Ground-based remote sensing of tropospheric HDO/H 2 O ratio profiles

Forschungszentrum Karlsruhein der Helmholtz-Gemeinschaft

NDACC H2O workshop, Bern, July 2006

Ground-based remote sensing of tropospheric HDO/H2O ratio profiles

by M. Schneider, F. Hase, and T. Blumenstock

Presentation and validation of an innovative retrieval approach PROFFIT (V9.4):Motivation: Ground-based FTIR allows monitoring of many trace speciesin the atmosphere. In addition, significant variability in isotopic compositioncan be measured for several species (e.g. O3, H2O, …). Whereas theretrieval of vertical profiles for individual species has reached a high levelof sophistication, codes still lack to support dedicated isotopic retrievalwork. published in Atmos.

Chem. Phys. Discuss., june 2006



The additional complication of isotopic retrieval

Target quantity: ratio of the isotopic species as function of altitude

Standard approach: the isotopic species are jointly retrieved from ameasured spectrum and the resulting profiles are compared.

Drawback: AKs of the two species can differ significantly. So the profiles are incompatible. The situation is equivalent to the case of comparing profiles of a certain species measured by different remote sensing instruments (yes, again Rodgers & Connor …).



How can VMR-ratios be constrained properly?

The advantages of a log-retrieval (Hase et al.,“intercomparison of retrieval codes…“, JQSRT, 2004):• exclude unreasonable (negative) VMRs• log-normal pdf is superior guess in case of large percentage variability.

These claims have been proven for H2O profile retrievals by careful intercomparison of Izaña FTIR results with sonde data (Schneider et al. ,“water vapour profiles by ground-based FTIR…“, ACP, 2006).

OE / TP methods constrain differences between components of the solution vector and the a-priori vector. Differences between log(VMR) components correspond to ratios of the VMR quantities:

log(a / b) = log(a) - log(b)



In the following the possibilities of the PROFFIT Ver. 9.4 isotope retrieval option will be presented:

• Theoretical estimation of precision of HDO/H2O ratios (common approach versus PROFFIT Ver. 9.4)

• Empirical validation and interpretation of HDO/H2O ratios retrieved from spectra measured at Izaña



The theoretical error estimation is performed in form of a full treatment (reason: important non-linearities):

1. Forward calculation of assumed HDO and H2O profile pairs

2. Introducing of errors (measurement noise, temperature profile, ILS, spectroscopic parameter, …)

3. Inversion of the simulated spectra

This „Monte Carlo“ error estimation works only if we apply a large ensemble of HDO and H2O profile pairs, which obey the real HDO/H2O statistics !



1110 1111 1112 1113

0

1

2

3

4

5

1117,9 1121 1122

1221

0

1

2

3

4

5

1252 1253 1325

abso

lute

radi

ance

s [W

/(cm

2 ste

rad

cm-1)]

meaurement retrieval residuum x10

H2O

x10

wavenumber [cm-1]

H2O

H2O H

2O

abso

lute

radi

ance

s [W

/(cm

2 ste

rad

cm-1)]

HDOCH

4

CH4

wavenumber [cm-1]

HDON

2O

CH4

HDO

Applied spectroscopic windows:

H2O

HDO



Detection of atmospheric δD variabilities:

The problem:

δD variabilities are very small (σ of 7%, compared to σ of H2O or HDO of 100%) → measurements have to be very precise !

Advantage:

HDO and H2O have some errors in common → these errors are eliminated when calculating the ratio.

Here we examine if a common approach (individual OE of HDO and H2O and subsequent calculation of the ratio) already provides for a satisfactory precision, or if the PROFFIT Ver. 9.4 isotope retrieval option (OE of the ratio itself) has to be applied.



Smoothing error

Correlation between real profiles and retrieved profiles in the absence of measurement noise and parameter errors:

5 10 15

5

10

15

real atmosphere [km]

retri

eved

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

D profiles: real vs. retrieved

5 10 15

5

10

15


retri

eved

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

common approach PROFFIT v9.4



5 10 15

5

10

15


retri

eved

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

D profiles: real vs. retrieved

5 10 15

5

10

15



retri

eved

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

Total error

Correlation between real profiles and retrieved profiles for a realistic error scenario:



Theoretical random error estimation (summary)

error source total column 2.3-4.3km 5.3-8.3km 6.4-10.0km

smoothing 2 (2) 4 (8) 31 (66) 52 (76)

measurement noise

2 (3) 3 (23) 17 (74) 12 (75)

phase error (ILS)

0 (0) 0 (1) 0 (1) 1 (4)

T profile 0 (0) 0 (8) 1 (22) 1 (14)

γ (correlated) 0 (1) 2 (9) 13 (17) 5 (11)

γ (inconsistent) 2 (2) 14 (51) 44 (92) 34 (63)

total 3 (4) 15 (57) 50 (99) 61 (100)



Total error

5

10

15

0 50 100

5

10

15

0 50 100


random error [%]

altit

ude

[km

]

total random errors

random error [%]

altit

ude

[km

]

Precision of common approach only allows us to retrieve the δD variability in the boundary layer.

The PROFFIT Ver. 9.4 isotope retrieval option improves the precision of δD in the boundary layer and enables the retrieval of the δD variability in the middle/upper troposphere.



Comparison of mean δD profiles as obtained by Ehhalt (1974) and by our measurements

5

10

15

-500 -400 -300 -200 -100

Ehhalt 1974 FTIR FTIR with assumed

line shape error

D [o/oo

]

altit

ude

[km

]

We find indications of an inconsistency in the line shape of HDO and H2O !

→ improving the quality of the spectroscopic data would further improve the precision of the retrieved δD values



-600

-500

-400

-300

-200

-100

0

-600

-500

-400

-300

-200

-100

0

single measurement 3 months running mean

D [0 / 00

]

2005 2006

lower troposphere (2.3-5.3km)

J F M A M J J A S O N D J F M

D [0 / 00

]

middle/upper troposphere (5.3-8.8km)

higher latitude or strongly descending airmasses

tropical airmasses

Remote sensing of water vapour isotope ratios above Izaña Observatory:

middle/upper tropospheric δD is a good tracer for origin of airmass

annual cycle in lower troposphere is correlated to sea surface temperature



Middle/upper tropospheric trajectories for 15th July and 17th September 2005:



Middle/upper tropospheric trajectories for 10th August and 30th November 2005:



Typical middle/upper tropospheric trajectories (3rd July 2005):

Conclusion of validation:

1. The PROFFIT Ver. 9.4 isotope retrieval option yields middle/upper tropospheric δD values with a precision of 50%

2. Middle/upper tropospheric δD gives us information about the static stability of the atmosphere where the airmass comes from (see also Gedzelman, 1988)



-400 -300 -200 -100 0-400

-300

-200

-100

0

IFS

125

HR

D [0 / 00

]

IFS 120M D [0/00

]

Comparison of measurements performed within 1 h

D(125HR)-D(120M): 10 +/- 28 0/00

-> empirically obtained error due to measurement noise: 19 0/

00

Consistency for measurements performed side-by-side with two different FTIR spectrometer:

for the middle/upper troposphere (5.3-8.8km)



-600

-500

-400

-300

-200

-100

0

IFS 120M IFS 125HR mean + std of JJA

D (5

.3-8

.8km

)

1999 2000 2001 2002 2003 2004 2005 2006

7 year time series of middle/upper tropospheric δD in the subtropical eastern Atlantic:



Summary and outlook:• PROFFIT Ver. 9.4 isotope retrieval option allows for the detection of middle/upper tropospheric δD (common approach doesn’t) !

• Middle tropospheric δD can also be retrieved from lower situated sites (not limited to mountain observatories)

• Method has a wide range of application: isotope ratios for O3, CH4, CO2 (among others); adaptation to satellite codes

• still to do: detailed interpretation of δD time series:

middle tropospheric δD as index for low static stability in the troposphere, correlation with NAO, ENSO ?



Thank You !!!!!



The additional complication of isotopic retrieval

Possible practical approaches:• Accept profiles as given, apply full intercomparison formalism when

comparing deduced ratios to model predictions.Elaborate post processing, no stand-alone FTIR data product!

• Iterative scheme: main isotope result serves as a-priori for the others.Scatter in ratios is reduced, but no reliable scheme, as main

isotope AKs imprinted on updated a-priori. Only partial back propagation of the full solution state to the main isotope!

(Forget about cycling back and forth: beware of “Rodgers trap”)

• Optimal scheme: Retrieve the full solution state, add expected correlation between isotopes as additional constraint!

Individual sensitivities of the isotopes are fully exploited in construction of the full solution state, OE solution for ratio results!



5

10

15

-0,9 -0,8 -0,7 -0,6 -0,5 -0,4 -0,3 -0,2 -0,1 0,0

D [0/00

]

ln([D/1000]+1)

altit

ude

[km

]

Ehhalt, 1974 (Queen Air)

Ehhalt, 1974 (Sabreliner)

applied in this work

-600 -500 -400 -300 -200 -100 0

Measurements of HDO/H2O ratios are rare. Most extensive dataset is from Ehhalt (1974):

5 10 15

5

10

15

altitude [km]

altit

ude

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

D interlevel correlations

median δD profile: inter-level correlation of δD values:

→ we derive HDO/H2O statistics from Ehhalt (1974) measurements and simulate an ensemble of 500 profile pairs



HDO/H2O ratio is commonly expressed in form of δD, which is the relative difference of the actual HDO/H2O ratio to a standard ratio (Rs= 3.1152 x 10-4):

2

s

[HDO]/H O]δD = 1000 -1R



Consistency for measurements performed within 6h

A pair of measurements performed within several hours is only partly representative for the highly variable atmospheric conditions encountered on different days or months. This consistency check should underestimate the overall precision.

lower troposphere: 31% precision → in disagreement with theoretical estimation of 15% ! Reason: in boundary layer exist large variabilities of δD on small time scales

middle/upper troposphere: 37% precision → in agreement with theoretical estimation of 50%

-500 -400 -300 -200 -100-500

-400

-300

-200

-100

-500 -400 -300 -200 -100-500

-400

-300

-200

-100N=57=30.9 % (est.:<15%)

D u

p to

6h

late

r [0 / 00

]

D [0/00]

lower troposphere (2.3-5-3km) middle/upper troposphere (5.3-8.8km)

D u

p to

6h

late

r [0 / 00

]

D [0/00]

N=57=37 % (est.:<50%)



-500 -400 -300 -200 -100 0-500

-400

-300

-200

-100

-500 -400 -300 -200 -100-150

-100

-50

0

50

100

150

^slope:x/x

^xx

retri

eved

D [0 / 00

]

real D [0/00

]

erro

r [0 / 00

]

retrieved D [0/00

]

-600 -400 -200 0 200-600

-400

-200

0

200

-600 -400 -200 0 200

-200

-100

0

100

200^

D re

triev

ed w

ith e

rror

[0 / 00]

D retrieved without error [0/00

]

offset: x-x

xx^

slope: x/x

^___

offs

et c

orre

cted

err

or [0 / 00

]

offset corrected D [0/00

]

Error estimation by means of linear least squares fit

Smoothing error:

no mean error (no offset) → no systematic error ?No ! Systematic sensitivity error (slope of regression line < 1) !

Spectroscopic pressure broadening coefficient error:

mean and systematic error



Errors are estimated by least squares fits between real and retrieved δD:

random error component:scattering around the regression curve:

_ 21reg

x

Smoothing error

5

10

15

0 50 100

5

10

15

0 50 100

random error [%]

altit

ude

[km

]


smoothing errors (random)

^random error [%]

altit

ude

[km

]



systematic error component:

difference of regression curve from diagonal: offset (mean error) and slope (sensitivity error)

Smoothing error

5

10

15

-25 0 25

5

10

15

-25 0 25

5

10

15

-100 -50 0 50

5

10

15

-100 -50 0 50


mean error [%]

altit

ude

[km

]

smoothing error (systematic)

mean

mean error [%]

altit

ude

[km

]

sensitivity error [%]

altit

ude

[km

]

sensitivity


altit

ude

[km

]



Measurement noise error

5

10

15

0 50 100

5

10

15

0 50 100

common approach PROFFIT v9.4measurement noise (random)

random error [%]

altit

ude

[km

]

random error [%]

altit

ude

[km

]

This error is not common for HDO and H2O → error is not eliminated in the ratio when common retrieval is applied

Inter-species constraint forces common response of HDO and H2O → reduced error in the ratio



Measurement noise error

Common approach no OE of δD → random error source produces systematic error!

PROFFIT isotope retrieval option provides for OE of δD → random error source produces only random errors

5

10

15

-25 0 25

5

10

15

-25 0 25

5

10

15

-50 0 50

5

10

15

-50 0 50

mean

measurement noise (systematic)

mean error [%]

altit

ude

[km

]


mean error [%]

altit

ude

[km

]

sensitivity


altit

ude

[km

]


altit

ude

[km

]



Temperature error

5

10

15

0 50

5

10

15

0 5

common approach PROFFIT v9.4temperature profile (random)

random error [%]

altit

ude

[km

]

random error [%]

altit

ude

[km

]

Systematic error of the common approach below 5%

Error in HDO and H2O similar → already in common approach δD error below 25%

Isotope retrieval option further equalises error responses of HDO and H2O → very small error in δD



ILS (instrumental line shape) error

5

10

15

0 5

5

10

15

0 5


phase error (random)

random error [%]

altit

ude

[km

]

random error [%]

altit

ude

[km

]

Systematic error of the common approach below 1%

Error in HDO and H2O common → already in common approach negligible δD error

Isotope retrieval option further equalises error responses of HDO and H2O → even smaller error in δD



Correlated errors of line shape of H2O and HDOThis systematic error source leads to random errors in δD due to non-linearities.

The error source is common for HDO and H2O, but HDO and H2O lines have different sensitivities, responses → error is not completely eliminated in ratio when common retrieval is applied

Inter-species constraint forces common response of HDO and H2O → reduced δD error for unsaturated H2O lines

For saturated lines sensitivities of HDO and H2O are too different → also large error when inter-species constraint is applied

5

10

15

0 50 100

5

10

15

0 50 100


random error [%]

altit

ude

[km

]

correlated error (random)

random error [%]

altit

ude

[km

]



Correlated errors of line shape of H2O and HDO

Systematic errors are larger for common approach

5

10

15

-50 0 50

5

10

15

-50 0 50

5

10

15

-50 0 50

5

10

15

-50 0 50

meancorrelated error (systematic)

mean error [%]

altit

ude

[km

]


mean error [%]

altit

ude

[km

]

sensitivity


altit

ude

[km

]


altit

ude

[km

]



Inconsistency in line shape of H2O and HDO (γ(HDO) 2% increased)

This systematic error source leads to random errors in δD due to non-linearities.

The error source is not common for HDO and H2O → large error when common retrieval is applied

Inter-species constraint forces common response of HDO and H2O → reduced error in ratio for unsaturated H2O lines, but still large

5

10

15

0 50 100

5

10

15

0 50 100


random error [%]

altit

ude

[km

]

inconsistencies in (random)

random error [%]

altit

ude

[km

]



Inconsistency in line shape of H2O and HDO (γ(HDO) 2% increased)

Systematic errors are larger for common approach

5

10

15

-100 -50 0 50 100

5

10

15

-100 -50 0 50 100

5

10

15

-100 0 100 200

5

10

15

-100 0 100 200

mean

inconsistencies in (systematic)

mean error [%]

altit

ude

[km

]


mean error [%]

altit

ude

[km

]

sensitivity


altit

ude

[km

]


altit

ude

[km

]



5 10 15

5

10

15

H2O profiles: sonde vs. FTIR

sonde atmosphere [km]

FTIR

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

5 10 15

5

10

15


FTIR

atm

osph

ere

[km

]-0,64-0,52-0,400,400,520,640,760,881,00

5 10 15

5

10

15


FTIR

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

5 10 15

5

10

15


LT s

lant

< 5

x1021

cm-2

L

T sl

ant <

2.5

x1022

cm-2


FTIR

atm

osph

ere

[km

]

-0,64-0,52-0,400,400,520,640,760,881,00

Empirical validation of H2O profiles: ptu-sonde vs. FTIR:

Inter-species constraint not only improves quality of δD profiles but also quality of H2O profile:

H2O retrieval benefits from additional information present in HDO lines

Detection of UT/LS H2O with the proposed lines ?For a definitive conclusion we need a larger ensemble of compared profiles (the correlation shown here bases on only 7 profiles) !



120M ↔ 125HR intercomparison (January, March, and April 2005)

Lower troposphere: for measurements within 1h 12% error → in slight disagreement with theoretical estimation of 3%.

Middle/upper troposphere: for measurements within 3h 21% error → in best agreement with theoretical estimation of 17%

-500 -400 -300 -200 -100-500

-400

-300

-200

-100

-500 -400 -300 -200 -100-500

-400

-300

-200

-100

lower troposphere (2.3-5-3km) middle/upper troposphere (5.3-8.8km)

D 1

20H

R [0 / 00

]

D 120M [0/00]

(N=33)N=17=12 % (est.: 3%)

D 1

20H

R [0 / 00

]

D 120M [0/00]

N=33=20.7% (est.: 17%)

This intercomparison allows an empirical assessment of the error due to measurement noise !

It confirms that the boundary layer δD value varies strongly on small timescales.



0 5 10 15 20 25 30-400

-350

-300

-250

-200

-150

JJA

mea

n +

std

of D

in M

T at

Izan

a [0 / 00

]

number of days with hurricane, trop. storm, or trop. depression south of 25°N and east of 50°W

Is there a correlation with hurricane activity ??

Documents

Ground-based remote sensing of tropospheric HDO/H 2 O ratio profiles