Uncertainty in Amazon vegetation productivity in CMIP6 projections driven by surface energy fluxes

Abstract. The Amazon basin rainforest is a critical component of the climate system, currently representing 25 % of terrestrial carbon gains and storing 150 to 200 billion tonnes of carbon. If and by which extent the Amazon rainforest will remain a net carbon sink is an open scientific question, motivated by the unexplained diversity across Earth System Model (ESM) results. Specifically, divergent responses are observed in Amazon vegetation productivity projections, especially under sustained global warming scenarios. We explore this inter-model diversity in projected Amazon vegetation in CMIP6 historical and ssp585 scenario simulations with thirteen ESM by explicitly accounting for the relative contributions of changes in the El Niño-Southern Oscillation (ENSO) and local mean-state climate changes. Our results demonstrate the dominant role of local mean-state climatic changes in shaping the response of the Amazon carbon cycle for 7 out of 13 ESM, with only a minor role for changes in ENSO and its teleconnection despite the strong inter-model diversity in representing ENSO. While temperature and water availability influence displays a high inter-model agreement, the most critical local processes determining uncertainty and divergence across ESM responses within the Amazon basin are the surface energy balance components, in particular shortwave incoming radiation and latent heat fluxes. We identify the main sources of model specificities in land scheme parameterizations, especially the incorporation of Phosphorous limitation, which leads to a stronger reduction of vegetation productivity under strong warming scenarios. We therefore advocate for increased focus from modelling groups towards a more accurate and consistent representation of surface radiative and turbulent fluxes in the Amazon region. Additionally, we hypothesize that a uniform incorporation of Phosphorous limitation across all the ESM may contribute to minimize the uncertainties. This dual approach can lead to more robust estimates of vegetation productivity within the Amazon basin across different climate change scenarios.