How does global warming affect European SOC stocks? A data-driven space-for-time approach

Abstract

Global warming effects on soil organic carbon (SOC) stocks are expected to be site specific but current process-based models still struggle to forecast spatially explicit long-term trends properly. This study estimated such long-term effects of global warming on European SOC stocks using a novel data-driven space-for-time approach. In principle, this approach estimated site-specific SOC stocks under future climate from SOC stocks in comparable soils but in regions that are already exposed to such climate today. About 20k observations of the LUCAS soil dataset were used to train a machine learning model that predicted SOC stocks from current climate as well as static environmental properties (e.g. geology, soil type, soil texture). Then, this SOC model was used to forecast future SOC stocks in Europe under various CMIP6 climate scenarios. Preliminary results suggest Europe’s top 20 cm of mineral soil to loose on average 2 to 6 Mg SOC ha-1 by the end of this century. But global warming-induced changes in SOC showed pronounced regional differences. SOC was anticipated to even rise under global warming in some areas, particularly in Northern European forest ecosystems. In vast parts of southern Europe, unprecedented future climate limited the applicability of the data-driven SOC model. This was the case for up to 49% of all sites in the most extreme climate scenario. In contrast, for the remaining 51% of sites in all climatic scenarios, equivalent "soil-climate twins" could be successfully located elsewhere in contemporary Europe. It is proposed that outcomes from data-driven space-for-time models could complement and act as cross-checks for process-based modelling outputs to gain confidence in long-term projections of SOC stocks under global warming.