稀疏植被区土壤表层介电特性分析及水盐信息提取研究

发布时间：2018-01-09 08:33

本文关键词：稀疏植被区土壤表层介电特性分析及水盐信息提取研究　出处：《新疆大学》2017年硕士论文　论文类型：学位论文

【摘要】：新疆维吾尔自治区地处我国西北地区,因水资源空间分布极不均衡,导致荒漠多、绿洲少,加之新疆大部分区域的降雨较少、气候干燥,生态环境极其脆弱。与此同时,新疆位于中纬度地区,光照资源丰富,农作物产量高且质量优,作为我国粮食后备储备基地,担负着粮食安全、社会安全、国家安全等重大战略。近年来,由于区域农田的不合理开发、地下水资源的滥用等现象,区域生态环境日益恶化,尤其在绿洲-荒漠交错区域,大面积土地发生退化、生态多样性骤减。造成土地退化的因素较多,而土壤水分匮乏、盐分含量高是绿洲-荒漠交错区域土地退化的重要因素。目前,新疆境内受盐渍化胁迫的耕地面积约占耕地总面积的1/3,土壤盐渍化现象已严重威胁到区域生态系统协调发展,抑制了农作物的正常生长、生产,带给绿洲经济的损失不容忽视。本文选择新疆典型绿洲——渭-库绿洲为研究区,该区域主要靠渭干河和库车河水系补给形成绿洲,绿洲稳定性受水资源影响较大。在西部大开发等相关政策的刺激下,渭-库绿洲境内县乡的社会经济水平逐年提高,农用地面积迅猛扩增,巨大的水量需求加剧了水资源空间分布的不均衡性,绿洲下游、绿洲周边生态系统持续恶化、稀疏植被覆盖区土地退化现象加剧,绿洲至荒漠区域间的缓冲区域日渐缩减。空间传感手段的发展实现了大区域、多时相的土壤信息提取技术,在相关技术手段快速发展的背景下,土壤信息监测精度逐步增高,尤其是高分辨率(光谱、空间)遥感数据更是为土壤信息监测精度提供了保障。本文以提取新疆渭-库绿洲局部地表土壤水分、盐分信息为研究目的,选择多极RadarSat-2雷达影像和国产高分一号多光谱影像为数据源,(1)以实测介电常数实部和虚部为基础,运用多种数学变换形式,建立土壤水分、盐分的介电常数逐步回归方程;(2)运用AIEM模型模拟一定入射角下,地表粗糙度、土壤水分与雷达后向散射系数间的拟合研究,并建立裸露地表土壤含水量经验模型;(3)优选GF-1的多种光谱指数,利用BP神经网络模型针对雷达数据和多光谱数据建立土壤盐分模型,并进行验证和精度评估。主要得出以下结论:1、土壤含水量对介电常数实部有较大的影响且相关性可达0.94,虚部不仅受含水量的影响,也受含盐量的影响;在盐分较大的情况下,虚部随着含水量的增加而增大,虚部与土壤盐分含量相关性较弱,仅0.45,但与含水量高度相关,相关性可达0.90。此外,利用介电常数的5种变换形式,所建立土壤水分实部模型精度较高,其中,倒数模型为土壤水分最优模型,建模精度的决定系数R2为0.82、验证精度R2为0.91;与此同时,运用介电常数虚部综合模型模拟土壤盐分,盐分指数:(水分+盐分)2为最优模型,建模精度的决定系数R2为0.84、验证精度R2为0.91。水分与实部、虚部均呈良好的线性关系,而盐分与虚部呈非线性关系。2、利用AIEM模型模拟了区域地表粗糙度和土壤水分情况,区域地表粗糙度与实际相符,绿洲-荒漠交错区、河道附近的粗糙度较大;土壤水分精度较高,决定系数R2达0.86,土壤含水量呈西高东低分布,绿洲边缘和河道附近土壤含水量较高。3、由于土壤盐分线性经验模型难以准确评估土壤盐分含量,因此结合多光谱辅助数据对后向散射信息进行分割。利用RadarSat-2四极化后向散射系数、及地表粗糙度和土壤水分的雷达提取数据、GF-1提取的光谱指数作数据集,并构建BP神经网络模型。经多次试验,BP神经网络模型较好地提取了区域土壤盐分信息,模型模拟精度达到了78.95%,因此从一定程度上验证了光学遥感与主动微波遥感结合进行盐渍区信息提取的有效性。综上所述,本研究通过主动微波遥感特征,利用AIEM物理模型建立了适用于裸露区土壤水分反演的经验模型;同时,结合多光谱数据,利用BP算法建立了土壤盐分模型,且模拟精度较高,准同步地表土壤水盐研究可为干旱区土壤信息提取工作提供一定的参考。
[Abstract]:The Xinjiang Uygur Autonomous Region is located in the northwest of China, because of the spatial distribution of water resources is extremely uneven, resulting in much less in most areas of Xinjiang oasis, the less rainfall, dry climate, the ecological environment is extremely fragile. At the same time, Xinjiang is located in the middle latitudes, the light is rich in resources, high crop yield and quality, as our food reserve base, responsible for food safety, social security, major national security strategy. In recent years, due to the unreasonable development of regional farmland, abuse of groundwater resources, the regional ecological environment is deteriorating, especially in the oasis desert ecotone, a large area of land degradation, ecological diversity caused more sharply. The factors of land degradation, and soil water scarcity, high salt content is an important factor in the oasis desert ecotone of land degradation. At present, the territory of Xinjiang by saline stress The cultivated land area accounts for about 1/3 of the total farmland, soil salinization has been a serious threat to the sustainable development of eco system, inhibit the normal growth of crop production, to oasis economic losses can not be ignored. This paper chooses Xinjiang oasis - Wei library oasis as the research area, the region mainly rely on the Weigan River and Kuche river water supply forms oasis, oasis stability affected by water greatly. The policy of western development and stimulation, social economic level Wei library oasis in the county increased year by year, agricultural land area of rapid expansion, huge water demand exacerbated the imbalance, the spatial distribution of water resources in Oasis downstream, continued to deteriorate around the oasis ecosystem, increasing land sparse vegetation degradation, a buffer zone between oasis to desert areas is increasingly reduced. The development space of the sensing region Time domain, soil information extraction phase, with the rapid development of related technology under the background of information monitoring accuracy of soil gradually increased, especially high resolution (spectral, spatial) remote sensing data is provided a guarantee for monitoring accuracy. Based on the soil information extraction of Xinjiang Wei library oasis local soil moisture, salinity information for the purpose of the study, select the multipole RadarSat-2 radar image and domestic high one multi spectral image as the data source, (1) to the measured dielectric constant of the real and imaginary parts as the foundation, using a variety of mathematical transformations, the establishment of the soil moisture, dielectric constant salinity regression equation; (2) using AIEM model a certain incident angle, surface roughness, soil moisture and radar to the fitting of scattering coefficient, and the establishment of bare soil moisture model; (3) multi spectral index optimization GF-1, using BP neural network Network model for radar data and multi spectral data to establish the model of soil salinity, and the validation and accuracy assessment. Main conclusions are as follows: 1. Soil moisture content on the real part of permittivity and the influence of correlation is the larger of the 0.94, the imaginary part is not only affected by water content, is also affected by salinity; in the situation of salt is large, the imaginary part increases with the increase of water content, and the imaginary part correlation of soil salt content is only 0.45, but is highly correlated with the water content, the correlation can reach 0.90. in addition, using 5 kinds of transformations of the dielectric constants of the established real soil moisture model is of high precision, the countdown model for optimal decision model of soil moisture, the modeling accuracy of coefficient R2 is 0.82, verify the accuracy of R2 is 0.91; at the same time, using the imaginary part of the dielectric constant of the integrated model simulation of soil salinity, salt index: (water + salt) 2 is the optimal model, modeling The accuracy of the determination coefficient of R2 is 0.84, R2 is 0.91. water and verify the accuracy of the real part, the imaginary part showed a good linear relationship, and the imaginary part of the non-linear relationship between salinity and.2, simulation of the regional surface roughness and soil moisture conditions using the AIEM model, the regional surface roughness and the actual match, the oasis desert crisscross District, near the river roughness larger; soil moisture with high precision, the coefficient of determination R2 reached 0.86, soil moisture in the East West High low distribution, the edge of oasis and the river near the soil moisture was higher than.3, because of the linear empirical model of soil salinity is difficult to accurately assess the soil salt content, so the combination of multi spectral data of auxiliary backscattering information segmentation. By using RadarSat-2 four polarization backscattering coefficient, data extraction and surface roughness and soil moisture on the radar, the spectral index GF-1 extracted data set, and the BP neural network model was built by multiple. Test, the BP neural network model used to extract regional soil salinity information model, simulation accuracy reached 78.95%, so to some extent, the verification of optical remote sensing and active microwave remote sensing combined with effective extraction of saline region information. In summary, this study through the characteristics of active microwave remote sensing, established empirical model exposed area soil moisture inversion using the physical model of AIEM; at the same time, combined with multi spectral data, established the soil salinity model by using BP algorithm, and the simulation precision, quasi synchronous study on soil water and salt surface can provide some reference for information extraction of soil in arid areas.

【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S156.41

【参考文献】