Tom Broeg

Agricultural soils are increasingly under pressure due to land use intensification and the ongoing effects of climate change. Current EU policies, such as the "Carbon Removals and Carbon Farming Regulation" (CRCF), aim to improve the resilience of cropland soils by carbon sequestration through climate-smart management. However, such regulations significantly increase the demand for spatiotemporal soil data to monitor and verify the effectiveness of carbon farming measures.In recent years, the analysis of remote sensing-based bare soil observations has been increasingly used to generate accurate, high-resolution maps of cropland properties, such as soil organic carbon (SOC). However, due to the lack of robust reference data and the slow-changing nature of SOC, validating temporal model performance remains challenging. In this study, we tested the extent to which spatiotemporal models based on satellite data can support wall-to-wall soil monitoring and provide information on the temporal variability of SOC in cropland soils.To achieve this, bare soil composites were derived from Landsat and Sentinel-2 data using a moving window approach and compared to Bavarian long-term soil monitoring data from 1986 to 2022. The results showed that while overall model performance was high, the validation of measured SOC trends yielded significantly lower accuracy, underlining the high uncertainty in predicting temporal soil carbon dynamics. While long-term analyses of 25+ years were necessary to detect significant SOC changes in most cases, the classification of the results revealed a low confusion rate between sites with increasing or decreasing SOC trends across the observation period.These findings are supported by recent results based on the repetition of the German agricultural soil inventory, currently being conducted at the Thünen Institute of Climate-Smart Agriculture. Although significant uncertainties remain in quantifying SOC dynamics within 10-year intervals, results can be improved by taking the plot-scale SOC variability into account. This preprocessing step not only improves the significance of spatiotemporal SOC models ("model-then-derive") but also allows for the direct prediction of SOC changes based on a "derive-then-model" approach.In summary, these results provide a first step toward an integrated soil monitoring system based on remote sensing and repeated soil sampling. While the findings demonstrate that it is possible to validate spatiotemporal SOC models using long-term sampling data, they also highlight the necessity of further improving the accuracy and applicability of the models. Based on our studies, we will further discuss the opportunities and challenges to independently validate SOC trends claimed by carbon farming schemes using remote sensing data.

Accurate maps of cropland soil organic carbon stocks (SOCS) and clay content are essential for climate-smart agriculture. Soil reflectance composites (SRC), derived from multispectral bare soil observations, offer a scalable approach to high-resolution soil mapping. While studies often focus on maximizing model performance, challenges remain regarding (1) the bias introduced by masking and excluding soil samples during SRC generation and (2) the accurate representation of the full range and distribution of soil properties in the resulting maps. Evaluating different SRC parameters, we found that commonly used indices such as the Normalized Burn Ratio 2 (NBR2) and the Normalized Difference Vegetation Index (NDVI) were significantly correlated with clay content and SOCS, respectively. These dependencies can lead to the systematic exclusion of high SOCS (>80 Mg ha-1) and clay (>30 mass%) samples during SRC generation, introducing bias in the resulting maps. Models trained solely on SRC bands failed to capture the full range of the training data, limiting the applicability of the soil property maps. While the inclusion of additional remote sensing features, such as spectral-temporal metrics and indices, significantly improved the prediction accuracy, the representation of the imbalanced samples remained challenging. We demonstrated that a combined framework of spatial data augmentation and majority undersampling was effective in improving the range and concordance correlation coefficient (CCC) of the predictions (SOCS = 0.82; Clay = 0.9). Our findings emphasize the importance of (1) evaluating excluded samples to identify potential SRC-induced bias, and (2) optimizing model predictions reflecting the observed data range to improve the reliability and usability of the resulting soil maps.

Soil monitoring requires accurate and spatially explicit information on soil organic carbon (SOC) trends and changes over time. Spatiotemporal SOC models based on Earth Observation (EO) satellite data can support large‐scale SOC monitoring but often lack sufficient temporal validation based on long‐term soil data. In this study, we used repeated SOC samples from 1986 to 2022 and a time series of multispectral bare soil observations (Landsat and Sentinel‐2) to model high‐resolution cropland SOC trends for almost four decades. An in‐depth validation of the temporal model uncertainty and accuracy of the derived SOC trends was conducted based on a network of 100 long‐term monitoring sites that were continuously resampled every 5 years. While the general SOC prediction accuracy was high (R² = 0.61; RMSE = 5.6 g kg⁻¹), the direct validation of the derived SOC trends revealed a significantly greater uncertainty (R² = 0.16; p < 0.0001), even though predicted and measured values showed similar distributions. Classifying the results into declining and increasing SOC trends, we found that 95% of all sites were either correctly identified or predicted as stable (p < 0.001), highlighting the potential of our findings. Increased accuracies for SOC trends were found in soils with higher SOC contents (R² = 0.4) and sites with reduced tillage (R² = 0.26). Based on the signal‐to‐noise ratio and temporal model uncertainty, we were able to show that the necessary time frame to detect SOC trends strongly depends on the absolute SOC changes present in the soils. Our findings highlight the potential to generate significant cropland SOC trend maps based on EO data and underline the necessity for direct validation with repeated soil samples and long‐term SOC measurements. This study marks an important step toward the usability and integration of EO‐based SOC maps for large‐scale soil carbon monitoring.

Precise knowledge about the soil organic carbon (SOC) content in cropland soils is one requirement to design and execute effective climate and food policies. In digital soil mapping (DSM), machine learning algorithms are used to predict soil properties from covariates derived from traditional soil mapping, digital elevation models, land use, and Earth observation (EO). However, such DSM models are trained for a specific dataset and region and have so far only allowed limited general statements to be made that would enable the models to be transferred to different regions. In this study, we test the transferability of SOC models for cropland soils using five different covariate groups: multispectral soil reflectance composites (satellite), soil legacy data (soil), digital elevation model derivatives (terrain), climate parameters (climate), and combined models (combined). The transferability was analyzed using data from two federal states in southern Germany: Bavaria and Baden-Wuerttemberg. First, baseline models were trained for each state with combined models performing best in both cases (R² = 0.68/0.48). Next, the models were transferred and tested with soil samples from the other state whose data were not used during model calibration. Only satellite and combined models were transferable, but accuracy declined in both cases. In the final step, models were trained with samples from both states (mixed-data models) and applied to each state separately. This process significantly improved the accuracies of satellite, terrain, and combined models, while it showed no effect on climate models and decreased the models based on soil covariates. The experiment underlines the importance of EO for the transfer and extrapolation of DSM models.