土壤盐分含量的遥感反演研究

Retrieval of soil salt content based on remote sensing

土壤盐碱化是干旱、半干旱农业区主要的土地退化问题，同时也是一个重要的环境问题。遥感技术能够快速、实时地提供盐碱地的性质、范围、盐碱程度等方面的信息。该文以河北省黄骅市为研究区，通过对实测的土壤光谱的分析，发现植被严重干扰了土壤对盐分含量光谱响应关系，同时，在451．42-593．79nm波长范围内的土壤反射率对土壤盐分含量较为敏感，在土壤光谱分析的基础上，建立了土壤盐分含量反演的相关统计模型。但由于遥感影像特征与土壤盐分含量之间存在较复杂的非线性关系，因此统计模型反演精度不够理想。因而，又尝试运用BP人工神经网络方法来反演土壤盐分含量。研究表明，BP人工神经网络模型具有很强的非线性拟合能力，与统计模型相比，其土壤盐分含量的反演精度有显著提高。

soil salinization is one of the most important problems of land degradation and the basic environmental problem in arid and semi-arid regions. The remote sensing technology can rapidly and timely provide the information about properties, geographical distribution and extent of soil salinization. Taking the city Huanghua of Hebei Province in China as the study area and through the analysis on the data of soil spectrum measured in field, it was found in this study that vegetation affects greatly the spectral response of soil for salt content, and at the same time the spectrum ranging from 451.42 nm to 593.79 nm is much sensitive to the variation in soil salt content and, therefore, based on the analysis of soil spectrum, the relevant statistic model for predicting soil salt content was constructed. However, due to rather complicated non-linear relations existed between image features and soil salt content, the results of soil salt content retrieved from the statistic model is not so ideal. For this reason, an artificial neural network model （BP model） was constructed and applied in the retrieval of soil salt content. Because of its superior ability for solving the non-linear problem, the BP model provided a much better accuracy in retrieval of soil salt content compared with the results from the statistic model.