黄河口地区环境脆弱性的遥感反演  

Mapping Environmental Vulnerability from ETM + Data in the Yellow River Mouth Area

作　　者：王瑞燕[1,2,3] 于振文[1] 夏艳玲[4] 王向锋[5] 赵庚星[2] 姜曙千[2] 

机构地区：[1]山东农业大学农学院,山东泰安271018 [2]山东农业大学资源与环境学院,山东泰安271018 [3]土肥资源高效利用国家工程实验室,山东泰安271018 [4]鲁东大学地理与规划学院 [5]垦利县国土资源局,山东垦利257500

出　　处：《光谱学与光谱分析》2013年第10期2809-2814,共6页Spectroscopy and Spectral Analysis

基　　金：山东省高等学校科技计划项目(J11LC11);山东省省自然科学基金项目(ZR2009BQ017);山东省博士后创新项目专项资金项目(201003026);国家自然科学基金项目(41271235)资助

摘　　要：环境脆弱性定量遥感研究,可以为环境脆弱性研究提供稳定的数据源支撑。通过遥感反演获取区域环境脆弱性的空间分布。从土壤和植被角度,构建了环境脆弱性综合评价指标体系,采用AHP-模糊评判方法计算采样点环境脆弱度,并将其分别与样点ETM+光谱反射率及其转换数据的相关关系进行分析,确定其敏感波段,在此基础上,采用传统回归方法、基于BP人工神经网络分析方法和支持向量机回归方法建立环境脆弱度的光谱反演模型,并采用该模型对研究区的环境脆弱度进行反演,得到环境脆弱性度时空分布图。结果表明,返青期NDVI、九月份NDVI以及返青期的亮度分量是环境脆弱度的ETM+敏感光谱参数,模型精度比较结果显示,除了支持向量机模型外,其他模型都达到了显著水平,其中以BP神经网络模型的精度最高,传统回归模型也可满足预测需要,但多元回归的模拟精度要高于一元回归模型。研究结果可为大空间尺度的卫星水平环境脆弱性遥感反演提供理论支持。The environmental vulnerability retrieval is important to support continuing data. The spatial distribution of regional enviromnental vulnerability was got through remote sensing retrieval. In view of soil and vegetation, the environmental vulnera- bility evaluation index system was built, and the environmental vulnerability of sampling points was calculated by the AHP-fuzzy method, then the correlation between the sampling points environmental vulnerability and ETM ＋ spectral reflectance ratio in- cluding some kinds of conversion data was analyzed to determine the sensitive spectral parameters. Based on that, models of cor- relation analysis, traditional regression, BP neural network and support vector regression were taken to explain the quantitative relationship between the spectral reflectance and the environmental vulnerability. With this model, the environmental vulnerabili- ty distribution was retrieved in the Yellow River Mouth Area. The results showed that the correlation between the environmental vulnerability and the spring NDVI, the September NDVI and the spring brightness was better than others, so they were selected as the sensitive spectral parameters. The model precision result showed that in addition to the support vector model, the other model reached the significant level. While all the multi-variable regression was better than all one-variable regression, and the model accuracy of BP neural network was the best. This study will serve as a reliable theoretical reference for the large spatial scale environmental vulnerability estimation based on remote sensing data.

关 键 词：环境脆弱性 ETM+ 遥感反演 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]