A framework for soil texture prediction based on remote sensing information from different climate patterns and crop growth periods

Soil texture is an important parameter representing the physical properties of soil, so accurate mapping of it is crucial for revealing the intrinsic soil properties. Using Sentinel-2 images from Youyi Farm, which is located in the third major black soil region of Northeast China, bare soil information, and crop growth information, the relationship between them and soil texture mapping accuracy under different annual climate patterns (2019–flooded, 2020–normal, and 2021–drought) was explored. The results indicated that (1) the highest mapping accuracy was obtained for sand, silt, and clay after the recursive feature elimination, with R² and RMSE values reaching 0.732/8.544%, 0.762/6.725%, and 0.612/1.925%, respectively. (2) In the flooded year, crop-growth information (NDVI, EVI) added in different months had a small effect on the mapping accuracy of clay, while sand and silt showed large fluctuations. (3) The results show that all bands of remote sensing imagery have major influences on all soil texture predictions and that crop growth information contributes relatively little to sand and silt but significantly influences clay predictions. This study offers new perspectives and methods for high-resolution mapping of soil texture and related soil property studies.