Comparison of Regional Simulation of Biospheric CO2 Flux from the Updated Version of CarbonTracker Asia with FLUXCOM and Other Inversions over Asia

Abstract

<p>There are still large uncertainties in the estimates of net ecosystem exchange of CO<sub>2</sub><br>(NEE) with atmosphere in Asia, particularly in the boreal and eastern part of temperate Asia. To<br>understand these uncertainties, we assessed the CarbonTracker Asia (CTA2017) estimates of the<br>spatial and temporal distributions of NEE through a comparison with FLUXCOM and the global<br>inversion models from the Copernicus Atmospheric Monitoring Service (CAMS), Monitoring<br>Atmospheric Composition and Climate (MACC), and Jena CarboScope in Asia, as well as<br>examining the impact of the nesting approach on the optimized NEE flux during the 2001–2013<br>period. The long‐term mean carbon uptake is reduced in Asia, which is −0.32 ± 0.22 PgC yr<sup>‐1</sup>,<br>whereas –0.58 ± 0.26 PgC yr<sup>‐1</sup> is shown from CT2017 (CarbonTracker global). The domain<br>aggregated mean carbon uptake from CTA2017 is found to be lower by 23.8%, 44.8%, and 60.5%<br>than CAMS, MACC, and Jena CarboScope, respectively. For example, both CTA2017 and CT2017<br>models captured the interannual variability (IAV) of the NEE flux with a different magnitude and<br>this leads to divergent annual aggregated results. Differences in the estimated interannual<br>variability of NEE in response to El Niño–Southern Oscillation (ENSO) may result from<br>differences in the transport model resolutions. These inverse models’ results have a substantial<br>difference compared to FLUXCOM, which was found to be –5.54 PgC yr<sup>‐1</sup>. On the one hand, we<br>showed that the large NEE discrepancies between both inversion models and FLUXCOM stem<br>mostly from the tropical forests. On the other hand, CTA2017 exhibits a slightly better correlation<br>with FLUXCOM over grass/shrub, fields/woods/savanna, and mixed forest than CT2017. The land<br>cover inconsistency between CTA2017 and FLUXCOM is therefore one driver of the discrepancy in<br>the NEE estimates. The diurnal averaged NEE flux between CTA2017 and FLUXCOM exhibits<br>better agreement during the carbon uptake period than the carbon release period. Both CTA2017<br>and CT2017 revealed that the overall spatial patterns of the carbon sink and source are similar, but<br>the magnitude varied with seasons and ecosystem types, which is mainly attributed to differences<br>in the transport model resolutions. Our findings indicate that substantial inconsistencies in the<br>inversions and FLUXCOM mainly emerge during the carbon uptake period and over tropical<br>forests. The main problems are underrepresentation of FLUXCOM NEE estimates by limited eddy<br>covariance flux measurements, the role of CO<sub>2</sub> emissions from land use change not accounted for<br>by FLUXCOM, sparseness of surface observations of CO<sub>2</sub> concentrations used by the assimilation<br>systems, and land cover inconsistency. This suggested that further scrutiny on the FLUXCOM and<br>inverse estimates is most likely required. Such efforts will reduce inconsistencies across various<br>NEE estimates over Asia, thus mitigating ecosystem‐driven errors that propagate the global<br>carbon budget. Moreover, this work also recommends further investigation on how the<br>changes/updates made in CarbonTracker affect the interannual variability of the aggregate and<br>spatial pattern of NEE flux in response to the ENSO effect over the region of interest.</p>