寒旱灌区土壤水盐时空变异规律与尺度效应研究

Soil moisture detection has been one of the research focuses in microwave RS, and soil salinity detection research is still at the primary stage. the study on the scale effect of water and salt has a wide range of challenge. This project takes Hetao Irrigation District which locates cold and arid zone as example,we analysize the cotents of soil water and salt,by ways of field trial,data analysis,RS estimation,and model building.Using much polarization, much angle SAR and Landsat image data in Irrigation District, fusing optical and microwave RS data, in order to compensate for the shortcoming of single data and method; we research on the relationship of the soil salt contents, dielectric constant, the radar backscattering coefficient and spectral reflectance, establishing salinity estimation experiential model; By means of selection of significant impact salinization band, analyzing the relationship of different polarization combinations and salt contents, eventually establishing the Artificial Neural Network model of the backscattering coefficient of optimal polarization combinations, salt significant impact band, Non-RS factors, with soil salt contents. Meanwhile building the Artificial Intelligence coupling empirical Model to estimate terrain coarse level and soil water contents. by using Geological Statistics ,Artificial Neural Kriging and Fractal Theory to study the scale effect problem.This research can enrich the RS evaluation theory and method of Water Science,providing scientific guidance for the water saving irrigation system optimization and soil salinization government at last.

土壤水分探测是微波遥感研究的重点之一，而对土壤盐分探测的研究仍处于初级阶段。土壤水盐空间尺度效应研究在水科学领域具有广泛的挑战性。本项目以寒旱区典型灌区-河套灌区为研究区，采用野外试验、数据分析、遥感反演建模相结合的方法定量估计灌区土壤水盐含量。利用灌区SAR 和Landsat遥感影像数据，融合光学与微波数据，弥补单一遥感数据和方法的不足；通过实验研究土壤含盐量与介电常数、雷达后向散射系数、光谱反射率的关系，建立盐分反演的经验模型；选择盐渍化影响显著波段，研究不同极化组合的土壤盐分响应关系，建立最优极化组合后向散射系数、盐渍化响应显著波段、非遥感因子与土壤含盐量的人工神经网络模型。同时构建遥感反演地表粗糙度与土壤水分的人工智能与经验耦合模型。并采用地质统计学、人工神经克立格与分形理论研究尺度效应问题。本研究成果将丰富水科学遥感反演的理论与方法，科学指导节水灌区灌溉制度优化管理与盐渍化防治。

土壤水分探测是微波遥感研究的重点之一，而对土壤盐分探测的研究仍处于初级阶段。本项目以寒旱区典型灌区-河套灌区内解放闸灌域为研究区，采用野外试验、数据分析、遥感反演建模相结合的方法定量估计灌域土壤水盐含量。利用多极化SAR雷达影像数据，通过实验研究土壤含水率、含盐量与雷达后向散射系数的关系，建立最优极化组合后向散射系数、非遥感因子与土壤水盐含量的人工神经网络模型，经验证后的人工神经网络模型可用于后期土壤水盐含量的雷达反演。同时进行了ETM+光学遥感反演农田土壤含水率的试验研究。本研究成果将丰富水科学遥感反演的理论与方法，科学指导节水灌区灌溉制度优化管理与盐渍化防治。