ABSTRACT This study aimed to improve the accuracy of soil salinity inversion by correcting satellite remote sensing data using unmanned aerial vehicle (UAV) observations. Multi‐source data, including UAV hyperspectral measurements and satellite imagery, were used to perform hierarchical correction during the study period, covering spectral reflectance, spectral indices and spectral model outputs. Two correction strategies—numerical regression and ratio‐averaging—were applied and compared across these three levels. The correction effects of numerical regression and ratio‐averaging methods varied across different levels. At the spectral reflectance level, the correction effect of ratio‐averaging was better than that of numerical regression. However, at the spectral index and spectral model levels, numerical regression performed better. Correction effects at different levels were ordered spectral model > spectral reflectance > spectral index. At the spectral model level, the R 2 of the satellite model following correction by numerical regression improved from 0.630 to 0.787. When the model was applied from training areas to validation areas, the R 2 of the original satellite model decreased by 0.112 (from 0.630 to 0.518). In contrast, following correction using UAV data alone, the R 2 of the satellite model decreased by 0.077 (from 0.787 to 0.710). The method developed in this study can effectively overcome the problem of low accuracy of soil salinity inversion with single satellite remote sensing data, and achieve more accurate and wider ranging soil salinity inversion, providing a reference for soil salinity monitoring and control.
Liu et al. (Thu,) studied this question.