changes in the optical aiming on the measured CH4 mixingratio was empirically estimated by deliberately creating a misalignment which caused 5%decrease in the laser intensity on the detector. This resulted in an apparent change of 23 ppb

103


CH4. Considering the laser intensity variations of about 0.35% over many hours during thefield campaign, the noise contribution to the analyzer precision was about 1.6 ppb at most.

The influence of changing cell temperature and pressure on the measured methane mixingratio by FMA has also been investigated. Even though the analyzer showed an excellentstability during the field test, it was not completely immune to changes in ambient conditionsand a temperature dependence of 0.84 ppb K�1 could be established. Additionally, the CH4

mixing ratio values systematically dropped by 51 ppb when switching from low to high flowmode. This, however, has no influence on typical eddy covariance measurements.

As a concluding remark, the QCLAS is more sensitive to temperature fluctuations thanthe FMA. However, under well controlled environment this dependence can be efficientlysuppressed. For both setups, the drifts due to ambient temperature changes are too slow tohave any significant impact on EC flux measurements, but they are likely to affect accuracyof absolute mixing ratio measurements in case of long-term monitoring applications.

Response time

In order to evaluate their responsiveness, the analyzers were placed in the field in theirtypical sampling configuration as used for the eddy-covariance flux measurements. TheFMA sample gas flow of about 30.5 slpm corresponds to 165 lpm gas flow at a nominal cellpressure of 19 kPa. Thus, the cell volume of 0.408 L is refreshed at approximately every0.15 s. In this configuration the data acquisition was set to 20 Hz. For the QCLAS system,the sample flow of 13.5 slpm corresponds to 171 lpm at a nominal cell pressure of 8 kPa.Thus, the cell volume of 0.5 L is refreshed at approximately every 0.18 s. To be consistentwith this physical response time, the spectral retrieval rate was set to 8 Hz, although dataacquisition rates of up to 20 Hz would have been possible.

The above reported values for the cell refresh rate were calculated based on the cell vol-ume divided by the actual pumping speed. The real response time (first order, 1=e) was alsoempirically determined in the field, through the entire sampling setup, by rapidly switchingfrom ambient air to calibration gas and fitting the rising/falling tail to an exponential func-tion. The instrumental response (0.23 s for QCLAS and 0.4 s for FMA) of this empiricalapproach is slightly slower than expected based on the well-mixed reactor model. This canbe explained by the presence of filter and drier elements in the sampling system which maycause some smearing. This is consistent with the cut-off frequency of about 2 Hz in thecospectra.

In summary, the response time of both setups was slightly longer than expected from themanufacturers’ specifications, which typically consider the cell volume and the sample flow.However, since this parameter is properly taken into account by the dampening correction,

104


it does not pose any significant limitation on the flux measurements.

5.3.2. CH4 mixing ratio and fluxes

Time series of the methane mixing ratios during the field campaign are shown in Fig. 5.5.The CH4 mole fraction values retrieved by the QCLAS were averaged to 1 min, calibratedand reported in dry air. The comparison of the data sets obtained from the two instruments isillustrated by a scatter plot with the associated histogram of their ratio. The combined effectof water dilution and pressure broadening experienced by the FMA led to a shifted andbroadened distribution profile of the histogram. This is because contrary to the FMA whichdirectly measured CH4 in humid air samples, the strategy for the QCLAS measurements wasto dry the air. Nevertheless, available air humidity measurements at the site allowed for anapproximate correction of water effects on the methane mole fraction values measured bythe FMA. It is evident that not drying the air only led to about 1% error at most in absolutemixing ratio during the entire field campaign, which is acceptable for most applications ifonly methane mixing ratios have to be measured, but not for EC flux measurements withoutappropriate corrections.

The excellent agreement of both analyzers for mixing ratios does not automatically implycomparable results with respect to flux measurements. Thus, we have also tested whether (i)the sensitivity of the two CH4 analyzers is sufficient to resolve the artificial flux produced byfumigation from a rectangular surface and (ii) the calculated flux agrees with the flux valueestimated based on the applied emission rates.

Figure 5.6 shows cospectra and power spectra obtained during a period with high artificialCH4 flux (102 nmol CH4 m�2 s�1). Since the two instruments gave very consistent resultsduring the entire field campaign, these spectra can be considered as representative for oursetups under stationary turbulence conditions. Contrary to the ideal Kaimal spectra, realspectra show systematic damping in the high frequency part, which must be corrected. Forthe QCLAS system, an empirical approach described by Ammann et al. (2006) was used.This correction can be parameterized as a function of the wind speed and will be specific tothe used set-up. For the FMA system, the correction was derived from the damped cospectrawhich are approximated by a statistical best fitted theoretical cospectra as suggested by Eug-ster and Senn (1995). The difference between the two methods was analyzed and quantifiedby calculating the damping loss correction for the QCLAS with the same approach as for theFMA. This would lead to an average reduction of correction of about 6% for the QCLASEC-system.

The agreement of measured cospectra with the idealized Kaimal cospectra (Kaimal et al.,

105


1972, with minor modifications as shown by Eugster and Senn, 1995) is remarkable for bothinstruments, despite the larger noise level of the QCLAS (see Fig. 5.6). It is also noteworthy,that the significantly larger white noise level of the QCLAS does not compromise its per-formance for flux measurements. This is not unexpected since the eddy covariance methodis a random-noise rejection method, that is, because random noise of the CH4 sensor is un-correlated with random noise of the wind sensor, the covariances and thus the cospectra areunaffected by pure random noise. What is however clearly seen is the fact that both systemsare not perfect at the highest frequencies and the cospectra differ from idealized conditionsas expected from first-order damping (Eugster and Senn, 1995).

The main goal of the fumigation experiment was to investigate whether the EC systemsadequately measure the simulated flux. Therefore, the calculated 30-min fluxes were com-pared to each other as well as against the fumigation flux (see Fig. 5.7). In order to comparethe measured fluxes with the emissions from the fumigation area, the footprint correctionaccording to Neftel et al. (2008) was applied. This correction was calculated for each sys-tem individually, as the towers were slightly apart from each other (0.7 m). Yet the footprintcorrection factors for the two flux systems differed mostly by less than 10%. Periods withless than 20% contribution from the artificial source were not included in further analysisbecause the uncertainties inherent to the footprint calculation were considered as too large.The agreement between the footprint corrected fluxes and the released trace gas flux wasremarkable and even fluctuations in the flux due to changes in wind direction occured simul-taneously. Nevertheless, the agreement with the known source was poor during these condi-tions, because the sudden shifts in wind direction might not be reflected precisely enough bythe footprint calculated with half-hourly averaged micrometeorological quantities. As soonas the wind direction was constant, the scatter of the measured and footprint weighted fluxeswas getting very small. The methane flux measured by the QCLAS system was systemati-cally slightly higher than the flux determined based on the FMA measurements. However,the mean difference was 5.9 nmol CH4 m�2 s�1 and thus well below the uncertainty thatone would expect for eddy covariance flux measurements. This is remarkable given thatthe relevant data included not only mixing ratio measurements but also all assumptions andprocedures used during flux calculation and the relevant footprint simulation.

We also recall that during the field campaign the sample was dried for QCLAS measure-ments, whereas the FMA was analyzing humid air. Thus, for an unbiased comparison, themethane flux retrieved by the FMA system had to be corrected for density fluctuations (WPLterm) as well as for cross-sensitivity effect according to the approach of Neftel et al. (2010).For illustration, we calculated term II of Eq. (5.6) for typical conditions experienced dur-ing our experiments. Taking for the water flux w0�0vD4mmol m�2 s�1, the average methane

106

5.4. Conclusions

mixing ratio ŒCH4�= 2 ppm (from Fig. 5.5) and the water effect quantified by the dilution ex-periment (from Fig. 5.3), the term II results in a flux of 5.6 nmol m�2 s�1. This conservativeestimate shows that the resulting correction is relatively small in our case compared to thelarge, artificially issued flux of 84 nmol m�2 s�1 (see Fig. 5.7). Under certain conditions –especially for high water vapour fluxes combined with low methane fluxes – the WPL andcross-sensitivity corrections might well dominate the flux measurements as has been shownby Neftel et al. (2010) for background N2O flux measurements.

5.4. Conclusions

This paper demonstrates the benefits of using infrared laser absorption spectroscopy for insitu, fast and high precision ambient methane mixing ratio measurements. The compactand cryogen-free instruments can be operated unattended in the field to provide continuousmeasurements of CH4 mixing ratios. Their fast response provides the opportunity to in-vestigate the temporal dynamics of the mechanisms controlling ecosystem-atmosphere CH4

exchange. Overall, both systems are performing remarkably well for flux measurements.To our knowledge, this is the first example of an intercomparison of direct eddy covariancemethane flux measurements and the first time that eddy covariance CH4 flux data has beenvalidated based on a known, artificial methane emission.

Detailed investigation of the instrumental response to various parameters revealed, inagreement with recent publications (Neftel et al., 2010; Chen et al., 2010), that infraredgas analyzers employing laser absorption techniques may show some response to the gas-matrix composition, especially to water vapour. The empirically found correlation of CH4

mixing ratio with sample gas humidity is in accordance with the theoretical expectationof the additional pressure broadening effect induced by the water on the CH4 absorptionline. Therefore, it is suggested that for eddy covariance measurements of small fluxes suchcross-sensitivity effects should be taken into consideration independently of the employedwavelength and spectroscopic technique if (1) the sample is not dried prior to analysis or (2)the corresponding corrections are not implemented in the retrieval algorithms.

Ecosystem research will certainly have high benefits of recent developments achieved ininfrared laser and detector technologies. This is particularly true for the quantum cascadelaser fabrication. The latest continuous-wave (cw) QCLs can operate at room-temperature(up to 50ıC C) and have much higher output power and, thus, they are expected to allow sig-nificant improvements in the sensitivity of both direct absorption and cavity enhanced spec-troscopy. It is also noteworthy, that the QCLAS used in this study simultaneously detectsN2O and NO2 with sufficient precision and time resolution for eddy covariance flux mea-

107


surements. Similarly, the most recent cavity enhanced analyzers include multiple species,e.g. CH4, CO2 and H2O, and perform dilution and line-broadening correction on the re-trieved mixing ratio values. Finally, for both QCLAS and FMA, new analyzers capable ofcarbon isotope specific on-line CH4 measurement are under development.

Acknowledgements BT thanks Albert Manninen for sharing the Matlab code to gener-ate absorption spectra which include the water broadening contribution. This studywas supported by the COST action 729 project “Assessment of nitrogen biosphere-atmosphere exchange based on novel quantum cascade laser technology” SBF Nr.C06.0017, and by the project MAIOLICA of the ETH Competence Center Environ-ment and Sustainability (CCES).

5.5. References

Ammann, C., Brunner, A., Spirig, C., and Neftel, A.: Technical note: Water vapour con-centration and flux measurements with PTR-MS, Atmos. Chem. Phys., 6, 4643–4651,doi:10.5194/acp-6-4643-2006, 2006.

Ammann, C., Flechard, C. R., Leifeld, J., Neftel, A., and Fuhrer, J.: The carbon budget ofnewly established temperate grassland depends on management intensity, Agr. Ecosyst.Environ., 121, 5–20, doi:10.1016/j.agee.2006.12.002, 2007.

Baer, D. S., Paul, J. B., Gupta, M., and O’Keefe, A.: Sensitive absorption measurements inthe near-infrared region using off-axis integrated-cavity-output spectroscopy, Appl. Phys.B, 75, 261–265, doi:10.1007/s00340-002-0971-z, 2002

Bain, W. G., Hutyra, L., Patterson, D. C., Bright, A. V., Daube, B. C., Munger, J. W.,and Wofsy, S. C.: Wind-induced error in the measurement of soil respi-ration using closed dynamic chambers, Agr. Forest Meteorol., 131, 225–232,doi:10.1016/j.agrformet.2005.06.004, 2005.

Bartlett, K. B. and Harriss, R. C.: Review and assessment of methane emissions from wet-lands, Chemosphere, 261(1–4), 261–320, doi:10.1016/0045-6535(93)90427-7, 1993.

Bowling, D. R., Miller, J. B., Rhodes, M. E., Burns, S. P., Monson, R. K., and Baer, D.: Soil,plant, and transport influences on methane in a subalpine forest under high ultravioletirradiance, Biogeosciences, 6, 1311–1324, doi:10.5194/bg-6-1311-2009, 2009.

Brown, L. R., Benner, D. C., Champion, J. P., Devi, V. M., Fejard, L., Gamache, R. R.,Gabard, T., Hilico, J. C., Lavorel, B., Loete, M., Mellau, G. C., Nikitin, A., Pine, A. S.,Predoi-Cross, A., Rinsland, C. P., Robert, O., Sams, R. L., Smith, M. A. H.,Tashkun, S. A., and Tyuterev, V. G.: Methane line parameters in HITRAN, J. Quant.Spectrosc. Ra., 82, 219–238, doi:10.1016/S0022-4073(03)00155-9, 2003.

108

5.5. References

Bubier, J. L. and Moore, T. R.: An ecological perspective on methane emissions from north-ern wetlands, Trends Ecol. Evol., 9, 460–464, doi:10.1016/0169-5347(94)90309-3, 1994.

Camarda, M., Gurrieri, S., and Valenza, M.: Effects of soil gas permeability and recirculationflux on soil CO2 flux measurements performed using a closed dynamic accumulationchamber, Chem. Geol., 265, 387–393, doi:10.1016/j.chemgeo.2009.05.002, 2009.

Chen, H., Winderlich, J., Gerbig, C., Hoefer, A., Rella, C. W., Crosson, E. R., Van Pelt, A.D., Steinbach, J., Kolle, O., Beck, V., Daube, B. C., Gottlieb, E. W., Chow, V. Y., Santoni,G. W., and Wofsy, S. C.: High-accuracy continuous airborne measurements of greenhousegases (CO2 and CH4) using the cavity ring-down spectroscopy (CRDS) technique, Atmos.Meas. Tech., 3, 375–386, doi:10.5194/amt-3-375-2010, 2010.

Eugster, W. and Pluss, P.: A fault-tolerant eddy covariance system for measuring CH4 fluxes,Agr. Forest Meteorol., 150, 841–851, doi:10.1016/j.agrformet.2009.12.008, 2010.

Eugster, W. and Senn, W.: A cospectral correction model for measurement of turbulent NO2

flux, Bound.-Lay. Meteorol., 74, 321–340, doi:10.1007/BF00712375, 1995.

Eugster, W., Zeyer, K., Zeeman, M., Michna, P., Zingg, A., Buchmann, N., and Emmeneg-ger, L.: Methodical study of nitrous oxide eddy covariance measurements using quan-tum cascade laser spectrometery over a Swiss forest, Biogeosciences, 4, 927–939,doi:10.5194/bg-4-927-2007, 2007.

Fowler, D., Hargreaves, K. J., Skiba, U., Milne, R., Zahniser, M. S., Moncrieff, J. B., Bev-erland, I. J., Gallagher, M. W., Ineson, P., Garland, J., and Johnson, C.: Measurements ofCH4 and N2O fluxes at the landscape scale using micrometeorological methods, Philos T.Roy. Soc. A, 351, 339–356, 1995.

Friborg, T., Christensen, T., and Søgaard, H.: Rapid response of greenhouse gas emissionto early spring thaw in a subarctic mire as shown by micrometeorological techniques,Geophys. Res. Lett., 24, 3061–3064, doi:10.1029/97GL03024, 1997.

Hendriks, D. M. D., Dolman, A. J., van der Molen, M. K., and van Huissteden, J.: A compactand stable eddy covariance set-up for methane measurements using off-axis integratedcavity output spectroscopy, Atmos. Chem. Phys., 8, 431–443, doi:10.5194/acp-8-431-2008, 2008.

Ibrom, A., Dellwik, E., Flyvbjerg, H., Jensen, N. O., and Pilegaard, K.: Strong low-pass fil-tering effects on water vapour flux measurements with closed-path eddy correlation sys-tems, Agr. Forest Meteorol., 147, 140–156, doi:10.1016/j.agrformet.2007.07.007, 2007.

Kaimal, J. C., Wyngaard, J. C., Izumi, Y., and Cote, O. R.: Spectral characteristics of surface-layer turbulence, Q. J. Roy. Meteorol. Soc., 98, 563–589, 1972.

Keppler, F., Hamilton, J. T. G., Braß, M., and Rockmann, T.: Methane emissions from ter-restrial plants under aerobic conditions, Nature, 439, 187–191, doi:10.1038/nature04420,2006.

109


Kormann, R., Muller, H., and Werle, P.: Eddy flux measurements of methane over the fen“Murnauer Moos”, 11ı110 E, 47ı390N, using a fast tunable diode laser spectrometer, At-mos. Environ., 35, 2533–2544, doi:10.1016/S1352-2310(00)00424-6, 2001.

Kroon, P. S., Hensen, A., Jonker, H. J. J., Zahniser, M. S., Van’t Veen, W. H., and Ver-meulen, A. T.: Suitability of quantum cascade laser spectroscopy for CH4 and N2O eddycovariance flux measurements, Biogeosciences, 4, 715–728, doi:10.5194/bg-4-715-2007,2007.

Matthias, A. D., Yarger, D. N., and Weinbeck, R. S.: A numerical evaluation ofchamber methods for determining gas fluxes, Geophys. Res. Lett., 5, 765–768,10.1029/GL005i009p00765, 1978.

McManus, J. B., Nelson, D. D., Herndon, S. C., Shorter, J. H., Zahniser, M. S., Blaser, S.,Hvozdara, L., Muller, A., Giovannini, M., and Faist, J.: Comparison of cw and pulsedoperation with a TE-cooled quantum cascade infrared laser for detection of nitric oxide at1900 cm�1, Appl. Phys. B, 85, 235–241, 2006.

Neftel, A., Flechard, C., Ammann, C., Conen, F., Emmenegger, L., and Zeyer, K.: Experi-mental assessment of N2O background fluxes in grassland systems, Tellus B, 59, 470–482,doi:10.1007/s00340-006-2407-7, 2007.

Neftel, A., Spirig, C., and Ammann, C.: Application and test of a simple tool for operationalfootprint evaluations, Environ. Pollut., 152, 644–652, doi:10.1016/j.envpol.2007.06.062,2008.

Neftel, A., Ammann, C., Fischer, C., Spirig, C., Conen, F., Emmenegger, L., Tuzson, B.,and Wahlen, S.: N2O exchange over managed grassland: application of a quantum cas-cade laser spectrometer for micrometeorological flux measurements, Agr. Forest Mete-orol., 150, 775–785, doi:10.1016/j.agrformet.2009.07.013, 2010, special Issue on EddyCovariance (EC) flux measurements of CH4 and N2O.

Nelson, D. D., Shorter, J. H., McManus, J. B., and Zahniser, M. S.: Sub-part-per-billiondetection of nitric oxide in air using a thermoelectrically cooled mid-infrared quantumcascade laser spectrometer, Appl. Phys. B, 75, 343–350, doi:10.1007/s00340-002-0979-4, 2002.

Rothman, L. S., Gordon, I. E., Barbe, A., et al.: The HITRAN 2008 molecular spectroscopicdatabase, J. Quant. Spectrosc. Rad., 110, 533–572, doi:10.1016/j.jqsrt.2009.02.013, 2009.

Smeets, C. J. P. P., Holzinger, R., Vigano, I., Goldstein, A. H., and Rockmann, T.: Eddycovariance methane measurements at a Ponderosa pine plantation in California, Atmos.Chem. Phys., 9, 8365–8375, doi:10.5194/acp-9-8365-2009, 2009.

Smith, K. A., Dobbie, K. E., Ball, B. C., Bakken, L. R., Sitaula, B. K., Hansen, S.,Brumme, R., Borken, W., Christensen, S., Prieme, A., Fowler, D., Macdonald, J. A.,Skiba, U., Klemedtsson, L., Kasimir-Klemedtsson, A., Degorska, A., and Orlanski, P.:Oxidation of atmospheric methane in Northern European soils, comparison with other

110

5.5. References

ecosystems, and uncertainties in the global terrestrial sink, Glob. Change Biol., 6, 791–803, doi:10.1046/j.1365-2486.2000.00356, 2000.

Varghese, P. L. and Hanson, R. K.: Collisional narrowing effects on spectral line shapesmeasured at high resolution, Appl. Optics, 23, 2376–2385, doi:10.1364/AO.23.002376,1984.

Verma, S. B., Ullman, F. G., Billesbach, D., Clement, R. J., Kim, J., and Verry, E. S.: Eddycorrelation measurements of methane flux in a northern peatland ecosystem, Bound.-Lay.Meteorol., 58, 289–304, doi:10.1007/BF02033829, 1992.

Webb, E., Pearman, G., and Leuning, R.: Correction of flux measurements for density effectsdue to heat and water vapour transfer, Q. J. Roy. Meteorol. Soc., 106, 85–100, 1980.

Werle, P. and Kormann, R.: Fast chemical sensor for eddy-correlation measure-ments of methane emissions from rice paddy fields, Appl. Optics, 40, 846–858,doi:10.1364/AO.40.000846, 2001.

Werle, P., Mucke, R., and Slemr, F.: The limits of signal averaging in atmospheric trace-gasmonitoring by tunable diode-laser absorption spectroscopy (TDLAS), Appl. Phys. B, 57,131–139, doi:10.1007/BF00425997, 1993.

111



Figure 5.1.: Schematic map of the experimental field setup. (1) position of both EC fluxsystems: sonic anemometers and sample tube inlets for QCLAS and FMA; (2)weatherproof housing with the FMA; (3) trailer housing the QCLAS; (4) and (5)alternate positions of fumigation grids for the two main wind directions: circlesindicate the position of the 30 release orifices, the dashed enveloping rectanglerepresents the source area used in the footprint calculation (assuming a constantaverage emission flux). The contour lines illustrate the flux footprint with 30%,50%, 70%, and 90% flux contribution for a northeasterly wind case on 31 March2009 11:00–14:30 CEST (uD3:8m s�1, z=LD�0:1).

112


0.001

0.01

0.1

1

10

Alla

n Va

rianc

e (p

pb2 )

12 4 6 8

102 4 6 8

1002 4 6 8

1000Integration Time (s)

2290

2280

2270

2260

CH

4 [pp

b]60x10350403020100

collected data time (s)

QCLAS FMA

Figure 5.2.: Precision and stability over time of both analyzers quantified by the Allan vari-ance technique. The top part shows the CH4 mixing ratio measured over 16 hby the two analyzers from a pressurized air cylinder. Data from the QCLAS areshown in light gray, those from the FMA are in dark gray. The bottom part isthe log-log plot of the sample variance as a function of the averaging time. Thedashed line indicates the white-noise behavior, while the black line is the vari-ance associated to the long-term measurement. The time series was also splitinto 30 min sequences and all individual Allan variance plots were calculated.The filled areas thus represent the envelope of all individual 30 min Allan vari-ance plots, indicating the range that typical 30 min Allan variances would fallinto.

113


2.09 2.09

2.08 2.08

2.07 2.07

2.06 2.06

2.05 2.05

2.04 2.04

2.03 2.03

2.02 2.02

2.01 2.01

CH

4 [pp

m]

2.52.01.51.00.50.0H2O [%]

2.09

2.08

2.07

2.06

2.05

2.04

2.03

2.02

2.01

CH

4 [pp

m]

16:00 17:00 18:00Time (hh:ss)

2.5

2.0

1.5

1.0

0.5

0.0

H2 O

[%]

y = 2.0853 - 0.024783(6)·x

CH4 H2O

Figure 5.3.: Instrumental response to varying water vapor content. Top graph shows themethane mixing ratio changes measured by QCLAS over time induced by thestepwise addition of water vapour. The same behaviour was observed for bothanalyzers. The linear dependence of the CH4 mixing ratio on the water vapourconcentration as measured by QCLAS is indicated on the bottom graph. Thegray dots show the original one-second data while the circles represent 20 minaverages (with 1 � error bars) for the individual dilution steps. The solid lineis a linear fit through the averaged data points. The dashed line indicates theapparent CH4 mixing ratio after correction of the volumetric dilution (Eq. 5.3)by water vapor. The real (dry air) mixing ratio (dotted line) can only be ob-tained when an additional correction factor (Eq. 5.5) is applied to account forthe pressure broadening effect.

114


-20

-15

-10

-5

0

5

10

CH

4 flu

x [n

mol

m-2

s-1]

543210H2O flux [mmol m-2s-1]

no correction (raw flux) density (WPL) correction only density and cross-sensitivity correction

Figure 5.4.: Scatter plot of the evaluated CH4 fluxes at different stages of corrections. Thesedata were recorded with the QCLAS at the same site during August 2008 whenwater vapour fluxes were high and no significant methane fluxes were expected.The dashed lines show linear regression curves (forced through zero) for thethree data-sets. The regression line slope for the data with “no correction” is�2:7 nmol CH4 (mmol H2O)�1.

115


3500

3000

2500

2000

CH

4 [pp

b]

27.03.2009 29.03.2009 31.03.2009 02.04.2009 04.04.2009 06.04.2009

3500

3000

2500

2000

FMA

CH

4 [pp

b]

3500300025002000QCLAS CH4 [ppb]

-1000

100

Res

idua

l

slope: 0.998 ± 3.47e-05 (R2 = 0.993)

b)

4000

3000

2000

1000

0

Nr.

of c

ount

s

1.031.021.011.000.990.980.97CH4 (QCLAS) /CH4 (FMA)

wet dry

c)

a)

Figure 5.5.: (a) Time series of the methane mixing ratios during the field campaign mea-sured by QCLAS. (b) Scatter plot between the data sets measured by QCLASand FMA, and (c) a histogram plot of the ratio between these two data sets illus-trating two situations: the ratio calculated using CH4 measured by the FMA inhumid air (gray bars) and the ratio when the FMA data were dilution corrected(black bars). The appended gaussian fit indicates that the ratio between the twodata sets has a nearly normal distribution.

116


Figure 5.6.: Comparison of spectra and cospectra during high flux conditions (102 nmol CH4

m�2 s�1, 31 March 2009, 11:00–14:38 CEST (218 records) with uD3:86m s�1).The gray line shows the idealized cospectrum for neutral and unstable condi-tions according to Kaimal et al. (1972).

117


01

23

wH2O

[mmo

l m!2

s!1 ]

Prec

ip [m

m]

01

2

050

100

footp

rint [%

]

*******************************************************

****

**

****

*

*******

**

*

***********

*

***********

**********

**********

**

**

**********************

*******************************

*

*******************************************************************************************************

***************************************

************

wdir [

deg]

018

036

0

fumigation area moved

F CH 4

[nmo

l m!2

s!1 ]

Mar 27 Mar 29 Mar 31 Apr 02

050

100

150

Period 1 Period 2 Period 3 Period 4

FMAQCLAStracer

"F C

H 4 [n

mol m

!2s!

1 ]

!50

050

Period 1 Period 2 Period 3 Period 4

FMA!tracer

QCLAS!tracer

FMA!QLAS

Figure 5.7.: Methane fluxes during the fumigation experiment. The top panel shows precip-itation and the water vapour flux. Wind direction and the proportion contributedby the fumigated area to the measured flux are displayed in the second panel. Inthe third panel, the footprint corrected fluxes are plotted for the fumigation pe-riods. The gray solid line represents the applied flux. Additionally, the boxplotillustrates the difference of the measured and the applied flux for each systemas well as the difference between the two instruments for each fumigation pe-riod. The upper and lower ends of the box are drawn at the quartiles, and thebar through the box at the median. The whiskers extend from the quartiles ˙1.5 the inter quartile range.

118


T abl

e5.

1.:C

hara

cter

istic

sof

the

two

EC

flux

syst

ems

and

resp

ectiv

ean

alyz

ers.

Para

met

ers

EC

syst

emw

ithFM

AE

Csy

stem

with

QC

LA

S

Ana

lyze

rTyp

eof

f-ax

isin

tegr

ated

dual

cwqu

antu

mca

scad

ela

ser

cavi

tyou

tput

spec

trom

eter

dire

ctab

sorp

tion

spec

trom

eter

Mea

sure

dsp

ecie

sC

H4

CH

4;H

2O;N

2O;N

O2

CH

4ab

sorp

tion

line

posi

tion

(µm

)1.

657.

84C

H4

line

inte

nsity

(�10�21

cm�1/m

olec

ule

cm�2)

0.8–

1.3a

24.6

Opt

ical

path

leng

th(m

)3300<L<3450

b76

Cel

lvol

ume

(L)

0.4

0.5

Cel

lpre

ssur

e(k

Pa)

198

Soni

can

emom

eter

type

Gill

Inst

rum

ents

R2A

(3-D

)G

illIn

stru

men

tsH

S(3

-D)

Sam

ple

line

inle

tpos

ition

0.25

mbe

low

soni

cce

nter

0.25

mbe

low

soni

cce

nter

Soni

cm

easu

rem

enth

eigh

t(m

)1.

21.

2Sa

mpl

ing

tube

leng

th(m

)6.

725

Sam

ple

gas

flow

atin

let(

slpm

)30

.513

.5Se

lect

edda

taac

quis

ition

rate

(Hz)

208

aT

here

are

four

clos

ely

spac

edC

H4

abso

rptio

nlin

es,w

hich

are

notr

esol

ved

byth

eFM

A.b

The

exac

tval

ueis

notk

now

n.T

helo

wer

limit

was

estim

ated

base

don

norm

alop

erat

ion

cond

ition

ofth

eFM

A,

i.e.m

easu

ring�2

ppm

CH

4at

ace

llpr

essu

reof

19kP

are

sults

inab

out1

6%ab

sorp

tion

sign

al.T

his

tran

slat

esin

3.3

kmco

nsid

erin

gth

elin

ein

tens

ities

and

assu

min

gno

optic

allo

ssw

ithin

the

cavi

ty.T

heup

perl

imit

isca

lcul

ated

from

the

repo

rted

cavi

tyri

ng-d

own

time

of11

.6µ

s.

119

This chapter was published in Atmospheric Measure-

ment Techniques Discussions, 2012, 5, 351–384,

doi:10.5194/amtd-5-351-2012.

6Flux correction for closed-path laserspectrometers without internal water

vapor measurements

Rebecca V. Hiller1, Christoph Zellweger2, Alexander Knohl3,

Werner Eugster1

1Institute of Agricultural Sciences, ETH Zurich, Zurich, Switzerland2Empa, Swiss Federal Laboratories for Materials Science and Technology, Laboratory for Air

Pollution and Environmental Technology, Dubendorf, Switzerland3Institute of Bioclimatology, Georg-August University of Gottingen, Gottingen, Germany

Abstract

Recently, instruments became available on the market that provide the possibil-ity to perform eddy covariance flux measurements of CH4 and many other tracegases, including the traditional CO2 and H2O. Most of these instruments em-ploy laser spectroscopy, where a cross-sensitivity to H2O is frequently observedleading to an increased dilution effect. Additionally, sorption processes at theintake tube walls modify and delay the observed H2O signal in closed-path sys-tems more strongly than the signal of the sampled trace gas. Thereby, a phaseshift between the trace gas and H2O fluctuations is introduced that dampens the

121

Chapter 6: Flux correction

H2O flux observed in the sampling cell. For instruments that do not provide di-rect H2O measurement in the sampling cell, transfer functions from externallymeasured H2O fluxes are needed to estimate the effect of H2O on trace gas fluxmeasurements. The effects of cross-sensitivity and the damping are shown foran eddy covariance setup with the Fast Greenhouse Gas Analyzer (FGGA, LosGatos Research Inc.) that measures CO2, CH4, and H2O fluxes. This instru-ment is technically identical with the Fast Methane Analyzer (FMA, Los GatosResearch Inc.) that does not measure H2O concentrations. Hence, we used mea-surements from a FGGA to derive a modified correction for the FMA account-ing for dilution as well as phase shift effects in our instrumental setup. Withour specific setup for eddy covariance flux measurements, the cross-sensitivitycounteracts the damping effects, which compensate each other. Hence, the newcorrection only deviates very slightly from the traditional Webb, Pearman, andLeuning density correction, which is calculated from separate measurements ofthe atmospheric water vapor flux.

6.1. Introduction

The eddy covariance technique (EC) is the established method to directly measure the ex-change of trace gases, such as CO2 and H2O, between the land surface and the atmosphere(Baldocchi, 2003). Recently, new instruments became available that are suitable to mea-sure fluxes of additional trace gases, such as CH4 or N2O (e.g. Eugster and Pluss, 2010;Smeets et al., 2009; Neftel et al., 2010; Hendriks et al., 2008; Kroon et al., 2007; Eugsteret al., 2007). However, by careful evaluation of the suitability of the instrument for the ECmethod, it becomes clear that some adaptations of the standard processing techniques, e.g.accounting for cross-sensitivity effects with H2O, are needed (McDermitt et al., 2010).

Most new instruments apply laser spectroscopy to measure trace gases and two generaltypes of instruments can be distinguished: (1) closed-path instruments pull air through asampling cell where the air is measured, and (2) open-path instruments measure the compo-sition of the air directly in the atmosphere. Both types of instruments measure the decay of alaser beam over a known distance (Baer et al., 2002; McDermitt et al., 2010). In closed-pathinstruments, the beam is reflected by mirrors to achieve a path length of several kilometers(Baer et al., 2002; Crosson, 2008). The definition of the spectral line assumes that the gasmatrix does not change. Atmospheric water content can however vary considerably and thusbroaden the spectral line. Hence, the instrument or the post-processing of the data have toaccount for this effect (Rella, 2010). This phenomenon was described for a Quantum Cas-

122

6.2. Background

cade Laser (QCL, Aerodyne Research Inc.) by Neftel et al. (2010), and for a Fast MethaneAnalyzer (FMA, Los Gatos Research Inc.) by Tuzson et al. (2010). Picarro Inc. discussedthe issue for their laser spectroscopy-based instrument G1301 in a white paper (Rella, 2010).The same artifact also influences open-path instruments such as the LI-7700 (Li-Cor) wherea combined correction accounts for the dilution as well as for the additional H2O interfer-ence on the measured CH4 flux (McDermitt et al., 2010). With closed-path instruments, onesolution would be to dry the air prior to sampling. However, this tends to have a negative im-pact on the frequency response and high-frequency performance of the sensor (Griffis et al.,2008). Alternatively, a correction procedure could solve the cross-sensitivity issue in a moreelegant way as will be discussed in this paper.

Since the FMA does not measure H2O, the correction of the density effect caused by watervapor molecules is essential to improve the accuracy of the flux measurements. Therefore,the H2O flux measured by a separate system is typically used to apply this correction. Thetraditional density flux correction by Webb et al. (1980) employs the water vapor flux mea-sured in the same sampling cell as the other gas of interest. The water vapor signal howevergets more strongly dampened from the inlet to the sampling cell (Ibrom et al., 2007a) than aless polar gas such as CH4. Thus, substituting the water vapor flux in the original correctionterm with the water vapor flux measured separately with a second instrument and at somedistance from the first instrument might overcorrect the density effect. To find out whetherthis is a small (negligible) or large and important effect was the goal of this study.

We used a Fast Greenhouse Gas Analyzer (FGGA, Los Gatos Research Inc.) to asses thephase shift and damping of the H2O signal with respect to the CH4 signal and in relationto an externally measured H2O flux. Based on these findings, two transfer functions willbe suggested to (1) correct the H2O flux measured by the FGGA to represent atmosphericconditions; and (2) its inverse function that allows to translate an externally measured H2Oflux to the conditions in the sampling cell which then can be used to correct the CH4 fluxfor the density fluctuations. This correction procedure can be used to correct CH4 fluxmeasurements performed with a FMA which does not provide internal H2O measurements.

6.2. Background

This study closely follows the concept of the density flux correction accounting for densityeffects due to heat and water vapor fluctuations that was introduced by Webb et al. (1980).We will shortly introduce the concept of the theory and highlight the importance in relationto this study. The original concept will be referred to as WPL correction thereafter.

123


For water vapor, the original correction for open-path analyzers reads (WPL Eq. 25):

E D .1C ��/fEraw C .�v=T /w0T0g ; (6.1)

with E [kg m�2 s�1] the corrected water vapor flux, � D mamv� 28:97=18 Œ�� the ratio of

molecular masses of dry air and water vapor, � D �v�aŒ�� the ratio of the vapor density and

the density of dry air, Eraw [kg m�2 s�1] the measured water vapor flux, �v [kg m�3] thedensity of dry air, T ŒK� the air temperature, and w0T0 [K m s�1] the sensible heat flux inkinematic units.

For trace gas fluxes, buoyancy effects from water vapor additionally influence the mea-sured flux. Thus, Webb et al. (1980) introduced the following correction (WPL Eq. 24):

F D Fraw C�F D Fraw C �.�c=�a/E„ ƒ‚ …water vapor

C .1C ��/.�c=T /w0T0„ ƒ‚ …heat

; (6.2)

where F [kg m�2 s�1] is the corrected trace gas flux and Fraw the measured one. �c [kg m�3]denotes the density of the particular trace gas under consideration. The curly brackets sum-marize the correction terms for density fluctuations caused by the buoyancy effect of watervapor and by sensible heat. Taken together, they constitute �F , the WPL correction term.

The sensitivity of the WPL correction itself largely depends on the ratio between the con-centration of the particular trace gas under consideration and the corresponding flux. LargestWPL corrections are expected for small fluxes at comparably high background concentra-tions (Ibrom et al., 2007b). If �c is factored out in Eq. (6.2), it yields:

F D Fraw C �c

n.�=�a/E„ ƒ‚ …water vapor

C�.1C ��/=T

�w0T0„ ƒ‚ …

heat„ ƒ‚ …environment

o: (6.3)

The term in the curly bracket to the right of �c is only influenced by the measurementsof the environment and does not depend on the density of the trace gas nor the trace gasflux. Therefore, the WPL correction (�F ), the difference between the measured and thecorrected flux, depends on the density of the trace gas multiplied by a term related to theenvironment conditions where the measurements was performed. This term includes thewater content of the air, the water vapor flux, air temperature, and the sensible heat flux. Therelative magnitude of the WPL correction compared to the measured flux can be assessedby the ratio �F=Fraw. This is equal to .�c � environment/=Fraw and hence the relative WPLcorrection is highest for high background concentrations (�c) combined with small fluxesFraw. Detto et al. (2011) presented in their Fig. 6 the influence of the latent heat flux on

124

6.2. Background

CH4 flux for a closed-path instrument, where the heat relatet component can be neglectedbased on the evindence that temperature fluctuations are almost completely removed in theclosed-path instruments. However, since the WPL correction is an additive term and not ascaling factor, one should be careful with expressing the WPL correction as a percentage ofthe measured flux. This becomes most obvious at small fluxes, where the WPL correctionmay lead to a sign change of the flux.

For closed-path instruments, the environment relevant for the WPL correction relates tothe environment in the sampling cell, not to the ambient environment where the intake ofthe gas inlet was placed. Therefore, the water vapor and sensible heat flux in the cell definethe magnitude of the WPL correction. In case of the sensible heat flux, the tubing from theinlet to the measurement cell and the waste heat of the instrument cancel the natural shorttime temperature fluctuations. Hence, w0T0 gets very small and the density effects due toheat transfer is negligible (Leuning and Judd, 1996). The correction term reduces to densityeffects due to water vapor. This term depends on various factors. If the air is dried prior tomeasurement or if concentrations are measured in units of dry mole fractions, this correctioncan be dismissed too. Otherwise, the term applies, but the instrumental setup may influencethe measured H2O flux and consequently also the magnitude of the correction to be applied.Damping of the H2O signal reduces the H2O flux in a closed-path analyzer and thus theWPL correction would be overestimated by using H2O flux estimates taken from a secondinstrument (Ibrom et al., 2007b). The investigation of such effects is important to obtainthe highest accuracy of trace gas flux measurements (Smeets et al., 2009). Signal dampingdue to amplitude effects (Massman and Clement, 2005) as well as phase shifts (Ibrom et al.,2007b) should be attributed.

The attenuation of the water vapor flux measured by a closed-path instrument dependson the relative humidity (Ibrom et al., 2007a; Mammarella et al., 2009). This however doesnot apply to CO2 and CH4 fluxes, since the molecules are not as sticky as H2O and hencedo not strongly interact with the intake hose’s inner wall. Thus, the transportation of theair through the instrument leads to a decoupling of the CO2 or CH4 and H2O signal (Ibromet al., 2007b). This decoupling consequently results in a reduction of the water vapor flux atthe time lag that was determined for the CO2 flux. Therefore, the WPL correction needs tobe adjusted in a way that the H2O flux in the cell is approximated, as for example shown byIbrom et al. (2007b), Smeets et al. (2009), and Massman (2005).

125


6.3. Methods

6.3.1. Instruments

The Fast Methane Analyzer (FMA) and the Fast Greenhouse Gas Analyzer (FGGA, bothLos Gatos Research Inc., Mountain View, CA, USA) are capable of measuring CH4 molefractions. Both instruments operate at high measurement frequencies up to 20 Hz as requiredfor eddy covariance measurements. The construction of the two instruments is identicalexcept for a second laser in the FGGA that measures CO2. H2O is also only measured bythe FGGA, but retrieved from the same laser absorption spectrum as CH4. With the help ofmeasurements with the FGGA, the influence of H2O in the sampling cell of the FMA can beassessed.

6.3.2. Dilution and cross-sensitivity experiment

Linearity and precision of the FMA and FGGA were tested in the laboratory. The instru-ments were connected to a common air source. Eleven concentration levels at regular inter-vals between 0 and 4 ppm for CH4 and 16 between 0 and 1500 ppm for CO2 were producedby diluting CH4 and CO2 with CH4 and CO2 free air, respectively. The concentration waskept constant for 20 min in case of CH4 and 15 min for CO2. After verifying the linearity ofthe measurements, the precision of the CH4 measurements was checked against two NationalOceanic & Atmospheric Administration (NOAA) standards (CA04580 and CA05316). H2Owas calibrated against a LI-610 dew point generator (Li-Cor, Lincoln, NB, USA). Further,dry air with a known CH4 and CO2 concentration was humidified to check for dilution ef-fects and cross-sensitivity of CH4 and CO2 measurements to H2O. H2O concentrations wereset to 9 different levels between 0 and 25 000 ppm and kept constant for 20 min. The airtemperature in the laboratory was controlled during all experiments.

6.3.3. Flux measurements

Site

The flux measurements presented in this study were acquired at the ETH agricultural re-search station Chamau (8ı 240 3800 E, 47ı 120 3700N, 400 m a.s.l., Switzerland) in summer2009. The station is located in the broad prealpine Reuss Valley. Two flux towers, a mo-bile and a long-term one, were situated on a temperate, intensively managed grassland withmixed ryegrass-clover vegetation raised for forage (Zeeman et al., 2010).

126

6.3. Methods

Instruments and setup

The mobile tower consisted of a Fast Greenhouse Gas Analyzer (FGGA, Los Gatos Re-search, Inc., Mountain View, CA, USA) measuring concentrations of CO2, CH4, and H2Oand a Solent R2A ultrasonic anemometer (Gill Instruments Ltd., Lymington, UK), there-after abbreviated as sonic. The sonic was installed at a height of 1.7 m with the inlet for theFGGA attached to the sonic boom 25 cm below the instrument head. The sample was drawnthrough a 6.7 m long tube with 8 mm inner diameter (Synflex-1300, Eaton Performance Plas-tics, Cleveland, OH, USA) followed by two serially mounted particle filters with integrateddroplet separator (SMC, Japan, model AF30-F03 with 5 µ m filter and model AFM30-F03with 0.3 µ m filter, respectively). The FGGA was connected to an external dry vacuum scrollpump (BOC Edwards XDS-35i, Crawly, UK) drawing 30.5 slpm to ensure turbulent flow inthe sampling tube. An embedded computer (Advantech ARK-3381, Taipei, Taiwan) runninga Linux system recorded the raw data at 20.8 Hz. Eugster and Pluss (2010) and Tuzson et al.(2010) provide a more detailed description of the system.

The long-term tower measured CO2 and H2O fluxes by an open-path Infrared Gas Ana-lyzer (IRGA, Li-7500, Li-Cor, Lincoln, NB, USA) and a Solent R3 ultra sonic anemometer(Gill Instruments Ltd., Lymington, UK) at 2.40 m above ground, about 5 m separated fromthe mobile tower. A more detailed description of this system can be found in Zeeman et al.(2010).

Additionally, several meteorological variables were measured: air temperature and rela-tive humidity of the air (HygroClip S3, Rotronic AG, Bassersdorf, Switzerland), and atmo-spheric pressure (LI-7500, Li-Cor, Lincoln, NB, USA).

Even though measured by the HygroClip, relative humidity values used in the evalua-tions were derived from the LI-7500 measurements (RHIRGA) due to higher accuracy of thisinstrument, whereas the HygroClip data were used for cross-checking the results.

Flux processing

The statistical software R (Version 2.12.1, R Development Core Team, 2010) was used forall analyses. This includes the processing of the flux data from the mobile system with theR package Reddy (under development and available from the authors upon request). CO2

and H2O fluxes of the long-term tower were calculated by the in-house software eth-flux

(Mauder et al., 2008) which is also available from the authors. In eth-flux and Reddy, atwo-way rotation of the wind vector was performed for each 30-min interval (McMillen,1988). The time lag between the vertical wind speed w and the scalar was determined foreach 30-min block such that the correlation maximized. If not stated differently, this time lag

127


was applied before calculating the respective flux. Fluxes were calculated as the covariancebetween the two time series following the rules of Reynold’s averaging (w0c0) (Baldocchi,2003). The CH4 fluxes at our site were often very small and close to the detection limit of thesystem. Thus the time lag between w and the CH4 concentration measurements was difficultto detect since no clear peak in the cross-correlation function was found. For periods witha detectable CH4 time lag, it was very similar to the one of CO2. Thus, we used the timelag found for CO2 to calculate the CH4 fluxes. The H2O flux measured by the FGGA wascalculated in two ways: (1) the time lag maximizing the correlation between w and the H2Osignal (w0H2O0FGGA) and (2) with the time lag found for CO2 (w0H2O0FGGA;lag.w0CO0

2/). Prior

to the flux calculations, all scalars of the FGGA were post-calibrated as described below.Flux data were filtered according the following criteria. For the mobile system, fluxes

were rejected if the cell pressure dropped below 100 Torr or the number of available trace gasmeasurements was below 20 000 of 36 000 within the 30-min averaging window. Further,flux records were removed if Reddy automatically detected the maximal or minimal time lagwithin the lag detection window, namely 10 and 40 records for w0CH04, 15 and 40 records forw0H2O0, and 10 and 30 records for w0CO02. For the long-term station, fluxes were dismissedif the automated gain control (AGC) value by the LI-7500, the so called window dirtiness,exceeded 70 % or was not a value dividable by 6.25. Both criteria are indicators for rainor dew disturbing the measurement path. Since the LI-7500 reports the AGC as a four-bitvariable, 16 different values between 0 and 100 can be reported which results in incrementalsteps of 6.25. Thirty-minute averages between these steps indicate temporal changes of theAGC value within the averaging period, e.g. introduced by slight rain or an insect passingthe sensor path during a short part of the averaging interval. Further, fluxes were rejectedwhen the friction velocity dropped below 0.08 m s�1 as suggested by Zeeman et al. (2010)for this site.

Cospectra and damping loss correction

Before the calculation of the cospectra, a two way rotation was applied to the wind vec-tors (McMillen, 1988) and a time lag was implemented to maximize the correlation of thetwo time series. The scalar concentrations were converted to dry mole fractions and a cor-rection for additional cross-sensitivity to H2O effects (see Sect. 6.4.1) was applied. Sinceno cospectral analysis function exists for R, we wrote a new function downloadable fromhttp://www.swissfluxnet.ch/R/. The procedure follows the suggestion by Stull (1988,p.310ff) and was double-checked against the algorithm for Fast Fourier Transformation sug-gested by Press (1988). Thereafter, a bandwidth averaging was performed on the output(Kaimal and Gaynor, 1983) and the high frequency flux losses were quantified following Eu-

128

6.3. Methods

gster and Senn (1995). A slight modification of the theoretical cospectral model introducedby Kaimal et al. (1972) is damped according to the inductance L Œs�1� in an alternating-current circuit to match the calculated cospectra. This corresponds to the standard first-orderdamping known from similar approaches, e.g. by Horst (1997). For unstable conditions, thecospectral approximation reads (Eq. 29 in Eugster and Senn, 1995):

f Cow;cL.f /

w0c0LD0

D

�10:53n=

�.1C 4�2f 2L2/.1C 13:3n/1:72

�; n 6 1:0

4:21n=�.1C 4�2f 2L2/.1C 3:8n/2:4

�; n > 1:0

�; �2 6 � 6 0 ;

(6.4)and for stable conditions (Eq. 30 in Eugster and Senn, 1995):

f Cow;cL.f /

w0c0LD0

D1

1C 4�2f 2L2�

0:81.n=n0/

1C 1:5.n=n0/2:1; 0 < � 6 2 (6.5)

with

n0 D

�0:23; �2 6 � 6 0

0:23.1C 6:4�/3=4; 0 < � 6 2

�: (6.6)

Here, � [-] denotes the Monin-Obukhov stability parameter and n the normalized frequencyn D f z=u with f [Hz] the natural frequency, z [m] the measuring height above zero-planedisplacement, and u [m s�1] the mean horizontal wind-speed along the mean wind direction.

Cospectra were calculated from 1-h blocks of high frequency data for the whole measure-ment period. The damping factorsLwere retrieved by a non-linear least squares fit (functionnls() in R) to the individually calculated cospectra. For further analyses, L was only usedfor periods when a detectable time lag was found within the predefined window and the fit ofthe idealized damped cospectrum on the calculated cospectrum was significant at p < 0:1.L was originally set constant over periods without system maintenance where the measuredtracer was NO2 (Eugster and Senn, 1995). Especially for the water vapor flux, additionaldampening of the signal is however observed depending on the relative humidity of the air(Ibrom et al., 2007a; Mammarella et al., 2009). Therefore, L was adjusted according to thefollowing empirical model with relative humidity RH as predictor:

L D aC b � RHC c � RH2 ; (6.7)

where a, b, and c were obtained by conditional linear regression fitting. To reduce the scatterin the data, the automatically determinedL values were binned into relative humidity classesof 5 % width. The model was fitted through the medians of each class. In constrast to Lfor the H2O fluxes, analyses of the empirically determined L for the CH4 and CO2 fluxesresulted in a constant value. The measured fluxes were corrected with the derived L for

129


damping losses following the method by Eugster and Senn (1995).

6.4. Results

6.4.1. Dilution and cross-sensitivity experiment

The linearity of the instruments was tested in a dilution experiment (data not shown). Com-paring the median of the measurements at each level against the given concentrations, R2

yields 0.999997 for the FMA and 0.999995 for the FGGA. This implies that a two-pointcalibration is sufficient to calibrate the instrument. CH4 was calibrated against two NOAAstandards (1.713 ppm and 1.905 ppm). The linear relationship through these two points (inppm) yielded:

CH4 D �0:0147 C 1:0002 � CH4;FMA

CH4 D �0:0024 C 0:9774 � CH4;FGGA :(6.8)

The calibration of the FGGA H2O signal determined the following polynomial of the secondorder (in ppm):

H2Oref D .�52:36 ˙ 8:01 / C

. 1:011 ˙ 0:002 / � H2OFGGA C

. 0:00000239 ˙ 0:00000007/ � H2O2FGGA ;

(6.9)

where H2OFGGA refers to the H2O humid mole fraction measured with the FGGA and H2Oref

to the dry mole fraction of the dew point generator, both measured in or converted to ppm.The uncertainty in the parameter estimate is given with ˙ one standard error and the R2

yields 0.999. The FGGA overestimates H2O concentrations at low concentrations up to�5000 ppm and thereafter overestimates them. For the following analysis, the water vaporsignal was post-calibrated according to the parameter fit in Eq. (6.9).

The dilution experiment described in Sect. 6.3.2 showed a drop of the CO2 and CH4

humid mole fractions that was triggered by the changes of humidity. Most of the drop canbe explained by the dilution of the air by water vapor (difference of the black and dark greypoints in Fig. 6.1). The dark grey points correspond to the dry mole fraction reported by theFGGA. However, a small part of the drop, the difference between dark and light grey pointsin Fig. 6.1, was not caused by dilution and hence must result from different processes. Rella(2010) suggested the following equation to calculate the dry mole fraction for the PicarroG1301 analyzer, here adapted for the FGGA instrument:

c D cFGGA=.1C a � H2OC b � H2O2/ ; (6.10)

130

6.4. Results

where H2O is the corrected H2O dry mole fraction measured by the FGGA and cFGGA themeasured humid mole fraction of the issued dry mole fraction c of CO2 or CH4. Table 6.1shows the empirical coefficients for Eq. (6.10) which combines the dilution effect as well asother water dependent processes influencing the measured concentration.

The additional water dependent processes affect the measured concentration in a similarway as the dilution effect. To correct the effect of density fluctuations caused by the dilution,the WPL correction (Eq. 6.2) can be applied. The dilution is a linear process while Eq. (6.10)additionally includes the square of H2O. The derivation of the WPL correction would beseverely complicated if the square term is included. Since the performance of a simple linearmodel, same as Eq. (6.10) but without term b�H2O2, is similar (compare corresponding R2

in Tables 6.1 and 6.2), we decided to use the latter for the further analysis. The details forthe individual gases and instruments can be found in Table 6.2.

An additional fit of Eq. (6.10) was determined for CH4 measured by the FMA withH2O taken from the measurements of the FGGA. The results only marginally differs fromFig. 6.1b (not shown). The results for the fit can be found in Table 6.1 and 6.2.

6.4.2. Damping loss correction

The automatically fitted damping factor L for w0H2O0FGGA;lag.w0CO02/

shows a clear depen-dency on the relative humidity (see Fig. 6.2). Such dependency ofL on the relative humiditywould also be found if the L for w0H2O0 at the time lag maximizing w0H2O0 were used (notshown). The increase in L with increasing relative humidity can by explained by sorptionand desorption effects. The variability of the w0H2O0 parallel increases with increasing L.This can be explained by lower water vapor fluxes at high relative humidity and hence a lesswell defined cospectrum. The automatic fits of L are less robust and result in a higher scat-tering. Since outliers would strongly influence the parameter fit for the relationship betweenL and the relative humidity, L was binned into relative humidity classes of 5 % width. Apolynomial of the 2nd order was fitted through the medians of the individual classes, whichyields:

Lw0H2O0FGGA;lag.w0CO0

2/

D 1:07 � 2:99 � 10�2 � RHIRGA C 3:62 � 10�4� RH2IRGA ; (6.11)

with RHIRGA [%] being the relative humidity obtained from the air temperature measured bythe HygroClip and the H2O concentration measured by the LI-7500 IRGA. This relationshipbetween L and the relative humidity was used in the flux processing to determine the damp-ing factor for the water vapor flux and correct the flux therewith for damping losses. Nodependency on relative humidity was found for the CO2 and CH4 fluxes (not shown). For

131


these two trace gases, L seems to be a constant as proposed in the original paper by Eugsterand Senn (1995).

6.4.3. Signal desynchronisation

Different time lags to maximize the covariance w0H2O0 and w0CO02 were found. Based onthe assumption that no time lag would be observed under ambient conditions, we guess thatthe signal of CO2 and H2O was desynchronized on the way from the inlet to the measure-ment cell. Figure 6.3 shows the histogram of the determined time lags between w and theparticular scalar concentration. In general, the peak of w0H2O0 moved towards longer lagscompared to w0CO02. Also the distribution of w0H2O0 lags was wider than the one for w0CO02.For w0CH04, the distribution of time lags was almost uniform because the flux at our site wasvery low. However, the comparison of the time lag distribution of w0CH04 in the two panelsin Fig. 6.3 reveals that there is a slight difference seen in the range where the w0H2O0 timelag was most frequent (20–30 records). For CH4 humid mole fractions, more cases with alag in this range were observed than for dry mole fractions. Since the CH4 flux was verylow at our site, the water-introduced fluctuations in the CH4 signal were sometimes strongerthan the natural variations of CH4, which resulted in lags for CH4 being identical to those ofw0H2O0.

For w0H2O0, the time lag clearly depended on the relative humidity (Fig. 6.4a) while itstayed constant for w0CO02. This explains the wider distribution of the peak for time lags ofw0H2O0 compared to w0CO02 in Fig. 6.3. Nevertheless, more records lacked a detectable timelag during conditions of high relative humidity for w0CO02, and lags of w0H2O0 were hardlydetectable at RHIRGA > 95%.

The ratio between w0H2O0FGGA;lag.w0CO02/

and w0H2O0FGGA decreased with increasing rela-tive humidity as a result of higher temporal separation of the optimal lags of CO2 and H2O(see Fig. 6.4b). The median observed w0H2O0 flux losses were increasing up to 30 % athighest relative humidity. However, w0H2O0 was smaller in these conditions than duringdryer periods, and thus the absolute flux difference was minimal. Looking at the cospectraof w0H2O0, the flux difference appears visible as an additional damping of the signal (seeFig. 6.5). This observation agrees well with the relative humidity dependent damping lossesfound for w0H2O0 (Eq. 6.11).

The H2O concentration measurements of the FGGA tend to be lower than those measuredby the LI-7500 IRGA. This ratio between H2OFGGA and H2OIRGA decreased with increasingrelative humidity (see Fig. 6.6). For the comparison, the H2O measurements from the FGGAwere converted from dry mole fractions to mmol m�3, the unit recorded by the IRGA. Points

132

6.4. Results

were ecluded from the analysis if the H2O concentration measured by the two instrumentsdiverted by more than 20 % (7 out of 933 points). A non-linear model fits the medians of the5 % relative humidity bin very well. The relationship yields:

H2OFGGA

H2OIRGAD 1:072 � 2:528 � 10�3 � RHIRGA C 1:146 � 10

�5� RH2IRGA : (6.12)

6.4.4. Flux comparison

Putting the previous findings together, the w0H2O0FGGA;lag.w0CO02/

should be scalable to matchthe atmospheric w0H2O0IRGA, e.g. measured by the LI-7500. Therefore, w0H2O0FGGA;lag.w0CO0

2/

was post-calibrated according to Eq. (6.9) and the damping loss correction applied withthe damping factor calculated from the relative humidity (Eq. 6.11). Compared to the rawflux, the corrected w0H2O0FGGA;lag.w0CO0

2/compares better with the w0H2O0IRGA. In addition

to the post-calibration of the water concentration and the application of the damping correc-tion, the flux was also corrected for the observed signal reduction with increasing relativehumidity (Eq. 6.12). The combination of all corrections reduced the difference betweenw0H2O0FGGA;lag.w0CO0

2/and w0H2O0IRGA considerably, where the latter contributed 5 % to the

total improvement of 47 % (see Fig. 6.7).

6.4.5. Adapted WPL correction

The information about the damping of w0H2O0 by the instrumental setup and the cross-sensitivity of other trace gas measurements to H2O provides the basis for the adapted WPLcorrection. This modified correction can be applied to fluxes measured by the FMA whereH2O is not measured. The heat term of the original WPL correction (see Eq. 6.2) is negli-gible for the FMA since it is a closed-path instrument (Leuning, 2005). The adapted WPLcorrection yields:

F D aslopeFraw C �Œ.aintercept C aslope�c/=�a�bE

de: (6.13)

The factors aslope and aintercept apply the post-calibration of the concentration measurementsas described in Sect. 6.4.1, Eq. (6.8). Since means are removed when calculating fluxes,the intercept can be omitted for the flux, but must be included in the post-calibration of thedensity of the trace gas.

The additional water dependent cross-sensitivities affect the measurements in a similarway as the dilution effect and can be expressed as an amplification of the dilution effect.The linear correction approach (Eq. 6.10 without term b�H2O2, see Table 6.2) implies that,

133


in case of CH4, the dilution effect is amplified by 18 %. Therefore, the water vapor flux Emust be multiplied with 1.18, denoted as factor b in Eq. (6.13).

Since w0H2O0 observed in the measurement cell controls the density flux effects causedby H2O, the atmospheric w0H2O0 needs to be modified such that it fits the observed fluxin the cell. Therefore, the corrections derived for w0H2O0FGGA;lag.w0CO0

2/were reverted and

applied to the atmospheric w0H2O0 to reconstruct the conditions in the cell. The dampingfactor L was calculated according to Eq. (6.11) and the damping effect assessed accordingto Eqs. (6.4) or (6.5), depending on the stability of the atmosphere. E in Eq. (6.13) wasdivided by the calculated damping effect d (Eq. 6.11) and by e that accounts for the reducedH2O concentration in the sampling cell (Eq. 6.12) to simulate w0H2O0FGGA;lag.w0CO0

2/.

The damping of the water vapor flux reduces the correction by about 20 % compared tothe original correction (see Fig. 6.8). However, this reduction is compensated by the spectralbroadening effect due to H2O. The final result compares well with the flux calculated fromdry mole fractions. In total, the clear CH4 uptake is reduced to a near zero flux after applyingall corrections.

6.5. Discussion

The manufacturer’s calibration of the FGGA and FMA agreed very well with the measure-ments against CH4 calibration standards as also noted by Detto et al. (2011). They foundvery small instrumental drifts for their FMA and FGGA and suggest that no post-calibrationis needed for eddy covariance applications. However, CH4 and CO2 measurements dependon H2O. Additional to the dilution effect, a cross-sensitivity on H2O was found. This occur-rence was already observed for a FMA (Tuzson et al., 2010) and other laser spectrometersof various brands (Chen et al., 2010; Tuzson et al., 2010; Neftel et al., 2010; McDermittet al., 2010). Such an interference is no problem as long as H2O is measured simultane-ously in the same measurement cell as the other trace gas. The effect can be quantified in adilution experiment as described in Sect. 6.3.2. Picarro Inc. and Licor Inc. published the fac-tors for their instruments Picarro G1301 and Licor Li-7700 (Rella, 2010; McDermitt et al.,2010). Meanwhile, new instruments by Los Gatos Research Inc. already correct for cross-sensitivity effects (FMA with serial numbers starting from 11-0049 and above and FGGAstarting from 11-0018) and hence no additional correction is needed. FMA and FGGA withserial numbers 10-0150 and below do not include water correction, but the instrument canbe upgraded by the manufacturer. Instruments with serial numbers inbetween might cor-rect or are user upgradable and Los Gatos Research Inc. will provide instrument specificinformation upon request (Robert Provencal, personal communication, 13 September 2011).

134

6.5. Discussion

Instruments that do not measure H2O perform best if additional measurements of w0H2O0

are available to correct for dilution and cross-sensitivity effects of H2O. The needed dilutionfactors can be derived from a dilution experiment added to the traditional WPL correction.Another option would be to dry the air sample before the analysis which might howeverintroduce negative effects to the frequency response and high-frequency performance of thesensor (Griffis et al., 2008).

Further, the conditions in the sampling cell are decisive for the WPL correction. The watervapor flux is damped by the measurement system and hence is reduced in the cell. Therefore,atmospheric water vapor flux would overcorrect the dilution effect. The damping strengthof the system depends on the tubing, filters, etc. and is therefore different for individualsetups. Additionally, environmental variables such as wind speed and atmospheric stabilityinfluence the damping strength (see Sect. 6.3.3). For water vapor, the damping also dependson relative humidity (Mammarella et al., 2009; Ibrom et al., 2007a,b). Smeets et al. (2009)already adapted the WPL correction following the suggestions by Ibrom et al. (2007b) forthis issue. They used a FMA that does not measure H2O and thus had no possibility tovalidate their assumption. Moreover, they did not address the additional cross-sensitivityeffects of CH4 humid mole fractions on H2O. Detto et al. (2011) stated that they found nodifference between the flux calculated from dry mole fractions and the flux corrected withthe water vapor flux derived from an external sensor (LI-7500) for their FGGA. This agreeswith our findings. Nevertheless, the original WPL term accounting for the dilution of the airsample by H2O is reduced. This reduction is however compensated by the cross-sensitivityof the measurements to H2O.

The dependency of the damping on relative humidity can be at least partially attributed tothe likeliness that the tube walls and filters on the way to the cell absorb and desorb watermolecules. For CO2 and CH4, no dependency was found and the system seems to be inertto these gases. Compared to other trace gases, the sorption processes of water prolong thetravelling time of H2O into the sampling cell. The different time lag introduces a phase shiftbetween H2O and e.g. CO2 which leads to an additional damping of w0H2O0FGGA;lag.w0CO0

2/at

the time lag of w0CO02.

We additionally observed a reduction of measured water vapor concentration with increas-ing relative humidity. The used particle filters are designed to retain liquid water, e.g. if theinstrument is operated under foggy conditions. Unfortunately, we found that the filter mate-rial itself is not hydrophobic enough and thus introduces a memory effect in the H2O signal.If the humidity of the atmosphere increases, the filter retains some water, even though theair is not saturated. Reversely, water in the filter reevaporates if the air dries. This might bea reason for the observed reduction in H2O concentrations measured by the FGGA.

135


6.6. Conclusions

Our results show that the damping of the water vapor signal as well as cross-sensitivityeffects to water affect the measured trace gas flux. While the damping reduces w0H2O0 ob-served in the cell, the cross-sensitivity effects counteract the damping. For our setup, thetwo processes coincidentally compensated each other. Hence, the original WPL correctionwould lead to the same result as our adapted approach which accounts for the individualprocesses that affect the water vapor flux. However, the damping of w0H2O0 is system spe-cific and depends on deployed filters etc. Therefore, our approach is not generally valid, butprovides at least an estimate for the order of magnitude of the described effects.

Acknowledgements This project was funded by the Competence Center Environment anSustainability at the ETH Domain (CCES). We thank Hans-Ruedi Wettstein, managerof the agricultural research station Chamau for access to the field site and SebastianWolf for his valuable comments on the manuscript.

6.7. References

Baer, D., Paul, J., Gupta, M., and O’Keefe, A.: Sensitive absorption measurements in thenear-infrared region using off-axis integrated-cavity-output spectroscopy, Appl. Phys. B,75, 261–265, 2002.

Baldocchi, D. D.: Assessing the eddy covariance technique for evaluating carbon dioxideexchange rates of ecosystems: past, present and future, Glob. Change Biol., 9, 479–492,2003.

Chen, H., Winderlich, J., Gerbig, C., Hoefer, A., Rella, C. W., Crosson, E. R., Van Pelt,A. D., Steinbach, J., Kolle, O., Beck, V., Daube, B. C., Gottlieb, E. W., Chow, V. Y.,Santoni, G. W., and Wofsy, S. C.: High-accuracy continuous airborne measurements ofgreenhouse gases (CO2 and CH4) using the cavity ring-down spectroscopy (CRDS) tech-nique, Atmos. Meas. Tech., 3, 375–386, doi:10.5194/amt-3-375-2010, 2010.

Crosson, E.: A cavity ring-down analyzer for measuring atmospheric levels of methane,carbon dioxide, and water vapor, Appl. Phys. B, 92, 403–408, 2008.

Detto, M., Verfaillie, J., Anderson, F., Xu, L., and Baldocchi, D.: Comparing laser-basedopen- and closed-path gas analyzers to measure methane fluxes using the eddy covariancemethod, Agr. Forest Meteorol., 151, 1312–1324, doi:10.1016/j.agrformet.2011.05.014,2011.

Eugster, W. and Pluss, P.: A fault-tolerant eddy covariance system for measuring CH4 fluxes,Agr. Forest Meteorol., 150, 841–851, doi:10.1016/j.agrformet.2009.12.008, 2010.

136

6.7. References

Eugster, W. and Senn, W.: A cospectral correction model for measurement of turbulent NO2flux, Bound.-Lay. Meteorol., 74, 321–340, 1995.

Eugster, W., Zeyer, K., Zeeman, M., Michna, P., Zingg, A., Buchmann, N., and Emmeneg-ger, L.: Methodical study of nitrous oxide eddy covariance measurements using quan-tum cascade laser spectrometery over a Swiss forest, Biogeosciences, 4, 927–939,doi:10.5194/bg-4-927-2007, 2007.

Griffis, T. J., Sargent, S. D., Baker, J. M., Lee, X., Tanner, B. D., Greene, J.,Swiatek, E., and Billmark, K.: Direct measurement of biosphere-atmosphere isotopicCO2 exchange using the eddy covariance techniques, J. Geophys. Res., 113, D08304,doi:10.1029/2007JD009297, 2008.

Hendriks, D. M. D., Dolman, A. J., van der Molen, M. K., and van Huissteden, J.: A compactand stable eddy covariance set-up for methane measurements using off-axis integratedcavity output spectroscopy, Atmos. Chem. Phys., 8, 431–443, doi:10.5194/acp-8-431-2008, 2008.

Horst, T. W.: A simple formula for attenuation of eddy fluxes measured with first-order-response scalar sensors, Bound.-Lay. Meteorol., 82, 219–233, 1997.

Ibrom, A., Dellwik, E., Flyvbjerg, H., Jensen, N. O., and Pilegaard, K.: Strong low-passfiltering effects on water vapour flux measurements with closed-path eddy correlationsystems, Agr. Forest Meteorol., 147, 140–156, 2007a.

Ibrom, A., Dellwik, E., Larsen, S. E., and Pilegaard, K.: On the use of the Webb-Pearman-Leuning theory for closed-path eddy correlation measurements, Tellus B, 59, 937–946,2007b.

Kaimal, J. C. and Gaynor, J.: The Boulder Atmospheric Observatory, J. Climate Appl. Me-teor., 22, 863–880, 1983.

Kaimal, J. C., Wyngaard, J. C., Izumi, Y., and Cote, O. R.: Spectral charac-teristics of surface-layer turbulence, Quart. J. Roy. Meteorol. Soc., 98, 563–589,doi:10.1002/qj.49709841707, 1972.

Kroon, P. S., Hensen, A., Jonker, H. J. J., Zahniser, M. S., van ’t Veen, W. H., and Ver-meulen, A. T.: Suitability of quantum cascade laser spectroscopy for CH4 and N2O eddycovariance flux measurements, Biogeosciences, 4, 715–728, doi:10.5194/bg-4-715-2007,2007.

Leuning, R.: Measurements of trace gas fluxes in the atmosphere using eddy covariance:WPL corrections revisited, in: Handbook of Micrometeorology, edited by: Lee, X., Mass-man, W., and Law, B., Vol. 29 of Atmospheric and Oceanographic Sciences Library, 119–132, Springer Netherlands, doi:10.1007/1-4020-2265-4 6, 2005.

Leuning, R. and Judd, M. J.: The relative merits of open- and closed-path analysersfor measurement of eddy fluxes, Glob. Change Biol., 2, 241–253, doi:10.1111/j.1365-2486.1996.tb00076.x, 1996.

137


Mammarella, I., Launiainen, S., Gronholm, T., Keronen, P., Pumpanen, J., Rannik, U., andVesala, T.: Relative humidity effect on the high-frequency attenuation of water vaporflux measured by a closed-path eddy covariance system, J. Atmos. Ocean. Technol., 26,1856–1866, doi:10.1175/2009JTECHA1179.1, 2009.

Massman, W.: Concerning the measurement of atmospheric trace gas fluxes with open-and closed-path eddy covariance system: The WPL terms and spectral attenuation, in:Handbook of Micrometeorology, edited by: Lee, X., Massman, W., and Law, B., 29 ofAtmospheric and Oceanographic Sciences Library, 133–160, Springer, The Netherlands,doi:10.1007/1-4020-2265-4 7, 2005.

Massman, W. and Clement, R.: Uncertainty in eddy covariance flux estimates resultingfrom spectral attenuation, in: Handbook of Micrometeorology, edited by Lee, X., Mass-man, W., and Law, B., 29 of Atmospheric and Oceanographic Sciences Library, 67–99,Springer, The Netherlands, doi:10.1007/1-4020-2265-4 4, 2005.

Mauder, M., Foken, T., Clement, R., Elbers, J. A., Eugster, W., Grunwald, T., Heusinkveld,B., and Kolle, O.: Quality control of CarboEurope flux data – Part 2: Inter-comparisonof eddy-covariance software, Biogeosciences, 5, 451–462, doi:10.5194/bg-5-451-2008,2008.

McDermitt, D., Burba, G., Xu, L., Anderson, T., Komissarov, A., Riensche, B., Schedlbauer,J., Starr, G., Zona, D., Oechel, W., Oberbauer, S., and Hastings, S.: A new low-power,open-path instrument for measuring methane flux by eddy covariance, Appl. Phys. B,1–15–15, 2010.

McMillen, R. T.: An eddy correlation technique with extended applicability to non-simpleterrain, Bound.-Lay. Meteorol., 43, 231–245, 1988.

Neftel, A., Ammann, C., Fischer, C., Spirig, C., Conen, F., Emmenegger, L., Tuzson, B., andWahlen, S.: N2O exchange over managed grassland: Application of a quantum cascadelaser spectrometer for micrometeorological flux measurements, Agr. Forest Meteorol.,150, 775–785, 2010.

Press, W. H.: Numerical recipes in C, Cambridge University Press, 1988.

R Development Core Team: R: A language and environment for statistical computing,R Foundation for Statistical Computing, Vienna, Austria, available at: http://www.

R-project.org/, ISBN 3-900051-07-0, 2010.

Rella, C.: Accurate greenhouse gas measurements in humid gas streams using the PicarroG1301 carbon dioxide/methane/water vapor gas analyzer, Picarro, Inc., Sunnyvale, Cali-fornia, 2010.

Smeets, C. J. P. P., Holzinger, R., Vigano, I., Goldstein, A. H., and Rockmann, T.: Eddycovariance methane measurements at a Ponderosa pine plantation in California, Atmos.Chem. Phys., 9, 8365–8375, doi:10.5194/acp-9-8365-2009, 2009.

138

6.7. References

Stull, R.: An introduction to boundary layer meteorology, Kluwer Academic Pub, 1988.

Tuzson, B., Hiller, R. V., Zeyer, K., Eugster, W., Neftel, A., Ammann, C., and Emmenegger,L.: Field intercomparison of two optical analyzers for CH4 eddy covariance flux measure-ments, Atmos. Meas. Tech., 3, 1519–1531, doi:10.5194/amt-3-1519-2010, 2010.

Webb, E., Pearman, G., and Leuning, R.: Correction of flux measurements for density effectsdue to heat and water vapour transfer, Quart. J. Roy. Meteorol. Soc., 106, 85–100, 1980.

Zeeman, M. J., Hiller, R., Gilgen, A. K., Michna, P., Pluss, P., Buchmann, N., and Eugster,W.: Management and climate impacts on net CO2 fluxes and carbon budgets of threegrasslands along an elevational gradient in Switzerland, Agr. Forest Meteorol., 150, 519–530, doi:10.1016/j.agrformet.2010.01.011, 2010.

139


Figure 6.1.: Results from the dilution experiment for CO2 (a) and CH4 (b). Moisture wasadded to the dry gas at 9 increasing levels, which were held constant over20 min. The dilution of the gas explains most of concentration drop (differencebetween black and dark grey points). However, a small portion cannot be at-tributed to this process (difference between dark and light grey points) and musthave a different source, which is also related to the water vapor concentration.The solid line indicates the mixing ratio of the supplied air.

140

6.7. References

40 50 60 70 80 90

0.5

1.0

1.5

2.0

2.5

L w′H

2O′ F

GG

A,la

g(w

′CO

2′) [

s−1]

Relative humidity [%]

●

● ●●●●

●●●

● ●

●

●●● ●

●

●●

●

●

●

●●

●●●●

●

●

●

●●

●●●

●

●●●●

●●

●

●

●●

●●

●

●

●

● ●

●●

●

●

● ●

●

●

●

● ●

●

●

●

●

●

●

●

●

●

●●

● ●●●●

●

●●

●●

●

●

● ●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●●●

●●●

●●

●●

●●

● ●●

●

●

●

●

●

●

●

●

●

●

● ●●

● ●●3.02 4.523.54

Figure 6.2.: Damping factor L for w0H2O0FGGA;lag.w0CO02/

related to relative humidity. Thegrey band shows the interquartile range for 5 % relative humidity bins with atleast 8 observations. This corresponds to 50 % of all observed values. The fineline connects the medians of the individual bins and the bold line indicates themodel fit through the medians.

141


10 20 30 40

0

50

100

150

200

Fre

quen

cy

Lag [#records]

w′CO2′

w′H2O′

w′CH4′

(a)10 20 30 40

0

50

100

150

200

Fre

quen

cy

Lag [#records]

w′CO2 dry′

w′H2O dry′

w′CH4 dry′

(b)

Figure 6.3.: Histograms of the time lags for w0CO02, w0H2O0, and w0CH04 within the searchwindow of 10–40 records (0.5–2 s). In (a), the time lags were determinedfrom humid mole fractions whereas in (b), dry mole fractions were used. Onlyrecords where all three fluxes showed a detectable time lag were selected andthus the number of observations is equal for all gases of the individual his-togram.

142

6.7. References

10

15

20

25

30

35

40

lag

[#re

cord

s]

Bins of rh [%]

30−40 50−60 70−80 90−100

lag(w′H2O′)lag(w′CO2′)

(a)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

w′H

2O′ la

g(w

′CO

2′)w

′H2O

′ lag(

w′H

2O′)

Bins of rh [%]

30−40 50−60 70−80 90−100

(b)

Figure 6.4.: (a): time lags of w0H2O0 depend on the relative humidity of the air whereasno such relation was found for w0CO02. (b): the relative difference between thewater vapor flux at the time lag of w0CO02 and at the optimal lag maximizingw0H2O0.

143


Frequency (Hz)

f Co w

,c (f

) [p

pm m

s−1

]

0.0001 0.001 0.01 0.1 1 10

0

10

20

30

40

50 optimal laglag w′CO2′

Figure 6.5.: Cospectrum of w0H2O0 for 26 July 2009 11:40–13:25 CET (217 records). Thesolid line shows the cospectrum of w and H2O at the time lag maximizingthe flux. The dashed line represents the cospectrum of w0H2O0 at the timelag of w0CO02. The total flux is slightly lower with considerable dampingin the higher frequencies. The relative humidity of the air was around 54 %(�640 mmol m�3).

144

6.7. References

40 50 60 70 80 90 100

0.850.900.951.001.051.101.15

H2O

FG

GA

H2O

IRG

A [−

]

Relative humidity [%]

●●

● ●

●

●

●●

●

●●

● ●●

●

●

●●

●●

●

●●

●●

●

●

●

●●

●

●

●●

●

●

●

●

●

●

●●

●●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●●

●●●

●

●

●

● ●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●●

●

●

● ●

●

●

●

●

●

●●

● ●

●

●●

●

●

●

●●●

●● ●

●

●●

●●●

●●●

●●

●●●

●

● ●●

●

●

●

● ●

●

●

●

●

●

●

●●

●

●●

●●

●●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●

●●

●

●●

● ●

●●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●●

●

●

●

●●

●

●

●

●

● ●

● ●

●

●

●

●

●●

●

●●

●

●

●

●

●

●

●

●

●●

●●

●

●

●

●

●

●

●

●

●●

●● ●

●

●●

●●●●

●

●●

●●

●

●

●

●●

●

●●●

●

●

●

●●●

●

●●●

●

●● ●

●●●

●●

● ●

●

●

●●

●●

●

●

●●

●

●●●

●

●●

●

●

●

●

●

●

●

●●

●

●●●

●

●●

●●●

●●

●

● ●●

●

●

●

●

●

●

●

●

●●

●●

●●●●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●

●

●

●●●●

●

●

●

●

●

●●

●●

●●

●

●

●

●

●

●

● ● ●●

●

●●

●

●●

●

●

●●

●

●●

●●

●

●

●●

● ●

●●

●●

●

●

●

●

●

●

●●

●

●

●

●

●●

●

●

●

●

●●

●

●

●●

● ●

●●

●●

●

●●

●●

●●

● ●

●

●

●

●●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●

●●

●

● ●

●

●

●

●

● ●

●

●

●

●●

●

●

●

● ●

●

● ●

●●

●

●

●

●

●

●

●●

●●

●●

●

●

●

●●

●●

●

●

●●●

● ●

●●

●

●

● ●

●

●●

●

●●

●

●●

●●

●●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●

●●

●

●

●

●●●

●

●

●

●

●

● ●

●

●●

●

●

●

●

●

●

●●● ●

●

●

●

●●

●

●

●

● ●●

●

●

●●

●●●●●

● ●●

●

●●

●●

●

●

●

●●

●

●●

●

●

●

●

●

●●

●● ●

●

●

●

●

●

●

●

●

●

●●

●

●●

●

●

●

●●

●

●

●

●

●

●

●●●

●●

●

●

●

●

●

●

●●

●

●

●●

● ● ●

●

●

●

●●

●

●

●

●

●

●

●

●

●●

●

●●

●●

●

●

●

●●

● ●

●

●

●●

●

●

●

●●

●

●

●●

●

●

●●

●

●

●

●

●

●

●

●

● ●

●

●●

●

●

●●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●●

●

●

●●

●

●●

●

●

●

●

●

●

●●

●

●

●

●

●

●●

●●

●●

●

●

●●

●

●●

●

●

●

●● ●

●

●●

●

●

●

●●●

● ● ● ●

●

●

●●

●●

●

●

●

●

●● ●

●

●●

●

●●●

●●

●●●

●

●

●● ●

●●●

●

●

●

●

●

●

●

●

●

●

●

●

●●●

●

●

● ●● ●

●

● ●

●●

● ●

●

●●

●

●

●

●

●

Figure 6.6.: Water vapor concentration ratio between the FGGA and LI-7500 decreases withhigher relative humidity of the air. The dots show the individual data points. Theband represents the interquartile range for 5 % relative humidity bins. The blackline indicates the polynomial fit of the second order through the class medianswhereas the white line connects the medians of the classes.

145


●

●

●

●

●

●

●

●

−4 −2 0 2 4 6

−2

0

2

4

6

8

w′H2O′FGGA [mmol m−2s−1]

w′H

2O′ IR

GA [m

mol

m−2

s−1]

●

●

●

●

●

●

●

●

raw FGGA flux

w′H2O′IRGA = 1.64 ⋅ w′H2O′FGGA

corrected FGGA flux

w′H2O′IRGA = 1.17 ⋅ w′H2O′FGGA

1:1 relationship

Figure 6.7.: Water vapor fluxes of the two systems. w0H2O0IRGA were binned into classesof 1 mmol m�2 s�1. The symbols denote the median values for both variableswithin the class. The cross hairs span the interquartile range. The dashed linerepresents the values before correcting w0H2O0FGGA;lag.w0CO0

2/for damping losses,

whereas the solid line denotes the results thereafter. The dotted line indicatesthe 1:1 relationship.

146

6.7. References

FC

H4 [

nmol

m−2

s−1]

−40

−20

0

20

40

−40

−20

0

20

40

−40

−20

0

20

40

−40

−20

0

20

40

−40

−20

0

20

40

−40

−20

0

20

40

FC

H4,

raw

+W

PL

+W

PL

(d⋅e

)

+b

⋅WP

L(d

⋅e)

FC

H4,

dry

median(FCH4,dry)

Figure 6.8.: Boxplots of the observed raw and corrected CH4 fluxes. From left to right, thefirst one shows the raw flux. The second boxplot shows the flux after applyingthe original WPL term. In the third one, the water vapor flux is damped as itwould be observed in the instrument whereas the forth also includes the spectralbroadening effect of the instrument. The last boxplot is considered the referenceflux: the flux when raw values are converted to dry mole fractions before thecalculation of the covariance. The solid line is drawn at zero and the dashed lineindicates the expected value, the median of the reference flux.

147


Table 6.1.: Fitted parameters for Eq. (6.10) (CO2, CH4, and H2O in ppm).

Parameter Unit Values for CO2;FGGA Values for CH4;FGGA Values for CH4;FMA

a Œppm�1��1:219 � 10�06

(˙2:169�10�09)�1:189 � 10�06

(˙1:480�10�09)

�1:219 � 10�06

(˙1:182 � 10�09)

b Œppm�2�1:229 � 10�12

(˙1:073�10�13)2:096 � 10�13

(˙7:318� 10�14)1:678 � 10�12

(˙5:846�10�14)R2 0.9928 0.9966 0.9992

148

6.7. References

Table 6.2.: Fitted parameters for Eq. (6.10) without term b�H2O2 (CO2, CH4, and H2O inppm).

Parameter Unit Values for CO2;FGGA Values for CH4;FGGA Values for CH4;FMA

a Œppm�1��1:195 � 10�06

(˙5:825�10�10)�1:184 � 10�06

(˙3:952�10�10)�1:186 � 10�06

(˙3:265�10�10)R2 0.9927 0.9966 0.9991

149

7General discussion

7.1. Synthesis

Comparing GHG fluxes between different spatial scales is a challenging task (Xiao et al.,2012) that is only possible if accurate flux measurements or estimates are available for thedifferent spatial scales. However, such analyses are amongst other requirements a prere-quisite to confidently estimate the GHG budget of regions or countries. Most current GHGinventories base on the bottom-up approach where fluxes are estimated from emission factorsthat rely on direct measurements of the individual processes. Many studies showed thatthese inventories do not agree with direct measurements of GHG concentrations in the air(Bergamaschi et al., 2005). Hence, the verification and the therewith possible improvementof the inventories is important to support policy makers to e.g. decide about most powerfulmeasures in reducing GHG emission.

This study aimed to contribute to a better understanding of greenhouse gas fluxes at theecosystem and regional scale with a special focus on the scalability of the fluxes betweenthe spatial scales.

Starting at the most detailed level, Chapter 6 quantifies the effect of H2O cross-sensitivityon the measurements of laser spectroscopy instruments on trace gas fluxes. This occurrenceleads to an intensification of the dilution effect and hence an enhanced density flux correctionis required to account for H2O introduced density changes of the air. In our instrumentalsetup, damping processes of the H2O signal counteract the additional dilution effect. Theseinclude for example sorption and desorption processes at the sidewalls of the intake tubesand in the filter with integrated droplet separator. Hence, the application of the originaldensity flux correction after Webb et al. (1980) calculated from the H2O flux measured by

151

Chapter 7: General discussion

another instrument might be inappropriate. Drying the air prior to sampling would overcomethis problem, but tends to have a negative side effect on the high-frequency performanceof the instrument (Griffis et al., 2008) and therewith on the flux. Therefore, we deriveda function to transfer the ambient water vapor flux to the one observed in the measurementcell. The cross-sensitivity effect of the instrument to water is compensated by the dampeningeffect and finally results in no change of the original density flux correction. This mayexplain why Detto et al. (2011) did not see a difference between fluxes calculated from drymole fractions and from humid mole fractions corrected for density effects by the originaldensity correction. The density flux correction itself is however important for small fluxesat comparably high background concentrations, as shown for our fluxes measured over agrassland. For large emission fluxes observed over a wetland ecosystem, the effect of thedensity flux correction is small, but still biases annual budget towards higher uptakes orlower emissions if not corrected for.

At the instrumental setup level, the question arises whether the ecosystem flux is ade-quately measured. Chapter 5 evaluates the performance of two completely different setupsto measure methane fluxes. Both setups were capable to accurately measure the prescribedflux simulated in a tracer experiment under field conditions. Two recent studies support thesefindings. Peltola (2011) compared the performance of a prototype of the LI-7700 (Li-CorInc., Lincoln, NB, USA), a G1301-f (Picarro Inc., Santa Clara, CA, USA), a TGA-100A(Campbell Scientific, Inc., Logan, Utah, USA) and a Fast Methane Analyzer (FMA, LosGatos Research Inc., Mountain View, CA, USA) and presented good agreement between themeasured fluxes of the individual instruments, especially between the fluxes derived withthe G1301-f and the FMA. Detto et al. (2011) reported similar results for a FMA and aprototype of the LI-7700. Hence, the new spectroscopic instruments enable accurate CH4

flux measurements by the eddy covariance technique for methane emissions from wetlandsand grazed pastures. Very small fluxes as for example observed uptake rates over grass-lands hit the detection limit of all setups and spectroscopic effects as well as the density fluxcorrection gain importance.

Direct measurement of the trace gas fluxes from a certain ecosystem type is not alwaysstraightforward as shown in Chapter 4. Especially in a heterogeneous landscape, smallpatches of a certain land type can add to a considerable area. A good example thereforeare cultivated turf grasses that cover 2% of the continental United States and are estimatedto take up 17 Tg C yr�1 (Milesi et al., 2005). Private backyards are however often too smallto successfully deploy eddy covariance measurements. Elsewise, they are likely influencedby nearby CO2 sources such as traffic, as they were at an over 1.5 ha large turf-grass fieldin the first-ring suburb of Minneapolis–Saint Paul, Minnesota, USA. Traffic contribution

152

7.1. Synthesis

was estimated based on winter fluxes (when vegetation was dormant), footprint analyses,and continuous traffic counts. The derived relationship between the footprint influenced bythe road and the traffic count allowed to successfully removing the traffic influence on themeasured CO2 flux. The annual CO2 budget was reduced by � 135 g C m�2 yr�1 for twosubsequent years. Hence, footprint models supply a powerful tool to partition fluxes fromdifferent sources. Peters et al. (2011) (Appendix A) even goes one step further and comparesthe water vapor flux measured at a tall tower (40 m) with the water vapor flux from the indi-vidual ecosystem types within the footprint of the tower measurement in the same suburbanarea as the turf-grass flux measurements were performed. A good agreement between thecomponent-based and total H2O flux was found with an overall imbalance of 3%.

While the study in Appendix A applied various measurement methods to estimate theindividual components, Chapter 3 compares different methods to measure CH4 emissionsfrom a mid-latitudes run-off-river hydropower reservoir. Chamber measurements only rep-resent a small surface area of the lake whereas eddy covariance measurements integrateover a larger part of the lake. Both methods are however not capable to directly assess thetotal CH4 emissions of the lake and their representativeness is questioned. Aircraft basedflux estimates result in similar (same order of magnitude) fluxes as observed by chambersthat were deployed on the same day and compare well with eddy covariance measurementsperformed in a separate study in the previous year. The higher fluxes from aircraft measure-ments compared to the chamber fluxes are likely influenced by the box model simplificationand therewith associated assumptions. Nevertheless, airborne flux measurements are capa-ble to detect potential large-area CH4 sources and provide a rough estimate of their quantity.

Finally, Chapter 2 compares two independent methods to calculate aircraft based fluxeswith a spatially explicit inventory. The simplified box model and the eddy covariance tech-nique yield fluxes at the same order of magnitude as the bottom-up inventory estimate. Whileeddy covariance fluxes tend to underestimate the emissions, the variability and the mean inthe box model fluxes are considerably higher. Since airborne eddy covariance measurementstend to be lower than surface flux measurements (Desjardins et al., 1989) and the simplifiedbox model is influenced by lateral advected concentration gradients, we expect the emis-sions to range somewhere between the flux estimates by the two approaches. Applying aless simplified box model including all flows through the sidewalls of the box could help tofurther constrain the model estimate.

This thesis contributes to a better understanding of greenhouse gas fluxes at the ecosystemand regional scale, including a critical assessment of the used method. Regional scale CH4

fluxes were calculated from aircraft measurements. While some uncertainties for the sim-plified and full-featured box model remain, eddy covariance fluxes are robust and compare

153


well with the spatially explicit inventory for the area of interests. Footprint models have beenshown to be a powerful tool to partition fluxes from different sources and associate fluxeswith their source region. A field comparison successfully showed the applicability of the twoCH4 flux measurement setups and compared the fluxes with artificial prescribed CH4 flux.Additionally, the influence of spectroscopic effects of H2O on the CH4 flux measurements isassessed for a Fast Methane Analyzer.

7.2. Outlook

Twenty flight days provide a solid data set to evaluate aircraft based flux measurements.Additional analyses and measurements could contribute to a more detailed understanding ofCH4 emissions in Switzerland.

So far, only agriculturally dominated landscapes and lake ecosystem were considered inthe analyses. While agroecosystems are the most important source in Switzerland, emis-sion from those should be contrasted against emissions from e.g. urban environments. Theevaluation of larger boxes as e.g. shown by Lehning et al. (1996) for O3 might give a moreintegrated picture on Swiss CH4 emissions. Additionally, small regional scale studies ascompleted for the Reuss Valley or Lake Wohlen can be deployed over other CH4 emittingareas such as wetlands to constrain their source strength.

Nevertheless, the analyses remain restricted to the areas covered by aircraft measure-ments. While the spatial resolution of the current satellite based CH4 observations is lowerthan from aircraft measurements, they provide a larger spatial extent and a regular samplinginterval. The larger spatial extent would be helpful to connect and compare the measure-ments with regional and global models. The regular sampling interval would also extend thetemporal frame of the observations.

The flights were restricted to daytime for several reasons. The most important one is thatthe boundary layer is stably stratified during the night and turbulent mixing is very low.Hence, the influence of sinks and sources at the surface is restricted to the lowest few tens tohundreds of meters and pronounced vertical concentration profiles develop. Airborne mea-surements very close to the surface are dangerous, especially by night when sight is low.On the other hand, the relatively shallow boundary layers allows ground based concentra-tion profile measurements from towers or with tethered balloon systems. These types ofmeasurements are currently conducted by the project FasMeF (Farm-scale Methane Fluxes)founded by the Swiss National Science Foundation.

CH4 fluxes estimated by the simplified box model approach scatter more than eddy co-variance fluxes calculated from the same data. Hence, further evaluation of the box model

154

7.3. References

approach is needed to reduce the uncertainties associated with this method. This could bedone in an additional field experiment where the sides of the box are constrained with directmeasurements from close to the ground through the total height of the boundary layer. Mea-surements at the top of the boundary layer could be used to estimate the vertical exchangewith the free troposphere.

Footprint models are a powerful tool to link concentration measurement with surfacefluxes and will also be applied to the aircraft data described in this thesis as part of anadditional project. The basis for the planned analysis is the high-resolution transport modelFLEXPART-COSMO developed by Empa. A preliminary case study shows promising re-sults that this analysis will contribute to the evaluation of the spatially highly resolved CH4

emission inventory. In a further step, the footprints can also be used to derive an adaptedinventory constrained by inverse modeling.

Especially the application of inversed models would profit from measured isotopic ratiosof ı13C and ı2H in CH4. The isotopic signal of natural processes and combustion differs(Miller, 2005) and hence can help to disentangle the different sources that contribute to themeasured concentration, as successfully shown by Fletcher et al. (2004)

7.3. References

Bergamaschi, P., M. Krol, F. Dentener, A. Vermeulen, F. Meinhardt, R. Graul, M. Ramonet,W. Peters, and E. J. Dlugokencky (2005). Inverse modelling of national and EuropeanCH4 emissions using the atmospheric zoom model TM5. Atmospheric Chemistry andPhysics 5(9), 2431–2460. doi:10.5194/acp-5-2431-2005.


Detto, M., J. Verfaillie, F. Anderson, L. Xu, and D. Baldocchi (2011). Comparinglaser-based open- and closed-path gas analyzers to measure methane fluxes using theeddy covariance method. Agricultural and Forest Meteorology 151(10), 1312–1324.doi:10.1016/j.agrformet.2011.05.014.

Fletcher, M., S. E., P. P. Tans, L. M. Bruhwiler, J. B. Miller, and M. Heimann (2004, Octo-ber). CH4 sources estimated from atmospheric observations of CH4 and its 13C/12C iso-topic ratios: 1. inverse modeling of source processes. Global Biogeochem. Cycles 18(4),GB4004. doi:10.1029/2004GB002223.

Griffis, T. J., S. D. Sargent, J. M. Baker, X. Lee, B. D. Tanner, J. Greene, E. Swiatek,and K. Billmark (2008). Direct measurement of biosphere-atmosphere isotopic CO2 ex-change using the eddy covariance techniques. Journal of Geophysical Research. 113(D8),D08304. doi:10.1029/2007JD009297.

155


Lehning, M., H. Richner, and G. L. Kok (1996). Pollutant transport over complex terrain:Flux and budget calculations for the POLLUMET field campaign. Atmospheric Environ-ment 30(17), 3027–3044. 10.1016/1352-2310(96)00007-6.

Milesi, C., S. Running, C. Elvidge, J. Dietz, B. Tuttle, and R. Nemani (2005). Mapping andmodeling the biogeochemical cycling of turf grasses in the united states. EnvironmentalManagement 36(3), 426–438. doi:10.1007/s00267-004-0316-2.

Miller, J. B. (2005). The carbon isotopic composition of atmospheric methane and its con-straint on the global methane budget. In Stable Isotopes and Biosphere Atmosphere Inter-actions, pp. 288–310. San Diego: Academic Press.

Peltola, O. (2011). Field intercomparison of four methane gas analysers suitable for eddycovariance flux measurements. Master’s thesis, University of Helsinki. pp. 74.

Peters, E. B., R. V. Hiller, and J. P. McFadden (2011). Seasonal contributions of vegeta-tion types to suburban evapotranspiration. Journal of Geophysical Research. 116(G1),G01003. doi:10.1029/2010JG001463.

Webb, E., G. Pearman, and R. Leuning (1980). Correction of flux measurements for densityeffects due to heat and water vapour transfer. Quarterly Journal of the Royal Meteorolog-ical Society 106(447), 85–100.

Xiao, J., J. Chen, K. J. Davis, and M. Reichstein (2012). Advances in upscaling of eddy co-variance measurements of carbon and water fluxes. Journal of Geophysical Research. 117,G00J01. doi:10.1029/2011JG001889.

156

This chapter was published in Journal of Geophysical Re-

search, 2011, 116, G01003, doi:10.1029/2010JG001463

ASeasonal contributions of vegetation

types to suburban evapotranspiration

Emily B. Peters1, Rebecca V. Hiller2, Joseph P. McFadden3

1Department of Ecology, Evolution, and Behavior and Institute on the Environment, University of

Minnesota, Saint Paul, Minnesota, USA2Institute of Plant, Animal and Agroecosystem Sciences, ETH Zurich, Zurich, Switzerland

3Department of Geography, University of California, Santa Barbara, California, USA

157

Seasonal contributions of vegetation types to suburbanevapotranspiration

Emily B. Peters,1 Rebecca V. Hiller,2 and Joseph P. McFadden3

Received 25 June 2010; revised 8 October 2010; accepted 26 October 2010; published 22 January 2011.

[1] Evapotranspiration is an important term of energy and water budgets in urban areasand is responsible for multiple ecosystem services provided by urban vegetation. Thespatial heterogeneity of urban surface types with different seasonal water use patterns (e.g.,trees and turfgrass lawns) complicates efforts to predict and manage urbanevapotranspiration rates, necessitating a surface type, or component‐based, approach. In asuburban neighborhood of Minneapolis–Saint Paul, Minnesota, United States, wesimultaneously measured ecosystem evapotranspiration and its main component fluxesusing eddy covariance and heat dissipation sap flux techniques to assess the relativecontribution of plant functional types (evergreen needleleaf tree, deciduous broadleaftree, cool season turfgrass) to seasonal and spatial variations in evapotranspiration.Component‐based evapotranspiration estimates agreed well with measured water vaporfluxes, although the imbalance between methods varied seasonally from a 20%overestimate in spring to an 11% underestimate in summer. Turfgrasses represented thelargest contribution to annual evapotranspiration in recreational and residential land usetypes (87% and 64%, respectively), followed by trees (10% and 31%, respectively), with therelative contribution of plant functional types dependent on their fractional cover and dailywater use. Recreational areas had higher annual evapotranspiration than residential areas(467 versus 324 mm yr−1, respectively) and altered seasonal patterns of evapotranspirationdue to greater turfgrass cover (74% versus 34%, respectively). Our results suggest thatplant functional types capture much of the variability required to predict the seasonalpatterns of evapotranspiration among cities, as well as differences in evapotranspirationthat could result from changes in climate, land use, or vegetation composition.

Citation: Peters, E. B., R. V. Hiller, and J. P. McFadden (2011), Seasonal contributions of vegetation types to suburbanevapotranspiration, J. Geophys. Res., 116, G01003, doi:10.1029/2010JG001463.

1. Introduction

[2] Evapotranspiration is an important term of energy andwater budgets in urban areas and it consequently plays a keyrole in management decisions related to conservation of waterresources, mitigation of urban heat islands, and reduction ofstorm water runoff. Managing these ecosystem services re-quires accurate assessments of urban evapotranspirationrates, and indeed water vapor fluxes have been measured incities representing different climates, geographical regions,and land use types [Balogun et al., 2009;Grimmond and Oke,1999; Moriwaki and Kanda, 2004; Nemitz et al., 2002;Offerle et al., 2006; Spronken‐Smith, 2002]. Collectively,these studies show that cities vary in magnitude and season-

ality of evapotranspiration due to differences in climate,irrigation, and vegetation cover. For example, densely built‐up city centers or downtown areas with high impervioussurface cover tend to have lower evapotranspiration rates andsmaller evaporative fractions [Grimmond et al., 2004;Nemitzet al., 2002] than residential areas with short buildings andhigh vegetation cover [Balogun et al., 2009;Grimmond et al.,1996, 2006; Moriwaki and Kanda, 2004].[3] Recent studies suggest that plant growth form is another

important factor controlling total evapotranspiration in urbanareas, with turfgrass‐dominated areas showing higher waterfluxes than tree‐dominated areas [Balogun et al., 2009;Kotani and Sugita, 2005; Offerle et al., 2006]. The spatialheterogeneity of urban landscapes, however, greatly com-plicates the extrapolation of these ecosystem‐scale mea-surements to other urban and suburban areas, as vegetationcomposition varies widely within and between cities. Themanagement of urban ecosystem services necessitates anapproach that quantifies urban evapotranspiration ratesaccording to the major land‐surface types present, hereafterreferred to as a component‐based approach. Partitioning ofecosystem water fluxes among stand types or species has

1Department of Ecology, Evolution, and Behavior and Institute on theEnvironment, University of Minnesota, Saint Paul, Minnesota, USA.

2Institute of Plant, Animal, and Agroecosystem Sciences, ETH Zurich,Zurich, Switzerland.

3Department of Geography, University of California, Santa Barbara,California, USA.

Copyright 2011 by the American Geophysical Union.0148‐0227/11/2010JG001463

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, G01003, doi:10.1029/2010JG001463, 2011

G01003 1 of 16

Appendix A: Vegetaton contributions to urban ET

158

been previously investigated in natural ecosystems [Baldocchiet al., 2000; Oishi et al., 2008; Paco et al., 2009], yet thisinformation is relatively lacking in urban ecosystems.[4] Plant functional types provide a potentially useful

approach for partitioning the seasonal patterns of urbanevapotranspiration, as they represent major physiologicaland biophysical differences among plant species [Reichet al., 1997] and they also can be uniquely identified inurban landscapes using high‐resolution satellite imagery[Tooke et al., 2009]. For example, evergreen needleleaftrees tend to have lower leaf‐level transpiration rates thandeciduous broadleaf trees [Givnish, 2002], while both treetypes tend to have deeper roots and access to additionalsources of water as compared to cool season grasses[Jackson et al., 1996; Ludwig et al., 2004]. Plant func-tional types also represent distinct seasonal patterns inphysiological activity, or phenology, among plant species.For example, cool season turfgrasses, which are ubiquitousacross temperate urban landscapes [Milesi et al., 2005],have higher physiological activity in the spring and fall[Zhang et al., 2007], while evergreen needleleaf anddeciduous broadleaf trees show peak function in midsummer[Givnish, 2002]. Many studies have quantified evapotrans-piration rates of turfgrasses [Feldhake et al., 1983, 1984;Zhang et al., 2007] and, to a lesser extent, of trees in urbanareas [Bush et al., 2008;Whitlow and Bassuk, 1988;Whitlowet al., 1992], yet no work has examined the combined effectsof these plant functional types on the seasonality of urbanevapotranspiration.[5] Here we report a 2 year study in which we made

independent measurements of total ecosystem evapotrans-piration and its main component fluxes to better understandhow vegetation influences the spatial and seasonal variationin suburban evapotranspiration. We measured ecosystemevapotranspiration using an eddy covariance system mounted40 m above ground, tree transpiration using heat dissipationsap flow sensors, and turfgrass evapotranspiration using aportable eddy covariance system mounted 1.35 m aboveground. Our main objectives were to: (1) determine the mag-nitude and seasonal patterns of suburban evapotranspiration;(2) evaluate how well scaled component‐based estimatesmatch measured ecosystem evapotranspiration rates; (3) deter-mine how the magnitude and seasonality of ecosystem evapo-transpiration varies between residential and recreational landuse types; and (4) examine the seasonal water use patterns ofvegetation types and their influence on ecosystem evapotrans-piration in recreational and residential land use types.

2. Methods

2.1. Site Information

[6] Our study was conducted in a first‐ring suburbanneighborhood in the Minneapolis–Saint Paul metropolitanarea in east central Minnesota, United States (44°59′N,93°11′W). Prior to rapid residential development in the1950s, the prominent land use types in the area werefarms and nurseries. The area has a cold temperate cli-mate with mean annual temperature of 7.4°C and meanannual precipitation of 747 mm. Due to its location nearthe center of the metropolitan region, the study area wasinfluenced by the urban heat island effect that has beendocumented for Minneapolis–Saint Paul [Sen Roy and

Yuan, 2009; Todhunter, 1996; Winkler et al., 1981].The relatively flat terrain made this landscape suitable foreddy covariance measurements from the KUOM broadcasttower located in the center of our approximately 7 hastudy area (Figure 1). The area immediately adjacent tothe tower consisted of residential land use (single‐family,detached housing) to the northwest and northeast and recrea-tional land use (golf course) to the southwest. Vegetated sur-faces consisted primarily of forested patches, isolated trees, andopen turfgrass lawns. Of the over 80 tree species identified inthis area, the dominant canopy species were Acer negundo(boxelder), Acer saccharinum (silver maple), Fraxinus penn-sylvanica (green ash), Gleditsia tricanthos (honey locust),Picea glauca (white spruce), Populus deltoides (cotton-wood), Quercus alba (white oak), and Ulmus americana(American elm), with a mean tree height of 12 m.[7] We selected four sites with stands of trees suitable for

sap flux measurements within our study area (Figure 1)[Peters et al., 2010]. Sites were selected to include repre-sentative tree species and sizes with the total number of siteslimited by numerous logistical constraints (e.g., equipment,landowner permission, site suitability, signal cable lengths,and access to power). Trees at all sites were grown underpark‐like conditions with open canopies and a turfgrassground cover. The lower branches of many trees werepruned, but canopies were mature and healthy with noobvious signs of disease or damage. Root systems were notlikely restricted by impervious surfaces, as the nearest roadsand sidewalks occurred beyond the drip line of all trees andoften at distances > 10 m from each tree. Surface soils at allsites were moderately compacted with an average bulkdensity of 1.43 g cm−3, as is typical of urban soils [Craul,1985]. Management regimes were considered low mainte-nance at all sites and included regular mowing, but had littleor no irrigation and no fertilizer use based on personalcommunication with the landowners.[8] We made eddy covariance measurements at a height

of 1.35 m over a 1.5 ha turfgrass field that was locatedwithin the footprint of the KUOM tall tower (Figure 1). Thissite was representative of low‐maintenance lawns in thearea, as it was mowed to a height of 7 cm approximatelyonce per week with clippings left to decompose on thesurface and was not irrigated or fertilized. The C3 coolseason turfgrass species, Poa pratensis (Kentucky blue-grass), Lolium perenne (perennial ryegrass) and Festucaarundinacea (tall fescue), dominated this site. Surface soilsat this site were slightly compacted with an average bulkdensity of 1.22 g cm−3. While turfgrass management variedamong lawns in our study area, numerous logistical con-straints associated with making eddy covariance measure-ments in an urban area (e.g., landowner permission,security, adequate fetch) restricted our ability to simulta-neously measure a highly managed lawn.

2.2. Meteorological Measurements

[9] To assess the environmental conditions in which ourstudy occurred, we measured a suite of meteorologicalvariables. At the turfgrass site, net radiation (RN) andincoming solar radiation (RS) were measured at 2 m aboveground using a four‐component net radiometer (modelCRN1, Kipp and Zonen, Delft, Netherlands). Ground heatflux (G) was measured at 5 cm depth at two locations at the

PETERS ET AL.: VEGETATION CONTRIBUTIONS TO URBAN ET G01003G01003

2 of 16

159

turfgrass site using heat flux plates (HFT‐3.1, Radiation andEnergy Balance Systems, REBS, Seattle, Washington,United States). An integrating platinum resistance temper-ature probe (STP‐1, REBS, Seattle, Washington, UnitedStates) was used to measure storage of heat in the upper5 cm of soil, which was added to the measurementsrecorded by the heat flux plates. Volumetric soil watercontent was measured at 10 cm depth (ECH2O, DecagonDevices, Inc., Pullman, Washington, United States) at theturfgrass site and calibrated relative to a set of gravimetricmeasurements made in the same soil type at the Universityof Minnesota climate station. These calibrations were con-ducted by oven‐drying soil cores of a known volume thatwere collected throughout a 2 week dry‐down period fol-lowing a precipitation event. We measured RN (CNR1,Kipp and Zonen, Delft, Netherlands) from the KUOMbroadcast tower at 150 m above ground. The 50% sourcearea [Schmid et al., 1991] of our net radiation measurementwas within 150 m of the tower (consisting of tree and grasspatches on the golf course, three small buildings with dri-veways, a two‐lane road intersection, and a small stormwater pond), while the 90% source area was within 475 m

of the tower (consisting of a mixture of golf course land useand several blocks of single‐family detached housing).Precipitation data were obtained from the University ofMinnesota climate station (<1 km from the KUOM towersite). Vapor pressure deficit (D) was calculated from thewater vapor concentration and air temperature measure-ments at the KUOM tall tower and turfgrass sites.

2.3. Measuring Suburban Evapotranspirationand Its Component Fluxes

[10] Latent (LE) and sensible (H) heat fluxes over thesuburban landscape weremeasured from the 40m level on theKUOM broadcast tower using an eddy covariance systemconsisting of a 3‐D sonic anemometer (CSAT3, CampbellScientific, Inc., Logan, Utah, United States) and a closed‐pathinfrared gas analyzer (LI‐7000, LI‐Cor, Lincoln, Nebraska,United States). The measurement height was > 2× the meanheight of the trees, which overtopped the roofs of the houses.The sonic anemometer and the filtered air inlet for the gasanalyzer were mounted on a 3 m boom that was orientedtoward the northwest (315°). The gas analyzer and the data‐acquisition system were housed in an insulated, thermostat-

Figure 1. Land cover classification derived from 2.4 m resolution QuickBird imagery for a 1 km radiussurrounding the KUOM tall tower (indicated by the letter K at the origin) in a suburban neighborhood ofMinneapolis–Saint Paul, Minnesota. The mobile tower in the turfgrass field is indicated by the letter T.The Lauderdale (L), Saint Paul (S), and Grove (G) sap flux sites are indicated by the letters correspondingto their names. The CTC sap flux site was located in a residential area 1.3 km south of the map edge and isnot shown. The white lines represent isopleths of the average relative contributions of different land areas,based on the footprint model of Kljun et al. [2004], of the measured fluxes at the 40 m level of the talltower for the years 2007 and 2008, following the approach presented by Rebmann et al. [2005]. Yellowlines delineate the residential and recreational land use areas that were used to bin the data for the analysesin Figure 8 and section 3.6. Stippled areas in the SE corner of the map indicate low herbaceous vegetationin experimental crop plots.


3 of 16


160

ically controlled, heated and ventilated rack‐mount computerenclosure at the base of the tower. The water vapor channel ofthe gas analyzer was calibrated using a dew point generator(LI‐640, LI‐Cor, Lincoln, Nebraska, United States) in anearby laboratory, with two different LI‐7000 instrumentscycled regularly between the lab and the tower. Air wassampled through a 9.5 mm I.D. pure FEP tubing at a rateof 23 SLPM and a bypass flow rate of 7.5 SLPM through4 mm I.D. pure PTFE tubing and was delivered to thegas analyzer using a system of needle valves, mass‐flowmeters, and two rotary vane vacuum pumps (Gast models1023 and 211 for the main and the bypass pumps,respectively, Benton Harbor, Michigan, United States).The digital data from the sonic anemometer was trans-mitted to the tower base using a pair of fiber optic mediaconverters linked by multimode fiber optic cable to avoidRF interference from the broadcast tower. Data collectedwith the sonic anemometer and infrared gas analyzer wererecorded at 10 Hz using in‐house software running on autility‐grade UNIX server at the tower base (V120, SunMicrosystems, Santa Clara, California, United States).[11] Thirty minute turbulent fluxes, including ecosys-

tem evapotranspiration (Etotal), were computed using aslightly modified version of eth‐flux [see Mauder et al.,2008]. Following Vickers and Mahrt [1997], a filter wasimplemented to remove spikes in the sonic data likely causedby data communication. A maximum of five values exceed-ing ± 6 standard deviations within a 300 s moving windowwere removed. This procedure was repeated three times foreach 30 min block of data. The time lag due to the transport ofair through the tubing was calculated using the peak of thecross‐correlation function between the vertical wind velocityand the water vapor concentration. Themean lag for the watervapor signals was 11 s. All subsequent analyses were per-formed using the R (version 2.8) statistical language package[R Development Core Team, 2010]. The sonic temperaturewas corrected following Schotanus et al. [1983]. High‐frequency losses in the fluxes were corrected followingEugster and Senn [1995] using a damping‐loss constantof 0.28 s−1. Data were screened to remove periods wheninstruments were being serviced and wind directions werefrom the southeast (90 to 180°) to avoid interference fromthe tower structure. The land use associated with thosewind directions was also atypical of the predominantly res-idential area because it included experimental crop plotscultivated by the University of Minnesota. Gaps in mea-surements from January 2007 to December 2008 represented48% of the data, due to unfavorable wind directions frombehind the tower (20%) and system maintenance or poweroutages (28%). We performed the minimum necessary level

of filtering to maximize the amount of data available forcomparisons of land use types. Due to the complexity of thesuburban land surface and our focus on comparing scaled‐upcomponent fluxes to measured Etotal, we did not gap‐fill thetall tower data for the analyses presented here.[12] Measured Etotal should balance against the sum of its

component fluxes such that:

Etotal ¼ ET þ EG þ EW ð1Þ

where ET is evapotranspiration from trees, EG is evapo-transpiration from turfgrass lawns, and EW is evaporationfrom open water (e.g., ponds and swimming pools). Thefocus of the KUOM flux study was to quantify fluxes from aresidential landscape, which consists of trees and predomi-nantly nonirrigated turfgrass lawns (Table 1). In carrying outthe scaling exercise reported here, we wanted to understandhow much of the imbalance between our scaled‐up com-ponent fluxes and Etotal measured by the tall tower could beaccounted for by processes and surface types we did notmeasure. Thus, we used modeled estimates of canopyinterception, EG from irrigated lawns, and understory EG toprovide a more robust assessment of our scaled‐up mea-surements. Each of the surface type fluxes included in ourscaled‐up Etotal estimates was either independently mea-sured or modeled as described below. Evaporation frombare soil and impervious surfaces was assumed to be neg-ligible because bare soil represented a small fraction of thestudy area (∼7%, much of which was under tree canopies,based on FIA plot data) and impervious surfaces effectivelyprevent evaporation from soils below.[13] ET is composed of tree transpiration (TT) and rainfall

interception loss from tree canopies (IT) such that:

ET ¼ TT þ IT ð2Þ

In our study area, Peters et al. [2010] found species’ dif-ferences in TT per unit canopy area were largely explainedby plant functional type differences in canopy structure andgrowing season length. Consequently, we used evergreenneedleleaf and deciduous broadleaf plant functional types toestimate ET at the ecosystem scale, such that:

ET ¼ ETeg þ ETdec ð3Þ

where ETeg is evapotranspiration from evergreen needleleaftrees and ETdec is evapotranspiration from deciduousbroadleaf trees.[14] TT was measured using Granier‐type heat dissipa-

tion sap flow [Granier, 1987; Lu et al., 2004] on a totalof 37 trees (17 trees in 2007 and 20 different trees in

Table 1. Cover Types by Percentage of the Residential Land Use Area, Recreational Land Use Area, and Flux Footprint ClimatologyAreaa

Suburban LandscapeDeciduousTree (%)

EvergreenTree (%)

NonirrigatedTurfgrass (%)

IrrigatedTurfgrass (%)

OpenWater (%)

ImperviousSurfaces (%)

Residential area 37 6 25 9 2 22Recreational area 19 1 11 63 2 5Footprint area 25 4 18 35 2 15

aPercent cover was determined from the land cover classification map shown in Figure 1 and does not include turfgrass under tree canopies. Thefootprint area was defined as the total area bounded by the 10% contour line (Figure 1).


4 of 16

161

2008) across four sites. Trees represented seven differentgenera (Fraxinus, Juglans, Picea, Pinus, Quercus, Tilia,and Ulmus) and two plant functional types (evergreenneedleleaf and deciduous broadleaf) [Peters et al., 2010].Measurement periods ran from May to November in 2007and April to November in 2008. The software packageBaseLiner (version 2.4.1, Hydro‐Ecology Group, DukeUniversity) was used to calculate sap flux density per con-ducting sapwood area. TT per unit canopy area was calcu-lated as the product of each tree’s sap flux density and theratio of cross‐sectional sapwood area to projected canopyarea. Cross‐sectional sapwood area was determined frommultiple cores per tree using visual estimates apparentimmediately after removal from the trunk [Lu et al., 2004].Adjustments were made for radial variation in sap flowaccording to Pataki et al. [2010].[15] IT was estimated using a tree‐based adaptation of the

Rutter canopy interception model [Rutter et al., 1975;Valente et al., 1997]. The crown of each tree was considereda closed canopy and interception loss calculated on a can-opy‐area basis. Following Wang et al. [2008], we modeledcrown storage capacity of both evergreen needleleaf anddeciduous broadleaf trees as a function of leaf area index(LAI) with a specific leaf storage of 0.2 mm. The seasonalpattern of LAI was modeled using piecewise logisticequations fit to stand‐level LAI measurements (LAI‐2000,LI‐Cor, Lincoln, Nebraska, United States) that were col-lected biweekly at sap flux sites from prior to leaf‐out toafter senescence in 2007 and 2008 [Peters and McFadden,2010]. Maximum LAI values for evergreen needleleaf anddeciduous broadleaf trees were determined from midsum-mer LAI measurements on each study tree in 2008 [Peterset al., 2010] and LAI ranges were determined from sea-sonal LAI patterns observed in homogeneous stands ofevergreen needleleaf and deciduous broadleaf trees in 2006[Peters and McFadden, 2010]. Modeled LAI of evergreenneedleleaf trees ranged from a minimum of 7.8 m2 m−2 inwinter to a maximum of 8.8 m2 m−2 in summer, whilemodeled LAI of deciduous broadleaf trees ranged from aminimum of 0 m2 m−2 in winter to a maximum of 5.5 m2 m−2

in summer. IT was modeled every half‐hour by calculatingpotential evapotranspiration (EP) in kg H2Om−2 s−1 using thePriestley‐Taylor equation [Priestley and Taylor, 1972]:

EP ¼ �D RN � Gð Þ� Dþ �ð Þ ð4Þ

where a is a constant of 1.26, RN is net radiation inMJ m−2 s−1, G is ground heat flux in MJ m−2 s−1, l islatent heat of vaporization in MJ kg−1, D is the slope of thesaturation vapor pressure temperature curve in kPa °C−1, andg is the psychrometric constant in kPa °C−1.[16] Through our subsequent analyses comparing mea-

sured and component‐based estimates of Etotal, we deter-mined it was important to include irrigated and nonirrigatedturfgrass vegetation types when scaling up EG to the sub-urban ecosystem. Because most tree canopies in our studyarea had a turfgrass understory, we also included twounderstory turfgrass vegetation types in our estimate of EG,such that:

EG ¼ EGnirr þ EGirr þ EGunder eg þ EGunder dec ð5Þ

where EGnirr is evapotranspiration from nonirrigated turf-grass lawns, EGirr is evapotranspiration from irrigated turf-grass lawns, EGunder_eg is evapotranspiration from turfgrassgrowing below evergreen needleleaf tree canopies, andEGunder_dec is evapotranspiration from turfgrass growingbelow deciduous broadleaf tree canopies.[17] EGnirr was measured using the eddy covariance

method over a 1.5 ha turfgrass field located in the footprintof the KUOM tall tower [Hiller et al., 2010]. Water vaporand sensible heat fluxes were measured using an eddycovariance system consisting of a 3‐D sonic anemometer(CSAT3, Campbell Scientific, Logan, Utah, Unites States)and an open‐path infrared gas analyzer (LI‐7500, LI‐Cor,Lincoln, Nebraska, United States) inclined at an angle of 30°toward the west (270°), both mounted at a height of 1.35 mabove the ground on a portable meteorological tripod (905,Met One Instruments, Grants Pass, Oregon, United States).Wind velocity, air temperature, and scalar concentrations ofwater vapor were recorded at 20 Hz using a data logger(CR5000, Campbell Scientific, Logan, Utah, United States)and fluxes were computed over 30 min periods with theeth‐flux software by applying a time lag, if needed, andcalculating the covariance between the vertical wind speedand the water vapor concentration. No correction for self‐heating of the LI‐7500 [Burba et al., 2008] was applied inthe water vapor flux calculations because the effect wasestimated to be < 1% during cold winter periods and negli-gible during the growing season period. The turbulence dataprocessing followed the same procedures as for the tall towerexcept that the open‐path gas analyzer data were correctedfor changes in air density and sensible heat following Webbet al. [1980] and high‐frequency losses in the fluxes werecorrected using Moore’s [1986] transfer functions for lineaveraging and sensor separation. Gaps in measurements fromJanuary 2007 to December 2008 represented 42% of thedata, due to unfavorable wind directions from behind thetower (0–180°), power outages, and unsatisfied quality cri-teria [Foken and Wichura, 1996]. Gaps in EGnirr fluxes werefilled using multiple linear regression models of RN and airtemperature with 2 week windows and separate daytime andnighttime fitted coefficients.[18] EGirr was calculated using the Priestly Taylor equa-

tion (equation (4)) with a equal to 0.87. Also known as thecrop coefficient, a ranges from 0.6 to 1.14 for nonstressed,cool season grasses [Brown et al., 2001; Carrow, 1995;Ervin and Koski, 1998; Zhang et al., 2007]. While a isknown to vary seasonally, the exact patterns depend on spe-cies, soil type, and environmental conditions. In an effort tomodel water loss from a typical irrigated lawn in our studyarea, we chose a median a value among those reported in theliterature.[19] EGunder_eg and EGunder_dec were modeled as linear

functions of the solar radiation passing through each treecanopy [Eastham and Rose, 1988; Feldhake et al., 1985].Radiation passing through the canopy (RV) in W m−2 wascalculated using the Beer‐Lambert law [Campbell andNorman, 1998] such that:

RV ¼ RSe�kLAI ð6Þ

where RS is incident solar radiation at the top of the canopyin W m−2 and k is an attenuation coefficient. LAI was


5 of 16


162

modeled separately for evergreen needleleaf and deciduousbroadleaf trees as described above. Following another urbanstudy [Wang et al., 2008], we set k equal to 0.7 for both treetypes. Calculated RV was used to predict EGunder_eg andEGunder_dec from monthly linear regression models of mea-sured EGnirr and RS (R

2 ranged from 0.76 to 0.94 across the2007 and 2008 growing seasons). EW, which included a fewsmall ponds, water traps on the golf course, and residentialswimming pools, was calculated using the Priestly Taylorequation (equation (4)) with a equal to 1.26.

2.4. Scaling Up Component Fluxes

[20] We assessed the land cover in the study area usingsatellite imagery, aerial photography, a Geographic Infor-mation System (GIS) land use database (ArcMap, version 9.3,ESRI, Redlands, California, United States), and an urbanforest inventory. A land cover classification was producedusing QuickBird (2.4 m resolution) multispectral imageryacquired on 26 July 2006 and leaf‐off, true color aerialimagery (0.15 m resolution) acquired by Ramsey County on8 April 2003. Preparation of the QuickBird imagery includedorthorectification using a 10 m digital elevation model,georectification using the Ramsey County survey controlnetwork, masking of water bodies using a GIS data layer, andnormalized difference vegetation index (NDVI) thresholdingto isolate the vegetated fraction of the image. The vegetationclasses were then extracted using the following steps. First,an unsupervised isodata classification was used to separateareas covered by trees, turfgrass, and shadow. Second, theshadowed area was masked and reclassified using a super-vised classifier. Third, the tree class was separated intodeciduous broadleaf and evergreen needleleaf tree classesusing a green and red normalized band ratio from the leaf‐offcolor aerial imagery. Fourth, irrigated and nonirrigatedturfgrass areas were separated using a minimum‐distancesupervised classification. Fifth, swimming pools were clas-sified using a simple ratio of the blue to the near infraredbands of the QuickBird imagery, within the residential landuse area. We assessed the accuracy of the land cover clas-sification using the subplot centers of 150 randomly selectedfield plots that we surveyed in 2005 and 2006 using the U.S.Forest Service Forest Inventory and Analysis (FIA) urbanforest protocol [U.S. Department of Agriculture, 2005]. Theoverall per‐pixel accuracy of the classification was 82%,with similar accuracies of ∼80% for the deciduous tree,evergreen tree, and turfgrass classes, and 93% for theimpervious surface class.[21] Footprint models have been widely used to estimate

the source area of eddy covariance measurements of Etotal

[Kljun et al., 2004; Schmid, 1994]; however, these modelsassume that measurements are made over extended homo-geneous surfaces and they have not been rigorously vali-dated in spatially complex urban landscapes [Vesala et al.,2008a]. Although our study area contains considerablespatial heterogeneity in surface types, it does not violatethese model assumptions as much as a densely built‐up citycenter or downtown area because it has a relatively flattopography, a dense tree canopy that overtops most of thebuildings, and lacks deep street canyons caused by tallbuildings. Therefore, to allow for comparisons betweenmeasured Etotal and scaled component fluxes, we used aparameterized version of a Lagrangian stochastic footprint

model [Kljun et al., 2004] to estimate the crosswind inte-grated footprint of each half‐hourly measurement of Etotal.We ran the footprint model using a fixed boundary layerdepth of 1000 m and a roughness length calculated from theeddy covariance measurements. As the Kljun et al. [2004]parameterization does not predict two‐dimensional fluxsource areas, we used the following approach based onBarcza et al. [2009] to obtain a spatial sample of the frac-tional cover of land surface types within the source area.First, we used the footprint model to calculate distance fromthe tall tower to the maximal contribution point as well asthe distances corresponding to every tenth percentile of thefootprint for each half‐hourly measurement of Etotal. Sec-ond, we plotted the 10 points of the flux footprint along themean wind direction on the land cover map in a GIS andcalculated the fractional cover of each component classwithin a 1 ha circular area (radius = 56 m) around eachpoint. Third, we calculated the average of the 10 sampledareas, each weighted by the crosswind‐integrated footprintcorresponding to that point, to obtain the final estimate ofthe fractional land cover that contributed to each half‐hourlyflux measurement. To estimate the fractional cover of turf-grass underneath the tree canopy, we calculated the differ-ence between the percent turfgrass cover from fieldmeasurements on FIA plots (which included understoryvegetation) and the percent turfgrass cover derived from thesatellite data (which could not detect the understory). Onaverage, 50% of the tree‐covered area had a turfgrassunderstory and this value was used to represent the frac-tional cover of understory turfgrass for the entire footprintarea.[22] For comparison with measured Etotal, component

fluxes were converted from a per cover area basis to a perground area basis by multiplying ETeg, ETdec, EGnirr, EGirr,EGunder_eg, EGunder_dec, and EW by the fractional cover esti-mates and summing the terms. Comparisons between mea-sured and component‐based Etotal methods were restricted tohalf‐hour periods in 2007 and 2008 when friction velocityu* > 0.2 m s−1 and scaled Monin‐Obukhov length zm/L ≥−200 and ≤1. These direct comparisons between thesummed component fluxes and measured Etotal from thetall tower were possible only on a half‐hourly basisbecause gaps in the tall tower time series (meaning welacked both the measured Etotal fluxes and the footprintinformation needed to scale up the component‐based Etotal)inhibited our ability to make comparisons over longer in-tegrations such as monthly or annual sums. We conductedan error analysis of the component‐based estimates of Etotal

by propagating errors associated with the land coverclassification, the measured and modeled component fluxes[Taylor, 1982]. EGnirr was assumed to have a relative errorof 20%, which is typical of eddy covariance measurementsin general [Baldocchi et al., 1988]. EGirr was assumed tohave a relative error of 16%, based on the published range ofa reported for cool season turfgrass species [Brown et al.,2001; Carrow, 1995; Ervin and Koski, 1998; Zhang et al.,2007]. For both evergreen and deciduous trees, TT wasassumed to have a relative error of 40%, based on the mea-sured variation in annual sums among individual trees. IT wasassumed to have a relative error of 40% and 23% for decid-uous and evergreen trees, respectively, based on measuredvariation in LAI. Because EW, EGunder_dec, and EGunder_eg


6 of 16

163

were minor components of Etotal and we did not have a strongquantitative basis for estimating their respective errors, theywere excluded from this error analysis.[23] To compare measured Etotal from a residential

neighborhood and a recreational land use area in a golfcourse, we delineated two regions of interest (Figure 1)based on the Ramsey County GIS database. We restrictedland use comparisons to periods when the 10% flux con-tribution fell exclusively within one of the two land usetypes during daytime conditions (RS > 10 W m−2). AnnualEtotal from recreational and residential land use areas wasdetermined by scaling component fluxes according to theirrespective land cover fractions shown in Table 1.

3. Results

3.1. Land Cover

[24] Open turfgrass lawns were the dominant land covertype in our suburban study area, representing 53% of the talltower footprint climatology, followed by 29% tree cover,and 15% impervious surface cover (Table 1). Of the tree‐covered area, 86% was composed of deciduous broadleafspecies and 14% of evergreen needleleaf species. 45% oftrees had a DBH < 20 cm and 74% of trees had a projectedcanopy area < 40 m2 (Figure 2). A turfgrass understory waspresent in 52% of tree‐covered areas. Vegetation covervaried considerably between the two dominant land usetypes, such that turfgrass cover dominated the recreationalarea and tree cover dominated the residential area. Therecreational area had 40% higher turfgrass cover, 54% moreirrigated turfgrass cover, 23% less tree cover, and 13% lessimpervious surface cover than the residential area. Thedifference in percent cover of irrigated turfgrass betweenresidential and recreational areas was confirmed with per-sonal observations during the FIA field inventory that mosthomes did not have automatic irrigation systems.

3.2. Environmental Conditions

[25] The seasonal patterns of RN, air temperature, and Dwere relatively similar between the 2 years of study, butthere were several periods with higher D in the summer of2007 compared to 2008. The seasonal patterns of precip-itation and soil moisture varied considerably between 2007and 2008 (Figure 3). Not only was total annual rainfall

higher in 2007 (749 mm) and more similar to the 30 year(1971–2000) average (747 mm) than 2008 (550 mm), theseasonal distribution of rainfall varied as well [NationalClimatic Data Center (NCDC), 2004]. In 2007, the

Figure 2. Probability density functions of tree (a) diameter at 1.4 m height (DBH) and (b) projectedcanopy area in a suburban area of Minneapolis–Saint Paul, Minnesota. Trees were sampled (n = 164)using FIA urban forest protocols and were located within the tall tower footprint area, defined as thetotal area bounded by the 10% contour line (Figure 1).

Figure 3. Environmental conditions in a suburban area ofMinneapolis–Saint Paul, Minnesota from 1 January 2007to 31 December 2008. (a) Daily totals of net radiation(RN) were measured at the 150 m level of the KUOM tower.(b) Mean daily air temperature (Tair) and mean daytimevapor pressure deficit (D) were measured at the 40 m levelof the KUOM tower. (c) Soil water content (SWC) wasmeasured at 10 cm depth at the turfgrass site. Precipitation(bars) was measured at a nearby (<1 km) climate station.Gaps in soil water content data indicate time periods whenthe instrumentation failed.


7 of 16


164

greatest amount of precipitation occurred in September(170 mm), whereas in 2008 it occurred in April (109 mm).August was the wettest month (103 mm) over the 30 yearaverage [NCDC, 2004]. Consequently, rainfall patterns in2007 were relatively more similar to the long‐term averagethan in 2008. Air temperature and D were also higher inlate spring and early summer in 2007 compared to 2008.Soil moisture reached a growing season (April–November)low of 17% on average in May of 2007 compared to a lowof 11% on average in August of 2008.

3.3. Ecosystem Evapotranspiration

[26] Etotal from the suburban landscape, measured fromthe 40 m level on the KUOM tower, varied seasonallyacross the 2 years of study (Figure 4). On average, daytime(RS > 10 W m−2) sums of Etotal were near zero in winter inboth 2007 and 2008 and showed increased rates from Aprilto October. In April and May, we observed higher daytimesums of Etotal on average in 2007, but in July and Augustwe observed higher daytime Etotal on average in 2008.Daily Etotal peaked in June of 2007 at an average rate of2.7 mm d−1 and in July of 2008 at an average rate of3.1 mm d−1. The prevailing northwest wind direction ledEtotal to be relatively more representative of residentialthan recreational land use areas (Figure 1). We considerthis systematic bias when interpreting these results.[27] In midsummer (June–August), the eddy covariance

system at the 40 m level of the tall tower had an energybudget closure of 83%. This imbalance did not vary withwind direction or between residential and recreational landuse types. To estimate the energy budget closure at the talltower we used G measured at the turfgrass site, which didnot contain any impervious surfaces. Therefore, it is likelythat G was an underestimate for the suburban landscape andthis would have contributed to the energy imbalance. Theaverage midday (1100–1500 h) evaporative fraction (LE/RN)

and Bowen ratio (H/LE) were 0.40 and 0.67, respectively, inmidsummer. We observed a higher midsummer energybudget closure of 94% with the 1.35 m portable eddycovariance system at the turfgrass site, and with a highermidday evaporative fraction (0.45) and lower Bowen ratio(0.58).

3.4. Component Evapotranspiration Fluxes

[28] The magnitude and seasonal patterns of evapotrans-piration varied considerably among the plant functional typesrepresented in this study. Across the 2008 growing season(April–November), average daily EGnirr was higher thaneither ETeg or ETdec on a per cover area basis (Figure 5a).EGnirr and ETeg were similar at the beginning and end ofthe growing season and both remained > 1 mm d−1 onaverage for at least 7 months in 2008, compared to only4 months for ETdec. All three plant functional typesshowed maximum daily evapotranspiration rates in June,when day length was at its maximum. Midsummer dailyevapotranspiration in 2008 also varied among the four

Figure 4. Daytime (RS > 10 W m−2) sums of ecosystemevapotranspiration (Etotal) per month measured at the 40 mlevel of the KUOM tower over a suburban area of Minnea-polis–Saint Paul, Minnesota. Dark lines represent mediandaytime sums of Etotal, the box represents the 75th and25th percentiles, and the dotted lines represent values withina 1.5 interquartile range. n varies from 4 to 20 days permonth. The gap in the fall of 2008 indicates a period whenthe instrumentation was out of service following a lighten-ing strike.

Figure 5. (a) Average daily sums of measured evapotrans-piration per cover area for nonirrigated turfgrass (squares,EGnirr), evergreen needleleaf trees (triangles, ETeg), anddeciduous broadleaf trees (circles, ETdec) per month acrossthe 2008 growing season in a suburban area of Minneapo-lis–Saint Paul, Minnesota. EGnirr was measured using aportable eddy covariance tower, and ETeg and ETdec werecalculated from heat dissipation sap flow measurementsand a canopy interception model. (b) Average summertime(June–August) daily sums of evapotranspiration per coverarea in 2008 for measured and modeled vegetation compo-nents. Evapotranspiration from irrigated turfgrass (EGirr)was modeled using the Priestley‐Taylor equation, andevapotranspiration from understory turfgrass (EGunder_eg

and EGunder_dec) was modeled as a linear function of thesolar radiation passing through the tree canopy, as esti-mated by the Beer‐Lambert law. Error bars are ± 1 stan-dard error, and n varies from 8 to 31 days per month.


8 of 16

165

turfgrass components represented in this study (Figure 5b)with average daily EGirr (3.10 mm d−1) higher than EGnirr

(2.99 mm d−1), EGunder_dec (0.07 mm d−1), and EGunder_eg

(0.01 mm d−1). The differences in evapotranspiration ratesamong plant functional types measured and modeled in2007 were consistent with these results (data not shown).[29] IT represented 27% and 32% of annual ETdec and

ETeg, respectively. On summer days with rain events, dailyETdec and ETeg were 44% and 22% higher, respectively, thanon days without rain events. Although precipitation eventswere originally filtered out of the EGnirr measurements, gap‐filled EGnirr showed a 22% decline on summer days withrain events. We were unable to separate evaporation andtranspiration components of turfgrass because the eddycovariance measurements used at the turfgrass site representonly total evapotranspiration.

3.5. Comparing Component‐Based and EddyCovariance Measured Etotal

[30] Compared to component‐based estimates of Etotal inwhich all open‐canopy turfgrass was assumed to be nonir-rigated (EGnirr), component‐based estimates that included a

modeled irrigated turfgrass (EGirr) vegetation type showedbetter agreement with measured Etotal, particularly in sum-mer when winds were from the southwest (R2 = 0.66 and0.73, respectively, Figure 6). Including EGirr reduced thesummer imbalance from an underestimate of 7% to 3% inthe southwest quadrant, but had little effect on the imbalancein spring and fall or when winds were from the northwest ornortheast. For the remainder of our analyses, all component‐based estimates of Etotal consequently included modeledEGirr.[31] Overall, component‐based estimates of Etotal under-

estimated measured Etotal by an average of 3%, with thecomponent‐based approach explaining 83% of the variationin measured fluxes when a linear regression model was fitthrough the origin (y = 1.03x, P < 0.001). Thirty minuteestimates of component‐based Etotal had a mean relativeerror of 21%. The imbalance between the two methods,however, varied seasonally with the component‐basedapproach resulting in a 20% overestimate of measuredfluxes in spring, an 11% underestimate in summer, and a 9%overestimate in fall (Figure 7). The largest overestimates ofmeasured Etotal tended to occur when winds were from the

Figure 6. Comparison of half‐hourly measured and scaled component sums of ecosystem evapo-transpiration (Etotal) in a suburban area of Minneapolis–Saint Paul, Minnesota. Component sums wereconstructed either (a) without a separate irrigated turfgrass component (EGirr) or (b) with EGirr. Etotal

was measured at the 40 m level of the KUOM tower. Data shown are from summer (June–August) in2007 and 2008 when winds were from the southwest. Lines represent the 1:1 line.

Figure 7. Comparison of half‐hourly measured and scaled component sums of ecosystem evapotrans-piration (Etotal) in a suburban area of Minneapolis–Saint Paul, Minnesota during (a) spring (April andMay), (b) summer (June–August), and (c) fall (September–November). Etotal was measured at the40 m level of the KUOM tower. Component sums were constructed as ET + EG + EW. Data shownare from 2007 and 2008. Lines represent the 1:1 line.


9 of 16


166

southwest, while underestimates tended to occur when windswere from the northwest or northeast and during periods ofhighRN (>800Wm−2), highH fluxes (>175Wm−2), and highD (>2 kPa). Scaling up TT differences according to xylemanatomy types (conifer, ring‐porous, and diffuse‐porous)instead of plant functional types reduced the overallimbalance such that measured Etotal was overestimated by1% and varied seasonally from a 25% overestimate inspring to a 6% underestimate in summer. The imbalancewas also relatively unaffected by the method used tosample the land cover fractions associated with each half‐hourly flux. Using only the maximum contribution point orthe nonweighted mean of the 10 footprint contributionpoints changed the overall imbalance by −0.007% and+0.003%, respectively.

3.6. Comparing Etotal From Two Suburban LandUse Types

[32] Measured Etotal varied in magnitude and seasonalitybetween the two suburban land use types represented in ourstudy (Figure 8). Recreational areas had higher averagedaytime Etotal than residential areas in the spring and fall, aswell as during the dry period in June 2007. In addition,recreational areas had a lower average midday Bowen ratioin midsummer than residential areas (0.47 and 0.88,respectively). Daytime Etotal was relatively similar betweenthe two land use types in late summer (August and Sep-tember) and in winter.[33] Annual sums of Etotal, which were based on the sum

of scaled component fluxes, were higher on average fromthe recreational area (467 mm yr−1) compared to the resi-dential area (324 mm yr−1) (Figure 9a). Despite large dif-ferences in total annual rainfall between 2007 and 2008, theinterannual variation in annual Etotal was relatively small forboth residential and recreational land uses. In 2007 and2008, annual Etotal from recreational areas represented 62%

and 85% of annual precipitation, respectively, while annualEtotal from residential areas represented 42% and 61% ofannual precipitation, respectively. EG was the largest com-ponent of annual Etotal in both recreational and residentialareas (87% and 64%, respectively), followed by ET (10%and 31%, respectively), and EW (3% and 5%, respectively).EGirr was a much larger component of annual EG in recre-ational than residential areas (83% versus 23%, respec-tively), while ETdec was the dominant component of annualET in both land use types (95% and 83%, respectively)

Figure 8. Average daytime ecosystem evapotranspiration(Etotal) per month measured over two land use types, recre-ational (open circles) and residential (solid circles), in a sub-urban area of Minneapolis–Saint Paul, Minnesota. Etotal wasmeasured at the 40 m level of the KUOM tower. Datashown were restricted to daytime (RS > 10 W m−2) periodswithout precipitation. Error bars are ± 1 standard error, andn varies from 17 to 270 half‐hourly data points per month.

Figure 9. (a) Annual precipitation and annual evapotrans-piration of scaled component fluxes, trees (ET), turfgrass(EG), and open water (EW), from the residential (res.) andrecreational (rec.) land use areas (Figure 1). ET includesevapotranspiration from evergreen needleleaf (ETeg) anddeciduous broadleaf trees (ETdec). EG includes evapotrans-piration from irrigated (EGirr), nonirrigated (EGnirr), andunderstory turfgrass (EGunder_eg and EGunder_dec). The averageproportional contribution of vegetation components, ETdec

(circles), ETeg (triangles), EGirr (open squares), and EGnirr

(solid squares), to ecosystem evapotranspiration (Etotal) from(b) residential and (c) recreational land use areas per month.Data shown include only daytime periods (RS > 10Wm−2) in2008, and n varies from 254 to 892 half‐hourly data pointsper month.


10 of 16

167

because deciduous trees were more abundant than evergreentrees in the suburban area we studied.[34] The contribution of different vegetation types to Etotal

in 2008, as determined from the sum of scaled componentfluxes, varied seasonally in both land use types (Figures 9band 9c). EGirr had the highest proportional contribution toEtotal in July, while the maximum contribution from EGnirr

occurred in spring (April and May) and fall (October andNovember). The largest contribution from ETdec occurred inAugust and September, while the highest contribution fromETeg, albeit small, was in April and November, particularlyin residential areas. Data for vegetation contributions in2007 were consistent with these results (data not shown).

4. Discussion

4.1. Suburban Evapotranspiration Rates

[35] Ecosystem evapotranspiration (Etotal) observed in thisstudy fell within the range of values represented for othersuburban residential areas around the world [Balogun et al.,2009; Grimmond and Oke, 1999, 2002; Moriwaki andKanda, 2004; Offerle et al., 2006; Spronken‐Smith, 2002;Vesala et al., 2008b]. The evaporative fraction (0.40) weobserved was most similar to that measured in older sub-urban areas in North America, such as Chicago [Grimmondet al., 1994]. Among suburban areas previously studied inNorth America, the fraction of available energy used forevaporation ranges from 0.22 to 0.46, with the highestvalues associated with recently developed exurban areasdominated by open turfgrass lawns and trees with smallcanopies and the lowest values associated with oldersuburbs dominated by mature tree canopies [Balogun et al.,2009]. With 29% tree cover and 53% turfgrass cover, thefootprint climatology area of our study had a higher vegeta-tion cover than is typical of North American cities[Grimmond and Oke, 2002; Nowak et al., 1996]. Although ithas been suggested that advected energy from impervioussurfaces can partially compensate for less vegetation coveroverall in urban compared to rural ecosystems [Oke, 1979],the Etotal measured in this study remained less than valuesmeasured in nonurban ecosystems of North America. Forexample, growing season Etotal rates > 3 mm d−1 are com-monly observed at hardwood forests in northern Wisconsin[Cook et al., 2004], Indiana [Schmid et al., 2000], Michigan[Schmid et al., 2003], North Carolina [Oishi et al., 2008], andTennessee [Wilson et al., 2001]. Additionally, the annualEtotal of 467 and 324 mm yr−1 found at our suburban site waslower than values reported for rainfed maize and soybeancrops in Illinois (547 and 660 mm yr−1, respectively) [Falgeet al., 2001], tallgrass prairie in Oklahoma (637 mm yr−1)[Burba and Verma, 2005], mixed deciduous forest inTennessee (571 mm yr−1) [Wilson et al., 2001], andmixed hardwood forest in North Carolina (604 mm yr−1)[Oishi et al., 2008].[36] The seasonality of suburban Etotal observed in this

study was more similar to patterns found in temperatehardwood forests from southern latitudes than northernlatitudes of North America. For example, in a mixeddeciduous forest in Oak Ridge, Tennessee, Wilson et al.[2001] observed Etotal rates that were significantly abovezero from April to October, similar to patterns found in ourstudy. Additionally, in a deciduous maple‐beech to oak‐

hickory transition forest in south central Indiana, Schmidet al. [2000] found increased Etotal from May to Octo-ber. In contrast, elevated Etotal occurred only from May toSeptember in an upland hardwood forest in northernWisconsin [Cook et al., 2004], and from late May toSeptember in a hardwood‐boreal transition forest in thelower peninsula of Michigan [Schmid et al., 2003]. Thesecomparisons suggest that the urban heat island effect, inaddition to the high percent cover of turfgrasses that areactive in the spring and fall, may have played an importantrole extending the growing season at our suburban studysite, relative to surrounding rural areas. This is consistentwith analyses that find increases in annual net primaryproductivity in urban areas, particularly in cold regions,which have been attributed to extended growing seasonscaused by the urban heat island effect [Imhoff et al., 2004].

4.2. Scaling Component‐Based Fluxes

[37] Our comparison of measured and component‐basedestimates of Etotal showed that component‐based approachescan capture much of the seasonal and spatial variability insuburban Etotal with a similar accuracy (79%) to that of eddycovariance measurements (∼80%) [Baldocchi et al., 1988].Similar to studies in forest ecosystems, we found thatcomponent‐based estimates of Etotal slightly underestimatedmeasured values, particularly at high flux rates in summer[Bovard et al., 2005; Hogg et al., 1997; Oishi et al., 2008;Wilson et al., 2001]. Unlike these forest studies, however,we found that component‐based estimates tended to over-estimate measured fluxes in spring and fall. Severalpotential sources of error may explain these discrepanciesbetween methods, including (1) errors associated with themeasurements or models used to estimate componentfluxes, (2) a mismatch in the spatial footprints associatedwith the measured and scaled fluxes, (3) undersamplingspecies’ differences in water use, and (4) the spatial het-erogeneity of urban microclimates.[38] The measurements and models used to estimate

components of Etotal in this study all contribute some degreeof uncertainty when scaling up Etotal. Previous studies fromforest ecosystems have found that heat dissipation sap fluxtechniques may underestimate transpiration, particularlyduring periods of high radiation [Bovard et al., 2005; Hogget al., 1997; Oishi et al., 2008; Wilson et al., 2001], andcould explain the component‐based underestimates weobserved during summer. The lack of energy balance clo-sure observed at the tall tower and turfgrass sites, althoughtypical of eddy covariance systems in general [Wilson et al.,2002], suggests a systematic underestimate of both Etotal andEG. While an underestimate of EG may help explain thesummer imbalances, it is contrary to explaining the over-estimates observed in spring and fall. Depending on theseasonal variation in LAI among trees in our study area,the modeled seasonal LAI patterns also represent a sourceof error in our estimates of IT, EGunder_eg, and EGunder_dec.As comparisons between component‐based and measuredestimates of Etotal were restricted to periods with highquality data and exclude most rainfall events, it is unlikelythat the IT component significantly contributed to theobserved imbalances. The seasonal pattern of understoryturfgrass evapotranspiration, however, is consistent withthe observed imbalances. EGunder_eg and EGunder_dec were


11 of 16


168

highest in spring and fall when tree LAI was low, andEGunder_eg and EGunder_dec were lowest in midsummer whentree LAI was high (data not shown). It is also likely thatusing a constant coefficient equal to 0.87 in the Priestley‐Taylor equation to model EGirr led to an overestimate ofEtotal in spring and fall and an underestimate in summer.Studies show that a varies seasonally in nonstressed, coolseason grasses, with lowest values in spring and fall andhighest values in summer [Brown et al., 2001; Ervin andKoski, 1998; Zhang et al., 2007].[39] Even in urban areas with relatively flat topography,

flux footprints are particularly difficult to assess due to thespatial heterogeneity of surface types and complex transportof scalars, and can lead to the mischaracterization of foot-print areas when scaling up component fluxes [Vesala et al.,2008a, 2008b]. While we did not attempt to replicate the2‐D source area associated with each measured half‐hourly value of Etotal, we did find that component‐basedestimates of Etotal were not greatly affected by differentmethods of sampling the land cover classification map.Consequently, we believe the misrepresentation of thefootprint area to be a relatively minor source of error inour component‐based estimates of Etotal.[40] Species‐rich ecosystems, such as ours, additionally

complicate efforts to scale up measurements made on only afew individuals, species, or sites, as water use can be highlyvariable among tree [Kumagai et al., 2005; Oren et al.,1998; Pataki and Oren, 2003] and turfgrass species andamong management regimes [Zhang et al., 2007]. Althoughthe effect was relatively small, including TT differencesamong xylem anatomy types resulted in a 4% reduction inthe overall imbalance compared to using the coarser plantfunctional type categories. Evapotranspiration from unmea-sured understory plants has additionally been suggested tocontribute to observed imbalances in forest ecosystems[Bovard et al., 2005; Oishi et al., 2008]. It is unlikely thatevapotranspiration from common suburban understory veg-etation types, such as woody shrubs, vegetable or flowergardens, contributed significantly to the Etotal in our studyarea because their cover represented less than 2% of thelandscape. Although the seven tree genera and one turfgrasslawn we studied provides a limited sampling of species’differences in water use, the relatively good match that wefound between component‐based and measured Etotal sug-gests that this group of plant functional types captured muchof the species’ variation in water use within the suburbanlandscape we studied.[41] Management practices, including irrigation, fertiliza-

tion, mowing height, and canopy shading can also signifi-cantly impact rates of water loss from turfgrass [Feldhake etal., 1983; Zhang et al., 2007]. The large differences in dailyevapotranspiration between our measured nonirrigatedturfgrass and our modeled values of irrigated and understoryturfgrass suggest that irrigation management effects werecaptured in seasonality, but fertilizer and mowing practiceswere not accounted for. Occasional irrigation by home-owners, particularly during periods of low soil moisture inmidsummer, was also not accounted for and may con-tribute to the observed underestimate of measured fluxes insummer.[42] Finally, microclimate variability in urban areas

[Bonan, 2000; Byrne et al., 2008; Peters and McFadden,

2010] may represent a relatively more important source oferror in urban component‐based estimates of Etotal than innatural ecosystems. The spatial heterogeneity of urban sur-face types with different energy balances and heat storagecapacities results in the advection, or lateral transport, ofheat from paved surfaces with high sensible heat fluxes tovegetated surfaces with high latent heat fluxes [Oke, 1979;Spronken‐Smith et al., 2000]. Advection can be an impor-tant energy source to irrigated urban parks and lawns,causing elevation of EG rates by up to 35% [Feldhake et al.,1983; Oke, 1979; Spronken‐Smith et al., 2000]. Trees grownover asphalt can transpire 30% more water than trees grownover turfgrass [Kjelgren and Montague, 1998], while sometree species initiate stomatal closure and have reduced ratesof water loss [Kjelgren and Montague, 1998; Montague andKjelgren, 2004]. In addition, sparsely planted trees can haveTT rates two to three times higher than those of denselyplanted trees [Hagishima et al., 2007] and windbreaks canreduce EG rates by 25% [Danielson and Feldhake, 1981].[43] Our study trees had relatively open growth forms

compared to forest‐grown trees, yet they were all grownunder more park‐like conditions as compared to trees inmore densely built‐up urban areas. Consequently, our TTmeasurements likely underestimate ET rates of trees growingin complete isolation or near paved surfaces. This idea wasfurther supported by the large underestimates of Etotal weobserved during periods of high net radiation, high sensibleheat fluxes, and high vapor pressure deficit from northwestand northeast wind directions. In contrast, the turfgrass sitewas a relatively large lawn that was less likely to have beenexposed to advected sources of energy compared to smallerresidential lawns surrounded by sidewalks, driveways, andstreets. It is difficult to determine whether our EG mea-surements provide a systematic overestimate or underesti-mate of water loss from residential yards. As microclimateeffects are an important source of uncertainty in predictingcomponent‐based Etotal in urban and suburban ecosystems,future studies should focus on how best to account for thisvariability when scaling.

4.3. Contributions of Trees and Turfgrassesto Suburban Etotal

[44] The relative contributions of trees and turfgrass toannual Etotal was driven largely by fractional cover and plantfunctional type differences in daily water use. Studies fromforest ecosystems similarly show that the relative contribu-tion of different tree species to Etotal is driven both byspecies’ abundances on the landscape, often measured ascontributions to total basal area or leaf area, and by species‐specific differences in water use [Tang et al., 2006;Wullschleger et al., 2001]. Despite their higher rates ofwater use than deciduous broadleaf trees, evergreen nee-dleleaf trees had a nearly negligible contribution to Etotal inour study area due their relatively small fractional cover ofthe landscape. Turfgrass, however, due its high cover andhigh daily EG rates across the growing season, representedthe largest proportional contribution (87% and 64% in rec-reational and residential areas, respectively) to annual Etotal

in our study area. Pasture grasses in natural savannah eco-systems, by contrast, represent a much smaller component(20–44%) of Etotal than trees [Baldocchi and Xu, 2007; Pacoet al., 2009]. While these studies are from savanna eco-


12 of 16

169

systems with similar percent tree cover to our study area, itis important to note they are from areas with a Mediterra-nean climate and very different species compositions.[45] The proportional contribution of different vegetation

types to suburban Etotal varied seasonally due to differentpatterns in physiological activity among plant functionaltypes. While evergreen needleleaf and deciduous broadleaftrees both have maximum physiological function in themiddle of summer in temperate ecosystems, the evergreenleaf habit and high cold tolerance allows evergreen needle-leaf trees to remain physiologically active over a longergrowing season than either deciduous broadleaf trees or coolseason turfgrasses [Catovsky et al., 2002; Givnish, 2002;Havranek and Tranquillini, 1995]. Correspondingly, weobserved the highest proportional contribution to Etotal fromdeciduous broadleaf trees in late summer and, albeit small,the highest contributions from evergreen needleleaf trees inApril and November when deciduous trees were leafless.Our observation of a midsummer low in the contribution toEtotal from nonirrigated turfgrass is consistent with themidsummer dormancy of cool season turfgrasses [Feldhakeet al., 1984; Fry and Huang, 2004; Zhang et al., 2007].

4.4. Land Use Comparisons

[46] Similar to other urban studies, we found that themagnitude of Etotal varied between land use types accordingto differences in the total cover and composition of vege-tation [Frey et al., 2010; Grimmond and Oke, 1999; Offerleet al., 2006; Spronken‐Smith, 2002; Vesala et al., 2008b].Seasonal patterns of Etotal were also highly influenced by thephenology of the dominant vegetation type. For example,the turfgrass‐dominated recreational area not only showedhigher annual Etotal than the deciduous tree‐dominatedresidential area, but it specifically showed higher rates ofEtotal in spring and fall when cool season turfgrasses aremost active. In addition, our study extends these previousfindings to show that turfgrass management, particularlythrough irrigation inputs, is another important factor influ-encing the spatial and seasonal variability of suburban Etotal.The relatively high evaporative fluxes we measured fromthe recreational area during the dry conditions in June of2007 suggest that increased irrigation during periods of lowsoil moisture can significantly alter Etotal from urban areas.While homeowners also maintain irrigated turfgrass lawns,management intensity can vary widely among individuallandowners [Larson et al., 2009]. To predict suburban Etotal,these results suggest it will be important to know not justthe total vegetation cover in urban landscapes, but also thefractional cover of different vegetation types and turfgrassmanagement practices. In addition, it is likely that changesin land use, vegetation composition, or turfgrass manage-ment practices, particularly related to irrigation use, willlead to changes in total water fluxes from urban and sub-urban ecosystems.[47] Given the growing interest among cities in using

“green infrastructure” to manage problems such as stormwater runoff and energy demand, it is important for urbanplanners and designers to consider the trade‐offs amongdifferent potential landscape configurations [Mitchell et al.,2008]. For example, in arid regions the water conservationbenefits of low water use species must be weighed againstthe cooling benefits of high water use species [Larson et al.,

2009; Shashua‐Bar et al., 2009]. In contrast, cities in mesicregions with frequent rainfall events may value the stormwater interception benefits of vegetation with large canopiesrelatively more than other ecosystem services related toEtotal [Wang et al., 2008]. Although we found that coolseason turfgrasses had higher rates of water loss thanevergreen needleleaf and deciduous broadleaf trees perunit cover area, the overall cooling effect of turfgrass wasnot necessarily higher than that of trees. This is becausetrees also provide cooling by intercepting solar radiationand shading the ground beneath canopies [Peters andMcFadden, 2010]. In the arid communal settlement ofMidreshet Ben Gurion, Israel, for example, Shashua‐Baret al. [2009] found that turfgrass lawns shaded by treecanopies resulted in both a greater reduction in air tem-perature and a 50% reduction in water use compared tounshaded turfgrass lawns.[48] By approximating runoff as the difference between

measured precipitation and annual component‐based Etotal,we estimate runoff to represent 26% and 48% of annualprecipitation on average from the recreational and residen-tial areas in our study, respectively. While these estimatesneglect irrigation inputs, changes in storage, and ground-water recharge and, thus, should be used with caution, theyfall within the range of runoff fractions (2–59%) previouslyreported for watersheds in the Minneapolis–Saint Paulmetropolitan area [Brezonik and Stadelmann, 2002]. Theseresults are in agreement with studies showing that increasedpervious surface area in urban regions reduces direct runoff[Brezonik and Stadelmann, 2002; Haase, 2009; Mitchellet al., 2008; Wang et al., 2008]. Although it is beyondthe scope of our study to offer specific planting recom-mendations for use in green infrastructure, this studyadvances our ability to quantify the relative contributionsof different vegetation types to suburban Etotal, an importantstep in evaluating decisions related to the seasonal manage-ment of urban water and energy budgets.[49] Despite the good agreement we found between

component‐based and measured Etotal in the suburban eco-system we studied, differences in species compositionamong cities make it complex to extrapolate our componentmeasurements of evapotranspiration to other urban andsuburban areas, particularly in more southern latitudes. Forexample, the turfgrass site in this study represented onlycool season turfgrass species, as warm season species arenot typically planted in the climate zone associated withMinnesota. Yet cool and warm season turfgrasses have largedifferences in evapotranspiration, with cool season speciesusing up to 45% more water during the growing seasoncompared to warm season species [Feldhake et al., 1983;Zhang et al., 2007]. Cities located in warmer climates withhigher frequencies of warm season turfgrasses will need toconsider these species’ differences in water use when scal-ing up and predicting urban Etotal.

5. Conclusions

[50] We quantified seasonal variations in evapotranspira-tion from a suburban landscape in the Upper Midwestregion of the United States, with peak rates > 3 mm d−1 insummer. The good agreement between our component‐based and measured Etotal, with an overall imbalance of 3%,


13 of 16


170

suggests that the major plant functional types captured mostof the spatial and temporal variability required to quantifyurban Etotal. Component‐based estimates overestimatedmeasured Etotal in spring (20%) and fall (10%) and under-estimated measured Etotal in summer (11%), likely due toerrors associated with the irrigated turfgrass model andvariability from microclimate effects. We found turfgrasswas the major contributor to annual Etotal from both recre-ational and residential land uses types (87% and 64%,respectively) due to a high fractional cover (74% and 34%,respectively) and higher average rates of water use per coverarea in midsummer than deciduous broadleaf or evergreenneedleleaf trees (3.0, 1.4, and 2.3 mm d−1, respectively). Themaximum contribution to Etotal from nonirrigated turfgrassoccurred in spring and fall, irrigated turfgrass in midsum-mer, deciduous broadleaf trees in late summer, and ever-green needleleaf trees in early spring and late fall.Turfgrass‐dominated recreational land use areas had higheraverage annual Etotal compared to deciduous tree‐dominatedresidential areas (467 and 324 mm yr−1, respectively), aswell as an altered seasonal pattern of Etotal with higherfluxes in spring and fall, and during a midsummer drought.These results suggest that a plant functional type approachcould be used to estimate and compare Etotal among differentcities and to evaluate the effects of changes in land use,vegetation composition, or management practices on urbanwater and energy budgets.

Notation

D vapor pressure deficit, kPa.RN net radiation, W m−2.RS incoming solar radiation, W m−2.RV below canopy solar radiation, W m−2.G ground heat flux, W m−2.LE latent heat flux, W m−2.H sensible heat flux, W m−2.

LAI leaf area index, m2 m−2.Etotal ecosystem evapotranspiration, mm time−1.ET evapotranspiration from trees, mm time−1.EG evapotranspiration from turfgrass, mm time−1.EW evaporation from water, mm time−1.EP potential evapotranspiration, mm time−1.TT tree transpiration, mm time−1.IT interception loss from tree canopies, mm time−1.

ETeg evapotranspiration from evergreen needleleaftrees, mm time−1.

ETdec evapotranspiration from deciduous broadleaftrees, mm time−1.

EGnirr evapotranspiration from nonirrigated turfgrass,mm time−1.

EGirr evapotranspiration from irrigated turfgrass,mm time−1.

EGunder_eg evapotranspiration from turfgrass under ever-green needleleaf trees, mm time−1.

EGunder_dec evapotranspiration from turfgrass under decidu-ous broadleaf trees, mm time−1.

[51] Acknowledgments. We thank ∼350 homeowners in Saint Paul,Roseville, Falcon Heights, and Lauderdale, the City of Lauderdale, and theUniversity of Minnesota for granting us permission to conduct this researchon their property; Brian Horgan for advice and access to the turfgrass site at

the University of Minnesota Turfgrass Research, Outreach, and EducationCenter; Larry Oberg for technical advice and access to the KUOM broad-cast tower; Dave Ruschy for data from the University of Minnesota climatestation; Natascha Kljun for sharing her footprint model code; and MatthiasZeeman for sharing his footprint mapping script. We are grateful for advicefrom Tim Griffis, Sarah Hobbie, Rebecca Montgomery, and Tracy Twine,and for assistance with field measurements, software development, andimage processing from Ahmed Balogun, Ben Freeman, Vikram Gowree-sunker, Cheyne Hadley, Vicki Kalkirtz, Rebecca Koetter, Mark McPhail,Julia Rauchfuss, Scott Shatto, and Yana Sorokin. This work was supportedby an Interdisciplinary Doctoral Fellowship to E.B.P. from the Universityof Minnesota Institute on the Environment and a grant to J.P.M. fromNASA Earth Science Division (NNG04GN80G).

ReferencesBaldocchi, D. D., and L. K. Xu (2007), What limits evaporation fromMediterranean oak woodlands ‐ The supply of moisture in the soil,physiological control by plants or the demand by the atmosphere?,Adv. Water Resour., 30(10), 2113–2122.

Baldocchi, D. D., B. B. Hicks, and T. P. Meyers (1988), Measuringbiosphere‐atmosphere exchanges of biologically related trace gaseswith micrometeorological methods, Ecology, 69, 1331–1340.

Baldocchi, D. D., B. E. Law, and P. M. Anthoni (2000), On measuring andmodeling energy fluxes above the forest floor of a homogeneous andheterogeneous conifer forest, Agric. For. Meteorol., 102, 187–206.

Balogun, A. A., J. O. Adegoke, S. Vezhapparambu, M. Mauder, J. P.McFadden, and K. Gallo (2009), Surface energy balance measure-ments above an exurban residential neighbourhood of Kansas City,Missouri, Boundary Layer Meteorol., 133(3), 299–321, doi:10.1007/s10546-009-9421-3.

Barcza, Z., A. Kern, L. Haszpra, and N. Kljun (2009), Spatial representa-tiveness of tall tower eddy covariance measurements using remote sens-ing and footprint analysis, Agric. For. Meteorol., 149(5), 795–807,doi:10.1016/j.agrformet.2008.10.021.

Bonan, G. B. (2000), The microclimates of a suburban Colorado (USA)landscape and implications for planning and design, Landsc. UrbanPlan., 49(3–4), 97–114.

Bovard, B. D., P. S. Curtis, C. S. Vogel, H. B. Su, and H. P. Schmid(2005), Environmental controls on sap flow in a northern hardwoodforest, Tree Physiol., 25(1), 31–38.

Brezonik, P. L., and T. H. Stadelmann (2002), Analysis and predictivemodels of stormwater runoff volumes, loads, and pollutant concentra-tions from watersheds in the Twin Cities metropolitan area, Minnesota,USA, Water Res., 36(7), 1743–1757.

Brown, P. W., C. F. Mancino, M. H. Young, T. L. Thompson, P. J.Wierenga, and D. M. Kopec (2001), Penman Monteith crop coeffi-cients for use with desert turf systems, Crop Sci., 41(4), 1197–1206.

Burba, G. G., and S. B. Verma (2005), Seasonal and interannual variabilityin evapotranspiration of native tallgrass prairie and cultivated wheat eco-systems, Agric. For. Meteorol., 135, 190–201.

Burba, G. G., D. K. McDermitt, A. Grelle, D. L. Anderson, and L. Xu(2008), Addressing the influence of instrument surface heat exchangeon the measurements of CO2 flux from open‐path gas analyzers, GlobalChange Biol., 14(8), 1854–1876.

Bush, S. E., D. E. Pataki, K. R. Hultine, A. G. West, J. S. Sperry, and J. R.Ehleringer (2008), Wood anatomy constrains stomatal responses to atmo-spheric vapor pressure deficit in irrigated, urban trees, Oecologia, 156(1),13–20, doi:10.1007/s00442-008-0966-5.

Byrne, L. B., M. A. Bruns, and K. C. Kim (2008), Ecosystem properties ofurban land covers at the aboveground‐belowground interface, Ecosys-tems, 11(7), 1065–1077, doi:10.1007/s10021-008-9179-3.

Campbell, G. S., and J. M. Norman (1998), An Introduction to Environ-mental Biophysics, 2nd ed., 286 pp., Springer, New York.

Carrow, R. N. (1995), Drought resistance aspects of turfgrasses in thesoutheast ‐ Evapotranspiration and crop coefficiencts, Crop Sci., 35(6),1685–1690.

Catovsky, S., N. M. Holbrook, and F. A. Bazzaz (2002), Coupling whole‐tree transpiration and canopy photosynthesis in coniferous and broad‐leaved tree species, Can. J. For. Res., 32(2), 295–309.

Cook, B. D., et al. (2004), Carbon exchange and venting anomalies inan upland deciduous forest in northern Wisconsin, USA, Agric. For.Meteorol., 126(3–4), 271–295, doi:10.1016/j.agrformet.2004.06.008.

Craul, P. J. (1985), A description of urban soils and their desired character-istics, J. Arboriculture, 11(11), 330–339.

Danielson, R. E., and C. M. Feldhake (1981), Urban lawn irrigation andmanagement practices for water saving with minimum effect on lawnquality, Completion Rep. 106, Off. of Water Res. and Technol., Dep.of the Inter., Washington, D. C.


14 of 16

171

Eastham, J., and C. W. Rose (1988), Pasture evapotranspiration under vary-ing tree planting density in an agroforestry experiment, Agric. WaterManage., 15(1), 87–105.

Ervin, E. H., and A. J. Koski (1998), Drought avoidance aspects and cropcoefficients of Kentucky bluegrass and tall fescue turfs in the semiaridwest, Crop Sci., 38(3), 788–795.

Eugster, W., and W. Senn (1995), A cospectral correction model formeasurement of turbulent NO2 flux, Boundary Layer Meteorol., 74,321–340.

Falge, E., et al. (2001), Gap filling strategies for defensible annual sums ofnet ecosystem exchange, Agric. For. Meteorol., 104(1), 43–69.

Feldhake, C. M., R. E. Danielson, and J. D. Butler (1983), Turfgrassevapotranspiration 1. Factors influencing rate in urban environments,Agron. J., 75(5), 824–830.

Feldhake, C. M., R. E. Danielson, and J. D. Butler (1984), Turfgrass evapo-transpiration 2. Responses to deficit irrigation, Agron. J., 76(1), 85–89.

Feldhake, C. M., J. D. Butler, and R. E. Danielson (1985), Turfgrassevapotranspiration ‐ responses to shade preconditioning, Irrig. Sci., 6(4),265–270.

Foken, T., and B. Wichura (1996), Tools for quality assessment of surface‐based flux measurements, Agric. For. Meteorol., 78, 83–105.

Frey, C. M., E. Parlow, R. Vogt, M. Harhash, and M. M. Abdel Wahab(2010), Flux measurements in Cairo. Part 1: In situ measurements andtheir applicability for comparison with satellite data, Int. J. Climatol.,doi:10.1002/joc.2140.

Fry, J., and B. Huang (2004), Applied Turfgrass Science and Physiology,John Wiley, Hoboken, N. J.

Givnish, T. J. (2002), Adaptive significance of evergreen vs. deciduousleaves: Solving the triple paradox, Silva Fennica, 36(3), 703–743.

Granier, A. (1987), Evaluation of transpiration in a Douglas‐fir stand bymeans of sap flow measurements, Tree Physiol., 3(4), 309–319.

Grimmond, C. S. B., and T. R. Oke (1999), Evapotranspiration rates inurban areas, in Impacts of Urban Growth on Surface Water andGroundwater Quality (Proceedings of IUGG 99 Symposium HS5),IAHS Publ., 259.

Grimmond, C. S. B., and T. R. Oke (2002), Turbulent heat fluxes in urbanareas: Observations and a local‐scale urban meteorological parameteriza-tion scheme (LUMPS), J. Appl. Meteorol., 41, 792–810.

Grimmond, C. S. B., C. Souch, R. Grant, and G. Heisler (1994), Local scaleenergy and water exchanges in a Chicago neighborhood, in Chicago’sUrban Forest Ecosystem: Results of the Chicago Urban Forest ClimateProject, edited by E. G. McPherson, D. J. Nowak, and R. A. Rowntree,Gen. Tech. Rep. NE‐186, pp. 41–51, Northeast. For. Exp. Stn., For.Serv., U.S. Dep. of Agric., Radnor, Pa.

Grimmond, C. S. B., C. Souch, and M. D. Hubble (1996), Influence of treecover on summertime surface energy balance fluxes, San Gabriel Valley,Los Angeles, Clim. Res., 6(1), 45–57.

Grimmond, C. S. B., J. A. Salmond, T. R. Oke, B. Offerle, and A. Lemonsu(2004), Flux and turbulence measurements at a densely built‐up site inMarseille: Heat, mass (water and carbon dioxide), and momentum,J. Geophys. Res., 109, D24101, doi:10.1029/2004JD004936.

Grimmond, C. S. B., B. Crawford, J. Hom, B. D. Offerle, M. Patterson, andD. Golub (2006), Carbon dioxide fluxes in a suburban area of a NorthAmerican City, Geophys. Res. Abstr., 8.

Haase, D. (2009), Effects of urbanisation on the water balance ‐A long‐termtrajectory, Environ. Impact Assess. Rev., 29(4), 211–219, doi:10.1016/j.eiar.2009.01.002.

Hagishima, A., K. I. Narita, and J. Tanimoto (2007), Field experimenton transpiration from isolated urban plants, Hydrol. Process., 21(9),1217–1222, doi:10.1002/hyp.6681.

Havranek, W. M., and W. Tranquillini (1995), Physiological processesduring winter dormancy and their ecological significance, in Ecophysiol-ogy of Coniferous Forests, edited by W. K. Smith and T. M. Hinckley,Academic, New York.

Hiller, R. V., J. P. McFadden, and N. K. Kljun (2010), Interpreting CO2fluxes over a suburban lawn: The influence of traffic emissions, Bound-ary Layer Meteorol., doi:10.1007/s10546-010-9558-0, in press.

Hogg, E. H., et al. (1997), A comparison of sap flow and eddyfluxes of water vapor from a boreal deciduous forest, J. Geophys. Res.,102(D24), 28,929–28,937.

Imhoff, M. L., L. Bounoua, R. DeFries, W. T. Lawrence, D. Stutzer, C. J.Tucker, and T. Ricketts (2004), The consequences of urban land transfor-mation on net primary productivity in the United States, Remote Sens.Environ., 89(4), 434–443.

Jackson, R. B., J. Canadell, J. R. Ehleringer, H. A. Mooney, O. E. Sala, andE. D. Schulze (1996), A global analysis of root distributions for terrestrialbiomes, Oecologia, 108(3), 389–411.

Kjelgren, R., and T. Montague (1998), Urban tree transpiration over turfand asphalt surfaces, Atmos. Environ., 32(1), 35–41.

Kljun, N., P. Calanca, M. W. Rotachhi, and H. P. Schmid (2004), Asimple parameterisation for flux footprint predictions, Boundary LayerMeteorol., 112(3), 503–523.

Kotani, A., and M. Sugita (2005), Seasonal variation of surface fluxesand scalar roughness of suburban land covers, Agric. For. Meteorol.,135(1–4), 1–21, doi:10.1016/j.agrformet.2005.09.012.

Kumagai, T., H. Nagasawa, T. Mabuchi, S. Ohsaki, K. Kubota, K. Kogi, Y.Utsumi, S. Koga, and K. Otsuki (2005), Sources of error in estimatingstand transpiration using allometric relationships between stem diameterand sapwood area for Cryptomeria japonica and Chamaecyparis obtusa,For. Ecol. Manage., 206(1–3), 191–195.

Larson, K. L., D. Casagrande, S. L. Harlan, and S. T. Yabiku (2009),Residents’ yard choices and rationales in a desert city: Social priori-ties, ecological impacts, and decision tradeoffs, Environ. Manage.,44(5), 921–937, doi:10.1007/s00267-009-9353-1.

Lu, P., L. Urban, and P. Zhao (2004), Granier’s thermal dissipation probe(TDP) method for measuring sap flow in trees: Theory and practice, ActaBot. Sin., 46(6), 631–646.

Ludwig, F., T. E. Dawson, H. H. T. Prins, F. Berendse, and H. de Kroon(2004), Below‐ground competition between trees and grasses may over-whelm the facilitative effects of hydraulic lift, Ecol. Lett., 7(8), 623–631,doi:10.1111/j.1461-0248.2004.00615.x.

Mauder, M., T. Foken, R. Clement, J. A. Elbers, W. Eugster, T. Grunwald,B. Heusinkveld, and O. Kolle (2008), Quality control of CarboEuropeflux data–Part 2: Inter‐comparison of eddy‐covariance software, Bio-geosciences, 5, 12.

Milesi, C., S. W. Running, C. D. Elvidge, J. B. Dietz, B. T. Tuttle, andR. R. Nemani (2005), Mapping and modeling the biogeochemicalcycling of turf grasses in the United States, Environ. Manage., 36(3),426–438, doi:10.1007/s00267-004-0316-2.

Mitchell, V. G., H. A. Cleugh, C. S. B. Grimmond, and J. Xu (2008),Linking urban water balance and energy balance models to analyse urbandesign options, Hydrol. Process., 22(16), 2891–2900, doi:10.1002/hyp.6868.

Montague, T., and R. Kjelgren (2004), Energy balance of six commonlandscape surfaces and the influence of surface properties on gasexchange of four containerized tree species, Sci. Hortic., 100(1–4),229–249, doi:10.1016/j.scienta.2003.08.010.

Moore, C. J. (1986), Frequency response corrections for eddy correlationsystems, Boundary Layer Meteorol., 37, 17–35.

Moriwaki, R., and M. Kanda (2004), Seasonal and diurnal fluxes of radia-tion, heat, water vapor, and carbon dioxide over a suburban area, J. Appl.Meteorol., 43, 1700–1710.

National Climatic Data Center (NCDC) (2004), Climatography of theUnited States, 1971–2000, Natl. Environ. Satell., Data, and Inf. Serv.,Natl. Oceanic and Atmos. Admin, U.S. Dep. of Commer., Asheville, N. C.

Nemitz, E., K. J. Hargreaves, A. G. McDonald, J. R. Dorsey, and D. Fowler(2002), Micrometeorological measurements of the urban heat budget andCO2 emissions on a city scale, Environ. Sci. Technol., 36(14), 3139–3146.

Nowak, D. J., R. A. Rowntree, E. G. McPherson, S. M. Sisinni, E. R.Kerkmann, and J. C. Stevens (1996), Measuring and analyzing urbantree cover, Landsc. Urban Plan., 36(1), 49–57.

Offerle, B., C. S. B. Grimmond, K. Fortuniak, and W. Pawlak (2006),Intraurban differences of surface energy fluxes in a central Europeancity, J. Appl. Meteorol. Climatol., 45(1), 125–136.

Oishi, A. C., R. Oren, and P. C. Stoy (2008), Estimating components of for-est evapotranspiration: A footprint approach for scaling sap flux measure-ments, Agric. For. Meteorol., 148(11), 1719–1732, doi:10.1016/j.agrformet.2008.06.013.

Oke, T. R. (1979), Advectively assisted evapotranspiration from irrigatedurban vegetation, Boundary Layer Meteorol., 17(2), 167–173.

Oren, R., N. Phillips, G. Katul, B. E. Ewers, and D. E. Pataki (1998), Scalingxylem sap flux and soil water balance and calculating variance: A methodfor partitioning water flux in forests, Ann. Sci. For., 55(1–2), 191–216.

Paco, T. A., T. S. David, M. O. Henriques, J. S. Pereira, F. Valente, J. Banza,F. L. Pereira, C. Pinto, and J. S. David (2009), Evapotranspiration from aMediterranean evergreen oak savannah: The role of trees and pasture,J. Hydrol., 369(1–2), 98–106, doi:10.1016/j.jhydrol.2009.02.011.

Pataki, D. E., and R. Oren (2003), Species differences in stomatal control ofwater loss at the canopy scale in a mature bottomland deciduous forest,Adv. Water Resour., 26(12), 1267–1278.

Pataki, D. E., H. R.McCarthy, E. Livtak, and S. Pincetl (2010), Transpirationof urban forests in the Los Angeles metropolitan area, Ecol. Appl., in press.

Peters, E. B., and J. P. McFadden (2010), Influence of seasonality and veg-etation type on suburban microclimates, Urban Ecosyst., doi:10.007/s11252-010-0128-5.

Peters, E. B., J. P. McFadden, and R. A. Montgomery (2010), Biologicaland environmental controls on tree transpiration in a suburban landscape,J. Geophys. Res., 115, G04006, doi:10.1029/2009JG001266.


15 of 16


172

Priestley, C. H. B., and R. J. Taylor (1972), On the assessment of surfaceheat flux and evaporation using large‐scale parameters, Mon. WeatherRev., 100(2), 81–92.

R Development Core Team (2010), R: A language and environment for sta-tistical computing, R Found. for Stat. Comput., Vienna, Austria.

Rebmann, C., et al. (2005), Quality analysis applied on eddy covariancemeasurements at complex forest sites using footprint modelling, Theor.Appl. Climatol., 80(2–4), 121–141, doi:10.1007/s00704-004-0095-y.

Reich, P. B., M. B. Walters, and D. S. Ellsworth (1997), From tropics totundra: Global convergence in plant functioning, Proc. Natl. Acad. Sci.U. S. A., 94, 13,730–13,734.

Rutter, A. J., A. J. Morton, and P. C. Robins (1975), Predictive model ofrainfall interception in forests, 2. Generalization of model and compari-son with observation in some coniferous and hardwood stands, J. Appl.Ecol., 12(1), 367–380.

Schmid, H. P. (1994), Source areas for scalars and scalar fluxes, BoundaryLayer Meteorol., 67, 293–318.

Schmid, H. P., H. A. Cleugh, C. S. B. Grimmond, and T. R. Oke (1991),Spatial variability of energy fluxes in suburban terrain, Boundary LayerMeteorol., 54, 249–276.

Schmid, H. P., C. S. B. Grimmond, F. Cropley, B. Offerle, and H. B. Su(2000), Measurements of CO2 and energy fluxes over a mixed hard-wood forest in the mid‐western United States, Agric. For. Meteorol.,103, 357–374.

Schmid, H. P., H. B. Su, C. S. Vogel, and P. S. Curtis (2003), Ecosystem‐atmosphere exchange of carbon dioxide over a mixed hardwood forest innorthern lower Michigan, J. Geophys. Res., 108(D14), 4417, doi:10.1029/2002JD003011.

Schotanus, P., F. T. M. Nieuwstadt, and H. A. R. Debruin (1983),Temperature‐measurement with a sonic anemometer and its applica-tion to heat and moisture fluxes, Boundary Layer Meteorol., 26(1), 81–93.

Sen Roy, S., and F. Yuan (2009), Trends in extreme temperatures in relationto urbanization in the Twin Cities metropolitan area, Minnesota, J. Appl.Meteorol. Climatol., 48(3), 669–679, doi:10.1175/2008JAMC1983.1.

Shashua‐Bar, L., D. Pearlmutter, and E. Erell (2009), The cooling effi-ciency of urban landscape strategies in a hot dry climate, Landsc. UrbanPlan., 92(3–4), 179–186, doi:10.1016/j.landurbplan.2009.04.005.

Spronken‐Smith, R. A. (2002), Comparison of summer‐ and winter‐timesuburban energy fluxes in Christchurch, New Zealand, Int. J. Climatol.,22, 979–992.

Spronken‐Smith, R. A., T. R. Oke, and W. P. Lowry (2000), Advection andthe surface energy balance across an irrigated urban park, Int. J. Climatol.,20(9), 1033–1047.

Tang, J. W., P. V. Bolstad, B. E. Ewers, A. R. Desai, K. J. Davis, and E. V.Carey (2006), Sap flux‐upscaled canopy transpiration, stomatal conduc-tance, and water use efficiency in an old growth forest in the Great Lakesregion of the United States, J. Geophys. Res., 111, G02009, doi:10.1029/2005JG000083.

Taylor, J. R. (1982), An Introduction to Error Analysis: The Study ofUncertainties in Physical Measurements, 2nd ed., Univ. Sci. Books,Sausalito, Calif.

Todhunter, P. E. (1996), Environmental indices for the Twin Cities metro-politan area (Minnesota, USA) urban heat island ‐ 1989, Clim. Res., 6(1),59–69.

Tooke, T. R., N. C. Coops, N. R. Goodwin, and J. A. Voogt (2009),Extracting urban vegetation characteristics using spectral mixture anal-ysis and decision tree classifications, Remote Sens. Environ., 113(2),398–407, doi:10.1016/j.rse.2008.10.005.

U.S. Department of Agriculture (2005), Forest inventory and analysisnational core field guide, vol. 2, FIA field methods for Phase 3 measure-ments, version 3.0, For. Serv., Washington, D. C.

Valente, F., J. S. David, and J. H. C. Gash (1997), Modelling interceptionloss for two sparse eucalypt and pine forests in central Portugal usingreformulated Rutter and Gash analytical models, J. Hydrol., 190(1–2),141–162.

Vesala, T., N. Kljun, Ü. Rannik, J. Rinne, A. Sogachev, T. Markkanen,K. Sabelfeld, T. Foken, and M. Y. Leclerc (2008a), Flux and concen-tration footprint modelling: State of the art, Environ. Pollut., 142, 14.

Vesala, T., et al. (2008b), Surface‐atmosphere interactions over complexurban terrain in Helsinki, Finland, Tellus, Ser. B, 60(2), 188–199,doi:10.1111/j.1600-0889.2007.00312.x.

Vickers, D., and L. Mahrt (1997), Quality control and flux sampling pro-blems for tower and aircraft data, J. Atmos. Oceanic Technol., 14(3),512–526.

Wang, J., T. A. Endreny, and D. J. Nowak (2008), Mechanistic simulation oftree effects in an urban water balance model, J. Am. Water Resour. Assoc.,44(1), 75–85, doi:10.1111/j.1752-1688.2007.00139.x.

Webb, E. K., G. I. Pearman, and R. Leuning (1980), Correction of fluxmeasurements for density effects due to heat and water vapour transfer,Q. J. R. Meteorol. Soc., 106, 85–100.

Whitlow, T. H., and N. L. Bassuk (1988), Ecophysiology of urban treesand their management ‐ The North American experience, HortScience,23(3), 542–546.

Whitlow, T. H., N. L. Bassuk, and D. L. Reichert (1992), A 3‐year study ofwater relations of urban street trees, J. Appl. Ecol., 29(2), 436–450.

Wilson, K., et al. (2002), Energy balance closure at FLUXNET sites, Agric.For. Meteorol., 113(1–4), 223–243.

Wilson, K. B., P. J. Hanson, P. J. Mulholland, D. D. Baldocchi, and S. D.Wullschleger (2001), A comparison of methods for determining forestevapotranspiration and its components: Sap‐flow, soil water budget,eddy covariance and catchment water balance, Agric. For. Meteorol.,106(2), 153–168.

Winkler, J. A., R. H. Skaggs, and D. G. Baker (1981), Effect of temperatureadjustments on the Minneapolis‐St. Paul urban heat‐island, J. Appl.Meteorol., 20(11), 1295–1300.

Wullschleger, S. D., P. J. Hanson, and D. E. Todd (2001), Transpirationfrom a multi‐species deciduous forest as estimated by xylem sap flowtechniques, For. Ecol. Manage., 143(1–3), 205–213.

Zhang, X., L. Hu, X. Bian, B. Zhao, F. Chai, and X. Sun (2007), The mosteconomical irrigation amount and evapotranspiration of the turfgrasses inBeijing City, China, Agric. Water Manage., 89(1–2), 98–104.

R. V. Hiller, Institute of Plant, Animal, and Agroecosystem Sciences,ETH Zurich, CH‐8092 Zurich, Switzerland.J. P. McFadden, Department of Geography, University of California,

Santa Barbara, CA 93106, USA.E. B. Peters, Institute on the Environment, University of Minnesota,

Saint Paul, MN 55108, USA. ([email protected])


16 of 16

173

Acknowledgements

The work presented in this thesis would not have been possible without the help of thefollowing:

Werner Eugster is not only my main supervisor, but he also introduced me to the fieldof micrometerology during my diploma thesis project. He connected me with Joe Mc-Fadden, who provided me the opportunity to experience life as a researcher in St. Paul,MN within his group. These experiences provoked me to continue, to commit a Ph.D.Back to a place I had known from before, Werner provided me with the freedom to ac-complish many side projects, including: construction of bicycle trailers to pull expensivelayer spectrometers over dirt roads through the countryside, and development of a newR-package to process eddy covariance data. I benefited from his technical experience,scientific endeavors, and understanding of computer science. I thank him for his never-ending support, many interesting discussions, and scientific input regarding my work.

Nina Buchmann who amazed me with her ability to keep an overview over the projectand take action when help was needed. I thank her for her effort and help, particularly inthe collection of all the data necessary for the analyses. But most of all, I thank her forher support and for giving me the opportunity to accomplish my Ph.D. within her group.

Bruno Neininger is not only my co-referent; he operated the flights and carefully pro-cessed the data. I thank him for letting me actively participate in the flight campaigns andallowing me to be part of the MetAir family. Although these days were always connectedwith a lot of work, I will always have fond memories. I appreciate that he took time toexplain me many details of the measurement system and gave me valuable input on thedata analysis.

During the flight campaigns Bruno Neininger, operator of MetAir, was supported byBoris Schneider and the pilots Dave Oldani, Lorenz Muller, Yvonne Schwarz, and MoritzIsler. Stephan Baum at the Max-Planck-Institute for Biogeochemistry in Jena analyzedthe flasks filled with air during the flights; and Dominik Brunner always contributedvaluable weather input to pick the best flight dates.

175

Acknowledgements

A special campaign in collaboration with EAWAG (Swiss Federal Institute of AquaticScience and Technology) lead to Lake Wohlen. I would like to thank Torsten Diem andDorte Carstens for sampling the lake fluxes with chambers and Tonya DelSontro for thediscussions regarding the lake fluxes.

Intensive ground based CH4- and N2O flux measurements in Chamau were performed incollaboration with EMPA (Swiss Federal Laboratories for Materials Science and Tech-nology). I would like to acknowledge Bela Tuzon, Kerstin Zeyer, and Lukas Emmeneggerfor their enormous effort to achieve the best possible measurements.

Hans-Ruedi Wettstein, leader of the ETH agricultural research station, and his team werealways anxious to support our research, including the timing of the slurry application inthe spring in accordance with our flights.

The last flight campaign was devoted to wetland emissions of the raised bog of Rothen-thurm. The ground based concentration measurement were performed by bicycle, ratherthan car. Thank you to Jacqueline Stieger for your amazing effort pulling the heavy trailerthrough the countryside!

Within the project framework of MAIOLICA, Silas Hobi and Elke Hodson helped toderive a spatially highly resolved CH4 inventory for Switzerland. In combination withthe additional inventory by Thomas Kunzle from Meteotest, they build a solid base tocompare with the aircraft measurements. The footprint calculation by Dominik Brunnerlinks the inventory with the measurements. Many thanks for all your work!

The technical accomplishment by Peter Pluss, Patrick Flutsch, and Thomas Baur resultedin a flying, plus a driving, CH4 analyzer. I would like to give many thanks to them fortheir help out in the field, repairing, and maintaining our stations.

The instrument inter-comparison initializes by Lukas Emmenegger and realized by Chri-stoph Zellweger allowed detailed performance analyzes and strengthened my confidencein our instruments.

During the field instrument inter-comparison for CH4 flux measurements in Oensingenand the analysis thereafter, I experienced real scientific teamwork with Bela Tuzon, Ker-stin Zeyer, Werner Eugster, Albrecht Neftel, Christof Ammann, and Lukas Emmenegger.

The many hops with a Fast Greenhous Gas Analyzer on board would not have beenpossible if Alexander Knohl was not so generous as to lend me his instrument. Thank

176

Acknowledgements

you also for the interesting discussion and scientific input, especially to the density fluxcorrection manuscript.

Joe McFadden gave me the opportunity to ’test’ the air of science at the University ofMinnesota for one year before committing to a Ph.D. Out of this work, a manuscriptoriginated in collaboration with Natascha Kljun. I appreciate all of the support and inputthat I received from both Joe and Natascha.

Many interesting and fruitful discussions rose from ’flux coffees’ with Matthias Zeeman,Patrick Sturm, Albin Hammerle, Sebastian Wolf, Sophia Etzold, Luth Merbold, DennisImer, and Jacqueline Stieger.

Matthias Zeeman introduced me to grassland flux processing, shared his experiences atthe field sites, provided many useful scripts to assemble something meaningful out of themany bytes and bits, and offered me many interesting, insightful discussions and advice.

I thank Sebastian Wolf, Dorte Bachmann, Susanne Burri, Anna Gilgen, Dennis Imer,Sabina Keller, and Jacqueline Stieger for the proofreading manuscripts and providing ofvaluable comments.

My office mates Susanne Burri, Fabienne Zeugin, Jacqueline Stieger, Stefan Trogisch,Florian Suter, and Nadine Ruhr provided a positive atmosphere including non-scientificdiscussions about anything and everything.

A special thank you goes to Dorte Bachmann, Susanne Burri, Lydia Gentsch, SabinaKeller, and Johanna Spiegel and for supporting and motivating me :-)

A special THANK YOU to my parents for their constant support and for providing mea solid home base. My mother took care of my physical well-being while my fathermanaged many technical discussions and ski tours with me, thus contributing to my life-work balance. I would like to acknowledge my siblings Maurus, Tobias, and Priska andmy friends for supporting me and having an open ear for my thoughts and fears, butespecially for diverting my focus to the life apart of science.

Work presented in this thesis is part of the project MAIOLICA (Modelling And exper-Iments On Land-surface Interactions with atmospheric Chemistry and climAte) and isfunded by the Competence Center Environment an Sustainability at the ETH Domain(CCES).

177

Documents

Rights / License: Research Collection In Copyright - Non ...5927/eth-5927-02.pdf · 2 budget of the lawn ecosystem had to be adjusted by ˇ135 g C m 2 s 1 to determine the natural