基于多源数据的土壤湿度提取方法研究

发布时间：2019-02-23 15:12

【摘要】：准确、大面积、实时的估算地表土壤湿度是“渤海粮仓”科技示范工程的一项重要内容。传统的土壤湿度获取是基于地面站点监测,精度高、时间连续性好,但获取的是点尺度数据;遥感具有宏观、动态、实时、多源等特点,通过反演得到的是像元尺度数据,精度需要地面实测数据进行验证。为了综合利用两者的优势,获取高精度、空间连续的土壤湿度数据,论文以以“渤海粮仓”山东项目区滨州市无棣县为研究区,利用地面观测网数据、地面同步实测数据和MODIS遥感影像数据,通过对土壤湿度数据插值方法、升尺度方法和遥感反演模型的研究,综合利用了不同数据在土壤湿度提取上的各自优势、取长补短,主要研究工作如下:(1)提出了综合考虑时间和空间信息的地面观测土壤湿度数据插值方法—BMA-I本文采用协同克里金法作为空间预报模型,基于LS-SVM算法的时序递推预测方法作为时序预报模型,基于BMA方法进行模型平均,建立了一种新的插值方法BMAI,并与单纯使用普通克里金插值法和协同克里金插值法进行了对比分析,改善了克里金方法所带来的平滑效应,在一定程度上更能反映真实的土壤湿度数据。(2)改进表观热惯量计算方法,升尺度土壤湿度首先,用地面观测网的站点实测地表土壤温度数据计算温差,通过MODIS地表反射率产品计算得到像元尺度ATI,然后,建立BMA-I法插值得到的土壤湿度数据与像元尺度ATI之间的函数关系,通过贝叶斯线性回归得到升尺度之后像元尺度的地表土壤湿度。(3)建立了基于改进的温度-植被干旱指数模型反演地表土壤湿度的方法利用MODIS遥感影像获取Ts与NDVI数据,构建特征空间,根据干湿边方程计算得到TVDI,建立升尺度获得的土壤湿度数据与TVDI的关系模型,反演得到区域地表土壤湿度,经验证,反演结果明显好于仅用少量地面实测数据建立反演模型。
[Abstract]:Accurate, large area and real time estimation of soil moisture is an important part of the demonstration project of Bohai granary. The traditional soil moisture acquisition is based on ground site monitoring, which has high accuracy and good time continuity, but it acquires point-scale data. Remote sensing has the characteristics of macroscopic, dynamic, real-time, multi-source, etc. By inversion, the pixel scale data are obtained, and the precision needs to be verified by the ground measured data. In order to make comprehensive use of the advantages of the two methods and to obtain soil moisture data with high precision and continuous space, the paper takes Wudi County, Binzhou City, Shandong project area as the research area, and uses the data of ground observation network. Based on the research of soil moisture interpolation method, scaling method and remote sensing inversion model, the ground synchronous measured data and MODIS remote sensing image data are studied. The advantages of different data in soil moisture extraction are synthetically utilized to complement each other. The main research work is as follows: (1) the interpolation method of soil moisture data based on the time and space information is put forward in this paper. In this paper, the cooperative Kriging method is used as the spatial prediction model. The recursive prediction method of time series based on LS-SVM algorithm is used as the prediction model of time series, and a new interpolation method, BMAI, is established based on the BMA method for model averaging. Compared with the common Kriging interpolation method and the cooperative Kriging interpolation method, the smoothing effect of the Kriging method is improved. To some extent, it can better reflect the real soil moisture data. (2) improve the calculation method of apparent thermal inertia. The pixel scale ATI, was calculated by MODIS surface reflectance product. Then the functional relationship between soil moisture data interpolated by BMA-I method and pixel scale ATI was established. Through Bayesian linear regression, the surface soil moisture at pixel scale after scaling is obtained. (3) based on the improved temperature-vegetation drought index model, the method of retrieving the surface soil moisture using MODIS remote sensing image is established to obtain Ts and NDVI data. The characteristic space is constructed and the relationship model between soil moisture data and TVDI obtained by TVDI, is established according to the dry and wet edge equation. The surface soil moisture of the region is retrieved and verified. The inversion results are better than only a small amount of ground measured data to establish the inversion model.
【学位授予单位】：山东农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S152.71

【参考文献】