Project
Modelling and mapping soil organic carbon in space and time using machine learning and mechanistic approaches
This research explores the distribution of soil organic carbon over time and space using a hybrid approach that combines machine learning and process-based models. The goal is to predict SOC at national and continental scales, evaluating the model performance and assessing the impact of factors such as spatial resolution, sampling density, and complexity. The results of this study will provide insights for sustainable land management to improve soil health. Case studies include the Netherlands and the European Union.
Soil organic carbon is the largest terrestrial carbon reservoir. It plays an important role in governing soil functions and global carbon cycling processes. However, approximately 60-70% of soils in the European Union are unhealthy, and 25-30% of soils experience organic carbon loss caused by current management practices, among others. Healthy soils are the foundation of the food system. About 95% of our food comes from terrestrial sources. The World Resources Report outlines that the global food demand is expected to rise as much as 50 percent by 2050, which emphasizes the pressure on food security and the need for sustainable land management. The European Commission aims to achieve healthy soils by 2050, and its mission is in line with other important European initiatives such as the Green Deal and the EU Farm-to-Fork Strategy. Within this context, questions arise about how we can obtain detailed soil information, especially for the spatial-temporal variation of soil organic carbon due to its importance in climate mitigation and the food system.
Introduction and background
Soil organic carbon is the largest terrestrial carbon reservoir. It plays an important role in governing soil functions and global carbon cycling processes. However, approximately 60-70% of soils in the European Union are unhealthy, and 25-30% of soils experience organic carbon loss caused by current management practices, among others. Healthy soils are the foundation of the food system. About 95% of our food comes from terrestrial sources. The World Resources Report outlines that the global food demand is expected to rise as much as 50 percent by 2050, which emphasizes the pressure on food security and the need for sustainable land management. The European Commission aims to achieve healthy soils by 2050, and its mission is in line with other important European initiatives such as the Green Deal and the EU Farm-to-Fork Strategy. Within this context, questions arise about how we can obtain detailed soil information, especially for the spatial-temporal variation of soil organic carbon due to its importance in climate mitigation and the food system.
Project description
This PhD study aims to build and compare several hybrid models which combine machine learning and process-based modelling to model the spatial and temporal distribution of soil organic carbon on a national and continental scales. The best performing hybrid model will be further used to assess the effect of spatial resolution, sampling density and complexity level of the process-based model on model performance. Finally, the hybrid model will be extended to forecast soil carbon variability under different future management and climate scenarios at the continental scale. The generated knowledge will be further utilized to support the promotion of soil health and sustainable land management. Case studies will focus on the Netherlands and the European Union.