Modeling of Greenhouse Gas Emissions from Peatlands in Germany: A Machine Learning Approach

The ITMS (Integriertes Treibhausgas Monitoring System) Sources and Sinks module, funded by the German Federal Ministry of Research, Technology and Space, develops modelling approaches to simulate greenhouse gas (GHG) fluxes in Germany at high spatial and temporal resolution. By integrating existing measurement data from national and Bavarian research initiatives with new field observations from natural, drained, and rewetted peatlands collected in the MODELPEAT project, we aim to refine statistical modeling approaches of peatland GHG exchange. While the current German national GHG inventory approach for landuse specific peatlands relies on functional relationships in dependency on water table depth and the type of organic soil (Tiemeyer et al. 2020), this project introduces a machine learning framework that leverages an extensive monthly dataset (approximately 190 site years) to capture peatland GHG dynamics in more detail. The poster presents the methodological implementation of a eXtreme Gradient Boosting (XGB) decision tree model, which incorporates predictors representing seasonal dynamics, vegetation activity, meteorological conditions, and management practices, along with initial findings. As the project progresses, the approach is aimed to be applied across Bavaria on a 30×30 m grid to generate spatially explicit simulations of peatland GHG fluxes (CO₂, CH₄, N₂O). This work is essential for identifying emission hotspots and supporting the development of effective mitigation strategies.