Soil three-phase data

This study analyzed the spectral sensitivity of soil structure indices and constructed spectral indices to extract characteristic information more effectively than traditional methods. Based on simulated datasets derived from ground-satellite resampling (Sentinel-2, Landsat 8 and MODIS), we compared the performance of single machine learning algorithms and the Stacked Ensemble Learning (SEL) framework. Finally, to address the difficulty in solid phase inversion, we proposed an indirect monitoring strategy based on the conservation of three-phase proportions. Our research provides a reliable technical reference for the non-destructive and accurate assessment of soil structure in complex environments.