Combining GOSAT XCO₂ observations over land and ocean to improve regional CO₂ flux estimates

Abstract

We used the GEOS‐Chem data assimilation system to examine the impact of combining Greenhouse Gases Observing Satellite (GOSAT) XCO2 data over land and ocean on regional CO2 flux estimates for 2010–2012. We found that compared to assimilating only land data, combining land and ocean data produced an a posteriori CO2 distribution that is in better agreement with independent data and fluxes that are in closer agreement with existing top‐down and bottom‐up estimates. Adding XCO2 data over oceans changed the tropical land regions from a source of 0.64 Pg C/yr to a sink of −0.60 Pg C/yr and produced a corresponding reduction in the estimated sink in northern and southern land regions by 0.49 Pg C/yr and 0.80 Pg C/yr, respectively. This highlights the importance of improved observational coverage in the tropics to better quantify the latitudinal distribution of the terrestrial fluxes. Based only on land XCO2 data, we estimated a strong source in northern tropical South America, which experienced wet conditions in 2010–2012. In contrast, with the land and ocean data, we estimated a sink for this wet region in the north, and a source for the seasonally dry regions in the south and east, which is consistent with our understanding of the impact of moisture availability on the carbon balance of the region. Our results suggest that using satellite data with a more zonally balanced observational coverage could help mitigate discrepancies in CO2 flux estimates; further improvement could be expected with the greater observational coverage provided by the Orbiting Carbon Observatory‐2.

A global synthesis inversion analysis of recent variability in CO<sub>2</sub> fluxes using GOSAT and in situ observations

“…They, however, used the SRON-KIT RemoTeC GOSAT retrieval with a known issue over the ocean, and concluded that adding global land and ocean observation bias correction terms to their inversion was needed to make the land-ocean flux split more realistic and to improve the seasonal cycle of CO 2 in the southern extratropics. In contrast, studies have found no noticeable bias in the ACOS B3.5 ocean glint XCO 2 retrievals relative to TCCON and a mean bias of only −0.06 ppm relative to HIPPO ; the B3.4 version we use is on average ∼ 0.2 ppm lower than B3.5 in 2010 (Deng et al, 2016). So although a small overall negative bias in the biascorrected ACOS B3.4 ocean data cannot be ruled out (and there could of course be larger negative biases on a regional scale, such as in the southern extratropics), we conclude that the land-ocean flux split in inversions using either in situ or GOSAT data is strongly influenced by error correlations and dependent on the prior uncertainties assumed.…”

“…For constraining fluxes at relatively high temporal resolution, observations are chosen that consist of discrete wholeair samples collected in glass flasks approximately weekly and continuous in situ tall tower measurements of CO 2 mole fraction from the NOAA ESRL Carbon Cycle Cooperative Global Air Sampling Network (Dlugokencky et al, 2013;Andrews et al, 2009) Tsutsumi et al, 2006). Both data sets are calibrated to the WMO-X2007 scale.…”

“…Making measurements since 2009, GOSAT is the first satellite in orbit designed specifically to measure column mixing ratios of CO 2 (as well as methane) with substantial sensitivity to the lower troposphere, close to surface fluxes. A number of modeling groups have conducted CO 2 flux inversions using synthetic GOSAT data (Liu et al, 2014) and actual data (Takagi et al, 2011(Takagi et al, , 2014Maksyutov et al, 2013;Basu et al, 2013;Saeki et al, 2013a;Deng et al, 2014Deng et al, , 2016Chevallier et al, 2014;Reuter et al, 2014;Houweling et al, 2015). Unlike in situ measurements, which are calibrated directly for the gas of interest, remote sensing involves challenges in precision and accuracy stemming from the measuring of radiance.…”

Coupling the Canadian Terrestrial Ecosystem Model (CTEM v. 2.0) to Environment and Climate Change Canada's greenhouse gas forecast model (v.107-glb)

“…There will always be an unsatisfactory length of analyses available for a TEM spin-up period whenever an operational weather forecast system is involved. Moreover, greenhouse gas assimilation systems are constrained (by time, computational expense, and the observing system) and thus often focus on a few years of study at one time (e.g., Deng et al, 2014Deng et al, , 2016. Thus, the challenge is to merge this small dataset into the spin-up procedure used for the TEM.…”

“…In the 4D-Var system, the adjoint of the GEOS-Chem model is used to optimize the fluxes. Details of the GEOS-Chem adjoint model are given in Henze et al (2007) and a description of its application for inverse modeling of atmospheric CO 2 is provided in Deng et al (2014Deng et al ( , 2016.…”