基于卫星遥感的内蒙古达拉特旗黄河南岸灌区土壤盐分时空演变分析

Analysis of Spatial and Temporal Evolution of Soil Salinity in Yellow River South Bank Irrigation Area of Dalate Banner, Inner Mongolia Based on Satellite Remote Sensing

摘要:

土壤盐渍化是全球干旱半干旱地区农业生产、生态环境与土地资源可持续发展的主要挑战之一。内蒙古达拉特旗黄河南岸灌区长期受到土壤盐渍化的威胁，研究土壤含盐量动态变化及其驱动因素对该地区农业生产、生态保护与水资源管理具有重要指导意义。基于Sentinel-2卫星数据和实地土壤含盐量数据，通过相关性分析（Pearson correlation coefficient，PCC）、变量投影重要性分析（Variable importance in projection analysis，VIP）、灰色关联度分析（Grey relational analysis，GRA），并结合反向传播（Backpropagation，BP）、随机森林（Random forest，RF）和支持向量机（Support vector machine，SVM）3种机器学习算法估算了裸土期和植被覆盖期土壤含盐量（Soil salinity content，SSC）。同时基于最优反演模型以及Landsat-5、Sentinel-2数据，反演并统计了2000—2024年间灌区土壤含盐量时空变化特征。最后分析了蒸发量、降水量、温度、真实水汽压等气象因素对土壤含盐量变化的驱动作用。结果表明：裸土期PCC-RF模型表现最佳（建模集R2=0.849，RMSE为0.118%，MAE为0.079%；验证集R2=0.753，RMSE为0.158%，MAE为0.116%）。时空变化特征表明2000—2008年盐渍土面积持续增加，而2009—2016年实施的水权改造政策有效遏制了盐渍土扩张，并将大面积盐渍土转换为非盐渍土。蒸发量、降水量、蒸降比和温度是研究区土壤含盐量变化的主要气象影响因素。研究结果可为灌区土壤盐渍化监测与治理提供科学依据。

Abstract:

Soil salinization is one of the major challenges to agricultural production, ecological environment and sustainable development of land resources in global arid and semi-arid regions. The irrigation area on the south bank of the Yellow River in Dalate Banner, Inner Mongolia has long been threatened by soil salinization. Studying the dynamic changes of soil salinity and its driving factors is of great guiding significance for agricultural production, ecological protection and water resource management in this area. Based on Sentinel-2 satellite data and field soil salinity data, the soil salinity content (SSC) during the bare soil period and the vegetation coverage period was estimated through Pearson correlation coefficient analysis (PCC), variable importance in projection analysis (VIP), grey relational analysis (GRA), and combined with three machine learning algorithms: backpropagation (BP), random forest (RF), and support vector machine (SVM). At the same time, based on the optimal inversion model and Landsat-5 and Sentinel-2 data, the spatio-temporal variation characteristics of soil salinity in the irrigation area from 2000 to 2024 were inverted and statistically analyzed. Finally, the driving effects of meteorological factors such as evaporation, precipitation, evapotranspiration ratio and temperature on soil salinity changes were analyzed. The results showed that the PCC-RF model performed best during the bare soil period (R2 of the modeling set was 0.849, RMSE was 0.118%, MAE was 0.079%;R2 of the validation set was 0.753, RMSE was 0.158%, MAE was 0.116%). The spatio-temporal variation characteristics indicated that the area of saline soil continued to increase from 2000 to 2008, while the water rights reform policy implemented from 2009 to 2016 effectively curbed the expansion of saline soil and converted a large area of saline soil into non-saline soil. Evaporation, precipitation, evapotranspiration ratio and temperature were the main meteorological influencing factors of soil salinity changes in the study area. The research result can provide a scientific basis for the monitoring and control of soil salinization in the irrigation area.