Mind the Gap: Reconciling tropical forest carbon flux estimates from Global Earth Observation and National Reporting

Abstract Background Different methods estimating the global anthropogenic land flux, which is dominated by forest-related activities, vary in magnitude and direction with respect to whether the land is a net source or sink. One reason for these variations is the extent to which methods consider land to be “managed”, thus contributing to the anthropogenic flux. Earth Observation (EO) datasets characterising spatio-temporal changes in land cover and carbon stocks provide an independent approach to flux estimations that can be compared against National Greenhouse Gas Inventories (NGHGIs) to support accurate and timely monitoring, reporting and verification capacity. Using Brazil as a primary case study, with additional analysis in Indonesia and Malaysia, we compare EO-based estimates of forest fluxes to NGHGIs. Results Between 2001 and 2020, the EO-derived estimates of all forest-related emissions and removals indicate that Brazil was a net sink of carbon (-0.2 GtCO 2 yr − 1 ), while Brazil’s NGHGI reports a net carbon source (+ 0.8 GtCO 2 yr − 1 ). After adjusting the Global EO estimate to accommodate the Brazilian NGHGI definition of managed forest and other assumptions used in the inventory’s methodology, the Global EO net flux became a source of + 0.6 GtCO 2 yr − 1 , comparable to the NGHGI. Remaining discrepancies are largely due to differences in carbon removal factors and forest types considered in the two datasets. In Indonesia, Global EO and NGHGI net flux estimates were similar (+ 0.6 GtCO 2 yr − 1 ), but in Malaysia, they differed in both magnitude and sign (NGHGI: -0.2 GtCO 2 yr − 1 ; Global EO: +0.2 GtCO 2 yr − 1 ). Spatially explicit datasets on forest types were not publicly available in either country, limiting the possibility of detailed analyses and adjustments. Conclusions By adjusting the Global EO dataset to improve comparability with carbon fluxes estimated for managed forests in the Brazilian NGHGI, initially diverging estimates were largely reconciled and remaining differences explained. Despite limited spatial data in Indonesia and Malaysia, our comparison indicated where differing approaches may explain uncertainties and inaccuracies. Our study highlights that comparing Global EO and NGHGIs is a useful exercise to improve both datasets towards greater accuracy and alignment, provided that sufficiently transparent and complete information is available for such detailed analyses.