Impact of merging of historical and future climate data sets on land carbon cycle projections for South America

This is the final version. Available on open access from Wiley via the DOI in this recordData availability statement: The CRU–JRA data set is available from the CEDA archive of atmospheric data. The HadGEM2-ES climate model is available as one of the CMIP5 models, held by the portal of the Earth System Grid Federation. The code of the JULES model is held at jules.jchmr.org, and its outputs that inform Figures 3-7 and Figure S4 are available upon request from the authors.Earth System Models (ESMs) project climate change, but they often contain biases in their estimates of contemporary climate that propagate into simulated futures. Land models translate climate projections into surface impacts, but these will be inaccurate if ESMs have substantial errors. Bias concerns are relevant for terrestrial physiological processes which often respond non-linearly (i.e. contain threshold responses) and are therefore sensitive to absolute environmental conditions as well as changes. We bias-correct the UK Met Office ESM, HadGEM2-ES, against the CRU–JRA observation-based gridded estimates of recent climate. We apply the derived bias corrections to future projections by HadGEM2-ES for the RCP8.5 scenario of future greenhouse gas concentrations. Focusing on South America, the bias correction includes adjusting for ESM estimates that, annually, are approximately 1 degree too cold, for comparison against 21st Century warming of around 4 degrees. Locally, these values can be much higher. The ESM is also too wet on average, by approximately 1 mm·day−1, which is substantially larger than the mean predicted change. The corrected climate fields force the Joint UK Land Environment Simulator (JULES) dynamic global vegetation model to estimate land surface changes, with an emphasis on the carbon cycle. Results show land carbon sink reductions across South America, and in some locations, the net land–atmosphere CO2 flux becomes a source to the atmosphere by the end of this century. Transitions to a CO2 source is where increases in plant net primary productivity are offset by larger enhancements in soil respiration. Bias-corrected simulations estimate the rise in South American land carbon stocks between pre-industrial times and the end of the 2080s is ∼12 GtC lower than that without climate bias removal, demonstrating the importance of merging historical observational meteorological forcing with ESM diagnostics. We present evidence for a substantial climate-induced role of greater soil decomposition in the fate of the Amazon carbon sink.Newton FundNatural Environment Research Council (NERC