Mapping soil total carbon using multisource remote sensing at the continental scale

Continental-scale soil total carbon (STC) mapping is essential for climate mitigation but faces significant spatial and methodological constraints. This study presents a novel approach that integrates optical, thermal infrared (TIR), and radar data from multiple satellite platforms to map STC across Europe by using the Google Earth Engine. Unlike previous studies that predominantly relied on optical sensors, we systematically evaluated the influence of the radar system parameters and multi-sensor integration on the prediction accuracy via seven modeling scenarios. In various scenarios that involve distinct data integration methods, we integrated the WoSIS soil profile database with the Random Forest algorithm to develop soil models. Our results demonstrate that cross-polarized (HV and VH) channels outperformed co-polarized (VV and HH) channels, the C-band surpassed the L-band data, and multifrequency approaches significantly improved the predictions. We also show that the synergistic integration of optical, TIR, and radar data substantially enhanced the model performance (R 2 = 0.56) compared with models based on a single sensor type; this integration also outperformed the model that used only conventional environmental covariates, including terrain and climate (R 2 = 0.52). Further improvement was achieved when these satellite data were combined with conventional covariates (R 2 = 0.63). Furthermore, satellite-derived variables from optical, thermal and radar sensors were identified as the most important category of input variables for STC predictions. The resulting continental-scale digital soil maps captured pronounced STC spatial variability and detailed STC spatial patterns across Europe. This study demonstrates that integrating synergistic remote sensing with cloud computing effectively enables continental-scale soil mapping. • Cross-polarized radar data outperforms co-polarized for soil carbon mapping. • C-band radar provides more accurate results than L-band radar. • Satellite observations surpass conventional covariates for soil carbon modeling. • Optical-thermal-radar integration enhances soil total carbon prediction at continental scale. • The GEE platform is a promising tool for modeling and mapping soil carbon.