TCCON GGG2020 switch to GEOS IT met products
Joshua L. Laughner
1
, Coleen M. Roehl
2
, Debra Wunch
3
, David F. Pollard
4
,
Pascal Jeseck
5
, Yao T ́e
5
, Rigel Kivi
6
, and Pauli Heikkinen
6
1
Jet Propulsion Laboratory, California Institute of Technology, Pasadena,
CA, USA
2
Division of Geologic and Planetary Sciences, California Institute of
Technology, Pasadena, CA, USA
3
Department of Physics, University of Toronto, Toronto, Canada
4
National Institute of Water and Atmospheric Research Ltd (NIWA), New
Zealand
5
Sorbonne Universit ́e, CNRS, MONARIS, UMR8233, F-75005 Paris, France
6
Space and Earth Observation Centre, Finnish Meteorological Institute,
Finland
Contact: Joshua Laughner, josh.laughner@jpl.nasa.gov
©
2024 Authors
Abstract
The GGG2020 TCCON data product relies on the Goddard GEOS product for a priori
meteorological information. In particular, it was developed with the GEOS FP-IT (Forward
Processing for Instrument Teams) product, as we required the longer record of FP-IT (circa
2000 on) compared to the more recent FP product (circa 2013 on). In 2023, we were informed
that the FP-IT product would be succeeded by the GEOS IT product. Due to the effort
required to perform a full reprocessing of the TCCON data set (
∼
30 sites with records up
to 10 years long), it is not practical at this time to redo the GGG2020 retrievals with the
GEOS IT input data. Therefore, we have made arrangements to switch our retrievals over
to the GEOS IT product starting at midnight UTC, 1 Apr 2024. This date was chosen
to coincide with the switch of OCO-2 and -3 processing to also use GEOS IT, rather than
GEOS FP-IT. This document describes the testing done to quantify the impact this change
will have on the GGG2020 data product.
1
1 Testing methodology
GGG2020 requires a priori meteorological and trace gas VMR profiles as input to the re-
trieval. The meteorological data is taken directly from GEOS FP-IT or IT, interpolated to
the TCCON site location, and the VMR profiles are either likewise interpolated (H
2
O, O
3
,
CO) or derived from those meteorological values (CO
2
, CH
4
, N
2
O, HF) [1]. To test the im-
pact of the GEOS FP-IT to GEOS IT change, the input files containing these meteorological
and VMR data were generated using both meteorology products for four months: Oct 2022,
Feb 2023, Apr 2023, and July 2023. These months were chosen to match the months used in
OCO-2 testing, so that comparisons between the impact on OCO-2 and TCCON retrievals
could be compared if necessary. They also provide one month per season to test if the impact
of the change in meteorology varies throughout the year. Specific TCCON sites were directly
asked to run retrievals for these months using both meteorology product. These sites were
chosen to provide examples in different parts of the world. All TCCON sites were invited to
participate if they had the time to. As of April 25, 2024, the participating sites are:
•
Pasadena, CA, USA,
•
Armstrong, CA, USA,
•
East Trout Lake, Saskatchewan, Canada,
•
Lauder, New Zealand,
•
Lamont, OK, USA,
•
Park Falls, WI, USA,
•
Paris, France, and
•
Sodankyl ̈a, Finland.
When calculating differences, observations from the two datasets (i.e. the retrievals done
using GEOS FP-IT and GEOS IT) were matched based on their observation time. All
differences reported (e.g. ∆ Xgas) are GEOS IT minus GEOS FP-IT.
2 Changes to retrieved Xgas values
On average, the change to most of the Xgas products is near zero (XCO being the exception),
relative to the typical magnitude of the Xgas values (Table 1, Figs. 1 & 2). While the
histograms in Figs. 1 and 2 do show that individual values have significantly more value,
the averages over the four months studied mostly regress toward zero. This is the expected
behavior; while the change in meteorology will certainly have some random variation at
2
Site
XCO
2
XwCO
2
XlCO
2
XCH
4
XCO XN
2
O XH
2
O XHDO
ppm
ppm
ppm
ppm
ppb ppb
ppm
ppm
Pasadena
-0.015 0.016
0.042
-0.000089 -8.3
-0.027
1.30
2.10
Armstrong -0.014 0.0080
0.055
0.000051 -1.98 -0.073
-0.96
-1.8
ETL
-0.022 0.015
0.018
0.00010
-1.5
0.092
0.36
-2.2
Lauder
-0.013 0.0095
-0.016 0.00016
-0.62 0.11
-1.16
-2.0
Lamont
-0.010 0.016
0.053
0.00042
-1.20 0.039
-3.6
-8.1
Park Falls -0.046 -0.062
0.044
-0.00011
-1.6
-0.040
-0.70
-2.8
Paris
-0.028 -0.022
0.044
0.000078 -6.7
-0.053
-3.0
-1.4
Sodankyl ̈a -0.018 0.016
0.069
-0.00010
-1.1
-0.012
1.5
1.9
Mean
-0.021 -0.0034 0.039 0.000064 -2.9 0.0047 -0.78 -1.8
Table 1: The mean differences in the six primary TCCON Xgas products for each of the
sites participating in the test. Units for each gas are given in the header. Note that not all
sites have data for each month of the test, therefore some of the variation between sites is
due to temporal sampling. The “Mean” row is the mean of the individual sites’ means.
each time that will impact the retrieval, we did not expect a systematic change in the
meteorological variables or the a priori VMR profiles derived from them that would result
in a systematic bias between the two tests. Additionally, the timeseries shown in Fig. 3 and
4 show that, again excepting XCO, the differences are generally consistent month to month.
The changes in XH
2
O are somewhat more variable in July, but this is likely due to more
water in the atmospheric column during the northern hemisphere’s summer.
XCO is the exception; it clearly decreases at the Pasadena, Paris, and (to a lesser extent)
East Trout Lake and Armstrong sites. The larger changes in XCO are due to the larger
impact of the change in meteorology on the CO a priori profiles. CO is one of the gases for
which the a priori VMR profile is interpolated directly from the GEOS product. As shown
in Fig. 5, the Pasadena site has a very large decrease in the CO priors towards the surface.
Similar but smaller decreases can be seen in the Paris and Armstrong profiles, particularly
in July. East Trout Lake also has a decrease in the July CO profiles, but of order 50 ppb,
so it is difficult to see on the scale of Fig. 5. We expect this is due to changes in the GEOS
CO emissions, as Pasadena and Paris are both urban sites, where decreases in the GEOS
CO emissions would clearly have strong impacts on the CO profiles. Armstrong, though not
an urban site, is close enough to Los Angeles that its CO priors experience influence from
Los Angeles emissions at the GEOS grid size. East Trout Lake by contrast is a very rural
site. However, the data where it shows a decrease in XCO occur in July 2023, when there
were major wildfires across much of Canada. Our hypothesis is that the change to GEOS
CO emissions also altered the fire CO emissions, leading to the
∼
50 ppb decrease in lower
atmosphere CO in the East Trout Lake July priors.
3 Changes in remaining a priori profiles
Figures 6 to 10 show the changes in a priori pressure, temperature, CO
2
, CH
4
, and N
2
O.
Pressure and temperature tend to decrease and increase, respectively, near the surface, but
3
Figure 1: Histograms of the changes in XCO
2
(top row), XwCO
2
(second row), XlCO
2
(third
row), and XCH
4
(bottom row) and their original values. The changes are shown in the left
column, the original values in the right. The histograms are normalized such that the sum
of all bins equals 1.
4
Figure 3: Timeseries of the changes in XCO
2
(top row), XwCO
2
(second row), XlCO
2
(third
row), and XCH
4
(bottom row) and their original values. The changes are shown in the left
column, the original values in the right.
6
Figure 5: Mean change in the CO a priori profiles for each of the four months in this study.
Note that not all sites have data for all months.
by relatively small amounts. Changes in the stratosphere are more variable. The CO
2
,
CH
4
, and N
2
O profiles, which derive from the pressure and temperature profiles (along
with potential vorticity) [1] likewise show minor mean changes near the surface, with more
variability in the stratosphere. Given that the troposphere mean changes are small, and the
stratospheric changes tend to oscillate around zero, it makes sense that the changes to these
prior quantities have little mean effect on the retrieved Xgas values.
4 Conclusions
The change from GEOS FP-IT to GEOS IT as the source of a priori meteorological infor-
mation for the TCCON GGG2020 data product will produce some random differences in
individual observations, but little to no systematic difference for XCO
2
, XwCO
2
, XlCO
2
,
XCH
4
, XN
2
O, XH
2
O, and XHDO. Therefore, satellite validation and carbon cycle studies
using these products need not be concerned about a step change in the data, so long as a
reasonable averaging window is used. Studies using the GGG2020 XCO product should be
aware that, especially in urban sites and fire plumes, there likely is a systematic difference.
XCO trends that cross the 1 Apr 2024 transition date should note that urban sites, such as
Pasadena and Paris, may see a 5 to 10 ppb decrease after that date due to the a priori data.
The TCCON algorithm team is aware of the desirability of reprocessing at least urban sites
8
Figure 6: Mean change in the pressure a priori profiles for each of the four months in this
study. Note that not all sites have data for all months.
Figure 7: As Fig. 6, but for temperature.
9
Figure 10: As Fig. 6, but for N
2
O.
with GEOS IT a priori information to remove this step change in XCO and reduce the
high bias in XCO coming from the urban CO a priori profiles. One current limitation is
that doing so would require downloading the full GEOS IT record, which is a prohibitively
large amount of data storage for our system. We will explore other options that would
overcome this limitation, such as remote access of the GEOS IT product through OpenDAP
or migrating to a larger system, in the future. Another limitation is that reprocessing only
some sites with different CO a priori profiles would introduce inconsistency in the GGG2020
product until all sites could accomplish said reprocessing. Thus, a full reconciliation of the
change in CO a priori profiles may be delayed until the next full TCCON reprocessing with
a major update to the GGG retrieval.
Acknowledgements
A portion of this research was carried out at the Jet Propulsion Laboratory (JPL), California Institute of Technology, under
a contract with NASA (80NM0018D0004). Government sponsorship is acknowledged. Support for Caltech TCCON sites and
partial support for JLL provided by NASA grants NNX17AE15G and 80NSSC22K1066.
References
[1] Joshua L. Laughner et al. “A new algorithm to generate a priori trace gas profiles for
the GGG2020 retrieval algorithm”. In:
Atmos. Meas. Tech.
16.5 (Mar. 2023), pp. 1121–
11