HySOC · Combining knowledge-guided process models and data-driven machine learning models for enhancing soil carbon modelling in space and time

Accurate modelling and mapping soil organic carbon (SOC) variations in space and time are crucial for supporting soil health restoration, guiding sustainable land management practices, and contributing to climate change mitigation strategies. Although data-driven approaches, such as machine learning (ML), has attracted many attention in soil mapping tasks for its powerful ability to learn from data, it is better at capturing soil spatial variation than soil temporal dynamics. By contrast, process-based models benefit from mechanistic knowledge to express physical, chemical and biological processes that govern SOC temporal changes. Therefore, integrating two types of models will promise means to represent physically plausible SOC dynamics while retaining the spatial prediction accuracy of ML models. However, a clear research gap remains in applying this concept and designing an appropriate hybrid modelling framework to quantify soil variations in space and time. This HySOC (Hybrid modelling of SOC) project aims to propose a new paradigm of space-time SOC modelling framework that hybrid knowledge-guided process models and data-driven ML models, for fulfilling the needs of the next generation of soil mapping methods and for supporting future carbon–climate feedback projections. By bridging two types of model, the project will go beyond the state of the art by providing high spatial and temporal resolution SOC maps and also the open-access hybrid model architecture, algorithms and calibrated model parameters, thus reducing the uncertainty on our understanding of European soil carbon changes in the past, present, and future. On top of model innovation, the outcomes of HySOC will further help to assess the sensitivity of European soil carbon to multiple global change drivers, and enhance the capacity to diagnose and improve the representations of soil variations in Earth system models, thereby contributing to the goals of the EU Soil Mission and the European Green Deal.