基于泛克里金插值法的Landsat8热红外数据的高空间分辨率地表温度反演

The Land Surface Temperature Retrieved from Landsat8 Thermal Infrared Data Based on the Universal Kriging Interpolation Method

地表温度(land surface temperature, LST)地表物理状态过程中的一个重要参数,目前反演高精度地表温度是一个重要的研究目标。本文根据热辐射传输方程,利用空间分辨率30m Landsat8传感器的热红外数据反演地表温度,并利用泛克里金插值法对反演温度的结果进行尺度转换,获取2 m分辨率的地表温度数据。本文以广东省河源市东江林场为例,对2013年10月21日、2014年10月24日及2015年10月27日的Landsat8进行温度反演,对其反演结果进行泛克里金插值,最终获取2 m高空间分辨率的温度数据。本文以广州东江林场2013年10月21日、2014年10月24日和2015年10月27日获取的landsat8为例,为了检验地表温度数据的尺度转换的合理性,本文利用莫兰系数和夏皮洛-威尔逊检验法检验地表温度是否空间自相关性以及是否为正态分布。三帧温度结果的莫兰系数分别是0.96,0.96,0.97,表明该三帧温度数据空间自相关;由夏皮洛-威尔逊检验法获取的p值分别为0.1857,0.09968,0.08534,均大于0.05,温度数据服从正态分布。对2015年泛克里金插值前后的温度值与实测温度进行比较,平均误差系数减少了5.63%。由此可见,本文利用泛克里金法插值法对Landsat8反演的温度进行尺度转换后获得的结果更为合理。

Land surface temperature(LST)is a key parameter in the surface physical state process.Many researches focused on retrieving for high precision LST.In this paper,the landsat8 thermal infrared data with 30 m spatial resolution were used to retrieve LST based on the thermal radiant transfer equation.Then,the universal Kriging interpolation method were applied for obtaining the surface temperature data with 2 m spatial resolution.This study was conducted in Dongjiang state forest farm of Heyuan City,Guangdong Province,China,using Landsat8 satellite images from 21 October 2013,24 October 2014 and 27 October 2015.Moran’I coefficient and normal distribution were performed to validate the reasonability of universal Kriging interpolation method for scaling up the temperature images.The Moran’I coefficient of three images are 0.96,0.96 and 0.97,which indicated that the temperature data were spatial autocorrelation.And,the p-value of Shapiro-Wilk test for three temperature image were respectively 0.1857,0.09968 and 0.08534,which implied that three temperature image obeyed normal distribution.The prediction accuracies of the 2 m spatial resolution surface temperature from the universal Kriging interpolation method and the 30 m spatial resolution surface temperature were compared based on field observations.The results showed that the universal Kriging interpolation method successfully estimated surface temperature in the whole study area with decreasing error of 5.63%.