基于遥感指数和PLSR的玛纳斯河流域绿洲盐分的定量分析

Quantitative analysis of soil salt based on remote sensing indicators and PLSR at the oasis of Manasi River Watershed

干旱区绿洲土壤盐渍化问题严重影响着区域的农业产出和生态稳定性。综合遥感信息定量反演地表土层盐分含量成为区域范围土壤盐分监测的重要方法。为了探究遥感影像提取玛纳斯河流域绿洲范围内土壤盐渍化程度精确性,本研究采用偏线性回归分析的方法对不同遥感指数组合得到的土壤盐分含量进行精度对比。结果表明:土壤盐分含量与除绿度指数以外的遥感指标相关性显著,其中NDVI指数和土壤盐分指数(SI)相关性分别达-0.76和0.74。结合遥感数据,利用PLRS法构建土壤盐分预测模型,R2达0.89,均方根误差为0.35,预测精度较高。模型分析结果显示,玛纳斯河流域绿洲区土壤盐分含量为轻度盐渍化水平,高盐渍化土壤主要分布于绿洲中部的水域周围和荒地内。

Soil salinization in arid area has a serious impact on agricultural yield and ecological stability. Predicting on salt content in the topsoil by comprehensive remote sensing quantitative retrieval turns out to be an important means for soil salt monitoring in soil salinization areas at present. In this paper, in order to investigate the accurate extent of soil salinization within the oasis of Manasi River Watershed of remote sensing image, partial linear regression analysis was adopted to compare the precision of soil salt content from different remote sensing index combinations. The result showed that soil salt is significantly correlated with remote sensing indices, except green index. NDVI has the most negative correlation with soil content(-0.76), and SI the most positive(0.74). Combined with remote sensing data, the prediction model of soil salt is established by PLRS. When R2 is 0.89 and the root mean square error is 0.35, the prediction is relatively preciser. The result shows that soil salinization in the oasis of Manas River Watershed is mild. And soil with high salinization mainly locates in unused land and around wetland in the middle of the oasis.