【摘要】 气温(Ta)是描述陆地气候环境的一个重要参数,其异常变化直接影响人类的生存环境,因此如何高精度地估算气温成为当前研究的热点。MODIS数据因其分辨率较低不能提供精细的地表信息,为此,本文以更高分辨率的Landsat8影像为数据源,结合自动气象站的气温数据,耦合经纬度、归一化植被指数、归一化建筑指数和改进的归一化水体指数等多种因子,建立了多窗口线性回归模型(Multi-Window Linear Regression Model,MWLR)。最后以浙江北部为研究区,使用MWLR模型对该地区冬季气温进行了估算,模型预测的RMSE在1.458~1.551℃之间,R~2在0.835~0.842之间,当窗口大小为3×3时取最优精度(RMSE=1.458℃,R~2=0.835),优于一般的空间内插方法。研究结果验证了利用MWLR模型和Landsat8影像进行气温估算的有效性,并提供了一种基于遥感数据在局部地区开展高精度和高分辨率气温估算的模型。更多还原

【Abstract】 Air temperature(Ta) is a key descriptor of the terrestrial environment,and how to estimate Ta of high accuracy is a hotspot since the abnormal change of Ta has influence on the living environment of human beings.Considering MODIS data cannot offer much precise surface information,in this paper,a Multi-Window Linear Regression Model(MWLR)is established using Landsat8 images of higher resolution by coupling latitude and longitude,normalized difference vegetation index,normalized difference building index and modified normalized difference water index with Ta obtained from automatic meteorological stations.Finally,the MWLR model is used to estimate Ta of northern Zhejiang Province in winter,and RMSEof prediction is between 1.458~1.551℃,meanwhile,R~2 reaches 0.835~0.842.The highest precision is achieved when the window size is 3×3(RMSE=1.458 ℃,R~2=0.835).In a word,the result shows the accuracy of estimating Ta based on MWLR model and Landsat8 data,which means our model is appropriate to develop high precision and resolution Ta estimation over local area using remote sensing data.更多还原