The SPARC water vapour assessment II: Comparison of ...€¦ · ter vapour cool the stratosphere but warm the troposphere (Solomon et al.,2010). Model simulations predict a ˘1K decrease

The SPARC water vapour assessment II: Comparison of stratosphericand lower mesospheric water vapour time series observed fromsatellites

Downloaded from: https://research.chalmers.se, 2020-07-06 14:15 UTC

Citation for the original published paper (version of record):Khosrawi, F., Lossow, S., Stiller, G. et al (2018)The SPARC water vapour assessment II: Comparison of stratospheric and lower mesospheric watervapour time series observed from satellitesAtmospheric Measurement Techniques, 11(7): 4435-4463http://dx.doi.org/10.5194/amt-11-4435-2018

N.B. When citing this work, cite the original published paper.

research.chalmers.se offers the possibility of retrieving research publications produced at Chalmers University of Technology.It covers all kind of research output: articles, dissertations, conference papers, reports etc. since 2004.research.chalmers.se is administrated and maintained by Chalmers Library

(article starts on next page)

Atmos. Meas. Tech., 11, 4435–4463, 2018https://doi.org/10.5194/amt-11-4435-2018© Author(s) 2018. This work is distributed underthe Creative Commons Attribution 4.0 License.

The SPARC water vapour assessment II: comparison ofstratospheric and lower mesospheric water vapour time seriesobserved from satellitesFarahnaz Khosrawi1, Stefan Lossow1, Gabriele P. Stiller1, Karen H. Rosenlof2, Joachim Urban3,†, John P. Burrows4,Robert P. Damadeo5, Patrick Eriksson3, Maya García-Comas6, John C. Gille7,8, Yasuko Kasai9, Michael Kiefer1,Gerald E. Nedoluha10, Stefan Noël4, Piera Raspollini11, William G. Read12, Alexei Rozanov4, Christopher E. Sioris13,Kaley A. Walker14, and Katja Weigel41Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research, Hermann-von-Helmholtz-Platz 1, 76344Eggenstein-Leopoldshafen, Germany2NOAA Earth System Research Laboratory, Global Monitoring Division, 325 Broadway, Boulder, CO 80305, USA3Chalmers University of Technology, Department of Space, Earth and Environment, Hörsalsvägen 11,41296 Gothenburg, Sweden4University of Bremen, Institute of Environmental Physics, Otto-Hahn-Allee 1, 28334 Bremen, Germany5NASA Langley Research Center, Mail Stop 401B, Hampton, VA 23681, USA6Instituto de Astrofísica de Andalucía (IAA-CSIC), Glorieta de la Astronomía, 18008 Granada, Spain7National Center for Atmospheric Research, Atmospheric Chemistry Observations and Modeling Laboratory, P.O. Box 3000,Boulder, CO 80307-3000, USA8University of Colorado, Atmospheric and Oceanic Sciences, Boulder, CO 80309-0311, USA9National Institute of Information and Communications Technology, Terahertz Technology Research Center, 4-2-1Nukui-kita, Koganei, Tokyo 184-8795, Japan10Naval Research Laboratory, Remote Sensing Division, 4555 Overlook Avenue Southwest, Washington, DC 20375, USA11Istituto di Fisica Applicata N. Carrara Del Consiglio Nazionale delle Ricerche (IFAC-CNR), Via Madonna del Piano, 10,50019 Sesto Fiorentino, Italy12Jet Propulsion Laboratory, 4800 Oak Grove Drive, Pasadena, CA 91109, USA13Environment and Climate Change Canada, Atmospheric Science and Technology Directorate, 4905 Dufferin St.,ON, M3H 5T4, Canada14University of Toronto, Department of Physics, 60 St. George Street, Toronto, ON, M5S 1A7, Canada†deceased, 14 August 2014

Correspondence: Farahnaz Khosrawi ([email protected])

Received: 29 January 2018 – Discussion started: 6 March 2018Revised: 25 June 2018 – Accepted: 12 July 2018 – Published: 25 July 2018

Abstract. Time series of stratospheric and lower meso-spheric water vapour using 33 data sets from 15 differentsatellite instruments were compared in the framework ofthe second SPARC (Stratosphere-troposphere Processes Andtheir Role in Climate) water vapour assessment (WAVAS-II).This comparison aimed to provide a comprehensive overviewof the typical uncertainties in the observational database

that can be considered in the future in observational andmodelling studies, e.g addressing stratospheric water vapourtrends. The time series comparisons are presented for thethree latitude bands, the Antarctic (80◦–70◦ S), the tropics(15◦ S–15◦ N) and the Northern Hemisphere mid-latitudes(50◦–60◦ N) at four different altitudes (0.1, 3, 10 and 80 hPa)covering the stratosphere and lower mesosphere. The com-

Published by Copernicus Publications on behalf of the European Geosciences Union.

4436 F. Khosrawi et al.: Comparison of H2O time series

bined temporal coverage of observations from the 15 satel-lite instruments allowed the consideration of the time period1986–2014. In addition to the qualitative comparison of thetime series, the agreement of the data sets is assessed quanti-tatively in the form of the spread (i.e. the difference betweenthe maximum and minimum volume mixing ratios among thedata sets), the (Pearson) correlation coefficient and the drift(i.e. linear changes of the difference between time series overtime). Generally, good agreement between the time serieswas found in the middle stratosphere while larger differenceswere found in the lower mesosphere and near the tropopause.Concerning the latitude bands, the largest differences werefound in the Antarctic while the best agreement was foundfor the tropics. From our assessment we find that most datasets can be considered in future observational and modellingstudies, e.g. addressing stratospheric and lower mesosphericwater vapour variability and trends, if data set specific char-acteristics (e.g. drift) and restrictions (e.g. temporal and spa-tial coverage) are taken into account.

Dedication to Jo Urban

We would like to dedicate this paper to our highly valuedcolleague Jo Urban, who would have certainly been the leadauthor of this study had he not passed away so soon. Withouthis devoted work on UTLS water vapour over many years,this work would not have been possible. In particular, theretrieval of water vapour from the SMR observations andthe combination of these data with other data sets to un-derstand the long-term development of this trace constituentcomprised a large part his life’s work. With his passing, welost not only a treasured colleague and friend, but also a lead-ing expert in the microwave and sub-millimetre observationcommunity.

1 Introduction

Water vapour is the most important greenhouse gas and playsa key role in the chemistry and radiative balance of the atmo-sphere. Any changes in atmospheric water vapour have im-portant implications for the global climate (Solomon et al.,2010; Riese et al., 2012) and need to be monitored and under-stood (Müller et al., 2016). Accurate knowledge of the watervapour distribution and its trends from the upper troposphereup to the mesosphere is therefore crucial for understandingclimate change and chemical forcing (Hegglin et al., 2013).

Water vapour is the source of the hydroxyl radical (OH)which controls the lifetime of shorter-lived pollutants, tropo-spheric and stratospheric ozone and other longer-lived green-house gases such as methane (Seinfeld and Pandis, 2006).Further, water vapour is an essential component of polarstratospheric clouds (PSCs) which play a key role in Antarc-tic and Arctic ozone depletion during winter and spring

(Solomon, 1999). Accordingly, water vapour has an impor-tant influence on stratospheric chemistry through its abil-ity to form ice particles. Dehydration, that is, the removalof water vapour from the gas phase, can either be a re-versible or an irreversible process depending on the lifetimeof water-containing particles and their size. However, ice par-ticles generally live long enough and grow sufficiently largeto fall and remove water vapour permanently from an airmass so that dehydration can generally be defined as an irre-versible process. Dehydration in the stratosphere is generallyobserved over the Antarctic during winter (e.g. Kelly et al.,1989; Vömel et al., 1995; Nedoluha et al., 2000, 2007) andto a lesser extent also over the Arctic (e.g. Fahey et al., 1990;Pan et al., 2002; Khaykin et al., 2013; Manney and Lawrence,2016) as well as at the tropical tropopause (e.g. Jensen et al.,1996; Read et al., 2004; Schiller et al., 2009).

In addition to its role in the Earth’s radiative budget andmiddle atmospheric chemistry, water vapour is an importanttracer for transport in the stratosphere and lower mesosphere.Dynamical circulations that can be diagnosed with watervapour in the middle atmosphere are the Brewer–Dobson cir-culation in the stratosphere and the pole-to-pole circulation inthe mesosphere (Brewer, 1949; Remsberg et al., 1984; Moteet al., 1996; Pumphrey and Harwood, 1997; Seele and Har-togh, 1999; Lossow et al., 2017a; Remsberg et al., 2018).In the stratosphere, the water vapour abundance is primarilygoverned by two main sources: (1) the transport from the tro-posphere through the tropical tropopause layer (TTL), wherethe minimum temperature (the so-called cold point temper-ature) determines how much water vapour enters the strato-sphere (Fueglistaler and Haynes, 2005); (2) the oxidation ofmethane, which is the only important chemical source of wa-ter vapour in the stratosphere (Bates and Nicolet, 1950; LeTexier et al., 1988).

A major research focus in relation to water vapour hasbeen on the detection and attribution of long-term changes instratospheric and mesospheric water vapour based on in situand remote sensing measurements (Oltmans and Hofmann,1995; Oltmans et al., 2000; Rosenlof et al., 2001; Nedoluhaet al., 2003; Scherer et al., 2008; Hurst et al., 2011; Heg-glin et al., 2014; Dessler et al., 2014). Many of these mea-surements have indicated an increase in stratospheric andmesospheric water vapour that has significant implicationsfor atmospheric temperature. Increases in stratospheric wa-ter vapour cool the stratosphere but warm the troposphere(Solomon et al., 2010). Model simulations predict a ∼ 1 Kdecrease in stratospheric temperature per decade along witha 0.5–1 ppmv increase of water vapour in the 21st century(Gettelman et al., 2010). Both the future cooling of the strato-sphere and the future increase in water vapour enhance thepotential for the formation of PSCs, which would have sig-nificant implications on Arctic and Antarctic dehydrationand ozone loss (Khosrawi et al., 2016; Thölix et al., 2016).The methane increase in the stratosphere can only explainpart of the observed water vapour changes (e.g. Rosenlof

Atmos. Meas. Tech., 11, 4435–4463, 2018 www.atmos-meas-tech.net/11/4435/2018/

F. Khosrawi et al.: Comparison of H2O time series 4437

et al., 2001; Hurst et al., 2011). A complete understandingof water vapour changes also requires good knowledge ofshort-term variability, such as the annual oscillation (AO) andsemi-annual oscillation (SAO) or the variations caused by thequasi-biennial oscillation (e.g. Schoeberl et al., 2008; Rems-berg, 2010; Kawatani et al., 2014; Lossow et al., 2017b).

In addition to an observed long-term increase in strato-spheric water vapour, pronounced drops have occasionallybeen observed. One drop (sometimes denoted as the millen-nium drop) occurred in 2000 (Randel et al., 2006; Schereret al., 2008; Solomon et al., 2010; Urban et al., 2012; Brinkopet al., 2016), with water vapour abundances starting to re-cover around 2004–2005 onwards. This decrease was causedby a reduced transport of water vapour across the tropicaltropopause in response to lower cold point temperatures.The exact driving mechanism is still in question, but hasbeen suggested to be due to variations of the QBO (quasi-biennial oscillation), ENSO (El Niño Southern Oscillation)and the Brewer–Dobson circulation that collectively acted inthe same direction lowering the tropopause temperatures. In2011 and 2012 another drop occurred, which however wasshorter-lived than the millennium drop (Urban et al., 2014).Recently, another sharp decrease was observed in connectionwith the QBO disruption and the unusual El Niño event in2015 and 2016 (Tweedy et al., 2017; Avery et al., 2017), butthis decrease has also already recovered.

Within the framework of the second SPARC water vapourassessment (WAVAS-II), we compared time series of strato-spheric and lower mesospheric water vapour derived from anumber of different satellite data sets. The time series com-parison was performed for the Antarctic (80◦–70◦ S), thetropics (15◦ S–15◦ N) and the Northern Hemisphere mid-latitudes (50◦–60◦ N) at four different altitudes (0.1, 3, 10and 80 hPa). This selection of latitude bands covers all threebasic climatic regions (i.e. tropics, mid-latitudes and polar re-gion) and allows the inclusion of all stratospheric WAVAS-IIdata sets in the comparison. The combined temporal cover-age of the 15 satellite instruments allows the considerationof the time period 1986–2014. This work aims to provide es-timates of the typical uncertainties in the time series fromsatellite observations that should be taken into account in ob-servational and modelling studies. A brief overview of thedata sets used in this study is provided in the next sectionfollowed by a description of the analysis approach in Sect. 3.In Sect. 4 the results are presented, focusing on the compar-ison of the de-seasonalised water vapour time series. Com-parison results for the absolute time series are given in theSupplement. Finally, our results will be summarised and con-clusions will be given in Sect. 5.

2 Data sets

For the comparison of water vapour products performedwithin the second SPARC WAVAS-II assessment, 40 data

sets (not including data sets of minor water vapour isotopo-logues) have been considered, primarily focusing on the timeperiod from 2000 to 2014 (Walker and Stiller, 2018). In thepresent study, we included all 33 data sets that have observa-tional coverage in the stratosphere. A list of these data setsis provided in Table 1, along with the effective time peri-ods available for analysis. In addition, this table provides thedata set labels and numbers used in the figures. Overall, datasets from the following 15 instruments have been considered(listed in alphabetical order): ACE-FTS, GOMOS, HALOE,HIRDLS, ILAS-II, MAESTRO, MIPAS, MLS (aboard theAura satellite, not the instrument on the Upper AtmosphereResearch Satellite – UARS), POAM III, SAGE II, SAGE III,SCIAMACHY, SMILES, SMR and SOFIE. For a number ofinstruments there are multiple data sets based on differentdata processors, measurement geometries, retrieval versionsand spectral signatures used to derive the water vapour in-formation. This especially holds for MIPAS, where 13 datasets have been included in this comparison. The MIPAS mea-surements are processed by four different processing cen-tres: (1) the University of Bologna (Dinelli et al., 2010),(2) the European Space Agency (ESA; Raspollini et al.,2013), (3) IMK/IAA (von Clarmann et al., 2009; Stiller et al.,2012a) and (4) Oxford (Payne et al., 2007). The four proces-sors differ in several respects, such as their choices of spec-tral ranges (so called micro-windows), the vertical grid onwhich the retrievals are performed (pressure or geometric al-titude), the choice of regularisation (and related to this, thevertical resolution), the choice of spectroscopic database, thesophistication of the radiative transfer (in particular, whetheror not non-local thermodynamic equilibrium, NLTE, emis-sions are considered) and whether or not any attempt is madeto account for horizontal inhomogeneities, and the a pri-ori and the assumed p–T profile. Indeed, the temperatureused might be a large source of error for species retrievedin LTE regions. Some of the different processing schemesalso make use of different level-1b data versions (here V5and V7) based on different ESA calibrations. The spread ofresults seen for MIPAS indicates how specific choices withina retrieval approach may influence the retrieval results. TheHALOE, POAM III and SAGE II data sets also include ob-servations before 2000. These were considered in the com-parisons, so that the combined temporal coverage of all datasets ranges from 1986 to 2014. A complete description ofthe data sets and their characteristics can be found in theWAVAS-II data set overview paper by Walker and Stiller(2018). In comparison to our previous SPARC WAVAS-IIpaper (Lossow et al., 2017b) the following two data re-lated changes have been made: (1) the ACE-FTS v3.5 andMAESTRO data sets have been extended from March 2013until December 2014 (see Table 1 of Lossow et al., 2017b).(2) The MIPAS ESA v7 data set has been completed. In theaforementioned study, this data set comprised only a sampleof 200 000 observations (instead of 1 800 000), though at the

www.atmos-meas-tech.net/11/4435/2018/ Atmos. Meas. Tech., 11, 4435–4463, 2018


time the temporal coverage on a monthly basis had alreadybeen completed.

3 Approach

3.1 Time series calculation

For the first step, we screened the individual data sets ac-cording to the criteria recommended by the data providers. Acomplete list of these criteria is given in the WAVAS-II dataset overview paper by Walker and Stiller (2018). After thescreening we interpolated the data onto a regular pressuregrid. This comprises 32 levels per pressure decade, whichcorresponds to a fine vertical sampling of about 0.5 km. Theuppermost level we consider is 0.1 hPa. The interpolated pro-files were then binned monthly and for the three latitudebands chosen: 80◦–70◦ S, 15◦ S–15◦ N and 50◦–60◦ N. Themonthly zonal means ya(t,φ,z) are given as

ya(t,φ,z)=1

no(t,φ,z)

no(t,φ,z)∑i=1

xi(t,φ,z). (1)

In the equation above xi(t,φ,z) describes the individualobservations that fall into a given time t (i.e. month) and lat-itude φ bin, no(t,φ,z) indicates their total number and z de-notes the altitude level. Before this calculation the data in thegiven bin were screened using the median and the median ab-solute difference (MAD, Jones et al., 2012) in an attempt toremove unrepresentative observations that occasionally oc-cur. Data points outside the interval 〈median[xi(t,φ,z)] ±7.5 MAD[xi(t,φ,z)]〉, with i = 1,. . . ,no(t,φ,z), were dis-carded, targeting the most prominent outliers (Jones et al.,2012; Lossow et al., 2017b). For a normally distributed dataset, 7.5 MAD corresponds to about 5σ . For individual datasets this concerned on average between 0.03 % and 3.2 %percent of the data in a given bin. Averaged over all data setstypically 0.6 % of the data in a given bin were removed bythis screening. In addition to the monthly zonal means, thecorresponding standard error εa(t,φ,z) was calculated by

εa(t,φ,z)=

√1

no(t,φ,z)[no(t,φ,z)− 1]

no(t,φ,z)∑i=1

[xi(t,φ,z)− ya(t,φ,z)

]2. (2)

To avoid spurious data, averages that are smaller than theircorresponding standard errors in an absolute scale were dis-carded. Also, monthly averages based on less than 20 ob-servations for dense data sets (e.g. HIRDLS, MIPAS, MLS,SCIAMACHY limb, SMILES-NICT and SMR) and less than5 observations for sparse data sets (e.g. ACE-FTS, GOMOS,HALOE, ILAS-II, MAESTRO, POAM III, SAGE II, SAGEIII, SCIAMACHY occultation and SOFIE) were not consid-ered any further. This is a slightly more relaxed approach

than used in the time series analysis by Lossow et al. (2017b),where a minimum of 20 observations was required for alldata sets. However, additional tests have shown that such aconservative criterion is not required for the sparser data sets.

In our analysis we consider both absolute time series andde-seasonalised time series. The ILAS-II and SMILES datasets cover less than one year, so that a de-seasonalisation isnot meaningful. There are multiple ways to achieve a de-seasonalisation. The most common and simplest approachis to calculate for a given calendar month the average overseveral years. Subsequently this average is subtracted fromthe individual months contributing to this climatological av-erage (i.e. average approach). This approach requires that adata set covers every calendar month at least twice. For theMIPAS V5H data sets this requirement is not fulfilled as theycover only 21 months. To accomplish a de-seasonalisationeven for these data sets a regression approach was used.Every data set was regressed with the following regressionmodel:

f (t,φ,z)= Coffset(φ,z)

+CAO1(φ,z) · sin(2πt/pAO)

+CAO2(φ,z) · cos(2πt/pAO)

+CSAO1(φ,z) · sin(2πt/pSAO)

+CSAO2(φ,z) · cos(2πt/pSAO). (3)

This model contained an offset as well as the annual oscil-lation (AO) and semi-annual oscillation (SAO). The AO andSAO are parameterised by orthogonal sine and cosine func-tions. f (t,φ,z) denotes the fit of the regressed time seriesand C are the regression coefficients of the individual modelcomponents. pAO = 1 year is the period of the annual oscil-lation; likewise pSAO = 0.5 years is the period of the semi-annual oscillation. In accordance to pAO and pSAO given inyears, the time t is here also used on a yearly scale. To calcu-late the regression coefficients we followed the method out-lined by von Clarmann et al. (2010) using the standard er-rors εa(t,φ,z) (their inverse squared) of the monthly zonalmeans as statistical weights. Autocorrelation effects and em-pirical errors (Stiller et al., 2012b) were not considered inthis regression. The de-seasonalised time series yd(t,φ,z),thus the anomalies for each time t , are then given as

yd(t,φ,z)= ya(t,φ,z)− f (t,φ,z). (4)

For the sake of simplicity we do not assign any error to theregression fit, so that the standard error of the de-seasonalisedtime series is given by

εd(t,φ,z)= εa(t,φ,z). (5)

3.2 Comparison parameters

To assess how the different time series compare betweentwo data sets or altogether we use a number of parameters,



Table 1. Overview over the water vapour data sets from satellites used in this study.

Instrument Data set Label Number Time period

ACE-FTS v2.2 ACE-FTS v2.2 1 03/2004–09/2010v3.5 ACE-FTS v3.5 2 03/2004–12/2014

GOMOS LATMOS v6 GOMOS 3 09/2002–07/2011

HALOE v19 HALOE 4 10/1991–11/2005

HIRDLS v7 HIRDLS 5 01/2005–03/2008

ILAS-II v3/3.01 ILAS-II 6 04/2003–08/2003

MAESTRO Research MAESTRO 7 03/2004–12/2014

MIPAS Bologna V5H v2.3 NOM MIPAS-Bologna V5H 8 07/2002–03/2004Bologna V5R v2.3 NOM MIPAS-Bologna V5R NOM 9 01/2005–04/2012Bologna V5R v2.3 MA MIPAS-Bologna V5R MA 10 01/2005–04/2012

ESA V5H v6 NOM MIPAS-ESA V5H 11 07/2002–03/2004ESA V5R v6 NOM MIPAS-ESA V5R NOM 12 01/2005–04/2012ESA V5R v6 MA MIPAS-ESA V5R MA 13 01/2005–04/2012ESA V7R v7 NOM MIPAS-ESA V7R 14 01/2005–04/2012

IMKIAA V5H v20 NOM MIPAS-IMKIAA V5H 15 07/2002–03/2004IMKIAA V5R v220/221 NOM MIPAS-IMKIAA V5R NOM 16 01/2005–04/2012IMKIAA V5R v522 MA MIPAS-IMKIAA V5R MA 17 01/2005–04/2012

Oxford V5H v1.30 NOM MIPAS-Oxford V5H 18 07/2002–03/2004Oxford V5R v1.30 NOM MIPAS-Oxford V5R NOM 19 01/2005–04/2012Oxford V5R v1.30 MA MIPAS-Oxford V5R MA 20 01/2005–04/2012

MLS v4.2 MLS 21 08/2004 – 12/2014

POAM III v4 POAM III 22 04/1998–11/2005

SAGE II v7.00 SAGE II 23 01/1986–08/2005

SAGE III Solar occultation v4 SAGE III 24 04/2002–06/2005

SCIAMACHY Limb v3.01 SCIAMACHY limb 25 08/2002–04/2012Lunar occultation v1.0 SCIAMACHY lunar 26 04/2003–04/2012Solar occultation – OEM v1.0 SCIAMACHY solar OEM 27 08/2002–08/2011Solar occultation – Onion peeling v4.2.1 SCIAMACHY solar Onion 28 08/2002–08/2011

SMILES NICT v2.9.2 band A SMILES-NICT band A 29 01/2010–04/2010NICT v2.9.2 band B SMILES-NICT band B 30 01/2010–04/2010

SMR v2.0 544 GHz SMR 544 GHz 31 11/2001–12/2014v2.1 489 GHz SMR 489 GHz 32 11/2001–08/2014

SOFIE v1.3 SOFIE 33 08/2007–09/2014

namely the spread (i.e. the difference between the maximumand minimum volume mixing ratios among the data sets),the (Pearson) correlation coefficient and the drift (i.e. linearchanges of the difference between time series over time). Inthe following subsections, the calculation of these parametersis described in more detail.

3.2.1 Spread

We define the spread as the difference between the maximumand minimum volume mixing ratio among the data sets at agiven time and place. As such, the spread is a simple mea-sure of the collective consistency among the time series fromthe different data sets. We have chosen this approach forthe spread calculation since for the other approaches basedon standard deviation or percentiles, assumptions have to bemade. However, we have also calculated the spread using



the other two approaches and derived qualitatively the sameresults as for the maximum–minimum calculation. Prior tothe spread calculation, we performed an additional screen-ing among the data sets to avoid unrepresentative spread es-timates. The screening is again based on the median andmedian absolute difference, as done before for the monthlyzonal mean calculation. Monthly zonal means outside theinterval 〈median[yp(t,φ,z)i]±7.5 MAD[yp(t,φ,z)i]〉 werenot considered, with i = 1, . . .,nd(t,φ,z) and nd(t,φ,z) de-noting the number of data sets at a given time, latitude andaltitude. The subscript p is used as a placeholder either forthe absolute or the de-seasonalised data. This screening re-moved overall 2.6 % of the data for the latitude band between80◦ and 70◦ S. For the tropical and the mid-latitude bands,respectively 3.6 % and 3.7 % of the data were removed. Sub-sequently, the spread was derived. We did not impose anyadditional criterion on the number of data sets available for aspread estimate to be valid (two data sets is the natural min-imum). However, for much of the 1990s the only availablesatellite data sets are HALOE and SAGE II. Since both in-struments provide solar occultation measurements, the num-ber of coincidences is limited. Thus, their time series do notconstantly overlap, there are many gaps in the spread. There-fore, we focus in the results section on the time period be-tween 2000 and 2014.

3.2.2 Correlation

To describe the consistency between two time series we em-ployed the correlation coefficient r(φ,z):

r(φ,z)= (6)∑nt (φ,z)i=1

[yp(ti ,φ,z)1 − yp(φ,z)1

]·[yp(ti ,φ,z)2 − yp(φ,z)2

]√∑nt (φ,z)i=1


]2·

√∑nt (φ,z)i=1


]2 ,with

yp(φ,z)1 =1

nt (φ,z)

nt (φ,z)∑i=1

yp(ti,φ,z)1, (7)

yp(φ,z)2 =1

nt (φ,z)

nt (φ,z)∑i=1

yp(ti,φ,z)2. (8)

The subscripts at the end of the variables refer to the twodata sets. p is again a placeholder for the absolute and de-seasonalised data. nt (φ,z) is the number of months the twotime series actually overlap, i.e. where both data sets yieldvalid monthly means. Correlation coefficients were only con-sidered if the overlap was at least 12 months. We did not per-form any significance analysis for the coefficients since wesimply want to show if the expected high correlation betweentwo time series exist.

3.3 Drift

As drift we consider the linear change of the difference be-tween two time series, which indicates if the longer-term

variation of the two time series is the same or not. The dif-ference time series was calculated as

1yd(t,φ,z)= yd(t,φ,z)1− yd(t,φ,z)2, (9)

where the subscripts at the end once more denote the twodata sets. As indicated by this equation the drift analysis fo-cuses on de-seasonalised time series. The standard error cor-responding to the difference time series is given by

1εd(t,φ,z)=

√εd(t,φ,z)

21+ εd(t,φ,z)

22. (10)

Due to the lack of appropriate covariance data, this calcu-lation omits any covariance between the different data sets.The difference time series were then regressed with a re-gression model containing an offset, a linear term (whichdescribes the drift) and the QBO parameterised by the Sin-gapore (1◦ N, 104◦ E) winds at 50 hPa (QBO1) and 30 hPa(QBO2) provided by Freie Universität Berlin (http://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/qbo.dat):

f (t,φ,z)= Coffset(φ,z)+Clinear(φ,z) · t

+CQBO1(φ,z) ·QBO1(t)

+CQBO2(φ,z) ·QBO2(t). (11)

The calculation of the regression coefficients followedagain the method by von Clarmann et al. (2010), us-ing the inverse square of the corresponding standard error1εd(t,φ,z) as weight. Here, unlike in the regression for thede-seasonalisation, auto-correlation effects and empirical er-rors were considered to derive optimal uncertainty estimatesfor the drifts. This consideration used the approach outlinedby Stiller et al. (2012b). We show drift results if the overlapperiod between the two time series is at least 36 months. Asoverlap period we define the time between the first and thelast month both data sets yield a valid monthly mean. We alsoprovide the information regarding how many months bothdata sets actually overlap, but we did not put any additionalconstraint on this quantity. In addition, we have performedtests with more advanced regression models, which yieldedqualitatively the same results.

4 Results

In this section, the results for the time series comparisonare presented. First, we provide an example (Fig. 1) of thetypical altitude–time distribution (contour time series) to de-scribe the general characteristics of the water vapour dis-tribution in the three latitude bands considered: Antarctic(80◦–70◦ S), tropics (15◦ S–15◦ N) and the Northern Hemi-sphere mid-latitudes (50◦–60◦ N). These latitude bands wereselected since these cover all three basic climatic regions andallow the inclusion of all stratospheric WAVAS-II data sets inthe comparison. Contour time series of water vapour in these


http://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/qbo.dat



three latitude bands derived from all of the data sets consid-ered in this study are provided in the Supplement (Figs. S1–S3). These figures give a good first overview of the altitudeand temporal coverage of the individual data sets and theirrepresentation of the characteristics of the water vapour dis-tribution at the three latitude bands.

The comparison of the time series is then performed quali-tatively for all data sets at the three latitude bands and at fourselected altitudes covering the stratosphere and lower meso-sphere (0.1, 3, 10 and 80 hPa). Subsequently, we assess theagreement of the data sets quantitatively in form of the spreadover all data sets as well as the correlations and drifts amongthe individual data sets. While the example is based on ab-solute data, the comparison results presented in this sectionwere derived from de-seasonalised data. The correspondingresults based on absolute data (except for the drift) are pro-vided in the Supplement.

4.1 General characteristics of the water vapour timeseries

Figure 1 shows contour time series of water vapour inthe Antarctic (80◦–70◦ S), tropics (15◦ S–15◦ N) and mid-latitudes (50◦–60◦ N) based on the MLS data set for the timeperiod 2004–2014. Here, the typical characteristics of thewater vapour distributions in these latitude regions becomevisible. The water vapour distribution in the polar regions(Fig. 1 top) is determined by the following three processes:(1) dehydration of the lower stratosphere during polar wintercaused by the sedimentation of ice containing polar strato-spheric cloud particles (Kelly et al., 1989; Fahey et al., 1990);(2) vertical transport of dry/moist air. During polar winter,dry air from the upper mesosphere descends within the polarvortex to the upper stratosphere, while during summer andearly autumn moist air from the upper stratosphere is trans-ported into the mesosphere; (3) enhanced production of wa-ter vapour by methane oxidation during summer due to thehigher insolation (Bates and Nicolet, 1950; Le Texier et al.,1988).

In the tropics (Fig. 1 middle), the most prominent fea-ture in the water vapour time series is the “atmospheric taperecorder” (Mote et al., 1996). This feature is a consequenceof the annual oscillation of dehydration (or freeze-drying) atthe tropical tropopause due to the annual oscillation of thetropical tropopause temperature. The tape recorder signal istransported upwards to about 15 hPa by the ascending branchof the Brewer–Dobson circulation and maintains its integritybecause of the subtropical mixing barrier in the lower strato-sphere. Around the stratopause (∼ 1 hPa) a pronounced semi-annual oscillation is found that is induced by an interplayof transport and momentum deposition of different types ofwaves (Hamilton, 1998).

The water vapour distribution in the mid-latitudes (Fig. 1bottom) is primarily influenced by transport within theBrewer–Dobson circulation and the overturning circulation

in the mesosphere. In the lower stratosphere, low volumemixing ratios are transported from the lower latitudes tothe mid-latitudes in late spring/early summer (Ploeger et al.,2013). Likewise, in the lower mesosphere the effect of up-welling in summer and downwelling in winter can be clearlyseen, as described for the Antarctic.

4.2 Qualitative time series comparisons

In the following, the time series from the different satellitedata sets are compared qualitatively. The time series in thethree considered latitudes bands cover generally the time pe-riod from 1991 to 2014 (0.1 hPa), from 1986 to 2014 (3 and10 hPa) and 1988 to 2014 (80 hPa). A necessary requirementfor the analyses of the de-seasonalised time series was a min-imum data set length of one year, ruling out some shorterdata sets (see Sect. 3.1). However, these data sets are con-sidered in the Supplement, where the time series in abso-lute terms derived from all satellite instruments consideredin this study are provided (Figs. S3–S6). Some data sets, e.g.the MAESTRO data set, only have coverage up to the lowestpressure level (80 hPa) considered here and thus these datacan only be found in bottom subfigures (Figs. 2–4 and S3–S6). Overall, 25 data sets have been considered in the com-parison for the Antarctic while 24 data sets have been consid-ered in the comparison for the tropics. In the Northern Hemi-sphere mid-latitudes, the best temporal and spatial coverageof the satellite data sets is found and therefore, 27 out of the33 satellite data sets are considered in this comparison.

4.2.1 Antarctic (80◦–70◦ S)

Figure 2 shows the de-seasonalised water vapour time se-ries for the southern polar latitudes. The HIRDLS, SCIA-MACHY (solar occultation) and SAGE III observations haveno coverage in this latitude region, while the GOMOS ob-servations’ coverage is too limited to allow derivation of de-seasonalised time series. In the de-seasonalised time series, aspread among the data sets can be found at the four altitudesconsidered in the comparison. The largest anomalies and thelargest spread are found at 0.1 hPa (up to ±2 ppmv), whilethe smallest anomalies and thus the smallest spread is foundat 3 hPa (generally in the range of ±0.4 ppmv).

At 0.1 hPa the time series start from 1991 onwards withHALOE, since SAGE II measurements are not available atthis altitude. Large differences in the seasonal variation ofthe de-seasonalised time series are found, resulting in a con-siderable spread among the data sets, larger than at other alti-tudes. Large anomalies (up to±2 ppmv), and thus large inter-annual variation, are found for the MIPAS-Oxford V5H,MIPAS-ESA V5R and MIPAS-ESA V7R data sets, whilequite small anomalies are found for both ACE-FTS data sets.These large anomalies in the above mentioned MIPAS datasets are a consequence of the pronounced (spiky) seasonalvariation in the absolute data (see Fig. S1 in the Supplement)



2 3 4 5 6 7 8

Water vapour / ppmv

| 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressu

re /

hP

a

80° S − 70° S

−80

−78

−76

−74

−72

−70

Me

an

la

titu

de

/ d

eg

ree

| 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressu

re /

hP

a

15° S − 15° N

−15−12−9−6−303691215

Me

an

la

titu

de

/ d

eg

ree

| 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressu

re /

hP

a

50° N − 60° N

50

52

54

56

58

60

Me

an

la

titu

de

/ d

eg

ree

(a)

(b)

(c)

Figure 1. Water vapour time series for the latitude bands 80◦ to 70◦ S (a), 15◦ S to 15◦ N (b) and 50◦ to 60◦ N (c) based on the MLS data.The light grey and white lines indicate the tropopause as derived from the MERRA reanalysis data. The black dots and the corresponding yaxes on the right show the average latitude of the monthly mean data. White areas indicate that there are no data.

that is difficult to be accounted for in the sinusoidal regres-sion used for the de-seasonalisation.

Decadal changes in water vapour are found in the de-seasonalised time series at 3 hPa. Several periods of watervapour increases are followed by water vapour decreases.Negative anomalies are found around 1992 while positiveanomalies are found around 1996 (HALOE). Water vapourthen shows positive anomalies again in ∼ 2003 (HALOE,POAM III, SAGE II), followed by a decrease in 2003–2004,which again is followed by a slight increase in water vapourthat lasts until 2010. From 2010 onwards water vapour re-mains unchanged. The last increase in water vapour is moststrongly pronounced in SMR 489 GHz indicating a drift inthe SMR 489 GHz data relative to the other data sets (seealso Sect. 4.5). A large spread between the de-seasonalisedtime series is found between 1999 and 2004 (mainly betweenPOAM III, SAGE II and SMR 489 GHz). Between 2005 and2014, good agreement between the de-seasonalised time se-ries is found. However, SMR 489 GHz has somewhat higheranomalies (from 2011 onwards) than the other satellite datasets.

At 10 hPa, the spread among the data sets is quite similarto that observed at 3 hPa, but the variability in water vapouris more pronounced. There is a decrease in the SAGE IIde-seasonalised water vapour time series of 1986–1990. Anincrease in the de-seasonalised water vapour time series isfound in POAM III around 2001. Also from 2009 onwardsthere seems to be a slight increase in water vapour in all

data sets. The SMR 489 GHz de-seasonalised time series at10 hPa is in good agreement with the de-seasonalised timeseries of the water vapour products derived from the othersatellite instruments. However, the SMR 489 GHz as well asthe SOFIE anomalies are low relative to MLS. This becomesquite obvious at the end of the time series (2012–2014), whenonly ACE-FTS, MLS, SMR 489 GHz and SOFIE were tak-ing measurements. Also, the influence of the QBO is clearlyvisible at this altitude level. Distinct positive anomalies arefound in 2007–2008, 2011 and 2013.

At 80 hPa the water vapour distribution is strongly influ-enced by dehydration (Sect. 4.1). The de-seasonalised timeseries at 80 hPa once again depict the spread between the in-dividual instruments in this latitude band. At 80 hPa similarresults as for 10 hPa are derived (except that here no long-term changes are visible). However, here the deviations be-tween HALOE and SAGE II are smaller than at 10 and 3 hPa.As at 10 hPa, a decrease in the anomalies of the SAGE IIde-seasonalised time series is found for 1986–1990. The de-seasonalised time series then remains constant until 1998(HALOE and SAGE II). From 1998 onwards the spread be-tween the data sets increases. There is an increase in theanomalies found in 2001, which is followed by a decrease,which lasts until 2004. Another decrease in water vapour isfound in 2009. At 80 hPa, POAM III shows stronger inter-annual variation and higher/lower anomalies than at 10 and3 hPa, depending on which year is considered.



1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

ACE−FTS v2.2 (75.5 S)

ACE−FTS v3.5 (76.2 S)

HALOE (72.7 S)

MAESTRO (74.7 S)

MIPAS−Bologna V5H (75.0 S)

MIPAS−Bologna V5R NOM (75.0 S)

MIPAS−Bologna V5R MA (75.0 S)

MIPAS−ESA V5H (75.0 S)

MIPAS−ESA V5R NOM (75.0 S)

MIPAS−ESA V5R MA (74.9 S)

MIPAS−ESA V7R (75.0 S)

MIPAS−IMKIAA V5H (75.0 S)

MIPAS−IMKIAA V5R NOM (75.0 S)

MIPAS−IMKIAA V5R MA (74.9 S)

MIPAS−Oxford V5H (75.1 S)

MIPAS−Oxford V5R NOM (75.1 S)

MIPAS−Oxford V5R MA (74.9 S)

MLS (75.2 S)

POAM III (74.6 S)

SAGE II (72.6 S)

SCIAMACHY limb (75.3 S)

SCIAMACHY lunar (74.6 S)

SMR 544 GHz (74.9 S)

SMR 489 GHz (75.2 S)

SOFIE (73.8 S)

Figure 2. De-seasonalised time series at four different altitudes considering the latitude band 80◦ to 70◦ S. In the legend the average latitudeof the individual time series is indicated, which was calculated in two steps. First, for an individual monthly mean the latitudes of all profilescontributing to it were averaged. Any altitude dependence due to missing or screened data was ignored in this step. Finally, the mean latitudesover the entire time series were averaged. The same anomaly range (y axis) has been used in all panels so that the differences in the anomalyand the spread can be more easily compared. On the x axis the ticks are given in the middle of the year.



1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

ACE−FTS v2.2 (0.3 S)

ACE−FTS v3.5 (0.3 S)

GOMOS (0.1 S)

HALOE (1.2 S)

HIRDLS (0.1 S)

MAESTRO (0.4 S)

MIPAS−Bologna V5H (1.2 S)

MIPAS−Bologna V5R NOM (0.0)

MIPAS−Bologna V5R MA (0.0)

MIPAS−ESA V5H (1.2 S)

MIPAS−ESA V5R NOM (0.1 N)

MIPAS−ESA V5R MA (0.1 N)

MIPAS−ESA V7R (0.1 N)

MIPAS−IMKIAA V5H (1.2 S)

MIPAS−IMKIAA V5R NOM (0.1 N)

MIPAS−IMKIAA V5R MA (0.1 N)

MIPAS−Oxford V5H (1.2 S)

MIPAS−Oxford V5R NOM (0.1 N)

MIPAS−Oxford V5R MA (0.2 N)

MLS (0.1 S)

SAGE II (0.3 S)

SCIAMACHY limb (2.6 S)

SMR 544 GHz (0.1 N)

SMR 489 GHz (1.0 S)

Figure 3. As Fig. 2, but considering the latitude band between 15◦ S and 15◦ N.

4.2.2 Tropics (15◦ S–15◦ N)

Figure 3 shows the de-seasonalised water vapour time se-ries for the tropics. The POAM III, SAGE III, SCIAMACHY(solar and lunar occultation) and SOFIE data sets have no

coverage in this latitude band. In the SAGE II time se-ries some data gaps occur which are due to the aftermathof the Pinatubo eruption (resulting in unrealistically highwater vapour values that were filtered out) as well as the“short events” between June 1993 and April 1994, when



too few measurements were available (Taha et al., 2004).In the tropics, good consistency between the data sets isfound except at 0.1 hPa, where again the spread betweenthe data sets is largest. At 0.1 hPa some data sets exhibitlarger anomalies (±1.2 ppmv; e.g. MIPAS-Oxford V5H andMIPAS-ESA V7R), while others exhibit rather small anoma-lies (±0.3 ppmv; e.g. ACE-FTS and MLS). The HIRDLS,GOMOS and MAESTRO (80 hPa) data sets show generallylarger anomalies and thus larger spread than the other satel-lite data sets. The de-seasonalised time series in the tropicsreflect the decadal changes in water vapour that have beendocumented in the literature, such as the drop in stratosphericwater vapour after 2000 and in 2012 (Randel et al., 2004,2006; Urban et al., 2014). Further, at 3 and 10 hPa, a vari-ability in water vapour on an approximate 2-year timescaleassociated with the QBO is clearly visible.

At 0.1 hPa the time series starts in 1991 with the HALOEdata set, which is also the only one available for these altitudeand latitude regions until 2001. The de-seasonalised time se-ries from HALOE shows an increase between 1992 and 1996followed by a period with rather constant anomalies thatlasts until 2001. Afterwards a decrease is visible until 2005.SMR 489 GHz observes, in contrast to HALOE, an increasein water vapour between 2001 and 2005. Therefore, at the be-ginning of the SMR 489 GHz record the anomalies at 0.1 hPaare clearly lower than those from HALOE or the other satel-lite data sets measuring from 2001 onwards. However, a largespread between the data sets is also found during this time pe-riod. A similar increase (but somewhat stronger) is found inthe MIPAS Oxford V5H data set between 2001 and 2003,but here the anomalies are higher than the ones from theother satellite data sets. While the MIPAS Oxford V5H andSMR 489 GHz data sets show increasing anomalies, the otherdata sets show decreasing anomalies. From 2006 onwardsall data sets show increasing anomalies. Between 2012 and2014, ACE-FTS, MLS and SMR 489 GHz are the only datasets covering this time period and deviations among themare quite visible. SMR 489 GHz anomalies are higher andshow larger inter-annual variability than ACE-FTS and MLS.MLS (together with ACE-FTS) exhibit generally the lowestanomalies (±0.3 ppmv) compared to the other satellite datasets at this altitude.

At 3 and 10 hPa the time series begins with SAGE II in1986. From 1991 onwards HALOE observations are alsoavailable. Both SAGE II and HALOE provide here a muchbetter representation of the temporal development of the wa-ter vapour time series and the inter-annual variability than inthe Antarctic since both data sets have a much better tempo-ral coverage in the tropics (see Figs. S1 and S2 in the Sup-plement). SAGE II shows somewhat larger anomalies thanHALOE. Generally, the de-seasonalised time series showgood agreement with each other at these two altitude lev-els (3 and 10 hPa). Further, at these altitude levels, the low-est anomalies and the lowest spread between the data sets isfound, especially at 10 hPa. The deviations between MLS (or

ACE-FTS) and SMR 489 GHz found during the time period2012–2014 are still evident at 3 hPa but to a much lesser ex-tent than at 0.1 hPa. At 3 hPa, inter-annual variations (withanomalies roughly on the order of±1 ppmv) due to the QBOare clearly visible. At 10 hPa this variability is far less ob-vious. Also, the differences between SMR 489 GHz and theother data sets measuring during the time period 2001–2005(SAGE II and HALOE) are found to a lesser extent at 3 hPa,but not at 10 hPa. The GOMOS data set exhibits large scat-ter. At 10 hPa the HIRDLS data set indicates stronger inter-annual variability than the other satellite instruments. Thislevel is the uppermost altitude where HIRDLS can be re-trieved and accordingly the data here are more uncertain.Both drops in water vapour, the one in 2001 and the one in2012, are clearly visible in the de-seasonalised time series at10 hPa. The latter one is strongly pronounced in the three re-maining data sets covering that time period (ACE-FTS v3.5,MLS and SMR 489 GHz). There is also a clear variability onan approximate 2-year timescale associated with the QBOvisible at this altitude level, although not at all times are asclearly pronounced as at 3 hPa.

Similar to the other three pressure levels, at 80 hPa rel-atively good agreement between SAGE II and HALOE isfound. However, SAGE II typically shows somewhat loweranomalies than HALOE. At 80 hPa, higher variability withlarger anomalies than at 10 and 3 hPa is found (generallyaround ±0.8 ppmv). The data sets agree well in terms of theinter-annual variation. The drops in 2000 and 2011 are con-sistently observed, as are the recoveries afterwards. This isalso true for the pronounced QBO in 2006–2008. In 2005 theMIPAS-Bologna V5R NOM and MIPAS-ESA V5R NOMdata sets show strong negative anomalies (up to −2 ppmv)which are not found in the other data sets. Similar behaviourof these data sets is found in 2011, when they show strongpositive anomalies (up to 1.6 ppmv), while in the other satel-lite data sets, anomalies up to only 0.4–0.8 ppmv are found.MAESTRO shows strong scatter, mainly because 80 hPais near the upper altitude limit of the MAESTRO watervapour retrieval. Another distinctive characteristic in the de-seasonalised time series at 80 hPa is the increase in watervapour that lasts until mid-2014 (ACE-FTS v3.5, MLS andSMR 544 GHz) which is anti-correlated with the time seriesat 10 hPa.

4.2.3 Northern mid-latitudes (50◦–60◦ N)

Figure 4 shows the de-seasonalised time series for the North-ern Hemisphere mid-latitudes. The GOMOS, SCIAMACHYlunar and SOFIE data sets have no coverage in this latituderegion. As for the other latitude bands the largest spreadbetween the satellite data sets is found at 0.1 hPa. This isaccompanied by large inter-annual variability. The ACE-FTS v3.5, MIPAS-Bologna V5H, MIPAS-Oxford V5H andSMR 489 GHz data sets are among the data sets showing thelargest inter-annual variability and also the largest anomalies



1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

0.1 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wate

r va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wate

r va

po

ur

an

om

aly

/ p

pm

v

3 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

10 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014−2

−1.6

−1.2

−0.8

−0.4

0

0.4

0.8

1.2

1.6

2

Wa

ter

va

po

ur

an

om

aly

/ p

pm

v

80 hPa

ACE−FTS v2.2 (55.6 N)

ACE−FTS v3.5 (55.8 N)

HALOE (54.6 N)

HIRDLS (55.0 N)

MAESTRO (55.7 N)

MIPAS−Bologna V5H (53.8 N)

MIPAS−Bologna V5R NOM (55.0 N)

MIPAS−Bologna V5R MA (55.1 N)

MIPAS−ESA V5H (53.8 N)

MIPAS−ESA V5R NOM (55.0 N)

MIPAS−ESA V5R MA (55.0 N)

MIPAS−ESA V7R (55.0 N)

MIPAS−IMKIAA V5H (53.8 N)

MIPAS−IMKIAA V5R NOM (55.0 N)

MIPAS−IMKIAA V5R MA (55.0 N)

MIPAS−Oxford V5H (53.8 N)

MIPAS−Oxford V5R NOM (55.0 N)

MIPAS−Oxford V5R MA (55.0 N)

MLS (54.7 N)

POAM III (57.0 N)

SAGE II (54.4 N)

SAGE III (54.7 N)

SCIAMACHY limb (55.2 N)

SCIAMACHY solar OEM (54.4 N)

SCIAMACHY solar Onion (54.4 N)

SMR 544 GHz (54.8 N)

SMR 489 GHz (55.0 N)

Figure 4. As Figs. 2 and 3, but here the time series for the latitude band between 50◦ and 60◦ N are shown.

at 0.1 hPa. The MIPAS-Oxford V5H data set covers the timeperiod of 2002–2004 and here the largest anomalies (exceed-ing 2 ppmv) are found. The largest negative anomalies arefound in 2005 and 2006 with −1.6 and −2 ppmv, respec-tively. The differences between ACE-FTS v3.5 and the other

satellite data sets become most pronounced at the end of thedata record when only SMR 489 GHz and MLS were stillmeasuring. Here, ACE-FTS v3.5 shows some larger variabil-ity. At this altitude, the drift in the SMR 489 GHz data setis again visible. The anomalies are typically more negative



0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

Water vapour spread / ppmv

| 2000 | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressure

/ h

Pa

2468101214161820

Num

ber

of data

sets

| 2000 | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressure

/ h

Pa

2468101214161820

Num

ber

of data

sets

| 2000 | 2001 | 2002 | 2003 | 2004 | 2005 | 2006 | 2007 | 2008 | 2009 | 2010 | 2011 | 2012 | 2013 | 2014 |

10−1

100

101

102P

ressure

/ h

Pa

2468101214161820

Num

ber

of data

sets

80° S − 70° S

15° S − 15° N

50° N − 60° N

Figure 5. The difference between the maximum and minimum volume mixing ratio among the different de-seasonalised data sets as afunction of time and altitude for the three latitude bands. The light grey and white lines indicate the tropopause as derived from MERRAreanalysis data. The right y axes and the corresponding red dots indicate the maximum number of data sets available for this analysis at agiven time considering all altitudes.

compared to the other data sets until 2004, while they aremore positive after 2012. The HALOE data set indicates anincrease in water vapour until about 1997 and a decrease af-terwards. There appears to be a decrease in water vapour forall data sets from 2007 to 2010, followed by a pronouncedincrease that lasts until early 2012.

At 3 hPa, the de-seasonalised time series show generallygood agreement, while at 10 hPa the best agreement is found.Differences at 3 hPa are that SMR 489 GHz exhibits loweranomalies during the time period 2001 to 2006 and higheranomalies than the other data sets from 2010 to 2014 andthat SAGE II shows higher anomalies than the other satelliteinstruments at the end of their data record (2004–2005). Dif-ferences at 10 hPa are found in the time period 2004–2008,when SAGE II and HIRDLS show stronger inter-annual vari-ability, and during 2010–2012, when SMR 489 GHz exhibitssomewhat higher anomalies than the other satellite data sets.In both altitude levels, an increase in water vapour between1992 and 2000 (10 hPa) and 1992 and 1998 (3 hPa) is found.The two water vapour drops that occurred after 2000 and in2011 in the tropics (Randel et al., 2004, 2006; Urban et al.,2014) are also visible at 10 hPa in the Northern Hemispheremid-latitudes, however with a temporal delay.

Although the inter-annual and decadal variability at 80 hPais low, some satellite data sets (MAESTRO, POAM III andSMR 544 GHz) show larger deviations from the other satel-lite data sets. In the MAESTRO data, high inter-annual vari-

ability is found with anomalies reaching up to 1.6 ppmv. Inthis altitude region, MAESTRO has its best temporal cov-erage in the mid-latitudes, but still 80 hPa is at the upperlimit of the MAESTRO measurements and therefore not ev-ery measured profile reaches that high up. This explains whyhigher variability (scatter) than in the other satellite datasets is found for the MAESTRO time series. POAM III ex-hibits much larger anomalies than the other satellite datasets (+1.2 ppmv compared to ±0.4 ppmv). Although thePOAM III anomalies decrease with time, they still remainhigher than the anomalies from the other satellite data sets.The differences between POAM III and the other satellitedata sets are caused by the limited temporal sampling (onlysummer months are measured) of POAM III in this latituderegion making the de-seasonalisation by regression appar-ently fail. In the SMR 544 GHz data set, larger inter-annualvariability is found, but with much smaller anomalies thanMAESTRO. In the SAGE II data, the anomalies are decreas-ing slightly in the time period 1987–2002. Further, there issome pronounced QBO alongside an overall increase from2004 to 2012.

Overall, in the Northern Hemisphere mid-latitudes, thelowest inter-annual variability is found, especially at 80 hPa.Similar to the comparisons in the Antarctic and tropics, thelargest inter-annual and decadal variability as well as thelargest spread between the data sets is found at 0.1 hPa. Thedrops in stratospheric water vapour after 2000 and in 2011



−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102

Pre

ssure

/ h

Pa

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102

Pre

ssure

/ h

Pa

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102

Pre

ssure

/ h

Pa

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102

Pre

ssure

/ h

Pa

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102

Correlation coefficient

Pre

ssure

/ h

Pa

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

10−1

100

101

102


Pre

ssure

/ h

Pa

ACE−FTS v2.2

ACE−FTS v3.5

GOMOS

HIRDLS

MAESTRO

MIPAS−Bologna V5R NOM

MIPAS−Bologna V5R MA

MIPAS−ESA V5R NOM

MIPAS−ESA V5R MA

MIPAS−ESA V7R

MIPAS−IMKIAA V5R NOM

MIPAS−IMKIAA V5R MA

MIPAS−Oxford V5R MA

MLS

SCIAMACHY limb

SCIAMACHY lunar

SCIAMACHY solar OEM

SCIAMACHY solar Onion

SMR 544 GHz

SMR 489 GHz

SOFIE

80° S − 70° S

15° S − 15° N

50° N − 60° N

Figure 6. Example correlations between de-seasonalised MIPAS-Oxford V5R NOM time series and those from other data sets. Results areonly shown when the two data sets have an overlap of at least 12 valid monthly means. The dashed orange lines indicate the four altitudesfor which the correlations between all data sets are shown in the following figures.



that are observed in the tropics are also found at 10 hPa inthe mid-latitudes, but with a temporal delay and to a lesserextent than in the tropics.

4.3 Spread assessment

In the following, the spread between the data sets is quan-titatively assessed to provide an estimate of the uncertaintyin the observational database. Figure 5 shows the differencebetween the maximum and minimum volume mixing ratioamong the different de-seasonalised water vapour data setsas a function of time and altitude for the three latitude bands:Antarctic, tropics and Northern Hemisphere mid-latitudes.The spread of the absolute time series is shown in the Supple-ment in Fig. S7. The spread is calculated for the years 2000–2014. Earlier years are not considered due to the lack of asufficient number of satellite instruments measuring duringthat time period. Before 2000 only HALOE, POAM III andSAGE II data were available which results in a too sparse andnot meaningful picture (similar to the gaps found for the earlyyears in Fig. 5). The spread estimates become more meaning-ful as more satellite data sets become available. This can beseen from Fig. 5 for 2002 onwards. For the years 2000–2001and 2012–2014 between two and four data sets were avail-able. In these cases the differences among the data sets arenot as pronounced and probably less meaningful than for theyears 2002–2012, when the majority of satellite instrumentswere measuring.

In all three latitude bands the spread is large at the highestand lowest altitude level considered in this study, which cor-respond to the upper troposphere/tropopause region and thelower mesosphere. The large spread in these altitude regionsis related to large uncertainties in the water vapour observa-tions (e.g. due to increased measurement noise) as well asto the variability of the atmosphere and its different repre-sentation in the individual data sets. In addition, large spreadis found in the Antarctic lower stratosphere (Fig. 5 top) inwinter and spring, when the water vapour distribution in thelower stratosphere is affected by dehydration and transportof low water vapour from the mesosphere into the strato-sphere (Sect. 4.1). In the tropics (Fig. 5 middle), the low-est spread compared to the other latitude bands is found. In-creased values are found here as in the other regions at thehighest and lowest levels. The spread is lowest in the timeperiod 2006 to 2010. Similar behaviour is found for the mid-latitudes (Fig. 5 bottom), also here the spread seems to belower around 10 hPa during the time period 2006–2010. Themid-latitudes show features similar to the tropics and po-lar regions. In the Northern Hemisphere mid-latitudes, thelargest spread occurs in the lower stratosphere, where lowwater vapour is found due to air masses that are freeze driedwhen entering the stratosphere in the tropics (atmospherictape recorder), and in the lower mesosphere due to the de-scent of air within the polar vortex.

4.4 Correlation assessment

To assess the temporal consistency between individual datasets, the correlation coefficients between all possible combi-nations of data sets are considered. In this section, the resultsfor the de-seasonalised time series are presented, while theresults for the absolute time series are given in the Supple-ment. We start by presenting an example correlation of theMIPAS-Oxford V5R NOM time series with those from theother data sets and then present all correlations in the formof matrices.

4.4.1 Correlation example

Figure 6 shows the correlation between the de-seasonalisedMIPAS-Oxford V5R NOM time series and those from theother data sets for the Antarctic, tropics and the NorthernHemisphere mid-latitudes. The largest spread in the corre-lation between the satellite data sets is found in the Antarctic(Fig. 6 top), also where the lowest correlation over all al-titude levels is found (rarely exceeding a correlation coeffi-cient of 0.8). MIPAS-ESA V5R NOM and MIPAS-ESA V7Rare among the data sets showing the highest correlation withMIPAS-Oxford V5R NOM over all altitude levels whilethe lowest correlation with MIPAS-Oxford V5R NOM isfound for SCIAMACHY lunar throughout most altitudes.The SOFIE and SMR 544 GHz data sets show very lowcorrelations (even negative for SOFIE) at the lowest alti-tude levels (below 10 hPa) as well as above 3 hPa (but hereSMR 489 GHz instead of SMR 544 GHz). In between thesealtitudes levels the SOFIE and SMR 489 GHz data sets showsimilar correlation to MIPAS-Oxford V5R NOM as the otherdata sets.

In the tropics (Fig. 6 middle), the correlation coefficientsvary between 0.8 and 1 for most data sets between 30 and1 hPa. Low correlations are found for all data sets between100 and 30 hPa, except the MIPAS-IMKIAA V5R NOMdata set, which shows a high correlation (> 0.8) up to 1 hPawith MIPAS-Oxford V5R NOM. The data sets that showthe lowest correlation with MIPAS-Oxford V5R NOM (evenin some occasions negative) are GOMOS and MAESTRO.These data sets thus deviate from the typical correlationof most other data sets. Above 60 hPa and above 25 hPathis is also true for HIRDLS and SMR 544 GHz, respec-tively. These two data sets show reasonable correlation withMIPAS-Oxford V5R NOM at the lowest altitude levels, butthen the correlation coefficients decrease rapidly with in-creasing altitude, most likely due to increased measurementnoise. At altitudes above 0.7 hPa the correlation decreases forall data sets and the spread between the data sets increases.For MIPAS-ESA V5R NOM, the correlation, although de-creasing, remains rather high with a correlation coefficient of0.7. The lowest correlation at 0.1 hPa is found for the ACE-FTS v2.2, ACE-FTS v3.5, MIPAS Bologna V5R NOM andMIPAS-Bologna V5R MA data sets.



25 20 17 20 17 20 20 17 20 17 24 23

25 22 25 22 25 25 22 25 22 39 37 26

19 19 19 18

56 70 59 70 70 59 70 59 70 66 41

58 59 58 58 59 58 59 59 55 40

21 21 20

69 83 83 69 83 69 83 78 43

69 69 70 69 70 70 66 43

83 69 83 69 83 78 43

21 20

69 83 69 83 78 43

69 70 70 66 43

20

69 83 78 43

70 66 43

114 58

57

(a) 0.1 hPa

Data set number

25

20

17

20

17

20

20

17

20

17

24

23

25

22

25

22

25

25

22

25

22

39

15

37

26

17

13

21

21

21

14

20

68

82

69

82

82

69

82

69

82

26

77

43

68

69

68

68

69

68

69

69

23

65

43

21

21

14

20

69

83

83

69

83

69

83

27

78

43

69

69

70

69

70

70

23

66

43

83

69

83

69

83

27

78

43

21

14

20

69

83

69

83

27

78

43

69

70

70

23

66

43

14

20

69

83

27

78

43

70

23

66

43

28

114

58

12

30 30

14 57

(b) 3 hPa

1 2 8 9 10 11 12 13 14 15 16 17 18 19 20 21 32 33

Data set number

3332262322212019181716151413121110

98421

Data

set num

ber

1 2 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 26 32 33

33

32

21

20

19

18

17

16

15

14

13

12

11

10

9

8

2

1

Data

set num

ber

25 20 17 20 17 20 20 17 20 17 24 14 23

25 22 25 22 25 25 22 25 22 39 15 21 37 26

17 13

21 21 21 14 20

68 82 69 82 82 69 82 69 82 26 33 77 43

68 69 68 68 69 68 69 69 23 27 65 43

21 21 14 20

69 83 83 69 83 69 83 27 34 78 43

69 69 70 69 70 70 23 27 66 43

83 69 83 69 83 27 34 78 43

21 14 20

69 83 69 83 27 34 78 43

69 70 70 23 27 66 43

14 20

69 83 27 34 78 43

70 23 27 66 43

28 48 114 58

12 12 30

15 30 14

57 24

57

(c) 10 hPa

25

20

20

20

20

20

24

25

15

25

12

25

25

25

25

39

40

13

17

14

15

21

21

21

14

12

21

49

82

82

80

82

82

42

82

29

49

49

49

49

50

35

50

28

21

21

14

12

21

83

81

83

83

42

83

29

81

83

83

42

83

29

21

14

12

21

81

81

42

81

29

14

12

21

83

42

83

29

48

121

39

12

14

33 60

20 39

(d) 80 hPa

1 2 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 26 31 32 33

3331252322211918161514121110

987421

Data

set num

ber

1 2 4 7 8 9 10 11 12 14 15 16 18 19 21 22 23 25 31 33

Data set number

33323126232221201918171615141312111098421

Data

set num

ber

−0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


1: ACE−FTS v2.2

2: ACE−FTS v3.5

4: HALOE

7: MAESTRO

8: MIPAS−Bologna V5H

9: MIPAS−Bologna V5R NOM

10: MIPAS−Bologna V5R MA

11: MIPAS−ESA V5H

12: MIPAS−ESA V5R NOM

13: MIPAS−ESA V5R MA

14: MIPAS−ESA V7R

15: MIPAS−IMKIAA V5H

16: MIPAS−IMKIAA V5R NOM

17: MIPAS−IMKIAA V5R MA

18: MIPAS−Oxford V5H

19: MIPAS−Oxford V5R NOM

20: MIPAS−Oxford V5R MA

21: MLS

22: POAM III

23: SAGE II

25: SCIAMACHY limb

26: SCIAMACHY lunar

31: SMR 544 GHz

32: SMR 489 GHz

33: SOFIE

Figure 7. The correlations between de-seasonalised time series in the latitude band between 80◦ and 70◦ S. The upper panel considers the0.1 hPa (a) and 3 hPa (b) pressure levels, while in the lower panel the results at 10 hPa (c) and 80 hPa (d) are shown. Only data sets yieldingany result at a given altitude are shown. Thus, the number of data sets can vary from altitude to altitude. Comparisons yielding no results areindicated by grey crosses. For comparisons with results (the coloured boxes) the number of months the two data sets actually overlap (i.e.both yield a valid monthly mean) are indicated.



25 19 14 19 14 19 19 14 19 14 24 24

25 20 25 20 25 25 20 25 20 44 42

16 16 16 16 12 34

21 21 21 21

69 82 69 82 83 69 83 69 83 80

68 70 68 69 70 69 70 70 68

21 21 21

68 82 82 68 82 68 82 79

68 69 70 69 70 70 68

82 68 82 68 82 79

21 21

69 83 69 83 80

69 70 70 68

21

69 83 80

70 68

118

Data set number

25

19

14

19

14

19

19

14

19

14

24

24

25

20

25

20

25

25

20

25

20

44

42

24

21

24

21

24

24

21

24

21

27

37

16

16

16

16

12

100

34

21

21

21

14

21

69

82

69

82

83

69

83

69

83

80

68

70

68

69

70

69

70

70

68

21

21

14

21

68

82

82

68

82

68

82

79

68

69

70

69

70

70

68

82

68

82

68

82

79

21

14

21

69

83

69

83

80

69

70

70

68

14

21

69

83

80

70

68 118 31

1 2 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 32

Data set number

3223212019181716151413121110

984321

Data

set num

ber

1 2 3 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 23 32

32

21

20

19

18

17

16

15

14

13

12

11

10

9

8

4

2

1

Data

set num

ber

25 13 19 14 19 14 19 19 14 19 14 24 24

13 25 20 25 20 25 25 20 25 20 44 42

25 22 25 22 25 25 22 25 22 28 38

16 16 16 16 12 90 34

34 21 33 21 33 34 21 34 21 39 36

21 21 21 14 21

69 82 69 82 83 69 83 69 83 80

68 70 68 69 70 69 70 70 68

21 21 14 21

68 82 82 68 82 68 82 79

68 69 70 69 70 70 68

82 68 82 68 82 79

21 14 21

69 83 69 83 80

69 70 70 68

14 21

69 83 80

70 68

118

31

25

13

24

19

14

19

19

19

14

19

24

25

25

13

37

24

19

24

24

24

19

24

42

30

43

25

22

25

25

25

22

25

28

39

39

16

16

16

16

51

30

35

13

34

20

33

32

34

21

34

39

39

39

22

17

22

22

22

17

22

36

28

37

21

21

21

20

21

68

82

81

83

69

83

83

83

83

67

67

68

69

68

69

69

69

21

21

20

21

81

82

68

82

82

82

82

81

68

81

81

81

81

21

20

21

69

83

83

83

83

69

70

70

70

20

21

83

83

83

93

122

17

21 117

1 2 3 4 5 8 9 10 11 12 13 14 15 16 17 18 19 20 21 23 32

31252321191817161514121110

98754321

Data

set num

ber

1 2 3 4 5 7 8 9 10 11 12 14 15 16 17 18 19 21 23 25 31

Data set number

32232120191817161514131211109854321

Data

set num

ber

−0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


1: ACE−FTS v2.2

2: ACE−FTS v3.5

3: GOMOS

4: HALOE

5: HIRDLS

7: MAESTRO




11: MIPAS−ESA V5H



14: MIPAS−ESA V7R







21: MLS

23: SAGE II

25: SCIAMACHY limb

31: SMR 544 GHz

32: SMR 489 GHz

(a) 0.1 hPa

(b) 3 hPa(c) 10 hPa

(d) 80 hPa

Figure 8. As Fig. 7, but here the results for the latitude band between 15◦ S and 15◦ N are shown.

In the Northern Hemisphere mid-latitudes (Fig. 6 bottom),the correlation coefficients vary between 0.4 and almost 1in the altitude region between 0.7 hPa and 10 hPa dependingon which data set is considered. The spread in the NorthernHemisphere mid-latitudes is almost as large as the spread in

the Antarctic. Very high correlation (correlation coefficientof around 0.9–1) between MIPAS-Oxford V5R NOM andthe other data sets is found at, for example, around 1 hPa forthe MIPAS-ESA V5R NOM and MIPAS-ESA V7R data sets.The lowest correlation between MIPAS-Oxford V5R NOM



57 47 36 48 39 48 48 39 48 39 54 53

57 45 58 48 58 58 48 58 48 92 87

12

21 21 21 20

66 82 69 82 82 69 82 69 82 77

66 67 66 66 67 66 67 67 63

21 21 20

69 83 83 69 83 69 83 78

69 69 70 69 70 70 66

83 69 83 69 83 78

21 20

69 83 69 83 78

69 70 70 66

20

69 83 78

70 66

113

Data set number

57

47

36

48

39

48

48

39

48

39

54

29

29

53

57

45

58

48

58

58

48

58

48

92

32

32

87

74

12

21

21

21

13

20

66

82

69

82

82

69

82

69

82

38

38

77

66

67

66

66

67

66

67

67

31

31

63

21

21

13

20

69

83

83

69

83

69

83

39

39

78

69

69

70

69

70

70

31

31

66

83

69

83

69

83

39

39

78

21

13

20

69

83

69

83

39

39

78

69

70

70

31

31

66

13

20

69

83

39

39

78

70

31

31

66

43

43

113

20

12

13

13

15

12

12

24

13

13

14

55

51 51

1 2 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 32

Data set number

322827242322212019181716151413121110

98421

Data

set num

ber

1 2 4 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 27 28 32

32

21

20

19

18

17

16

15

14

13

12

11

10

9

8

4

2

1

Data

set num

ber

57 28 47 36 48 39 48 48 39 48 39 54 29 29 53

30 57 45 58 48 58 58 48 58 48 92 32 32 87

74 12

33 18 34 21 34 34 21 34 21 39 18 18 32

21 21 21 13 20

66 82 69 82 82 69 82 69 82 38 38 77

66 67 66 66 67 66 67 67 31 31 63

21 21 13 20

69 83 83 69 83 69 83 39 39 78

69 69 70 69 70 70 31 31 66

83 69 83 69 83 39 39 78

21 13 20

69 83 69 83 39 39 78

69 70 70 31 31 66

13 20

69 83 39 39 78

70 31 31 66

43 43 113

20 12 13 13 15

12 12 24

13 13 14

55 51

51

57

28

50

47

36

47

47

48

36

48

54

54

29

29

58

30

73

57

45

57

57

58

43

58

92

64

32

32

92

42

15

27

33

18

33

33

34

20

34

39

39

18

18

39

47

36

47

47

48

35

48

70

54

25

25

69

21

21

21

13

17

21

66

82

82

82

63

82

82

76

38

38

82

66

66

66

62

66

67

61

31

31

67

21

21

13

17

21

82

82

63

82

82

76

38

38

82

82

63

82

82

76

38

38

82

21

13

17

21

63

83

83

77

39

39

83

63

64

59

31

31

64

13

17

21

83

77

39

39

83

87

43

43

121

20

12

12

13

13

16

22

12

12

29

12

13

13

15

50

50

108

55

55 55

1 2 4 5 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 27 28 32

3128272524232221191817161514121110

9875421

Data

set num

ber

1 2 4 5 7 8 9 10 11 12 14 15 16 17 18 19 21 22 23 24 25 27 28 31

Data set number

322827242322212019181716151413121110985421

Data

set num

ber

−0.9 −0.8 −0.7 −0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


1: ACE−FTS v2.2

2: ACE−FTS v3.5

4: HALOE

5: HIRDLS

7: MAESTRO




11: MIPAS−ESA V5H



14: MIPAS−ESA V7R







21: MLS

22: POAM III

23: SAGE II

24: SAGE III

25: SCIAMACHY limb

27: SCIAMACHY solar OEM

28: SCIAMACHY solar Onion

31: SMR 544 GHz

32: SMR 489 GHz

(a) 0.1 hPa

(b) 3 hPa(c) 10 hPa

(d) 80 hPa

Figure 9. As Figs. 7 and 8, but considering the latitude band between 50◦ and 60◦ N.

and the other data sets is found above 1 hPa for the two ACE-FTS data sets while the SMR 489 GHz data set shows a ratherlow correlation throughout the entire altitude region consid-ered in this study. Below 10 hPa the lowest correlations (evennegative correlations) are found for HIRDLS, MAESTRO,

SCIAMACHY limb and SMR 544 GHz data sets. These datasets also deviate from the usual spread in correlation of thedata sets.



4.4.2 Correlation matrices

The correlation of all data sets is given in Figs. 7–9 in form ofmatrix plots for the three latitude bands and four altitude lev-els. In addition to the correlation coefficient, the number ofmonths of overlap between the time series is given (requiringa minimum of 12 months; see Sect. 3.2.2). The same figuresfor the correlation of the absolute time series are given in theSupplement (Figs. S8–S10). The correlation matrix shown inFig. 7 gives a good overview over the temporal consistencyof all data sets in the Antarctic. The correlations between thedata sets are generally positive (green), but in some casesnegative correlations (red) are found, for example, in thecase of the correlation between the MIPAS-IMKIAA V5Hand POAM III data sets at 10 hPa or that between the MLSand SCIAMACHY lunar data sets at 3 hPa. However, in thesetwo cases, the number of overlapping months is not that high(14 and 28) and this may explain the low correlation betweenthese data sets. An example of where a negative correlation isfound despite the high number of overlapping months (70) isthe correlation between the MIPAS-Bologna V5R NOM andMLS data sets at 0.1 hPa. An example of a high number ofoverlapping months (114) and high correlation coefficient isthe correlation between the MLS and SMR 489 GHz data setsat 10 hPa. Nevertheless, although in the Antarctic the corre-lation is generally positive, the correlation coefficient rarelyexceeds 0.5. An exception is the 3 hPa level, where a gen-erally high correlation among the MIPAS data sets is found.Similar behaviour between the MIPAS data sets is found at10 hPa.

In Fig. 8 the correlation matrix for the tropics is shown.The large spread between the data sets we found in Fig. 6at 0.1 hPa is also reflected in the correlations among all datasets. The same holds for the good correlations that are foundat 3 and 10 hPa. An exception here is the GOMOS dataset that shows negative correlations with all instruments at3 hPa, but the number of overlapping months is rather low.At 80 hPa the spread between the data sets is not as large asat 0.1 hPa, but still larger than at 3 and 10 hPa. At 80 hPaoccasionally negative correlations are found. This primar-ily concerns comparisons involving the GOMOS, HALOE,MAESTRO and MIPAS-Oxford V5H data sets. The low-est (negative) correlation is found between SMR 489 GHzand SAGE II data sets, but here the number of overlappingmonths (21) was also rather low.

The correlation matrix shown in Fig. 9 gives a goodoverview of the temporal consistency of all data sets inthe mid-latitudes. The majority of the correlations are pos-itive, but for some comparisons negative correlation isfound. One such example is the correlation between theMIPAS Bologna V5H and SMR 489 GHz data sets at 3 hPa.However, again the number of overlapping months was ratherlow and may explain the negative correlation between thesedata sets. An example of negative correlation, despite a highnumber of overlapping months, is found between MIPAS-

Bologna V5R NOM and MIPAS-Bologna-V5R-MA withMLS at 0.1 hPa. The correlation of these two data sets withthe other data sets is also generally low at 0.1 hPa. Also, forthe two ACE-FTS data sets the correlation of most data sets isoften low despite a sufficient number of overlapping months.Positive correlations are found for the ACE-FTS v2.2/v3.5data sets in comparison to the MIPAS-IMKIAA V5R MA,MIPAS-Oxford V5R MA, MLS and SMR 489 GHz. Thehighest correlation at 0.1 hPa is found between the two ACE-FTS data sets and between ACE-FTS v2.2 and MLS. At 3and 10 hPa generally high correlations among the MIPASdata sets are found. At 10 hPa the correlation of HIRDLSwith some data sets is high, but low with the other data sets.At 80 hPa low correlations between MAESTRO and all otherinstruments are found.

In summary, a high number of overlapping months doesnot necessarily guarantee a good correlation between twodata sets, but generally the chances are quite high if this isthe case. On the other hand, if data sets overlap only for alow number of months, good agreement between these datasets can still be found. Therefore, for assessing the agree-ment between two data sets, both the number of overlappingmonths and the correlation coefficient should be taken intoaccount. The correlation assessment again confirms what wefound before from the qualitative time series comparison,namely that the best agreement between the satellite data setsis found in the tropics, while in the Antarctic and NorthernHemisphere mid-latitudes a large spread between the datasets is found. Generally, the lowest correlations are found inthe Antarctic. Further, in each latitude band the correlation islower in the lower stratosphere and lower mesosphere than inthe middle stratosphere.

4.5 Drift assessment

In addition to the spread and correlations, the drifts amongthe satellite data sets are considered. As drift we considerthe linear change of the difference between two time series,which indicates if the longer-term variation of the two timeseries is the same or not (Sect. 3.3). As before, we start withan example. In Fig. 10 the drifts between the de-seasonalisedtime series of the SMR 489 GHz and all other data setsare shown for the Northern Hemisphere mid-latitudes (leftpanel) as well as the corresponding significance level (rightpanel). The significance level is given by the absolute ratioof the drift to the drift uncertainty. We consider a drift as sta-tistically significant when the significance level is larger than2σ (corresponding to the 95 % confidence level).

4.5.1 Drift example

Figure 10 shows that below 20 hPa large drifts (up to2.5 ppmvdecade−1 and even higher) are found betweenSMR 489 GHz and the other satellite data sets. In the al-titude region between 20 and 1 hPa, good consistency be-



−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5

10−1

100

101

102

Drift / ppmv ⋅ decade−1

Pre

ssu

re /

hP

a

−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5

10−1

100

101

102


Pre

ssu

re /

hP

a

0 1 2 3 4 5 6 7 8 9 10

10−1

100

101

102

Significance level / σ0 1 2 3 4 5 6 7 8 9 10

10−1

100

101

102

Significance level / σ

ACE−FTS v2.2 (78/53)

ACE−FTS v3.5 (123/87)

HALOE (43/12)

MAESTRO (123/49)

MIPAS−Bologna V5R NOM (84/77)

MIPAS−Bologna V5R MA (67/63)

MIPAS−ESA V5R NOM (84/78)

MIPAS−ESA V5R MA (76/66)

MIPAS−ESA V7R (84/78)

MIPAS−IMKIAA V5R NOM (84/78)

MIPAS−IMKIAA V5R MA (76/66)

MIPAS−Oxford V5R NOM (84/78)

MIPAS−Oxford V5R MA (76/66)

MLS (120/113)

POAM III (39/15)

SAGE II (45/24)

SAGE III (37/14)

SCIAMACHY limb (116/100)

SCIAMACHY solar OEM (108/51)

SCIAMACHY solar Onion (108/51)

SMR 544 GHz (153/141)

(a) (b)

Figure 10. The left panel shows the drifts between the de-seasonalised time series of the SMR 489 GHz data set and the other data sets. In theright panel the corresponding significance levels of the drift estimates are shown and the 2σ level is marked by a vertical line. This exampleconsiders the latitude band between 50◦ and 60◦ N. In the legend, the first number given in parentheses indicates the overlap period (overall altitudes) of the two data sets, i.e. the time between the first and the last month during which the data sets yield a valid monthly mean.Results are only shown here when this time period is at least 36 months. The second number indicates the number of months for which bothdata sets actually yield a valid monthly mean.

tween the satellite data sets is found despite the differenttime periods of measurements. The smallest drifts, rangingfrom about 0 to 0.5 ppmvdecade−1, are found around 20 hPa.The drifts consistently increase with altitude and maximisearound 0.4 hPa. Above 1 hPa the drifts of SMR 489 GHzvary between about 0.75 and 1.5 ppmvdecade−1 dependingon which data set the SMR 489 GHz data set is comparedto, but decrease with altitude towards 0.1 hPa. The driftsrange here between 0 and 1.25 ppmvdecade−1. The driftsbetween SMR 489 GHz and the other satellite data sets arein most cases significant at the 2σ uncertainty level as canbe seen from Fig. 10 (right panel). Larger drifts betweenSMR 489 GHz and the other data sets that obviously devi-ate from the majority of data sets are found for the compari-son to the POAM III, SAGE II, SAGE III and HALOE datasets. However, this is due to the fact that for these data setsnot only the overlap period with SMR 489 GHz is relativelyshort (4 years, 2001–2005), but also the number of monthsfor which both data sets actually yield a valid monthly meanis small (see numbers given in figure legend). Additionally,these drifts are in most cases not statistically significant atthe 2σ uncertainty level.

4.5.2 Drift matrices

In Figs. 11–13 the drift estimates between the time seriesof all data sets are summarised as matrix plots for the threelatitude bands and four altitudes. In the matrix plots, data setsare only shown if they yield any result at a given altitude.The drift estimates are based on the difference time seriesbetween the data sets given on the x axis and the data setsgiven on the y axis. Additional information that is given inthe matrix plots includes the overlap period of the two datasets, how many months the two data sets actually overlap andif the drift is significant or not at the 2σ uncertainty level aswell as the corresponding significance level for significantdrift.

In the Antarctic (Fig. 11), almost no significant driftsare found between the satellite data sets at the two low-est altitude levels (80 and 10 hPa). An exception here isthe MAESTRO data set which shows a significant (nega-tive) drift of −2 to −3 ppmvdecade−1 (significance levelup to 3.7) and POAM III which shows a significant pos-itive drift (2 to 3 ppmvdecade−1) compared to SAGE IIand SMR 544 GHz (at 80 hPa). While the overall time pe-riod MAESTRO had overlap with other data sets was suffi-ciently long (> 85 months), the number of coincident monthsfor these data sets was rather low (9 months). Further, at80 hPa, a significant negative drift is found between some



77−25 66−20 66−17 66−20 66−17 66−20 66−20 66−17 66−20 66−17 73−24

2.5

77−23 37−11

85−25 85−22 85−25 85−22 85−25 85−25 85−22

2.3

85−25 85−22 121−39

2.9

124−37 85−26

85−56 86−70 85−59 86−70 86−70 85−59 86−70 85−59 86−70

2.5

84−66 56−41

2.5

87−58

2.2

87−59

2.2

87−58

2.3

87−58

2.1

87−59

3.0

87−58

3.2

87−59 87−59

2.3

73−55 56−40

87−69

3.4

87−83

2.5

87−83

2.9

87−69 87−83 87−69

2.7

87−83 84−78 56−43

87−69 87−69 87−70 87−69 87−70 87−70 76−66 56−43

87−83 87−69 87−83 87−69 87−83

3.5

84−78 56−43

87−69 87−83 87−69 87−83

4.8

84−78 56−43

2.5

87−69 87−70 87−70

3.1

76−66

2.1

56−43

87−69 87−83 84−78

2.5

56−43

87−70 76−66 56−43

6.4

120−114 85−58

2.2

84−57

Data set number

2.277−25

2.966−20

2.766−17

4.066−20

4.666−17

2.666−20

66−20

3.066−17

3.466−20

2.666−17

73−24

4.971−11

2.177−23

37−11

4.785−25

3.185−22

5.285−25

5.785−22

3.385−25

3.685−25

4.285−22

4.085−25

3.285−22

121−39

96−15

4.2124−37

85−26

76−17

156−13

87−68

2.487−82

2.287−69

3.987−82

4.587−82

87−69

87−82

87−69

6.287−82

2.785−26

2.484−77

2.056−43

2.487−68

2.787−69

2.987−68

2.887−68

87−69

87−68

87−69

4.787−69

85−23

3.173−65

2.056−43

87−69

7.487−83

7.387−83

87−69

2.087−83

2.187−69

8.187−83

3.185−27

84−78

2.956−43

5.287−69

5.987−69

4.287−70

87−69

3.287−70

7.387−70

2.385−23

76−66

2.956−43

87−83

87−69

87−83

87−69

2.887−83

85−27

3.784−78

56−43

3.087−69

2.687−83

2.887−69

87−83

85−27

4.884−78

56−43

2.287−69

87−70

3.887−70

85−23

2.776−66

56−43

2.687−69

3.387−83

85−27

3.084−78

56−43

3.887−70

85−23

2.976−66

56−43

85−28

6.2120−114

85−58

76−12

2.248−30

2.7108−30

48−14 84−57

1 2 9 10 12 13 14 16 17 19 20 21 32 33

Data set number

33

32

26

23

22

21

20

19

17

16

14

13

12

10

9

4

2

1

Data

set num

ber

1 2 4 9 10 12 13 14 16 17 19 20 21 22 23 26 32 33

33

32

21

20

19

17

16

14

13

12

10

9

2

1

Data

set num

ber

77−25 66−20 66−172.1

66−20 66−17 66−20 66−20 66−173.2

66−20 66−17 73−24 71−11 73−14 77−23 37−11

85−25 85−22 85−25 85−22 85−25 85−25 85−222.7

85−25 85−224.4

121−39 96−15 121−21 124−37 85−26

76−17 156−13

87−68 87−82 87−69 87−822.1

87−82 87−69 87−82 87−69 87−82 85−26 75−332.3

84−77 56−43

87−68 87−69 87−68 87−68 87−69 87−68 87−69 87−69 85−23 63−27 73−65 56−43

87−69 87−832.5

87−83 87−693.1

87−83 87−69 87−83 85−27 76−34 84−78 56−43

87−692.3

87−69 87−70 87−69 87−70 87−70 85−23 63−27 76−66 56−43

87−83 87−694.1

87−83 87−692.4

87−83 85−27 76−342.2

84−78 56−43

87−695.0

87−83 87−693.0

87−83 85−27 76−342.3

84−78 56−43

87−69 87−70 87−70 85−23 63−27 76−66 56−43

87−69 87−83 85−27 76−34 84−78 56−43

87−70 85−23 63−27 76−66 56−43

85−28 121−483.7

120−1142.2

85−58

76−12 39−123.0

48−30

97−15 108−30 48−14

147−57 85−24

84−57

2.7

77−25

71−11

66−20

60−8

66−20

66−20

66−20

66−20

73−24

60−6

77−25

3.7

120−15

85−25

85−12

85−25

85−25

85−25

85−25

121−39

72−7

125−40

61−13

76−17

97−9

2.8

40−10

3.1

85−9

2.1

85−8

2.8

85−9

2.7

85−9

2.5

85−9

3.0

85−9

3.2

115−14

3.2

72−6

2.1

120−15

49−5

87−49

87−82

87−82

87−80

87−82

87−82

2.6

86−42

87−82

54−29

87−49

87−49

87−49

87−49

87−50

86−35

87−50

2.3

54−28

87−83

87−81

87−83

87−83

86−42

87−83

2.4

54−29

87−81

87−83

87−83

86−42

87−83

54−29

87−81

87−81

86−42

87−81

2.4

54−29

87−83

86−42

87−83

2.1

54−29

2.6

89−48

124−121

78−39

2.4

76−12

37−14

2.3

48−33 114−60

53−20 78−39

1 2 4 9 10 12 13 14 16 17 19 20 21 22 23 26 31 32 33

33

31

25

23

22

21

19

16

14

12

10

9

7

4

2

1

Data

set num

ber

1 2 4 7 9 10 12 14 16 19 21 22 23 25 31 33

Data set number

33

32

31

26

23

22

21

20

19

17

16

14

13

12

10

9

4

2

1

Data

set num

ber

−3 −2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 2 3


1: ACE−FTS v2.2

2: ACE−FTS v3.5

4: HALOE

7: MAESTRO





14: MIPAS−ESA V7R





21: MLS

22: POAM III

23: SAGE II

25: SCIAMACHY limb

26: SCIAMACHY lunar

31: SMR 544 GHz

32: SMR 489 GHz

33: SOFIE

(a) 0.1 hPa

(b) 3 hPa(c) 10 hPa

(d) 80 hPa

Figure 11. Drifts between the different data sets in the latitude band between 80◦ and 70◦ S at four specific altitudes. The drift estimates arebased on the difference time series between the data sets given on the x axis and the data sets given on the y axis. Additional information isgiven in the result boxes: the overall time period the two data sets overlap, how many months the data sets actually overlap (upper left corner)and if the drifts are significant (green frame) or not significant (slant) at the 2σ uncertainty level. The significance level is given in the lowerright corner in cases where the drift is significant.



MIPAS data sets and SOFIE. At 10 hPa, a significant (posi-tive) drift (0.8 ppmv decade−1) is found between the MIPAS-Oxford V5R NOM and ACE-FTS v2.2 data sets (signif-icance level of 3.2) and of 2 ppmvdecade−1 between theSMR 489 GHz and POAM III data sets (significance level3.0). Additionally, significant drifts are found between differ-ent MIPAS data sets relative to SMR 489 GHz and betweenthe MLS and SMR data sets. At 3 hPa most drifts are sig-nificant. Most MIPAS data sets exhibit significant positivedrifts relative to the ACE-FTS (significance level up to 5.7)and MLS (significance level up to 8.1) data sets. While in thecomparisons to the ACE-FTS data sets the actual number ofoverlapping months is limited, this is not the case in the com-parison to MLS. As before, for the SMR 489 GHz data setsignificant positive drifts are found (significance level up to4.8) relative to most other data sets. A large variety of driftsis found at 0.1 hPa, but in most cases the drift is not signif-icant. Data sets for which most drifts are significant at thisaltitude level are SMR 489 GHz (> 2 ppmvdecade−1, signif-icance level up to 6.4) and MIPAS-Bologna V5R MA (sig-nificance level up to 3.2).

In the tropics (Fig. 12), larger drifts are foundthan in the Antarctic, especially at 0.1 hPa. Here, mostdrifts are significant. Significant drifts are found for theMIPAS-Bologna V5R NOM, MIPAS-Bologna V5R MA,MIPAS-ESA V5R, MIPAS-IMKIAA V5R NOM, MIPAS-Oxford V5R NOM and SMR 489 GHz data sets. Forexample, for MIPAS-Bologna V5R NOM and MIPAS-Bologna V5R MA drift (significance level up to 6.5) in com-parison to most other satellite data sets is found. For MIPAS-Bologna V5R NOM this is also the case at 3 hPa (signifi-cance level up to 9.8). Large negative drifts are found forGOMOS (>−2.5 ppmvdecade−1, significance level up to3.9) compared to most data sets. Also for SMR 489 GHzsignificant positive drifts (up to ∼ 1 ppmvdecade−1, signif-icance level up to 8.5) for almost all data sets are found at3 hPa. Good consistency is found among the MIPAS datasets. The drifts are low and in most cases not significant.An exception here is MIPAS-Oxford V5R NOM (∼ 0.6–1 ppmvdecade−1, significance level up to 9.8). For the trop-ics the best agreement among the data sets is found at 10 hPa.In most cases the drift is not significant and in cases wherethe drift is significant the drifts are relatively low with 0.2–0.4 ppmvdecade−1. Larger drifts are found at this altitudefor GOMOS (up to −3 ppmvdecade−1) and HIRDLS (up to−2 ppmvdecade−1). For GOMOS the drifts are significant inmost cases (significance level up to 4.3), while this is not thecase for HIRDLS.

At 80 hPa a wide variety is found. Some data sets showpositive drift, some negative. In some cases the drift is sig-nificant and in other cases not. For example, a positivedrift (2 ppmvdecade−1) relative to almost all data sets isfound for MIPAS-Bologna V5R NOM (significance levelup to 6.4). For the HIRDLS data set a significant posi-tive drift (also ∼ 2 ppmvdecade−1) is found compared to

MIPAS-IMKIAA V5R NOM, MIPAS-IMKIAA-V5R MAand MIPAS-Oxford V5R NOM (significance level 2.0–4.6).A large drift (> 3 ppmvdecade−1) at this altitude level isfound for MIPAS-ESA V5R MA compared to MIPAS-IMKIAA V5R NOM (significance level 4.8). Also theMIPAS-Oxford V5R NOM shows significant drifts com-pared to a number of data sets.

The patterns of the estimated drifts in the Northern Hemi-sphere mid-latitudes shown in Fig. 13 are quite similar tothe drifts in the tropics and Antarctica. However, the es-timated change in ppmvdecade−1 seems to be somewhatlower in the mid-latitudes than in the tropics or Antarctic.The highest variety is again found at 0.1 hPa. Similar tothe tropics significant drifts are found, e.g., for the MIPAS-Bologna V5R NOM and MIPAS-Bologna V5R MA (up to−2 ppmvdecade−1, significance level up to 3.9) data setsrelative to the SMR 489 GHz data set. At 3 hPa, for mostdata sets the drifts are small and/or not significant. Signif-icant negative drifts are found for both ACE-FTS data setsand for SMR 489 GHz. For SMR 489 GHz drift is found rel-ative to most other data sets which is also in most cases sig-nificant. At 10 hPa HIRDLS shows pronounced drifts com-pared to the other data sets. However, these drifts are notsignificant except for the comparison with MLS (drift of3 ppmvdecade−1, significance level 2.3). Otherwise for mostdata sets the drifts are small and/or not significant at 10and 80 hPa. Exceptions are HIRDLS (−2 ppmvdecade−1)and MAESTRO (−1 ppmvdecade−1), which show negativedrift at 80 hPa. For HIRDLS in most cases the drift is sig-nificant (significance level up to 4.1), but for MAESTROin most cases not. For MIPAS-Bologna-V5R NOM signifi-cant positive drifts are found for all instruments that are inmost cases around 0.2–0.4 ppmvdecade−1, but higher com-pared to HIRDLS (significance level 4.1), MAESTRO (sig-nificance level 2.2), SCIAMACHY limb (significance level10.6) and SCIAMACHY solar OEM (significance level 6.6).Other data sets for which drifts are found compared to mostother data sets are SCIAMACHY limb, SCIAMACHY so-lar Onion and SMR 489 GHz.

5 Summary and conclusions

In the framework of the second SPARC water vapour as-sessment, time series of stratospheric and lower mesosphericwater vapour derived from satellite observations were com-pared. The comparison results presented comprise 33 datasets from 15 satellite instruments. These comparisons pro-vide a comprehensive overview of the typical uncertaintiesin the observational database that should be considered inthe future in observational and modelling studies addressingstratospheric and lower mesospheric water vapour variabilityand trends.

The time series comparison was performed for three lat-itude bands: the Antarctic (80–70◦ S), the tropics (15◦ S–



76−25 66−19 58−14 66−19 58−14 66−19

2.6

66−19 58−14

2.6

66−19 58−14 72−24 76−24

2.9

86−25

3.3

78−20 86−25 78−20 86−25

2.9

86−25 78−20 86−25 78−20 123−44

6.0

124−42

47−34

87−69 86−82

3.6

87−69

2.0

86−82

6.5

87−83

4.1

87−69

5.7

87−83

2.8

87−69

4.8

87−83

6.4

87−80

85−68

4.0

87−70 85−68

6.5

87−69

5.1

87−70

4.8

87−69

3.5

87−70

4.3

87−70

6.3

87−68

2.6

85−68 86−82

6.3

86−82 85−68

7.4

86−82 85−68

2.2

86−82

3.0

84−79

2.8

85−68

2.2

87−69 87−70

2.3

87−69 87−70 87−70

2.4

87−68

9.1

86−82 85−68

7.4

86−82 85−68

2.7

86−82

3.8

84−79

6.2

87−69 87−83 87−69

6.3

87−83 87−80

2.9

87−69 87−70 87−70

4.4

87−68

2.2

87−69

3.3

87−83 87−80

87−70 87−68

6.6

120−118

Data set number

76−25

3.9

60−5

66−19

66−19

66−19

66−19

58−14

4.1

66−19

58−14

72−24

2.4

76−24

72−6

86−25

78−20

86−25

78−20

86−25

86−25

78−20

3.3

86−25

78−20

2.1

123−44

8.8

124−42

3.6

62−24

3.7

61−21

3.6

62−24

3.9

61−21

3.5

62−24

3.6

62−24

3.9

61−21

3.3

62−24

3.9

61−21

2.2

82−27

2.3

106−37

164−100

3.0

47−34

2.8

87−69

4.7

86−82

87−69

5.9

86−82

4.5

87−83

87−69

9.8

87−83

2.3

87−69

5.1

87−83

2.1

87−80

85−68

2.2

87−70

85−68

87−69

87−70

5.4

87−69

87−70

87−70

4.5

87−68

2.5

85−68

2.7

86−82

86−82

85−68

8.6

86−82

85−68

86−82

3.4

84−79

3.6

85−68

87−69

87−70

7.6

87−69

87−70

2.5

87−70

3.6

87−68

86−82

2.5

85−68

4.9

86−82

85−68

86−82

5.0

84−79

87−69

5.8

87−83

87−69

87−83

5.1

87−80

6.4

87−69

87−70

87−70

4.3

87−68

6.2

87−69

5.7

87−83

6.5

87−80

87−70

3.9

87−68

8.5

120−118 44−31

1 2 4 9 10 12 13 14 16 17 19 20 21 32

Data set number

32

23

21

20

19

17

16

14

13

12

10

9

4

3

2

1D

ata

set num

ber

1 2 3 4 9 10 12 13 14 16 17 19 20 21 23 32

32

21

20

19

17

16

14

13

12

10

9

4

2

1

Data

set num

ber

76−25 60−5 36−13 66−19 58−14 66−19 58−14

2.3

66−19 66−19 58−14

2.2

66−19 58−14

2.3

72−24

2.6

76−24

72−6 36−13 86−25 78−20 86−25 78−20 86−25 86−25 78−20 86−25 78−20 123−44 124−42

3.7

62−25

4.5

61−22

3.7

62−25

4.5

61−22

3.7

62−25

4.0

62−25

4.3

61−22

3.9

62−25

4.4

61−22

2.4

82−28 106−38

2.9

135−90 47−34

38−34 38−21 37−33 38−21 37−33 38−34 38−21 38−34 38−21 38−39

2.4

38−36

87−69

2.8

86−82 87−69

3.2

86−82

3.4

87−83 87−69

3.9

87−83 87−69 87−83 87−80

2.8

85−68 87−70

3.6

85−68

5.1

87−69 87−70 87−69 87−70

2.6

87−70

2.6

87−68

3.1

85−68 86−82 86−82 85−68

5.8

86−82

2.4

85−68 86−82

2.3

84−79

4.1

85−68

5.8

87−69 87−70 87−69 87−70

2.5

87−70

2.3

87−68

86−82

2.9

85−68

4.5

86−82

2.7

85−68 86−82

3.3

84−79

5.2

87−69

4.8

87−83

4.4

87−69

2.8

87−83

4.2

87−80

87−69 87−70 87−70

2.8

87−68

87−69

2.9

87−83 87−80

87−70

2.1

87−68

120−118

2.0

44−31

76−25

60−5

36−13

76−24

66−19

58−14

2.3

66−19

66−19

66−19

58−14

66−19

72−24

76−25

76−25

2.9

72−6

36−13

126−37

2.9

86−24

78−19

2.6

86−24

2.9

86−24

86−24

78−19

86−24

123−42

96−30

127−43

72−6

2.1

62−25

61−22

62−25

62−25

62−25

61−22

62−25

82−28

106−39

106−39

4.1

105−51

39−30

47−35

36−13

4.2

38−34

38−20

37−33

37−32

3.3

38−34

2.0

38−21

4.6

38−34

38−39

38−39

38−39

86−22

78−17

86−22

86−22

86−22

78−17

86−22

122−36

96−28

126−37

2.2

87−68

2.2

86−82

3.7

86−81

6.0

87−83

3.3

87−69

6.4

87−83

5.5

87−83

4.9

87−83

5.8

87−83

85−67

85−67

87−68

87−69

2.6

87−68

87−69

87−69

87−69

86−81

4.4

86−82

85−68

4.9

86−82

4.1

86−82

3.2

86−82

4.0

86−82

4.8

55−8

55−8

55−8

55−8

4.5

86−81

85−68

5.2

86−81

2.8

86−81

2.1

86−81

3.3

86−81

2.5

87−69

5.1

87−83

87−83

2.4

87−83

87−83

3.9

87−69

87−70

87−70

87−70

3.4

87−83

4.3

87−83

2.9

87−83

92−93

124−122 44−21 116−117

1 2 3 4 5 9 10 12 13 14 16 17 19 20 21 23 32

31

25

23

21

19

17

16

14

13

12

10

9

7

5

4

3

2

1

Data

set num

ber

1 2 3 4 5 7 9 10 12 13 14 16 17 19 21 23 25 31

Data set number

32

23

21

20

19

17

16

14

13

12

10

9

5

4

3

2

1

Data

set num

ber

−3 −2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 2 3


1: ACE−FTS v2.2

2: ACE−FTS v3.5

3: GOMOS

4: HALOE

5: HIRDLS

7: MAESTRO





14: MIPAS−ESA V7R





21: MLS

23: SAGE II

25: SCIAMACHY limb

31: SMR 544 GHz

32: SMR 489 GHz

(a) 0.1 hPa

(b) 3 hPa(c) 10 hPa

(d) 80 hPa

Figure 12. As Fig. 11, but here for the tropics, i.e. between 15◦ S and 15◦ N.

15◦ N) and the Northern Hemisphere mid-latitudes (50◦–60◦ N) at four altitudes levels (0.1, 3, 10, 80 hPa) coveringthe stratosphere and lower mesosphere. The combined tem-poral coverage of observations from the 15 satellite instru-ments allows consideration of the time period 1986–2014. In

addition to the qualitative comparison of the time series, aquantitative comparison was provided based on the spread,correlation and drift between the individual time series.

The qualitative time series comparison shows that thelargest differences between the de-seasonalised time series



68−47 68−36 68−48 68−39 68−48 68−48 68−39 68−48 68−39 72−54

2.5

78−53

84−57

2.5

84−45 84−58 84−48 84−58 84−58 84−48 84−58 84−48 123−92

3.7

123−87

43−12

87−66

2.7

87−82 87−69 87−82

4.5

87−82

2.3

87−69

3.5

87−82 87−69 87−82

3.9

84−77

87−66 87−67 87−66

3.1

87−66

2.6

87−67

2.6

87−66 87−67 87−67

3.0

67−63

87−69

3.9

87−83 87−83 87−69 87−83 87−69 87−83 84−78

87−69 87−69 87−70 87−69 87−70 87−70

2.3

76−66

4.9

87−83 87−69

3.4

87−83 87−69 87−83

2.8

84−78

87−69 87−83 87−69

2.0

87−83 84−78

87−69 87−70 87−70 76−66

87−69 87−83 84−78

87−70 76−66

4.3

120−113

Data set number

78−57

6.168−47

2.168−36

9.068−48

4.868−39

4.868−48

3.068−48

3.368−39

8.868−48

2.268−39

2.372−54

76−29

76−29

3.378−53

6.484−57

84−45

7.284−58

4.584−48

3.284−58

3.384−58

2.584−48

6.084−58

84−48

5.6123−92

86−32

86−32

7.8123−87

86−9

2.0166−74

43−12

87−66

87−82

87−69

8.487−82

4.087−82

87−69

87−82

2.087−69

6.987−82

4.776−38

6.676−38

3.284−77

87−66

2.287−67

2.787−66

87−66

87−67

87−66

87−67

2.787−67

62−31

62−31

4.767−63

87−69

12.587−83

4.987−83

87−69

87−83

87−69

8.187−83

5.276−39

6.976−39

2.584−78

5.287−69

3.687−69

2.887−70

87−69

3.587−70

5.787−70

62−31

3.062−31

2.976−66

87−83

87−69

8.787−83

87−69

87−83

76−39

3.076−39

4.984−78

87−69

4.587−83

87−69

87−83

76−39

3.276−39

5.084−78

87−69

87−70

2.687−70

62−31

2.462−31

4.376−66

87−69

6.787−83

5.376−39

6.576−39

2.484−78

87−70

62−31

2.362−31

4.776−66

84−43

3.884−43

7.1120−113

2.237−12

36−13

36−13

39−15

37−8

45−24 37−14

108−55

7.3108−51

6.8108−51

1 2 4 9 10 12 13 14 16 17 19 20 21 32

Data set number

32

28

27

24

23

22

21

20

19

17

16

14

13

12

10

9

4

2

1D

ata

set num

ber

1 2 4 9 10 12 13 14 16 17 19 20 21 22 23 24 27 28 32

32

21

20

19

17

16

14

13

12

10

9

4

2

1

Data

set num

ber

78−57 37−28 68−47 68−36 68−48 68−39 68−48 68−48 68−39 68−48 68−39 72−542.7

76−293.2

76−29 78−53

38−30 84−57 84−45 84−58 84−48 84−58 84−58 84−48 84−58 84−48 123−92 86−32 86−32 123−87

86−9 166−74 43−12

38−33 38−18 38−34 38−21 38−34 38−34 38−21 38−34 38−212.3

38−39

87−66 87−695.0

87−825.9

87−82 87−69 87−82 87−694.0

87−822.6

76−383.5

76−382.2

84−77

2.187−66 87−67

3.187−66

3.387−66 87−67 87−66 87−67

2.687−67 62−31 62−31

2.367−63

87−694.2

87−834.7

87−83 87−692.1

87−83 87−693.2

87−83 76−392.5

76−392.3

84−78

87−69 87−69 87−70 87−69 87−70 87−70 62−31 62−313.1

76−66

3.887−83 87−69

3.087−83 87−69 87−83 76−39 76−39

2.684−78

2.687−69

4.687−83 87−69 87−83 76−39 76−39

4.384−78

87−69 87−70 87−70 62−31 62−313.0

76−66

87−693.3

87−832.4

76−392.9

76−39 84−78

87−70 62−31 62−312.5

76−66

84−43 84−43 120−113

87−20 37−12 36−13 36−13 39−15

37−8 45−24

37−14

108−552.6

108−51

3.2108−51

4.178−57

2.137−28

78−50

4.168−47

68−36

68−47

68−47

68−48

2.068−36

68−48

72−54

3.978−54

76−29

3.576−29

78−58

2.538−30

3.4129−73

2.084−57

84−45

84−57

2.184−57

3.284−58

3.180−43

84−58

3.8123−92

7.994−64

86−32

5.186−32

4.6129−92

86−9

125−42

43−15

37−27

4.138−33

2.838−18

2.538−33

2.538−33

2.438−34

38−20

2.638−34

2.838−39

38−39

2.338−39

2.284−47

84−36

84−47

84−47

84−48

80−35

84−48

2.6123−70

94−54

83−25

83−25

2.4129−69

3.187−66

6.387−82

6.787−82

7.887−82

5.687−63

4.687−82

6.987−82

10.687−76

3.376−38

6.676−38

5.187−82

87−66

87−66

2.187−66

3.287−62

87−66

87−67

5.387−61

62−31

62−31

2.187−67

87−82

2.987−82

3.087−63

2.487−82

87−82

9.487−76

76−38

3.676−38

2.787−82

2.987−82

2.987−63

3.187−82

87−82

9.487−76

76−38

4.276−38

2.587−82

87−63

6.787−83

3.787−83

8.287−77

76−39

4.176−39

87−83

3.187−63

2.987−64

3.187−59

2.862−31

62−31

87−64

2.887−83

9.887−77

76−39

4.776−39

3.287−83

8.592−87

84−43

4.284−43

3.4124−121

87−20

37−12

2.136−12

36−13

36−13

39−16

37−8

2.745−29 38−15

108−50

108−50

2.7116−108

2.8108−55

2.4108−55

3.8108−55

1 2 4 5 9 10 12 13 14 16 17 19 20 21 22 23 24 27 28 32

3128272524232221191716141210

975421

Data

set num

ber

1 2 4 5 7 9 10 12 14 16 17 19 21 22 23 24 25 27

28 31Data set number

32282724232221201917161413121095421

Data

set num

ber

−3 −2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 2 3


1: ACE−FTS v2.2

2: ACE−FTS v3.5

4: HALOE

5: HIRDLS

7: MAESTRO





14: MIPAS−ESA V7R





21: MLS

22: POAM III

23: SAGE II

24: SAGE III

25: SCIAMACHY limb

27: SCIAMACHY solar OEM

28: SCIAMACHY solar Onion

31: SMR 544 GHz

32: SMR 489 GHz

(a) 0.1 hPa

(b) 3 hPa(c) 10 hPa

(d) 80 hPa

Figure 13. As Figs. 11 and 12, but here the results for the latitude band between 50◦ and 60◦ N are shown.

are in the Antarctic and in the lower mesosphere (0.1 hPa)and tropopause region (80 hPa). In the stratosphere (3 and10 hPa) and the tropics, good agreement between the satel-lite data sets was found. These differences were quantita-tively confirmed by the correlation assessment, where the

best agreement between the satellite data sets was also foundin the tropics, while in Antarctic and Northern Hemispheremid-latitudes, large spread between the data sets was found.Generally, the lowest correlations between the individualdata sets were found in the Antarctic. In each latitude band



the correlation was lower in the lower stratosphere and lowermesosphere than in the middle stratosphere.

There are multiple factors that give rise to the observeddifferences between the individual data sets. A thorough dis-cussion on this is given in Lossow et al. (2017b). From thisstudy we know that the most important contributions arisefrom differences in temporal and spatial sampling, the influ-ence of clouds or NLTE effects. Other factors include sys-tematic differences, for example calibration problems. How-ever, for the time series comparison we would rank samplingbiases and systematic errors as the most important reasonsfor the differences as was discussed by Toohey et al. (2013)based on trace gas climatologies.

The reason why the largest differences between the datasets are found in the tropopause region, in the lower meso-sphere and in the Antarctic is that these are also the locationswhere the highest variability in water vapour is found. Giventhe limited vertical resolution of the satellite data sets, tro-pospheric influences start to play a role near the tropopause.Sampling differences become more pronounced due to thelarge variability, e.g. due to the fact that the satellite ob-servations are influenced differently by clouds. In the lowermesosphere, diurnal variation becomes more important. Thesatellite data sets do not have the same local time coverage.For example there is the influence of NLTE effects in mostMIPAS data sets except MIPAS-IMKIAA V5R MA, wherethese NLTE effect are explicitly considered. Larger devia-tions in the lower mesosphere occur, e.g. in the case of theMIPAS NOM data sets, which are close to their upper re-trieval limit there, and thus more uncertain.

Less agreement between the data sets was found for theAntarctic, especially in the lower stratosphere in winterand spring when dehydration occurs. Large differences be-tween the data sets were found in both the absolute and de-seasonalised data. In the absolute data, these differences areprimarily caused by differences in the influence of clouds onthe measurements. However, sampling biases can also playa role. In the de-seasonalised data some differences betweenthe data sets could be related to the de-seasonalisation ap-proach used in our study (e.g. POAM III). Since the dehy-dration is more a seasonal phenomenon, and accordingly isless characterised by a sinusoidal behaviour, the usage of si-nusoidal functions for the de-seasonalisation is not optimal.Instead, the average approach (see Sect. 3.1) would be moresuitable for de-seasonalisation in this region.

In addition to the assessment of the spread and correla-tions, the drifts between the individual data sets were also as-sessed, which indicates if the longer-term variations (drifts)of two time series are the same or not. From the driftcomparison we found that the drift patterns are quite sim-ilar for the three latitude bands considered. The drifts arehighest at the highest and lowest considered altitude lev-els (0.1 and 80 hPa). The majority of significant drifts werefound in the tropics (the latitude region with the lowestspread/variability), which makes the drift detection consid-

erably easier. Further, it is possible that some of the drifts(especially for the low-density samplers) are caused by sam-pling biases (Damadeo et al., 2018). The same drift approachas used here has been used by Lossow et al. (2018) to cal-culate drifts from profile-to-profile comparisons (using coin-cident data). However, no statistically significant differencewas found between the two sets of drifts in 95 % of the com-parisons.

Further, from the drift assessment we found that theMIPAS data sets show positive drifts relative to the ACE-FTS data sets in the Antarctic and Northern Hemisphere mid-latitudes at 3 hPa. Interestingly, no drifts of MIPAS relativeto ACE-FTS are found in the tropics. The reason for this iscurrently not understood. The drifts found in the MIPAS datasets are consistent with the time dependence unaccounted forin the correction coefficient for the non-linearity in the de-tector response function used in the data sets based on cali-bration version 5 (Walker and Stiller, 2018). Some improve-ment is seen in the MIPAS ESA V7R NOM data set, wherea time dependence of the correction coefficient is imple-mented, though not at all altitudes. Additionally, even driftsamong the different MIPAS data sets were found. This mightbe related to the different retrieval choices (as well as to theusage of different micro-windows) by the different proces-sors and to sampling differences between the NOM and MAobservations. Further, from the drift comparison, we foundthat the SMR 489 GHz data set shows a significant drift rela-tive to the other data sets, except at around 10 hPa. The driftsof the SMR 489 GHz data set are largest at around 50 and0.5 hPa with approximately 1.5 and > 2 ppmvdecade−1, re-spectively depending on the data set used for comparison.

Further, within this assessment study we encountered thefollowing difficulties in our analyses using the HIRDLS,GOMOS and MAESTRO data sets. The GOMOS time seriesexhibit larger scatter from month to month (coverage only inthe tropics for de-seasonalised data here) despite extendedscreening (Walker and Stiller, 2018), resulting in low corre-lations to the other data sets and pronounced negative driftsat 10 and 3 hPa. The quality of the HIRDLS data set dete-riorates towards 10 hPa, resulting in low correlations, largeranomalies and larger drifts. However, the drifts were mostlynot statistically significant. It should be noted here that inaddition to correcting for the effects of the obstruction inthe optics, changes in the calibration were made within theHIRDLS mission (Gille et al., 2008, 2012). This change incalibration may also have an influence on the drift estimates.The MAESTRO data set encounters large uncertainty (noise)at 80 hPa (in the correlations and drifts) which is related tothe vicinity to the uppermost limit of these retrievals. Similarbehaviour is also found for the SCIAMACHY limb and theSMR 544 GHz data sets.

Nevertheless, although the water vapour data sets havebeen thoroughly assessed in this study it is difficult or ratherimpossible to decide which data set is most suitable for fu-ture modelling and observational studies. This can only be



answered with respect to the specific scientific application towhich the data set is intended to be applied. For future stud-ies, e.g. on water vapour trends, we can state that the data setsthat provide the longest measurement record with high spa-tial and temporal coverage have an advantage over the oneswhich provide only observations in specific latitude bandsand/or altitude regions. For data sets that show drift relativeto other data sets (e.g. SMR 489 GHz), a drift has to be takeninto account, and data sets that are simply too short (less than1 year; e.g. ILAS-II and SMILES) cannot be used for trendstudies at all. Thus, from our assessment we find that mostdata sets can be considered in future observational and mod-elling studies, e.g addressing stratospheric and lower meso-spheric water vapour variability and trends, if data set spe-cific characteristics (e.g. an instrument drift) and restrictions(e.g. spatial and temporal coverage) are taken into account.

Data availability. Data are available upon request.

The Supplement related to this article is available onlineat https://doi.org/10.5194/amt-11-4435-2018-supplement.

Author contributions. The study was designed by JU, FK and SLwith contributions from the WAVAS-II core members GS, KR,JCG, MK, GEN, WGR and KAW. FK wrote the manuscript andSL performed the analyses and contributed to the writing of themanuscript. JU performed the first version of the analyses. GShelped with the selection of results to be presented in the paper.KHR contributed to the discussion of the results. Satellite data usedin this study were provided by GPS, JPB, RD, PE, MGC, JG, YK,MK, SN, PR, WGR, AR, CS, KAW and KW. Valuable commentson the manuscript were provided by GPS, KHR, RPD, PR, MG, SN,CS, JG, AR and KW.

Competing interests. The authors declare that they have no com-peting interests.

Special issue statement. This article is part of the special issue“Water vapour in the upper troposphere and middle atmosphere: aWCRP/SPARC satellite data quality assessment including biases,variability, and drifts (ACP/AMT/ESSD inter-journal SI)”. It doesnot belong to a conference.

Acknowledgements. The Atmospheric Chemistry Experiment(ACE), also known as SCISAT, is a Canadian-led mission mainlysupported by the Canadian Space Agency and the Natural Sciencesand Engineering Research Council of Canada. We would like tothank the European Space Agency (ESA) for making the MIPASlevel-1b data set available. MLS data were obtained from theNASA Goddard Earth Sciences and Information Center. Work atthe Jet Propulsion Laboratory, California Institute of Technology,was done under contract with the National Aeronautics and SpaceAdministration. SCIAMACHY spectral data have been providedby ESA. The work on the SCIAMACHY water vapour dataproducts has been funded by DLR (German Aerospace Center)and the University of Bremen. The SCIAMACHY limb watervapour data set v3.01 is a result of the DFG (German ResearchCouncil) Research Unit “Stratospheric Change and its Role forClimate Prediction” (SHARP) and the ESA SPIN (ESA SPARCInitiative) project and were partly calculated using resources of theGerman HLRN (High-Performance Computer Center North). Wewould like to thank Mark Hervig for providing the SOFIE data andTakafumi Sugita for providing the ILAS-II data. We acknowledgethe HALOE science team and the many members of the HALOEproject for producing and characterising the high-quality HALOEdata set. Further, we would like to thank Ellis Remsberg for valu-able comments on the manuscript. Stefan Lossow was funded bythe SHARP project under contract STI 210/9-2. We want to expressour gratitude to SPARC and WCRP (World Climate ResearchProgramme) for their guidance, sponsorship and support of theWAVAS-II programme. We acknowledge support by DeutscheForschungsgemeinschaft and Open Access Publishing Fund ofKarlsruhe Institute of Technology.

The article processing charges for this open-accesspublication were covered by a ResearchCentre of the Helmholtz Association.

Edited by: Stefan BuehlerReviewed by: Hugh C. Pumphrey and one anonymous referee

References

Avery, M. A., Davis, S. M., Rosenlof, K. H., Ye, H., and Dessler,A. E.: Large anomalies in lower stratospheric water vapour andice during the 2015-2016 El Niño, Nat. Geosci., 10, 405–409,https://doi.org/10.1038/ngeo2961, 2017.

Bates, D. R. and Nicolet, M.: The photochemistry of at-mospheric water vapor, J. Geophys. Res., 55, 301–327,https://doi.org/10.1029/JZ055i003p00301, 1950.

Brewer, A. W.: Evidence for a world circulation provided bythe measurements of helium and water vapour distributionin the stratosphere, Q. J. Roy. Meteor. Soc., 75, 351–363,https://doi.org/10.1002/qj.49707532603, 1949.

Brinkop, S., Dameris, M., Jöckel, P., Garny, H., Lossow, S., andStiller, G.: The millennium water vapour drop in chemistry–climate model simulations, Atmos. Chem. Phys., 16, 8125–8140,https://doi.org/10.5194/acp-16-8125-2016, 2016.

Damadeo, R. P., Zawodny, J. M., Remsberg, E. E., and Walker, K.A.: The impact of nonuniform sampling on stratospheric ozonetrends derived from occultation instruments, Atmos. Chem.


https://doi.org/10.5194/amt-11-4435-2018-supplement

https://doi.org/10.1038/ngeo2961

https://doi.org/10.1029/JZ055i003p00301

https://doi.org/10.1002/qj.49707532603

https://doi.org/10.5194/acp-16-8125-2016


Phys., 18, 535–554, https://doi.org/10.5194/acp-18-535-2018,2018.

Dessler, A. E., Schoeberl, M. R., Wang, T., Davis, S. M., Rosenlof,K. H., and Vernier, J.-P.: Variations of stratospheric water vaporover the past three decades, J. Geophys. Res., 119, 12588–12598,https://doi.org/10.1002/2014JD021712, 2014.

Dinelli, B. M., Arnone, E., Brizzi, G., Carlotti, M., Castelli, E.,Magnani, L., Papandrea, E., Prevedelli, M., and Ridolfi, M.:The MIPAS2D database of MIPAS/ENVISAT measurements re-trieved with a multi-target 2-dimensional tomographic approach,Atmos. Meas. Tech., 3, 355–374, https://doi.org/10.5194/amt-3-355-2010, 2010.

Fahey, D. W., Kelly, K. K., Kawa, S. R., Tuck, A. F., andLoewenstein, M.: Observations of denitrification and dehydra-tion in the winter polar stratospheres, Nature, 344, 321–324,https://doi.org/10.1038/344321a0, 1990.

Freie Universität Berlin (http://www.geo.fu-berlin.de/met/ag/strat/produkte/qbo/qbo.dat, last access: 20 July 2018.

Fueglistaler, S. and Haynes, P. H.: Control of interannual andlonger-term variability of stratospheric water vapor, J. Geophys.Res., 110, D24108, https://doi.org/10.1029/2005JD006019,2005.

Gettelman, A., Hegglin, M. I., Son, S., Kim, J., Fujiwara, M.,Birner, T., Kremser, S., Rex, M., Añel, J. A., Akiyoshi, H.,Austin, J., Bekki, S., Braesicke, P., Brühl, C., Butchart, N., Chip-perfield, M., Dameris, M., Dhomse, S., Garny, H., Hardiman,S. C., Jöckel, P., Kinnison, D. E., Lamarque, J. F., Mancini, E.,Marchand, M., Michou, M., Morgenstern, O., Pawson, S., Pitari,G., Plummer, D., Pyle, J. A., Rozanov, E., Scinocca, J., Shep-herd, T. G., Shibata, K., Smale, D., Teyssèdre, H., and Tian,W.: Multimodel assessment of the upper troposphere and lowerstratosphere: Tropics and global trends, J. Geophys. Res., 115,D00M08, https://doi.org/10.1029/2009JD013638, 2010.

Gille, J., Barnett, J., Arter, P., Barker, M., Bernath, P., Boone, C.,Cavanaugh, C., Chow, J., Coffey, M., Craft, J., Craig, C., Di-als, M., Dean, V., Eden, T., Edwards, D. P., Francis, G., Halvor-son, C., Harvey, L., Hepplewhite, C., Khosravi, R., Kinnison, D.,Krinsky, C., Lambert, A., Lee, H., Lyjak, L., Loh, J., Mankin,W., Massie, S., McInerney, J., Moorhouse, J., Nardi, B., Pack-man, D., Randall, C., Reburn, J., Rudolf, W., Schwartz, M., Ser-afin, J., Stone, K., Torpy, B., Walker, K., Waterfall, A., Watkins,R., Whitney, J., Woodard, D., and Young, G.: High ResolutionDynamics Limb Sounder: Experiment overview, recovery, andvalidation of initial temperature data, J. Geophys. Res., 113,D16S43, https://doi.org/10.1029/2007JD008824, 2008.

Gille, J., Cavanough, C., Halvorson, C., Hartsough, C., Nardi, B.,Rivas, M., Khosravi, R., Smith, L., and Francis, G.: Final cor-rection algorithms for HIRDLS version 7 data, Proc. SPIE 8511,Infrared Remote Sensing and Instrumentation XX, 85110K (24October 2012), https://doi.org/10.1117/12.930175, 2012.

Hamilton, K.: Dynamics of the tropical middle atmosphere: A tuto-rial review, Atmos.-Ocean, 36, 319–354, 1998.

Hegglin, M. I., Tegtmeier, S., Anderson, J., Froidevaux, L., Fuller,R., Funke, B., Jones, A., Lingenfelser, G., Lumpe, J., Pendlebury,D., Remsberg, E., Rozanov, A., Toohey, M., Urban, J., von Clar-mann, T., Walker, K. A., Wang, R., and Weigel, K.: SPARC DataInitiative: Comparison of water vapor climatologies from inter-national satellite limb sounders, J. Geophys. Res., 118, 11824,https://doi.org/10.1002/jgrd.50752, 2013.

Hegglin, M. I., Plummer, D. A., Shepherd, T. G., Scinocca, J. F.,Anderson, J., Froidevaux, L., Funke, B., Hurst, D., Rozanov,A., Urban, J., von Clarmann, T., Walker, K. A., Wang, H. J.,Tegtmeier, S., and Weigel, K.: Vertical structure of stratosphericwater vapour trends derived from merged satellite data, Nat.Geosci., 7, 768–776, https://doi.org/10.1038/ngeo2236, 2014.

Hurst, D. F., Oltmans, S. J., Vömel, H., Rosenlof, K. H., Davis,S. M., Ray, E. A., Hall, E. G., and Jordan, A. F.: Strato-spheric water vapor trends over Boulder, Colorado: Analysisof the 30 year Boulder record, J. Geophys. Res., 116, D02306,https://doi.org/10.1029/2010JD015065, 2011.

Jensen, E. J., Toon, O. B., Pfister, L., and Selkirk, H. B.: Dehy-dration of the upper troposphere and lower stratosphere by sub-visible cirrus clouds near the tropical tropopause, Geophys. Res.Lett., 23, 825–828, https://doi.org/10.1029/96GL00722, 1996.

Jones, A., Walker, K. A., Jin, J. J., Taylor, J. R., Boone, C.D., Bernath, P. F., Brohede, S., Manney, G. L., McLeod, S.,Hughes, R., and Daffer, W. H.: Technical Note: A trace gasclimatology derived from the Atmospheric Chemistry Experi-ment Fourier Transform Spectrometer (ACE-FTS) data set, At-mos. Chem. Phys., 12, 5207–5220, https://doi.org/10.5194/acp-12-5207-2012, 2012.

Kawatani, Y., Lee, J. N., and Hamilton, K.: Interannual varia-tions of stratospheric water vapor in MLS observations andclimate model simulations, J. Atmos. Sci., 71, 4072–4085,https://doi.org/10.1175/JAS-D-14-0164.1, 2014.

Kelly, K. K., Tuck, A. F., Murphy, D. M., Proffitt, M. H., Fahey,D. W., Jones, R. L., McKenna, D. S., Loewenstein, M., Podolske,J. R., Strahan, S. E., Ferry, G. V., Chan, K. R., Vedder, J. F.,Gregory, G. L., Hypes, W. D., McCormick, M. P., Browell, E. V.,and Heidt, L. E.: Dehydration in the lower Antarctic stratosphereduring late winter and early spring, 1987, J. Geophys. Res., 94,11317–11357, https://doi.org/10.1029/JD094iD09p11317, 1989.

Khaykin, S. M., Engel, I., Vömel, H., Formanyuk, I. M., Kivi, R.,Korshunov, L. I., Krämer, M., Lykov, A. D., Meier, S., Nae-bert, T., Pitts, M. C., Santee, M. L., Spelten, N., Wienhold,F. G., Yushkov, V. A., and Peter, T.: Arctic stratospheric de-hydration – Part 1: Unprecedented observation of vertical re-distribution of water, Atmos. Chem. Phys., 13, 11503–11517,https://doi.org/10.5194/acp-13-11503-2013, 2013.

Khosrawi, F., Urban, J., Lossow, S., Stiller, G., Weigel, K.,Braesicke, P., Pitts, M. C., Rozanov, A., Burrows, J. P., andMurtagh, D.: Sensitivity of polar stratospheric cloud formation tochanges in water vapour and temperature, Atmos. Chem. Phys.,16, 101–121, https://doi.org/10.5194/acp-16-101-2016, 2016.

Le Texier, H., Solomon, S., and Garcia, R. R.: The role of molec-ular hydrogen and methane oxidation in the water vapour bud-get of the stratosphere, Q. J. Roy. Meteor. Soc., 114, 281–295,https://doi.org/10.1002/qj.49711448002, 1988.

Lossow, S., Garny, H., and Jöckel, P.: An “island” in the strato-sphere – on the enhanced annual variation of water vapourin the middle and upper stratosphere in the southern trop-ics and subtropics, Atmos. Chem. Phys., 17, 11521–11539,https://doi.org/10.5194/acp-17-11521-2017, 2017a.

Lossow, S., Khosrawi, F., Nedoluha, G. E., Azam, F., Bramstedt, K.,Burrows, John. P., Dinelli, B. M., Eriksson, P., Espy, P. J., García-Comas, M., Gille, J. C., Kiefer, M., Noël, S., Raspollini, P., Read,W. G., Rosenlof, K. H., Rozanov, A., Sioris, C. E., Stiller, G. P.,Walker, K. A., and Weigel, K.: The SPARC water vapour assess-


https://doi.org/10.5194/acp-18-535-2018

https://doi.org/10.1002/2014JD021712

https://doi.org/10.5194/amt-3-355-2010

https://doi.org/10.5194/amt-3-355-2010

https://doi.org/10.1038/344321a0



https://doi.org/10.1029/2005JD006019

https://doi.org/10.1029/2009JD013638

https://doi.org/10.1029/2007JD008824

https://doi.org/10.1117/12.930175

https://doi.org/10.1002/jgrd.50752

https://doi.org/10.1038/ngeo2236

https://doi.org/10.1029/2010JD015065

https://doi.org/10.1029/96GL00722

https://doi.org/10.5194/acp-12-5207-2012

https://doi.org/10.5194/acp-12-5207-2012

https://doi.org/10.1175/JAS-D-14-0164.1

https://doi.org/10.1029/JD094iD09p11317

https://doi.org/10.5194/acp-13-11503-2013

https://doi.org/10.5194/acp-16-101-2016

https://doi.org/10.1002/qj.49711448002

https://doi.org/10.5194/acp-17-11521-2017


ment II: comparison of annual, semi-annual and quasi-biennialvariations in stratospheric and lower mesospheric water vapourobserved from satellites, Atmos. Meas. Tech., 10, 1111–1137,https://doi.org/10.5194/amt-10-1111-2017, 2017b.

Lossow, S., Kiefer, M., Walker, K. A., Bertaux, J.-L., Blanot,L., Hervig, M., Russell, J. M., Gille, J. C., Sugita, T., Sioris,C. E., Dinelli, B. M., Papandrea, E., Raspollini, P., García-Comas, M., Stiller, G. P., von Clarmann, T., Dudhia, A., Read,W. G., Nedoluha, G. E., Zawodny, J. M., Weigel, K., Rozanov,A., Azam, F., Bramstedt, K., Noël, S., Urban, J., Eriksson, P.,Murtagh, D. P., Sagawa, H., Kasai, Y., Hurst, D. F., and Rosenlof,K. H.: The SPARC water vapour assessment II: Profile-to-profile comparisons of stratospheric and lower mesospheric wa-ter vapour data sets obtained from satellite, to be submitted toAtmos. Meas. Tech., 2018.

Manney, G. L. and Lawrence, Z. D.: The major stratospheric finalwarming in 2016: dispersal of vortex air and termination of Arc-tic chemical ozone loss, Atmos. Chem. Phys., 16, 15371–15396,https://doi.org/10.5194/acp-16-15371-2016, 2016.

Mote, P. W., Rosenlof, K. H., McIntyre, M. E., Carr, E. S.,Gille, J. C., Holton, J. R., Kinnersley, J. S., Pumphrey, H. C.,Russell, J. M., and Waters, J. W.: An atmospheric taperecorder: The imprint of tropical tropopause temperatures onstratospheric water vapor, J. Geophys. Res., 101, 3989–4006,https://doi.org/10.1029/95JD03422, 1996.

Müller, R., Kunz, A., Hurst, D. F., Rolf, C., Krämer, M.,and Riese, M.: The need for accurate long-term measure-ments of water vapor in the upper troposphere and lowerstratosphere with global coverage, Earths Future, 4, 25–32,https://doi.org/10.1002/2015EF000321, 2016.

Nedoluha, G. E., Bevilacqua, R. M., Hoppel, K. W., Daehler, M.,Shettle, E. P., Hornstein, J. H., Fromm, M. D., Lumpe, J. D.,and Rosenfield, J. E.: POAM III measurements of dehydration inthe Antarctic lower stratosphere, Geophys. Res. Lett., 27, 1683–1686, https://doi.org/10.1029/1999GL011087, 2000.

Nedoluha, G. E., Bevilacqua, R. M., Gomez, R. M., Hicks,B. C., Russell, J. M., and Connor, B. J.: An evaluation oftrends in middle atmospheric water vapor as measured byHALOE, WVMS, and POAM, J. Geophys. Res., 108, 4391,https://doi.org/10.1029/2002JD003332, 2003.

Nedoluha, G. E., Benson, C. M., Hoppel, K. W., Alfred,J., Bevilacqua, R. M., and Drdla, K.: Antarctic dehydra-tion 1998-2003: Polar Ozone and Aerosol MeasurementIII (POAM) measurements and Integrated Microphysics andAerosol Chemistry on Trajectories (IMPACT) results withfour meteorological models, J. Geophys. Res., 112, D07305,https://doi.org/10.1029/2006JD007414, 2007.

Oltmans, S. J. and Hofmann, D. J.: Increase in lower-stratospheric water vapour at a mid-latitude Northern Hemi-sphere site from 1981 to 1994, Nature, 374, 146–149,https://doi.org/10.1038/374146a0, 1995.

Oltmans, S. J., Vömel, H., Hofmann, D. J., Rosenlof, K. H., andKley, D.: The increase in stratospheric water vapor from bal-loonborne, frostpoint hygrometer measurements at Washington,D.C., and Boulder, Colorado, Geophys. Res. Lett., 27, 3453–3456, https://doi.org/10.1029/2000GL012133, 2000.

Pan, L. L., Randel, W. J., Nakajima, H., Massie, S. T., Kan-zawa, H., Sasano, Y., Yokota, T., Sugita, T., Hayashida, S.,and Oshchepkov, S.: Satellite observation of dehydration in

the Arctic Polar stratosphere, Geophys. Res. Lett., 29, 1184,https://doi.org/10.1029/2001GL014147, 2002.

Payne, V. H., Noone, D., Dudhia, A., Piccolo, C., and Grainger,R. G.: Global satellite measurements of HDO and implica-tions for understanding the transport of water vapour intothe stratosphere, Q. J. Roy. Meteor. Soc., 133, 1459–1471,https://doi.org/10.1002/qj.127, 2007.

Ploeger, F., Günther, G., Konopka, P., Fueglistaler, S., Müller, R.,Hoppe, C., Kunz, A., Spang, R., Grooß, J.-U., and Riese, M.:Horizontal water vapor transport in the lower stratosphere fromsubtropics to high latitudes during boreal summer, J. Geophys.Res., 118, 8111–8127, https://doi.org/10.1002/jgrd.50636, 2013.

Pumphrey, H. C. and Harwood, R. S.: Water vapour and ozone in themesosphere as measured by UARS MLS, Geophys. Res. Lett.,24, 1399–1402, https://doi.org/10.1029/97GL01158, 1997.

Randel, W. J., Wu, F., Oltmans, S. J., Rosenlof, K., andNedoluha, G. E.: Interannual changes of stratospheric watervapor and correlations with tropical tropopause temperatures,J. Atmos. Sci., 61, 2133–2148, https://doi.org/10.1175/1520-0469(2004)061<2133:ICOSWV>2.0.CO;2, 2004.

Randel, W. J., Wu, F., Vömel, H., Nedoluha, G. E., andForster, P.: Decreases in stratospheric water vapor after2001: Links to changes in the tropical tropopause and theBrewer-Dobson circulation, J. Geophys. Res., 111, D12312,https://doi.org/10.1029/2005JD006744, 2006.

Raspollini, P., Carli, B., Carlotti, M., Ceccherini, S., Dehn, A.,Dinelli, B. M., Dudhia, A., Flaud, J.-M., López-Puertas, M.,Niro, F., Remedios, J. J., Ridolfi, M., Sembhi, H., Sgheri, L.,and von Clarmann, T.: Ten years of MIPAS measurements withESA Level 2 processor V6 – Part 1: Retrieval algorithm and di-agnostics of the products, Atmos. Meas. Tech., 6, 2419–2439,https://doi.org/10.5194/amt-6-2419-2013, 2013.

Read, W. G., Wu, D. L., Waters, J. W., and Pumphrey, H. C.: De-hydration in the tropical tropopause layer: Implications fromthe UARS Microwave Limb Sounder, J. Geophys. Res., 109,D06110, https://doi.org/10.1029/2003JD004056, 2004.

Remsberg, E.: Observed seasonal to decadal scale responses inmesospheric water vapor, J. Geophys. Res., 115, D06306,https://doi.org/10.1029/2009JD012904, 2010.

Remsberg, E., Russell, J. M., Gordley, L. L., Gille, J. C., andBailey, P. L.: Implications of the stratospheric water vapor dis-tribution as determined from the Nimbus 7 LIMS experiment,J. Atmos. Sci., 41, 2934–2948, https://doi.org/10.1175/1520-0469(1984)041<2934:IOTSWV>2.0.CO;2, 1984.

Remsberg, E., Damadeo, R., Natarajan, M., and Bhatt, P.:Observed responseses of mesospheric water vapor to so-lar cycle and dynamical forcings, J. Geophys. Res., 123,https://doi.org/10.1002/2017JD028029, 2018.

Riese, M., Ploeger, F., Rap, A., Vogel, B., Konopka, P.,Dameris, M., and Forster, P.: Impact of uncertainties inatmospheric mixing on simulated UTLS composition andrelated radiative effects, J. Geophys. Res., 117, D16305,https://doi.org/10.1029/2012JD017751, 2012.

Rosenlof, K. H., Chiou, E.-W., Chu, W. P., Johnson, D. G., Kelly,K. K., Michelsen, H. A., Nedoluha, G. E., Remsberg, E. E.,Toon, G. C., and McCormick, M. P.: Stratospheric water va-por increases over the past half-century, Geophys. Res. Lett., 28,1195–1198, https://doi.org/10.1029/2000GL012502, 2001.


https://doi.org/10.5194/amt-10-1111-2017

https://doi.org/10.5194/acp-16-15371-2016

https://doi.org/10.1029/95JD03422

https://doi.org/10.1002/2015EF000321

https://doi.org/10.1029/1999GL011087

https://doi.org/10.1029/2002JD003332

https://doi.org/10.1029/2006JD007414

https://doi.org/10.1038/374146a0

https://doi.org/10.1029/2000GL012133

https://doi.org/10.1029/2001GL014147

https://doi.org/10.1002/qj.127


https://doi.org/10.1029/97GL01158

https://doi.org/10.1175/1520-0469(2004)061<2133:ICOSWV>2.0.CO;2

https://doi.org/10.1175/1520-0469(2004)061<2133:ICOSWV>2.0.CO;2

https://doi.org/10.1029/2005JD006744

https://doi.org/10.5194/amt-6-2419-2013

https://doi.org/10.1029/2003JD004056

https://doi.org/10.1029/2009JD012904

https://doi.org/10.1175/1520-0469(1984)041<2934:IOTSWV>2.0.CO;2

https://doi.org/10.1175/1520-0469(1984)041<2934:IOTSWV>2.0.CO;2

https://doi.org/10.1002/2017JD028029

https://doi.org/10.1029/2012JD017751

https://doi.org/10.1029/2000GL012502


Scherer, M., Vömel, H., Fueglistaler, S., Oltmans, S. J., and Stae-helin, J.: Trends and variability of midlatitude stratosphericwater vapour deduced from the re-evaluated Boulder bal-loon series and HALOE, Atmos. Chem. Phys., 8, 1391–1402,https://doi.org/10.5194/acp-8-1391-2008, 2008.

Schiller, C., Grooß, J.-U., Konopka, P., Plöger, F., Silva dosSantos, F. H., and Spelten, N.: Hydration and dehydration atthe tropical tropopause, Atmos. Chem. Phys., 9, 9647–9660,https://doi.org/10.5194/acp-9-9647-2009, 2009.

Schoeberl, M. R., Douglass, A. R., Newman, P. A., Lait, L. R.,Lary, D., Waters, J., Livesey, N., Froidevaux, L., Lambert, A.,Read, W., Filipiak, M. J., and Pumphrey, H. C.: QBO and an-nual cycle variations in tropical lower stratosphere trace gasesfrom HALOE and Aura MLS observations, J. Geophys. Res.,113, D05301, https://doi.org/10.1029/2007JD008678, 2008.

Seele, C. and Hartogh, P.: Water vapor of the polar mid-dle atmosphere: Annual variation and summer meso-sphere conditions as observed by ground-based mi-crowave spectroscopy, Geophys. Res. Lett., 26, 1517–1520,https://doi.org/10.1029/1999GL900315, 1999.

Seinfeld, J. H. and Pandis, S. N.: Atmospheric chemistry andphysics: From air pollution to climate change, John Wiley andSons, New York, 2006.

Solomon, S.: Stratospheric ozone depletion: A reviewof concepts and history, Rev. Geophys., 37, 275–316,https://doi.org/10.1029/1999RG900008, 1999.

Solomon, S., Rosenlof, K. H., Portmann, R. W., Daniel,J. S., Davis, S. M., Sanford, T. J., and Plattner, G.: Con-tributions of stratospheric water vapor to decadal changesin the rate of global warming, Science, 327, 1219–1223,https://doi.org/10.1126/science.1182488, 2010.

Stiller, G. P., Kiefer, M., Eckert, E., von Clarmann, T., Kellmann,S., García-Comas, M., Funke, B., Leblanc, T., Fetzer, E., Froide-vaux, L., Gomez, M., Hall, E., Hurst, D., Jordan, A., Kämpfer,N., Lambert, A., McDermid, I. S., McGee, T., Miloshevich, L.,Nedoluha, G., Read, W., Schneider, M., Schwartz, M., Straub,C., Toon, G., Twigg, L. W., Walker, K., and Whiteman, D. N.:Validation of MIPAS IMK/IAA temperature, water vapor, andozone profiles with MOHAVE-2009 campaign measurements,Atmos. Meas. Tech., 5, 289–320, https://doi.org/10.5194/amt-5-289-2012, 2012a.

Stiller, G. P., von Clarmann, T., Haenel, F., Funke, B., Glatthor, N.,Grabowski, U., Kellmann, S., Kiefer, M., Linden, A., Lossow, S.,and López-Puertas, M.: Observed temporal evolution of globalmean age of stratospheric air for the 2002 to 2010 period, At-mos. Chem. Phys., 12, 3311–3331, https://doi.org/10.5194/acp-12-3311-2012, 2012b.

Taha, G., Thomason, L. W., and Burton, S. P.: Comparisonof Stratospheric Aerosol and Gas Experiment (SAGE) IIversion 6.2 water vapor with balloon-borne and space-based instruments, J. Geophys. Res., 109, D18313,https://doi.org/10.1029/2004JD004859, 2004.

Thölix, L., Backman, L., Kivi, R., and Karpechko, A. Yu.: Vari-ability of water vapour in the Arctic stratosphere, Atmos.Chem. Phys., 16, 4307–4321, https://doi.org/10.5194/acp-16-4307-2016, 2016.

Toohey, M., Hegglin, M. I., Tegtmeier, S., Anderson, J., Añel,J. A., Bourassa, A., Brohede, S., Degenstein, D., Froidevaux,L., Fuller, R., Funke, B., Gille, J., Jones, A., Kasai, Y., Krüger,K., Kyrölä, E., Neu, J. L., Rozanov, A., Smith, L., Urban, J.,Clarmann, T., Walker, K. A., and Wang, R. H. J.: Charac-terizing sampling biases in the trace gas climatologies of theSPARC Data Initiative, J. Geophys. Res., 118, 11847–11862,https://doi.org/10.1002/jgrd.50874, 2013.

Tweedy, O. V., Kramarova, N. A., Strahan, S. E., Newman, P. A.,Coy, L., Randel, W. J., Park, M., Waugh, D. W., and Frith,S. M.: Response of trace gases to the disrupted 2015–2016quasi-biennial oscillation, Atmos. Chem. Phys., 17, 6813–6823,https://doi.org/10.5194/acp-17-6813-2017, 2017.

Urban, J., Murtagh, D. P., Stiller, G., and Walker, K. A.: Evolu-tion and Variability of Water Vapour in the Tropical Tropopauseand Lower Stratosphere Region Derived from Satellite Measure-ments, in: Advances in Atmospheric Science and Applications,vol. 708, ESA Special Publication, 8, 2012.

Urban, J., Lossow, S., Stiller, G., and Read, W.: An-other drop in water vapor, EOS Trans., 95, 245–246,https://doi.org/10.1002/2014EO270001, 2014.

Vömel, H., Oltmans, S. J., Hofmann, D. J., Deshler, T., andRosen, J. M.: The evolution of the dehydration in the Antarc-tic stratospheric vortex, J. Geophys. Res., 100, 13919–13926,https://doi.org/10.1029/95JD01000, 1995.

von Clarmann, T., Höpfner, M., Kellmann, S., Linden, A., Chauhan,S., Funke, B., Grabowski, U., Glatthor, N., Kiefer, M., Schiefer-decker, T., Stiller, G. P., and Versick, S.: Retrieval of temperature,H2O, O3, HNO3, CH4, N2O, ClONO2 and ClO from MIPASreduced resolution nominal mode limb emission measurements,Atmos. Meas. Tech., 2, 159–175, https://doi.org/10.5194/amt-2-159-2009, 2009.

von Clarmann, T., Stiller, G., Grabowski, U., Eckert, E., and Or-phal, J.: Technical Note: Trend estimation from irregularly sam-pled, correlated data, Atmos. Chem. Phys., 10, 6737–6747,https://doi.org/10.5194/acp-10-6737-2010, 2010.

Walker, K. A. and Stiller, G. P.: The SPARC water vapour assess-ment II: Data set overview, to be submitted to Atmos. Meas.Tech., 2018.


https://doi.org/10.5194/acp-8-1391-2008

https://doi.org/10.5194/acp-9-9647-2009

https://doi.org/10.1029/2007JD008678

https://doi.org/10.1029/1999GL900315

https://doi.org/10.1029/1999RG900008

https://doi.org/10.1126/science.1182488

https://doi.org/10.5194/amt-5-289-2012

https://doi.org/10.5194/amt-5-289-2012

https://doi.org/10.5194/acp-12-3311-2012

https://doi.org/10.5194/acp-12-3311-2012

https://doi.org/10.1029/2004JD004859

https://doi.org/10.5194/acp-16-4307-2016

https://doi.org/10.5194/acp-16-4307-2016


https://doi.org/10.5194/acp-17-6813-2017

https://doi.org/10.1002/2014EO270001

https://doi.org/10.1029/95JD01000

https://doi.org/10.5194/amt-2-159-2009

https://doi.org/10.5194/amt-2-159-2009

https://doi.org/10.5194/acp-10-6737-2010

Documents

The SPARC water vapour assessment II: Comparison of ...€¦ · ter vapour cool the stratosphere but warm the troposphere (Solomon et al.,2010). Model simulations predict a ˘1K decrease