Spatio-temporal inversion of saline-alkali land in the oasis and desert region of the Hexi Corridor by fusing measured and remote sensing spectra

Saline-alkali land poses a critical challenge to ecological security, agricultural production, and sustainable development. Accurately assessing its spatio-temporal characteristics is of significant importance. In this study, the high-precision measured spectra were fused with the spatio-temporal continuous remote sensing spectra. Multidimensional spectral indices were constructed, the optimal set of indices was selected, and multiple machine learning models were applied to achieve high-precision, large-area, and long-term regional inversion in the desert and oasis region of the Hexi Corridor from 2000 to 2024. Subsequently, the spatio-temporal characteristics of saline-alkali land at different scales were revealed, and its primary driving factors were identified. Results indicated that the correlation between the fusion spectral indices and soil salinity was significantly enhanced. Among all inversion models, XGBoost model provided the highest inversion accuracy (training set R 2 = 0.895, validation set R 2 = 0.796). Moreover, saline-alkali land was more severe in the west and north, while relatively slighter in the east and south. In desert areas, saline-alkali land was extensive but slight, whereas in oasis areas, it was limited but severe. Regional soil salinization levels predominantly showed a declining trend. Saline-alkali land grade transitions primarily occurred between non-saline-alkali land and slightly saline-alkali land. Among all factors, precipitation and terrain exerted the strongest direct and indirect negative influences on soil salinity, respectively. This study establishes a new framework for high-precision saline-alkali land inversion and provides a scientific basis for regional water resource management, agricultural production layout, and ecological environment governance. • Fuse measured and remote sensing spectra to enhance the correlation with salinity. • Construct and screen multidimensional spectral indices. • Achieve high-accuracy, large-scale, and long-term saline-alkali land inversion. • More than 60% of the land is affected to varying degrees by saline-alkali land. • Precipitation exerted the strongest negative effect (-0.531) on soil salinity.