Abstract
Agricultural soils are essential for food security, but are increasingly affected by soil degradation and the impacts of climate change. The sequestration of soil organic carbon (SOC) is recognized as an effective strategy to improve soil quality, mitigate climate change, and increase resilience to droughts and extreme weather events. EU policies such as the Carbon Removals and Carbon Farming (CRCF) regulation aim to promote climate-smart agricultural practices but require robust and cost-efficient systems to verify SOC changes. Remote sensing offers the potential to derive large-scale information on SOC status and trends. However, its application for soil monitoring is currently limited by the lack of validation for the temporal model accuracy. This dissertation investigates the potential of multispectral soil reflectance composites (SRC), derived from Sentinel-2 and Landsat time series, for large-scale soil mapping and spatiotemporal SOC modeling. The results show that SRCs are strong predictors of cropland soil properties and enable the generation of high-resolution SOC maps. Correlations between SRC bands and soil properties such as SOC and clay are consistent with findings from soil spectroscopy, supporting the calibration of robust and transferable models. Using high-quality reference samples from the German Agricultural Soil Inventory, it was demonstrated that the selection of spectral indices and thresholding values significantly affects the spectral quality and bias of the generated SRC. In areas with heterogeneous soil conditions, the prediction accuracy can be further improved by calibrating local sub-models to account for regional variability in the soil signal. Based on harmonized Landsat and Sentinel-2 time series from 1986 to 2022, a spatiotemporal model was developed to assess the potential of remote sensing data to detect SOC changes in cropland soils. For the first time, SOC trend predictions were validated using repeated measurements from long-term soil monitoring sites. While model accuracy was not sufficient to replace direct sampling, the validation showed a low confusion rate between increasing and decreasing SOC trends, underlining the potential of satellite-based models for long-term soil monitoring. The findings suggest that remote sensing can improve the cost-efficiency of SOC monitoring by enabling targeted sampling at locations where predicted trends diverge from expected effects of climate-smart management. Further research should aim to (1) improve the validity of modeled SOC dynamics by integrating additional soil data on bulk density and vertical SOC distribution, and (2) further advance the methods for the retrieval of soil reflectance to increase the signal-to-noise ratio and shorten the required time frames to detect significant SOC changes.
Author and committee
dc:creator, dc:contributor.*- Author
-
- Brög, Tom
Identifiers
dc:identifier.*- Identifier
- hdl:10900/172864