基于光学和雷达多源遥感的于田绿洲土壤盐渍化时空分析

Soil salinization is currently a prominent global environmental problem. Spatiotemporal variation of soil salinization in Yutian Oasis was explored, and the advantages of radar remote sensing in excavated soil salinity were investigated to monitor and evaluate the temporal and spatial variabilities of soil salinization in arid oasis. Based on Phased Array type L- band Synthetic Aperture Radar 2 (PALSAR-2), Sentinel-1 polarimetric synthetic aperture radar data, and Landsat 8 Operational Land Imager (Landsat 8 OLI) multi-source dataset, the optimal backscattering characteristics of radar images were selected, and optical images were combined based on principal component analysis. Finally, the random forest method was used to classify the images. Quantitative extraction of soil salinization information in Yutian Oasis and the spatial and temporal variation of soil salinization were analyzed. Results showed that (1) under the random forest classification approach, the total accuracy of optical images in each year was 80.36%, and the kappa coefficient was 0.77. The classification accuracy of optical images combined with radar images was higher than that of optical images, the total accuracy was 85.62%, and the kappa coefficient was 0.82. (2) From 2015 to 2021, the area of salinized soil in Yutian Oasis was mainly distributed in the north of the study area, interlaced edge of the oasis, and desert. (3) The average annual variation of saline land area from 2015 to 2021 was -1120.55 hm2·a-1, and the change rate was -10.67%. The salinization of Yutian Oasis generally showed a downward trend and was mainly mild to moderate saline land.