Due to global warming and rising carbon dioxide concentration in the air, global carbon dynamics are changing. The biggest terrestrial organic carbon storage is soil where soil organic carbon (SOC) mineralization plays a crucial rule in the carbon cycle and determines whether soil is a CO2 source or sink. Multiple factors are influencing the mineralization, but it is not totally clear how, especially with changing climate conditions. Hence it is important to understand spatial and temporal patterns of carbon dynamics.
A lot of predictors have already been determined based on theories and empirical studies, but they all depend on certain assumptions. Here, we present a new approach to model SOC mineralization with a machine learning algorithm. The model will be based on remote sensing data (soil maps, climate data, land use) and field measurements of heterotrophic soil respiration and its major drivers (soil moisture and temperature). With this data the algorithm will “learn” the relationship between carbon mineralization, climate, soil type and character within one meter instead of relying on assumptions. Hence carbon4D will provide the first near real-time modelling framework of SOC mineralization rates and soil CO2 efflux in all 4 dimensions, which will provide new insights into patterns and controlling factors.