The extending archive of the Greenhouse Gases Observing Satellite (GOSAT) measurements (now covering about 6 years) allows increasingly robust statistics to be computed, that document the performance ...of the corresponding retrievals of the column-average dry air-mole fraction of CO2 (XCO2). Here, we demonstrate that atmospheric inversions cannot be rigorously optimal when assimilating current XCO2 retrievals, even with averaging kernels, in particular because retrievals and inversions use different assumption about prior uncertainty. We look for some practical evidence of this sub-optimality from the view point of atmospheric inversion by comparing a model simulation constrained by surface air-sample measurements with one of the GOSAT retrieval products (NASA's ACOS). The retrieval-minus-model differences result from various error sources, both in the retrievals and in the simulation: we discuss the plausibility of the origin of the major patterns. We find systematic retrieval errors over the dark surfaces of high-latitude lands and over African savannahs. More importantly, we also find a systematic over-fit of the GOSAT radiances by the retrievals over land for the high-gain detector mode, which is the usual observation mode. The over-fit is partially compensated by the retrieval bias-correction. These issues are likely common to other retrieval products and may explain some of the surprising and inconsistent CO2 atmospheric inversion results obtained with the existing GOSAT retrieval products. We suggest that reducing the observation weight in the retrieval schemes (for instance so that retrieval increments to the retrieval prior values are halved for the studied retrieval product) would significantly improve the retrieval quality and reduce the need for (or at least reduce the complexity of) ad-hoc retrieval bias correction.
The global land and ocean carbon sinks have increased proportionally with increasing carbon dioxide emissions during the past decades. It is thought that Northern Hemisphere lands make a dominant ...contribution to the global land carbon sink; however, the long-term trend of the northern land sink remains uncertain. Here, using measurements of the interhemispheric gradient of atmospheric carbon dioxide from 1958 to 2016, we show that the northern land sink remained stable between the 1960s and the late 1980s, then increased by 0.5 ± 0.4 petagrams of carbon per year during the 1990s and by 0.6 ± 0.5 petagrams of carbon per year during the 2000s. The increase of the northern land sink in the 1990s accounts for 65% of the increase in the global land carbon flux during that period. The subsequent increase in the 2000s is larger than the increase in the global land carbon flux, suggesting a coincident decrease of carbon uptake in the Southern Hemisphere. Comparison of our findings with the simulations of an ensemble of terrestrial carbon models over the same period suggests that the decadal change in the northern land sink between the 1960s and the 1990s can be explained by a combination of increasing concentrations of atmospheric carbon dioxide, climate variability and changes in land cover. However, the increase during the 2000s is underestimated by all models, which suggests the need for improved consideration of changes in drivers such as nitrogen deposition, diffuse light and land-use change. Overall, our findings underscore the importance of Northern Hemispheric land as a carbon sink.
The extending archive of the Greenhouse Gases Observing Satellite (GOSAT) measurements (now covering about 6 years) allows increasingly robust statistics to be computed, that document the performance ...of the corresponding retrievals of the column-average dry air-mole fraction of CO2 (XCO2). Here, we demonstrate that atmospheric inversions cannot be rigorously optimal when assimilating current XCO2 retrievals, even with averaging kernels, in particular because retrievals and inversions use different assumption about prior uncertainty. We look for some practical evidence of this sub-optimality from the view point of atmospheric inversion by comparing a model simulation constrained by surface air-sample measurements with one of the GOSAT retrieval products (NASA's ACOS). The retrieval-minus-model differences result from various error sources, both in the retrievals and in the simulation: we discuss the plausibility of the origin of the major patterns. We find systematic retrieval errors over the dark surfaces of high-latitude lands and over African savannahs. More importantly, we also find a systematic over-fit of the GOSAT radiances by the retrievals over land for the high-gain detector mode, which is the usual observation mode. The over-fit is partially compensated by the retrieval bias-correction. These issues are likely common to other retrieval products and may explain some of the surprising and inconsistent CO2 atmospheric inversion results obtained with the existing GOSAT retrieval products. We suggest that reducing the observation weight in the retrieval schemes (for instance so that retrieval increments to the retrieval prior values are halved for the studied retrieval product) would significantly improve the retrieval quality and reduce the need for (or at least reduce the complexity of) ad-hoc retrieval bias correction.
Atmospheric concentration measurements are used to adjust the daily to monthly budget of fossil fuel CO2 emissions of the Paris urban area from the prior estimates established by the Airparif local ...air quality agency. Five atmospheric monitoring sites are available, including one at the top of the Eiffel Tower. The atmospheric inversion is based on a Bayesian approach, and relies on an atmospheric transport model with a spatial resolution of 2 km with boundary conditions from a global coarse grid transport model. The inversion adjusts prior knowledge about the anthropogenic and biogenic CO2 fluxes from the Airparif inventory and an ecosystem model, respectively, with corrections at a temporal resolution of 6 h, while keeping the spatial distribution from the emission inventory. These corrections are based on assumptions regarding the temporal autocorrelation of prior emissions uncertainties within the daily cycle, and from day to day. The comparison of the measurements against the atmospheric transport simulation driven by the a priori CO2 surface fluxes shows significant differences upwind of the Paris urban area, which suggests a large and uncertain contribution from distant sources and sinks to the CO2 concentration variability. This contribution advocates that the inversion should aim at minimising model-data misfits in upwind-downwind gradients rather than misfits in mole fractions at individual sites. Another conclusion of the direct model-measurement comparison is that the CO2 variability at the top of the Eiffel Tower is large and poorly represented by the model for most wind speeds and directions. The model's inability to reproduce the CO2 variability at the heart of the city makes such measurements ill-suited for the inversion. This and the need to constrain the budgets for the whole city suggests the assimilation of upwind-downwind mole fraction gradients between sites at the edge of the urban area only. The inversion significantly improves the agreement between measured and modelled concentration gradients. Realistic emissions are retrieved for two 30-day periods and suggest a significant overestimate by the AirParif inventory. Similar inversions over longer periods are necessary for a proper evaluation of the optimised CO2 emissions against independent data.
The variational formulation of Bayes' theorem allows inferring CO 2 sources and sinks from atmospheric concentrations at much higher time-space resolution than the ensemble or analytical approaches. ...However, it usually exhibits limited scalable parallelism. This limitation hinders global atmospheric inversions operated on decadal time scales and regional ones with kilometric spatial scales because of the computational cost of the underlying transport model that has to be run at each iteration of the variational minimization. Here, we introduce a physical parallelization (PP) of variational atmospheric inversions. In the PP, the inversion still manages a single physically and statistically consistent window, but the transport model is run in parallel overlapping sub-segments in order to massively reduce the computation wall-clock time of the inversion. For global inversions, a simplification of transport modelling is described to connect the output of all segments. We demonstrate the performance of the approach on a global inversion for CO 2 with a 32 yr inversion window (1979-2010) with atmospheric measurements from 81 sites of the NOAA global cooperative air sampling network. In this case, we show that the duration of the inversion is reduced by a seven-fold factor (from months to days), while still processing the three decades consistently and with improved numerical stability.
Atmospheric CO2 inversions estimate surface carbon fluxes from an optimal fit to atmospheric CO2 measurements, usually including prior constraints on the flux estimates. Eleven sets of carbon flux ...estimates are compared, generated by different inversions systems that vary in their inversions methods, choice of atmospheric data, transport model and prior information. The inversions were run for at least 5 yr in the period between 1990 and 2010. Mean fluxes for 2001-2004, seasonal cycles, interannual variability and trends are compared for the tropics and northern and southern extra-tropics, and separately for land and ocean. Some continental/basin-scale subdivisions are also considered where the atmospheric network is denser. Four-year mean fluxes are reasonably consistent across inversions at global/latitudinal scale, with a large total (land plus ocean) carbon uptake in the north (-3.4 Pg C yr-1 (±0.5 Pg C yr-1 standard deviation), with slightly more uptake over land than over ocean), a significant although more variable source over the tropics (1.6 ± 0.9 Pg C yr-1 ) and a compensatory sink of similar magnitude in the south (-1.4 ± 0.5 Pg C yr-1 ) corresponding mainly to an ocean sink. Largest differences across inversions occur in the balance between tropical land sources and southern land sinks. Interannual variability (IAV) in carbon fluxes is larger for land than ocean regions (standard deviation around 1.06 versus 0.33 Pg C yr-1 for the 1996-2007 period), with much higher consistency among the inversions for the land. While the tropical land explains most of the IAV (standard deviation ~ 0.65 Pg C yr-1 ), the northern and southern land also contribute (standard deviation ~ 0.39 Pg C yr-1 ). Most inversions tend to indicate an increase of the northern land carbon uptake from late 1990s to 2008 (around 0.1 Pg C yr-1 , predominantly in North Asia. The mean seasonal cycle appears to be well constrained by the atmospheric data over the northern land (at the continental scale), but still highly dependent on the prior flux seasonality over the ocean. Finally we provide recommendations to interpret the regional fluxes, along with the uncertainty estimates.
Negative trends of carbon monoxide (CO) concentrations are observed in the recent decade by both surface measurements and satellite retrievals over many regions of the globe, but they are not well ...explained by current emission inventories. Here, we analyse the observed CO concentration decline with an atmospheric inversion that simultaneously optimizes the two main CO sources (surface emissions and atmospheric hydrocarbon oxidations) and the main CO sink (atmospheric hydroxyl radical OH oxidation). Satellite CO column retrievals from Measurements of Pollution in the Troposphere (MOPITT), version 6, and surface observations of methane and methyl chloroform mole fractions are assimilated jointly for the period covering 2002–2011. Compared to the model simulation prescribed with prior emission inventories, trends in the optimized CO concentrations show better agreement with that of independent surface in situ measurements. At the global scale, the atmospheric inversion primarily interprets the CO concentration decline as a decrease in the CO emissions (−2.3 % yr−1), more than twice the negative trend estimated by the prior emission inventories (−1.0 % yr−1). The spatial distribution of the inferred decrease in CO emissions indicates contributions from western Europe (−4.0 % yr−1), the United States (−4.6 % yr−1) and East Asia (−1.2 % yr−1), where anthropogenic fuel combustion generally dominates the overall CO emissions, and also from Australia (−5.3 % yr−1), the Indo-China Peninsula (−5.6 % yr−1), Indonesia (−6.7 % y−1), and South America (−3 % yr−1), where CO emissions are mostly due to biomass burning. In contradiction with the bottom-up inventories that report an increase of 2 % yr−1 over China during the study period, a significant emission decrease of 1.1 % yr−1 is inferred by the inversion. A large decrease in CO emission factors due to technology improvements would outweigh the increase in carbon fuel combustions and may explain this decrease. Independent satellite formaldehyde (CH2O) column retrievals confirm the absence of large-scale trends in the atmospheric source of CO. However, it should be noted that the CH2O retrievals are not assimilated and OH concentrations are optimized at a very large scale in this study.
The Global Carbon Budget 2018 (GCB2018) estimated by the atmospheric CO
2 growth rate, fossil fuel emissions, and modeled (bottom‐up) land and ocean fluxes cannot be fully closed, leading to a ...“budget imbalance,” highlighting uncertainties in GCB components. However, no systematic analysis has been performed on which regions or processes contribute to this term. To obtain deeper insight on the sources of uncertainty in global and regional carbon budgets, we analyzed differences in Net Biome Productivity (NBP) for all possible combinations of bottom‐up and top‐down data sets in GCB2018: (i) 16 dynamic global vegetation models (DGVMs), and (ii) 5 atmospheric inversions that match the atmospheric CO
2 growth rate. We find that the global mismatch between the two ensembles matches well the GCB2018 budget imbalance, with Brazil, Southeast Asia, and Oceania as the largest contributors. Differences between DGVMs dominate global mismatches, while at regional scale differences between inversions contribute the most to uncertainty. At both global and regional scales, disagreement on NBP interannual variability between the two approaches explains a large fraction of differences. We attribute this mismatch to distinct responses to El Niño–Southern Oscillation variability between DGVMs and inversions and to uncertainties in land use change emissions, especially in South America and Southeast Asia. We identify key needs to reduce uncertainty in carbon budgets: reducing uncertainty in atmospheric inversions (e.g., through more observations in the tropics) and in land use change fluxes, including more land use processes and evaluating land use transitions (e.g., using high‐resolution remote‐sensing), and, finally, improving tropical hydroecological processes and fire representation within DGVMs.
Key Points
Top‐down and bottom‐up estimates of net land‐atmosphere CO
2 fluxes agree well globally but show important mismatches at regional scales
Regional mismatches are dominated by differences between inversions and interannual variability in CO
2 fluxes
Mismatches between top‐down and bottom‐up data sets are explained by sensitivity to climate and by uncertainty in land use change forcing
Increasing atmospheric carbon dioxide (CO2) is the principal driver of anthropogenic climate change. Asia is an important region for the global carbon budget, with 4 of the world's 10 largest ...national emitters of CO2. Using an ensemble of seven atmospheric inverse systems, we estimated land biosphere fluxes (natural, land-use change and fires) based on atmospheric observations of CO2 concentration. The Asian land biosphere was a net sink of -0.46 (-0.70-0.24) PgC per year (median and range) for 1996-2012 and was mostly located in East Asia, while in South and Southeast Asia the land biosphere was close to carbon neutral. In East Asia, the annual CO2 sink increased between 1996-2001 and 2008-2012 by 0.56 (0.30-0.81) PgC, accounting for ∼35% of the increase in the global land biosphere sink. Uncertainty in the fossil fuel emissions contributes significantly (32%) to the uncertainty in land biosphere sink change.