【作者】 马驰；

【导师】 姜琦刚；

【作者基本信息】 吉林大学 ， 地质资源与地质工程， 2011， 博士

【摘要】 我国是受土地盐碱化危害最严重的国家之一。上个世纪中后期,由于全球变化所导致的气候干旱化以及水、土、生物资源被过量地开发利用等原因,使松辽平原生态环境恶化,土地盐碱化和草原退化等问题日益严重,土地盐碱化迅速扩大,生态环境受到严重破坏,甚至形成不毛之地,严重影响地区的可持续发展。遥感技术作为一种快速、准确获取空间信息的技术手段,对土地盐碱化监测、治理及合理开发利用,维持区域生态可持续发展等方面具有重要意义。本文利用人机交互解译方法对1975年MSS影像、1990年TM影像、2001年ETM影像进行盐碱化信息提取,并利用地理信息系统分析,获得近30年松辽平原盐碱土变化规律。对研究区内ASTER遥感影像、TM遥感影像利用6S模型进行大气校正,研究盐碱土在ASTER影像、TM影像上的光谱特征,结合土壤采样化验数据,对松辽平原土壤含盐量进行定量反演。为了提高反演精度,研究中对反射率进行了包括对数、倒数、微分等形式的多种数学变换。结果表明,这些变换可以在不同程度上提高反演精度。为了分离盐碱土,文中探讨了3种分类方法:最大似然值分类、人工网络神经元分类、决策树分类。统计结果表明,决策树分类可以获得更高的分类精度。本文试利用环境减灾卫星(HJ-1A)的HSI高光谱数据进行定量反演土壤含盐量的实验。研究过程中,利用FLAASH模型对高光谱数据进行大气校正,并对校正后影像的反射率及其变换形式与土壤实测含盐量进行单波段相关分析,确定盐碱在HSI高光谱中的敏感波段;为了消除高光谱波段间高相关性影响、提高反演精度,文中使用偏最小二乘回归分析(PLSR)的方法,提取反射率及其变换形式图像的最大解释盐分含量变化的主成分,预测研究区内土壤全盐含量及盐碱离子含量。为了研究区内盐碱土的数量变化和空间转移规律,本文利用解译与分类结果,将研究区内几种主要地类生成马尔科夫转移矩阵,分析盐碱土时空变化的特点与规律。最后以景观生态学为指导,运用景观指数法,对该区景观格局变化进行定量分析,揭示了景观格局变化的特征和过程。更多还原

【Abstract】 China is one of the most serious salinization countries. In the late of last century, for the drought caused by global climate change and over-exploitation of water, soil and biological resources, problem of environment degradation, soil salinization and grass degradation grows more and more serious in Songliao Plain, such as expansion of soil salinization, seriously damaged of ecological environment, which has serious impact on the local sustainable development. As a rapid and accurate method to get spatial information, RS technology has an important meaning on monitoring, management and utilization of soil salinization, maintaining regional ecological sustainable development.Salinization information can be extracted from MSS image in 1975, TM image in 1990 and ETM image in 2001 by interactive method, and the salinization change of nearly 30 years in Songliao Plain can be obtained based on the analysis of GIS. Corrected ASTER images and TM images of the study area using 6S atmospheric correction model to study the spectral characteristics of ASTER image and TM image, and with the soil sampling data soil salinity was quantitatively inversed in Songliao Plain. To improve the retrieval accuracy, Research conducted on the reflectivity including the number of countdown, the form of a variety of differential mathematical transformation. The results show that these transformations can improve the retrieval accuracy. In order to isolate soil, three kinds of classification methods: the maximum likelihood value of the classification, artificial neural networks and decision tree classification are discussed. Results show that the decision tree classifier can achieve higher classification accuracy. THSI paper tries to use the HSI hyper-spectral data of environmental mitigation Satellite (HJ-1A) to quantitative inversed soil salinity. In the course of the study, used FLAASH model for atmospheric correction on hyperspectral data, and carried out correlation analysis of single-band on the corrected image and its transformation to determine the salinity sensitive band in the HSI. To eliminate the high correlation impact between the spectral bands and improve the retrieval accuracy, used partial least squares regression (PLSR) method to extract the image reflectance and its biggest transformation in the form to explain the main component of salt content, and predicted local soil salt content and salt ion content.In order to study the number variation of salinization and transfer rules, using the result of interpretation and classification, several main class will be generated in Markov transition matrix, analyzed the temporal and spatial variation characteristics of soil salinity and the law. Finally, in the guidance of landscape ecology, using the landscape index, carried on the quantitative analysis on changes in the landscape pattern, revealed the characteristics and process of landscape pattern.更多还原