Earth System Models are not capturing observed tropical forest carbon dynamics

The land surface is absorbing carbon from the atmosphere. Tropical forests play a key role in this carbon uptake. Recent analyses of 565 long-term forest inventory plots across Africa and Amazonia show that structurally intact tropical forest are a large carbon sink, but that this sink has recently saturated and is projected to be in long-term decline. Here we compare these results with estimates from the two most recent generations of Earth System Models, CMIP5 (19 models) and CMIP6 (17 models). We show that, while CMIP5 and CMIP6 are of similar skill, they do not reproduce the observed 1985-2014 carbon dynamics. The pan-tropical net sink from inventory data is 0.99 Pg C yr^-1 (95% CI 0.7–1.3) between 2000–2010, the best sampled decade, double the CMIP6 multimodel-mean of 0.45 Pg C yr^-1 (95% CI 0.35–0.55) over the same decade. The observed saturating and declining sink beginning in Amazonia in the 1990s and Africa in the 2010s is not captured by the models, which show modest increases in sink strength. The future pan-tropical net sink from the statistical model decreases by 0.23 Pg C per decade (95% CI 0.09–0.39) until the 2030s, while CMIP6 multimodel-means project an increasing carbon sink under all scenarios (0.01–0.03 Pg C per decade; 95% CI 0.00–0.06) except the low-CO₂ scenario (-0.02 Pg C per decade; 95% CI -0.01–0.03). CMIP models balance positive CO₂ fertilisation with negative responses to higher temperatures and droughts on carbon gains from tree growth similarly to observations, but the modelling of carbon losses from tree mortality does not correspond well to the inventory data. Reason for the model-observation differences is the treatment of mortality in models. Integrated research programs combining continued tropical forest monitoring and targeted experiments are needed to reduce the uncertainties in this key carbon cycle feedback in next-generation models.