The power of integrating proximal and high-resolution remote sensing for mapping SOC stocks in agricultural peatlands

Background and aims Soil electrical conductivity (ECa) data derived from electromagnetic induction (EMI) is valuable for estimating peat thickness and soil organic carbon stocks (SOC stocks ). However, generating ECa maps locally using geostatistics limits the coverage area. This study explores the use of digital soil mapping (DSM) with random forest (RF) and universal kriging (UK) to create high-resolution ECa maps from field survey EMI data. The objective is to enhance the predictive accuracy of SOC stocks models in peatlands by incorporating these ECa maps as environmental variables. Methods Three scenarios were evaluated, combining different environmental variables and modelling techniques for ECa mapping. Scenario 1 used spectral indices from RapidEye satellite data and RF. Scenario 2 included spectral indices and terrain derivatives from LiDAR, with RF. Scenario 3 integrated spectral indices, terrain derivatives from LiDAR, and UK. Afterwards, we evaluated the effectiveness of adding ECa maps as environmental variables for SOC stocks mapping. Finally, we incorporated ECa maps from scenario 2 and RF in three ways: (a) scenario 2 variables only, (b) ECa 2 with scenario 2 variables, and (c) ECa 3 with scenario 2 variables. Results Scenarios 2 (ECa 2 ) and 3 (ECa 3 ) outperformed scenario 1 (ECa 1 ). The inclusion of ECa maps significantly improved the accuracy of SOC stocks models. Conclusion Our study demonstrates that DSM, combined with RF and UK techniques, enables the generation of high-resolution ECa maps from field surveys in peatlands. Incorporating these ECa maps as environmental variables enhances the accuracy of SOC stocks mapping, providing valuable insights for peatland management and carbon stock estimation. Graphical abstract