Accurate quantification of terrestrial carbon dynamics is essential for assessing ecosystem–climate feedbacks and informing climate mitigation in the Anthropocene. China, as both the world’s largest emitter and a region of rapid ecological change, plays a key role in the global carbon cycle. Yet uncertainties in atmospheric forcing datasets remain a significant source of uncertainty in land surface model simulations and are rarely assessed systematically. Here, we assess China’s terrestrial carbon budget (1979–2014) using the Community Land Model (CLM5.0) driven by three widely used meteorological datasets: CRUNCEP, the Global Soil Wetness Project Phase 3 (GSWP3), and the China Meteorological Forcing Dataset (CMFD), and evaluated against more than 800 FLUXNET site-months and nine eddy covariance towers. Forcing choice strongly alters the magnitude and trend of carbon fluxes, and China’s terrestrial ecosystems acted as either a weak carbon sink or a net source, depending on the forcing. GSWP3 performed best overall in simulating gross primary productivity (GPP), CRUNCEP relatively poorly, and CMFD best in high-altitude and cold-dry regions. Total ecosystem carbon storage was estimated at 86.30–90.00 PgC, primarily in soil (84.1%) and vegetation (15.9%). Interannual GPP variability was mainly controlled by precipitation (29.4%), followed by temperature (17.2%), while shortwave radiation had negative effects (11.5%). Shapley additive explanations analysis further showed that moisture-related variables (precipitation and humidity) dominate interannual GPP variability. By combining process-based modeling and machine learning, this study shows how uncertainties in meteorological inputs affect terrestrial carbon dynamics. These findings underscore the limitations of single-forcing simulations in Earth system modeling and highlight the need for improved meteorological inputs to support robust carbon budgeting and climate neutrality goals.

Read original article

Read original article