Mapping the soil organic matter content in a typical black-soil area using optical data, radar data and environmental covariates.

Soil organic matter (SOM) plays an extremely important role in soil formation, soil fertility, environmental protection and the sustainable development of agriculture and forestry, especially in areas with fertile black soil. Remote sensing technology has unique advantages in SOM mapping and can quickly and accurately obtain SOM spatial distribution maps at a low cost. However, the current methods available for fusing radar data, optical data and environmental covariates for use in SOM mapping are far from sufficient. The study area is Youyi Farm, a typical black-soil area in Northeast China. Multiyear monthly composites of the bare-soil periods (April and May) from 2019 to 2022 in the study area were created in the Google Earth Engine (GEE), including VH-polarized and VV-polarized Sentinel-1 radar data, 10 spectral bands of Sentinel-2 data, elevation and slope terms representing terrain covariates, and multiyear average precipitation and temperature data representing climate covariates. SOM was mapped in the study area based on 188 cultivated-layer (0–20 cm) sampling points combined with a random forest regression algorithm, and the SOM mapping accuracies of different strategies were evaluated. The results show that 1) the SOM mapping accuracy obtained using Sentinel-1 data combined with Sentinel-2 data was improved slightly compared to that obtained using Sentinel-2 data only, with improvements observed mainly in the prediction accuracies of areas with SOM contents below 5 %; 2) the highest accuracy was obtained by using the May composite image, while the April and May composite images reduced the accuracy; 3) adding environmental covariates greatly improved the SOM mapping accuracy, with the highest R2 being 0.696 and the lowest RMSE being 0.712 %; and 4) the importance of climate covariates was found to be higher than that of terrain covariates. This study expands the existing SOM mapping research using radar and optical data and evaluates the effects of different environmental covariates on the SOM mapping results. • We propose a SOM prediction model combining different data sources. • An appropriate time window is critical to SOM prediction. • Adding Sentinel-1 can slightly improve the accuracy of SOM prediction. • Climate covariates have proved more useful for SOM mapping. [ABSTRACT FROM AUTHOR]