Evaluation of a New 30m Soil Organic Carbon Density (Socd) Dataset for Pan-Europe over Türkiye, Where No Samples Were Included During Model Training

Soil Organic Carbon (SOC) is a vital component of soil that represents carbon derived from organic sources, enhances soil fertility, improves aggregate stability (necessary for creating pore spaces that store water beneficial for plants), and plays a critical role in atmospheric carbon dioxide exchange. To address this need, the 30 m resolution Landsat-based Analysis Ready Data (ARD) SOC dataset for Pan-Europe (2000-2022) was published in 2024 in the context of the AI4SoilHealth project funded by the European Commission. Although this dataset includes Türkiye, it was developed without incorporating training samples from Türkiye during the model training process, which means that their spatial representativeness is currently unknown. Therefore, the objective of this study is to assess the validity of this ARD SOC data across Türkiye using field measurements. In addition to this quantitative analysis, by calculating the Landsat-based indices used for this global SOC map and employing Kernel Quantile Regression (QR) as in the proposed model, we identified key indices that have sensitivity to the regeneration process during in-situ measurements, providing information on their relevance and sensitivity for monitoring soil variations.Although no training samples from Türkiye were included in the global dataset preparation, the AI4SoilHealth SOC data is strongly correlated at medium SOC levels. The correlation, however, weakens for lower and higher SOC values, especially those observed after regenerative practices.