喀斯特城市地表温度影响因素的遥感反演与分析

发布时间：2018-07-16 16:39

【摘要】：近年来随着城市化进程的加快,大量自然地表逐渐被人工不透水地表替代,地表覆盖发生变化影响城市热环境,产生越来越显著的城市热岛现象等热效应。城市热环境可用地表温度来描述,不透水面、植被、水体等地表覆盖类型是影响城市热环境的主要因素,因此定量分析地表温度影响因素对改善城市生态环境具有重要意义。传统的实验观测法很难获取整个城市地表温度的空间数据,而遥感模型反演法提供了快速获取城市大面积地面温度唯一快速、有效的手段。为了满足热红外遥感成像模拟的计算精度的需要,BP神经网络在估算子像元地表温度时,综合不透水面、植被、水体等多种地表参量。实验证明：尽管估算值与实际反演或实地测量温度存在差距,综合多种地表参量的计算精度高。因此,通过遥感图像和地表温度的研究,就能够为缓解桂林市热岛效应提供理论基础和技术支持。针对桂林喀斯特城市的快速发展所引起日益严重的城市热岛现象,对覆盖研究区的LandsatTM卫星图像利用支持向量机SVM提取土地利用信息,利用缨帽变换提取土壤亮度指数、绿度植被指数、湿度指数等地表参数,利用模型提取归一化植被指数NDVI、比值植被指数RVI、修改型土壤调整指数MSAVI等植被指数和水体指数MNDWI,利用Artis单窗算法估算热红外波段像元尺度地表温度,将地表温度的影响因素作为BP神经网络输入估算30m空间分辨率的亚像元地表温度,分析1989-2006年桂林城区土地利用变化、缨帽变换特征分量变化、植被参数变化、水体指数变化对地表温度的影响机理。结果表明,SVM可提高喀斯特城市土地利用信息遥感分类的精度,可有效地动态监测喀斯特城市土地利用的变化；SVM的地物分类精度和Kappa系数最高,总体分类精度为91.7%,超过90%,Kappa系数为0.827,明显高于人工神经网络、决策树和最大似然法的分类结果；1989-2006年桂林城区土地利用类型发生了很大变化,建筑物面积大幅度增加,而农业用地面积大幅度减少,较小水体的面积萎缩甚至完全消失。缨帽变换的绿度植被指数GVI与地表温度的相关性最大,相关系数最高,为0.8907,可作为最优植被参数来分析植被覆盖状况对地表温度的影响；地表温度与植被指数NDVI、RVI、MSAVI、GVI等均呈明显的负相关关系。研究表明：地表温度随水体指数数值的增加而降低,即地表温度与水体指数呈负相关关系。从1989年至2006年,桂林喀斯特城市规模的扩展导致水体面积由占研究区总面积的4.6%降至4.4%,减少量占总面积的0.2%、占总水体面积的4.3%；研究数据表明：桂林市区的两江四湖能有效改善中心城区的生态环境质量、缓解城市热岛现象。
[Abstract]:In recent years, with the acceleration of urbanization, a large number of natural surface is gradually replaced by artificial impermeable surface. The change of surface cover affects the urban thermal environment and produces more and more obvious urban heat island phenomenon and other thermal effects. Urban thermal environment can be described by surface temperature. Surface cover types such as impermeable surface, vegetation and water body are the main factors affecting urban thermal environment. Therefore, quantitative analysis of the influencing factors of surface temperature is of great significance to the improvement of urban ecological environment. The traditional experimental observation method is very difficult to obtain the spatial data of the whole city surface temperature, but the remote sensing model inversion method provides the only fast and effective method to get the large area surface temperature of the city quickly. In order to satisfy the accuracy of thermal infrared remote sensing imaging simulation, BP neural network is used to estimate the surface temperature of sub-pixel, which includes impermeable surface, vegetation, water body and other surface parameters. The experimental results show that, although there is a gap between the estimated value and the actual temperature inversion or field measurement, the calculation accuracy of synthesizing various surface parameters is high. Therefore, the study of remote sensing images and surface temperature can provide theoretical basis and technical support for alleviating the heat island effect in Guilin. In view of the increasingly serious urban heat island phenomenon caused by the rapid development of karst city in Guilin, the Landsat TM satellite image covering the study area is extracted from the land use information by support vector machine (SVM), and the soil luminance index is extracted by the tassel hat transform. Green vegetation index, humidity index and other surface parameters, The normalized vegetation index (NDVI), the ratio vegetation index (RVI), the modified soil adjustment index (MSAVI) and the water body index (MNDWI) were extracted by using the model. The Artis single window algorithm was used to estimate the thermal infrared band pixel surface temperature. The influence factors of land surface temperature were used as input of BP neural network to estimate the subpixel surface temperature of 30 m spatial resolution. The land use change, tassel transform characteristic component change and vegetation parameter change of Guilin urban area from 1989 to 2006 were analyzed. The influence mechanism of water index change on surface temperature. The results show that SVM can improve the accuracy of remote sensing classification of karst urban land use information, and can effectively dynamically monitor the changes of urban land use. The SVM has the highest classification accuracy and Kappa coefficient. The overall classification accuracy is 91.7, and the Kappa coefficient of more than 90 is 0.827, which is obviously higher than that of artificial neural network. The classification results of decision tree and maximum likelihood method have changed greatly from 1989 to 2006 in Guilin urban area, and the area of buildings has greatly increased. However, the area of agricultural land was greatly reduced, and the area of smaller water bodies shrank or even disappeared completely. The green vegetation index (GVI) of tasseled cap transform has the highest correlation with the surface temperature (0.8907), which can be used as the optimal vegetation parameter to analyze the effect of vegetation coverage on the surface temperature. There was a significant negative correlation between surface temperature and vegetation index (NDVI), RVI, MSAVI, GVI, etc. The results show that the surface temperature decreases with the increase of water body index, that is, there is a negative correlation between surface temperature and water body index. From 1989 to 2006, the expansion of karst city scale in Guilin resulted in the decrease of water body area from 4.6% to 4.44%, which accounted for 0.2% of the total area and 4.3% of the total water area. The data show that the two rivers and four lakes in Guilin can effectively improve the ecological environment quality and alleviate the phenomenon of urban heat island.
【学位授予单位】：广西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP79

【相似文献】