遥感地表温度空间分辨率降尺度研究

摘要

针对地表温度（LST）存在时空分辨率之间的矛盾这一难题，在已有的研究基础之上，对 NDVI 分层回归降尺度方法进行改进，利用各类别层次低分辨率 LST 同多个特征参数间的可决系数（R 2）筛选出最优尺度因子以替换 NDVI 指数。以北京市为研究区，基于 TM 影像计算得到多个特征参数和 LST 数据，将模拟的低分辨率LST 降尺度到多个分辨率层次，同目标 LST 进行精度验证。结果表明：（1）在全局和各类别层次上，根据 R 2筛选出的最优尺度因子均为 UI 指数；（2）在不同分辨率层次上，与全局策略和 NDVI 分层回归方法相比，UI 分层回归方法的降尺度精度均最高；（3）随着分辨率的变化，在各特征参数与 LST 间方程系数的变化程度中，SAVI 和NDBI 要明显低于 UI 和 NDVI，NDVI 的变化程度略低于 UI 指数。%Aiming at the problem that the spatial and temporal resolution of land surface temperature have the contradiction with each other,based on the existing research,this research made improvements on the NDVI stratified regression downscaling method.This paper used the coefficient of determination (R 2 )between low resolution LST and multiple related characteristic parameters as the optimal scaling factor’s screening index on each category level,and NDVI was replaced with the optimal scaling factor.Taking Beijing city as the research area,based on the Landsat TM images to calculate the multiple characteristic parameters and LST data,and downscaling the simulated low resolution LST to multiple resolution levels.The downscaled LST validated against the target LST.Results showed that:(1)in the global and each category level,the optimal scaling factors are all UI index according to the R 2 evaluation index;(2)in different resolution levels,compared with the global strategy and NDVI stratified regression downscaling method,the selected UI index stratified regression downscaling method proposes the highest precision;(3 )as the change of resolution,among the changing degree of fitted equation coefficients between different characteristic parameters and the LST,SAVI and NDBI index are clearly lower than UI and NDVI index,and the degree of change of NDVI is less slightly than UI index.