amt-9-669-2016.pdf

Atmos. Meas. Tech., 9, 669–682, 2016

www.atmos-meas-tech.net/9/669/2016/

doi:10.5194/amt-9-669-2016

Improved retrieval of gas abundances from near-infrared solar

FTIR spectra measured at the Karlsruhe TCCON station

M. Kiel

, D. Wunch

, P. O. Wennberg

, G. C. Toon

, F. Hase

, and T. Blumenstock

Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology, Karlsruhe, Germany

Department of Environmental Science and Engineering, California Institute of Technology, Pasadena, CA, USA

Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA

Correspondence to:

M. Kiel (matthaeus.kiel@kit.edu)

Received: 16 October 2015 – Published in Atmos. Meas. Tech. Discuss.: 23 November 2015

Revised: 19 January 2016 – Accepted: 16 February 2016 – Published: 29 February 2016

Abstract.

We present a modified retrieval strategy for solar

absorption spectra recorded by the Karlsruhe Fourier Trans-

form Infrared (FTIR) spectrometer, which is operational

within the Total Carbon Column Observing Network (TC-

CON). In typical TCCON stations, the 3800–11 000 cm

−

spectral region is measured on a single extended Indium

Gallium Arsenide (InGaAs) detector. The Karlsruhe setup

instead splits the spectrum across an Indium Antimonide

(InSb) and InGaAs detector through the use of a dichroic

beam splitter. This permits measurements further into the

mid-infrared (MIR) that are of scientific interest, but are not

considered TCCON measurements. This optical setup in-

duces, however, larger variations in the continuum level of

the solar spectra than the typical TCCON setup. Here we in-

vestigate the appropriate treatment of continuum-level vari-

ations in the retrieval strategy using the spectra recorded in

Karlsruhe. The broad spectral windows used by TCCON re-

quire special attention with respect to residual curvature in

the spectral fits. To accommodate the unique setup of Karl-

sruhe, higher-order discrete Legendre polynomial basis func-

tions have been enabled in the TCCON retrieval code to fit

the continuum. This improves spectral fits and air-mass de-

pendencies for affected spectral windows. After fitting the

continuum curvature, the Karlsruhe greenhouse gas records

are in good agreement with other European TCCON data

sets.

1 Introduction

Global climate change is a major research topic of today’s

environmental sciences. Human activities, such as burning

of fossil fuels, are the key drivers of the continuing increase

of atmospheric greenhouse gases and the gases involved in

their chemical production (Peters et al., 2013). Long-term

measurements of the atmospheric composition provide the

experimental data to quantify sinks and sources which are of

utmost importance to understand the anthropogenic impact

on global warming (Olsen and Randerson, 2004).

The Total Carbon Column Observing Network (TCCON)

provides measurements of column-averaged abundances of

greenhouse gases. TCCON is a ground-based network of

Fourier Transform Infrared (FTIR) spectrometers initiated in

2004 by the California Institute of Technology, Pasadena,

USA (Wunch et al., 2011). The stationary high-resolution

FTIR spectrometers measure total columns of CO

, CO,

, N

O, H

O, HF and other atmospheric gases. Precise

and accurate column abundances are retrieved from near in-

frared (NIR) solar absorption spectra using direct sunlight.

TCCON measurements are tied to the World Meteorologi-

cal Organization (WMO) scale via in situ aircraft measure-

ments flown over TCCON sites (Washenfelder et al., 2006;

Deutscher et al., 2010; Wunch et al., 2010; Messerschmidt

et al., 2011; Geibel et al., 2012). For the greenhouse gases

and CH

, TCCON achieves an accuracy and precision

in total column measurements of about 0.2 % which is nec-

essary to gain information about sinks and sources and for

satellite validation (Rayner and O’Brien, 2001). Currently,

about 23 globally distributed sites are affiliated with TC-

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

670

M. Kiel et al.: Improvement of the retrieval used for Karlsruhe TCCON data

CON. The network aims to improve global carbon cycle stud-

ies and to provide a primary validation data record of various

gaseous atmospheric components for retrievals from space-

based instruments. TCCON instruments measure the same

quantities in the same spectral region as satellite-borne in-

struments – e.g., the Orbiting Carbon Observatory 2 (OCO-2,

Frankenberg et al., 2015), the Scanning Imaging Absorption

Spectrometer (SCIAMACHY, Frankenberg et al., 2006) and

the Greenhouse Gases Observing Satellite (GOSAT, Morino

et al., 2011). Hence, for the validation, TCCON provides an

ideal data set.

The TCCON strives to attain the best site-to-site precision

and accuracy possible. Systematic biases that are consistent

throughout the network are fully accounted for by scaling

the TCCON retrieval results to the WMO scale via aircraft

and AirCore profiles (Wunch et al., 2010). Thus, the TC-

CON sets guidelines to ensure that the instrumentation at

each site is as similar as possible, and that the retrieval soft-

ware, including the spectroscopic line lists and line shapes,

is identical for each site. For example, if a particular site

used a different spectroscopic line list from the rest of the

sites, the network consistency would decrease even if that

line list is an improvement over the original. There are, how-

ever, several site-specific differences that can cause a degra-

dation in the TCCON’s consistency: differing instrument line

shapes (ILS) between instruments (Hase, 2012), laser sam-

pling errors (LSE) that differ between instruments (Messer-

schmidt et al., 2010), and differing optical component re-

sponses between instruments (beam splitters, detectors, fil-

ters, mirror coatings, etc.). The impacts of ILS differences are

mitigated by requiring that all instruments maintain a near-

perfect ILS at each TCCON station. The impacts of the LSE

are minimized by applying a correction to the TCCON inter-

ferograms (Dohe et al., 2013; Wunch et al., 2015). TCCON

partners typically use very similar optical components, and

detectors which addresses the last of these issues. This ap-

proach to standardizing the optical components is imperfect,

but the differences between spectra from different sites are

generally small. The Karlsruhe system, however, has a sig-

nificantly different optical setup, designed to allow for au-

tomated mid-infrared NDACC (Kurylo, 1991) and TCCON

measurements to be made from the same system. As shown

here, using the standard TCCON retrieval approach for this

setup causes biases of nearly 1 ppm in XCO

, which exceeds

the precision requirements of the network.

In this paper, we discuss the particular instrumental setup

of the Karlsruhe FTIR spectrometer and point out differences

from the standard TCCON setup. We identify difficulties in

the standard TCCON data processing when analyzing solar

absorption spectra recorded by the Karlsruhe spectrometer

and present a strategy for the Karlsruhe data set that improves

its consistency with respect to TCCON.

2 Instrumentation

The Karlsruhe TCCON FTIR spectrometer was initiated in

2009 at the Karlsruhe Institute of Technology (KIT) – Cam-

pus North (49.1

◦

N, 8.4

◦

E, 110 m a.s.l.). Karlsruhe is an

extensive urban region in central Europe and experiences

an oceanic, mild climate similar to most cities in the mid-

western part of Europe. The flat terrain is a favorable scene

for nadir-looking satellite overpasses as well as model stud-

ies. Solar spectra are acquired by operating a Bruker IFS

125HR spectrometer (Bruker Optics, Germany). The auto-

mated instrument is housed in an air-conditioned 20 ft sea

transport certified shipping container. The spectrometer fea-

tures a precise cube-corner Michelson interferometer con-

taining a semi-transparent calcium fluoride (CaF

) beam

splitter and a linearly moving scanner. An InSb detector cov-

ers the spectral range from 1900–5250 cm

−

and an InGaAs

detector covers the 5250–11 000 cm

−

spectral range. The

InSb diode is cryogenically cooled using a liquid nitrogen

(LN

) microdosing autofill cooling system (Norhof, Nether-

lands). A dichroic mirror (Optics Balzers Jena GmbH, Ger-

many) is installed with a cut-on wavenumber of 5250 cm

−

The instrument features a camera-based solar tracker de-

veloped by KIT (Gisi et al., 2011) with gold-coated optics

to minimize photon noise induced by the visible spectrum.

TCCON measurements are routinely recorded at a maxi-

mum optical path difference (OPD

max

) of 45 cm leading to

a spectral resolution of 0.02 cm

−

. In addition, solar spectra

are also recorded at OPD

max

64 cm and OPD

max

120 cm

leading to spectral resolutions of 0.014 and 0.0075 cm

−

, re-

spectively.

2.1 Differences to the standard TCCON setup

The Karlsruhe full spectral range is 1900–11 000 cm

−

, mea-

sured simultaneously with InSb and InGaAs detectors. The

typical TCCON spectral range is 3800–16 000 cm

−

, mea-

sured simultaneously with InGaAs and silicon (Si) detectors

(Washenfelder et al., 2006). To measure all TCCON gases,

the spectral range of the TCCON measurements must in-

clude 3800–11 000 cm

−

; only the oxygen A- and B-bands

are measured on the Si detector above 11 000 cm

−

, and

these retrievals are not part of the standard set of TCCON

retrievals (Wunch et al., 2011). The spectra from the Si de-

tector are important, however, because they are used to cal-

culate and correct for any LSE in the system (Wunch et al.,

2015), to study aerosols, and the oxygen A- and B-bands are

necessary for comparison with satellites, which cannot use

the oxygen band at 7885 cm

−

The Karlsruhe setup splits the 3800–11 000 cm

−

spec-

tral range for TCCON-style measurements across the InSb

and InGaAs detectors using a dichroic beam splitter that re-

flects the MIR spectral domain and transmits the NIR spec-

tral range. The cut-on of the dichroic (5250 cm

−

) is be-

tween two atmospheric windows separated by H

O absorp-

Atmos. Meas. Tech., 9, 669–682, 2016

www.atmos-meas-tech.net/9/669/2016/

M. Kiel et al.: Improvement of the retrieval used for Karlsruhe TCCON data

671

tion bands. In the Earth’s atmosphere, this spectral region is

strongly saturated such that no loss of information arises by

splitting the incoming beam into parts of MIR and NIR radi-

ation at the chosen wavenumber.

For TCCON measurements of CO, N

O and HF which

absorb in the 3800–4800 cm

−

region, a narrowband spectral

filter transmitting from 3800–5250 cm

−

is mounted in front

of the InSb diode, yielding higher signal-to-noise ratios and

minimizing any detector nonlinearity. A spectrum recorded

by the Karlsruhe instrument and a typical TCCON spectrum

recorded by the Park Falls spectrometer is depicted in Fig. 1.

The operation of the InSb diode provides additional spec-

tral coverage to wavenumbers as low as 1900 cm

−

when

using other narrowband filters. Additional gases absorb in

this region, including NO, O

, HCl, HCN, C

, C

, NO

and C

. Additionally, the fundamental absorption bands of

OCS and CO are in this region, making it the preferred spec-

tral region for retrievals of OCS and CO. The optical setup

also provides spectra of H

O, HDO, CH

and N

O in the

MIR bands. MIR measurements are performed following the

guidelines of the Network for the Detection of Atmospheric

Composition Change – Infrared Working Group (NDACC-

IRWG) in addition to the TCCON measurements in the NIR.

2.2 Impact of the optical setup on solar spectra

The combination of the dichroic beam splitter and InSb opti-

cal filter in the Karlsruhe FTIR instrument induces stronger

variations of the continuum in solar spectra than a stan-

dard TCCON FTIR setup. The lower panel of Fig. 1 shows

a Karlsruhe spectrum simultaneously recorded by the InSb

diode and the InGaAs diode. There are clear differences

in the shape of the spectrum between 3900–5250 cm

−

be-

tween the Karlsruhe instrument (Fig. 1, bottom panel) and

a typical TCCON instrument (Fig. 1, upper panel). Karl-

sruhe spectra contain an oscillating overall envelope which

is not present in standard TCCON spectra. Retrieved gases

within this spectral region are CO (center wavenumber (cw)

in cm

−

: 4233.0, 4290.4), N

O (cw: 4395.2, 4430.1, 4719.5),

HF (cw: 4038.95) and several H

O and HDO narrow spectral

windows.

Smaller, but significant differences in the continua are also

present in the 5250–11 000 cm

−

range. The Karlsruhe sig-

nal remains high with an oscillating overall envelope while

the Park Falls signal decreases smoothly with increasing

wavenumbers. This region contains the O

(cw: 7885.0)

spectral window which is used to calculate column-averaged

dry-air mole fractions (DMFs) of the target gases.

3 Analysis and data processing

Within TCCON, the recorded interferograms are processed

and analyzed with the GGG2014 Software Suite which in-

cludes GFIT, a non-linear least-squares spectral fitting al-

gorithm (Wunch et al., 2015). In general, all TCCON sites

4000

5000

6000

7000

8000

9000

10000

11000

Wavenumber

[

−

]

0.0

0.1

0.2

0.3

0.4

0.5

Signal (arbitrary units)

(

)

4000

5000

6000

7000

8000

9000

10000

11000

Wavenumber

[

−

]

0.00

0.05

0.10

0.15

0.20

Signal (arbitrary units)

ut-on dichroic

(5250cm )

−

(

)

(

)

Figure 1.

Upper panel: typical TCCON spectrum recorded by the

Park Falls instrument which operates an extended InGaAs detec-

tor, marked are the spectral regions of the main gases of interest;

lower panel: typical Karlsruhe spectrum recorded by the InSb and

InGaAs diode. The coverage of the full spectral range from 3800–

10 000 cm

−

is realized by the simultaneous operation of the two

diodes.

use the same software and retrieval analysis strategy to mini-

mize algorithmic biases between sites. The calibration of the

spectral radiances, exact modeling of the far line wing con-

tributions, and continuum transmission variability will cause

consistent errors for all TCCON stations, because the line

shape and continuum models are identical, thus negligibly

impacting the TCCON precision. However, surface tempera-

tures and signal-to-noise ratios can differ significantly from

site to site, and therefore these errors are much more impor-

tant to minimize.

3.1 Impact of the optical setup on spectral fits

Ideally, spectral residuals (the difference between the com-

puted and measured spectrum) should have no structure,

and consist only of the random noise associated with the

signal-to-noise ratio of the measured spectrum. We show that

residuals for CH

(cw: 5938.0, 6076.0), CO

(cw: 6339.5),

CO (cw: 4290.4), N

O (cw: 4719.5) and O

(cw: 7885.0)

show significant broad structure when fitted with the standard

GGG2014 TCCON retrieval, which fits only a scalar contin-

uum level and linear continuum tilt. Figure 2 shows spectral

fits and residuals for one particular Karlsruhe measurement

for the O

and N

O spectral windows. The residuals in the

www.atmos-meas-tech.net/9/669/2016/

Atmos. Meas. Tech., 9, 669–682, 2016

672

M. Kiel et al.: Improvement of the retrieval used for Karlsruhe TCCON data

0.02

0.00

0.02

Residual

7750

7800

7850

7900

7950

8000

Wavenumber

[

−

]

0.0

0.2

0.4

0.6

0.8

1.0

Transmittance

asurement

Model

0.05

0.00

0.05

Residual

4680

4690

4700

4710

4720

4730

4740

4750

4760

Wavenumber

[

−

]

0.0

0.2

0.4

0.6

0.8

1.0

Transmittance

Measurement

Model

Figure 2.

Spectral fits for a particular Karlsruhe spectrum: upper

panel, spectral fit and residual for O

(cw: 7885.0); lower panel,

spectral fit and residual for N

O (cw: 4719.5).

spectral window have the shape of a higher-order polyno-

mial while the N

O residual has a single extremum. Spectral

fits for the other affected target gases and spectral windows

are depicted in Fig. A1 in the Appendix.

Continuum curvature is related to our choice of optical

filters and dichroic, and is not atmospheric in nature. To

demonstrate this, we show that curvature exists in labora-

tory measurements using a black-body cavity at 1000

◦

as a source. The Karlsruhe FTIR instrument is not evac-

uated, therefore cavity measurements contain some atmo-

spheric absorption lines mainly from H

O in the laboratory

air (see Fig. A2, Appendix). Nevertheless, the curved resid-

uals from measurements with the black body have a similar

shape to residuals of atmospheric measurements. Figure 3

shows spectral fits of the O

and N

O spectral windows us-

ing black-body cavity measurements. For O

, residuals fol-

low the shape of a higher-order polynomial as seen for at-

mospheric measurements (see Fig. 2). Residuals within the

O spectral range follow the same parabolic shape as for

atmospheric measurements. This holds for all affected spec-

tral windows (see Fig. A3, Appendix). Hence, curvatures in

the residuals are due to the optical setup.

0.01

0.00

0.01

Residual

7750

7800

7850

7900

7950

8000

Wavenumber

[

−

]

0.98

1.00

1.02

Transmittance

Measurement

Model

0.02

0.00

0.02

Residual

4680

4690

4700

4710

4720

4730

4740

4750

4760

Wavenumber

[

−

]

0.98

1.00

1.02

Ttransmittance

Measurement

Model

Figure 3.

Same as Fig. 2 but for Karlsruhe cavity spectra. Both

residuals follow the same shape as seen for atmospheric measure-

ments.

4 Fitting in the continuum level

The standard GGG2014 retrieval strategy fits a level and a tilt

to the continuum of a spectral window. However, GFIT has

also the ability to fit an

th-order discrete Legendre polyno-

mial basis function to the continuum (Wunch et al., 2015).

This continuum fit option is meant to fit curvatures in the

continuum of the spectrum that are caused by instrumental

features that cannot be neglected in the data processing.

We invoke the higher-order continuum level fit option in

GGG2014. We determine the basis function order,

, for ev-

ery affected spectral window individually using spectral fits

of cavity measurements since their residuals are free from

atmospheric absorptions. Different continuum basis function

orders are tested to achieve the best fit in the continuum level.

An example of how the continuum fit improves residuals of

atmospheric spectral fits for the O

and N

O spectral win-

dows is given in Fig. 4. Spectral fits for all affected target

gases are depicted in Fig. A4 in the Appendix.

4.1 Impact of continuum fits on air-mass dependence

Air-mass-dependent retrieval biases must be accounted for,

as they can be aliased into the seasonal cycle and affect

the time series from different sites at different latitudes dif-

ferently. There are numerous factors that induce air-mass-

dependent artefacts, including continuum curvature.

Atmos. Meas. Tech., 9, 669–682, 2016

www.atmos-meas-tech.net/9/669/2016/

M. Kiel et al.: Improvement of the retrieval used for Karlsruhe TCCON data

673

0.02

0.00

0.02

Residual

7750

7800

7850

7900

7950

8000

Wavenumber

[

−

]

0.0

0.2

0.4

0.6

0.8

1.0

Transmittance

Measurement

Model

0.05

0.00

0.05

Residual

4680

4690

4700

4710

4720

4730

4740

4750

4760

Wavenumber

[

−

]

0.0

0.2

0.4

0.6

0.8

1.0

Transmittance

Measurement

Model

Figure 4.

Same as Fig. 2 but using the GGG2014 higher-order con-

tinuum fit option. For O

(upper panel)

5 was applied while

for N

O (lower panel)

3 was used.

Using cavity-ratioed spectra as a reference, we show that

implementing our continuum curvature fitting scheme sig-

nificantly reduces the air-mass-dependent biases caused by

the curvature. Our cavity-ratioed reference spectra are pro-

duced by dividing atmospheric spectra by a high signal-to-

noise ratio, reduced-resolution (0.05 cm

−

) black-body cav-

ity spectrum (1000

◦

C). This ratio eliminates broadband fea-

tures caused by the optics in the resulting calibrated atmo-

spheric spectra.

The impact of a continuum level fit on the air-mass de-

pendence is elaborated via a case study using Karlsruhe data

on 18 May 2014 when high air-mass values up to seven are

reached during the measurement day. In Fig. 5, the air-mass

dependence compared to the reference retrieval using cavity-

ratioed atmospheric spectra is depicted for the O

spectral

window. Running the standard GGG2014 TCCON retrieval

strategy (i.e., fitting only the continuum level and tilt), an

overall bias of

−

0.10 % results along with an air-mass de-

pendence leading to a relative difference of

−

0.15 % between

the reference run and the standard TCCON retrieval strategy

for air-mass values between six and seven. In comparison,

the air-mass dependence for column abundances from the re-

trieval when a higher-order continuum fit is applied shows

neither a significant air-mass dependence nor a significant

bias (0

04 %).

Airmass

0.15

0.10

0.05

0.00

0.05

∆

[

]

Cont inuum fit

−

reference

St andard GGG2014

−

reference

Figure 5.

Air-mass dependence for the O

(cw: 7885.0) spectral

window retrieved by the standard GGG2014 TCCON retrieval strat-

egy and using a higher-order continuum fit. As a reference, cavity-

ratioed atmospheric spectra are used for the standard GGG2014 re-

trieval setup.

In general, applying a higher-order Legendre polynomial

fit improves the air-mass dependence for CO

, CO, N

O, O

and CH

(cw: 6076.0) (see Fig. A5, Appendix).

There is no clear improvement for CH

(cw: 5938.0). On

the one hand, the overall bias is reduced for small air-mass

values. On the other hand, a stronger air-mass dependence

is induced by applying the higher-order continuum level

fit. Nevertheless, since the majority of the Karlsruhe mea-

surements are recorded between air-mass values of one and

two, the retrieval strategy with a higher-order continuum fit

seems to improve the air-mass dependence compared to the

standard GGG2014 retrieval. The remaining air-mass depen-

dence is most likely due to spectroscopic errors.

4.2 Impact of continuum fits on column-averaged

DMFs

The higher-order continuum fit improves spectral fits as well

as the air-mass dependence. It is also important to note that

the computed DMFs are changed. DMFs are computed by

ratioing the column abundance of the gas of interest by O

and multiplying by the assumed atmospheric DMF of O

(0.2095). Since O

is significantly impacted by continuum

curvature, the DMFs of all gases will change compared to

the standard GGG2014 retrieval strategy. The relative mean

difference is (0

132

010) % for the O

spectral window.

Therefore, DMFs of target gases change by 0.132 % when no

higher-order continuum fit is applied in the second retrieval

strategy (H

O, HF, and HCl). For all other target gases, any

differences are due to the change in retrieved O

abundances

and changes in abundances retrieved of the target gas itself.

An overview of the differences of all affected gases is given

in Table 1.

www.atmos-meas-tech.net/9/669/2016/

Atmos. Meas. Tech., 9, 669–682, 2016