统计模型与滤波算法的地表温度重建方法探讨

Study on reconstruction of LST based on the statistical model and filtering algorithm

针对当前基于被动微波遥感重建地表温度的统计方法难以实现大面积复杂下垫面区域数值重建的问题,提出了基于统计模型与滤波算法联合的地表温度重建方法。从时间序列角度探索地表温度与地表亮温的相关性,建立二者之间的统计模型,不需要进行地物分类,能有效避免下垫面复杂度对重建精度的影响;遍历像元,实现对大面积区域的数值重建。此外,采用滤波算法对基于统计模型的结果进行改正,利用地表温度时间序列的周期性进一步控制反演误差。针对MODIS地表温度产品重建的实验结果表明:所提算法精度明显提高,可用于各类下垫面覆盖区域的地表温度重建。

In view of the problem that the exist statistical methods of land surface temperature recon- struction based on passive microwave remote sensing are hard to achieve temperature values reconstruction in large-scale area with complex underlying surface, a land surface temperature reconstruction method based on statistical model and filtering algorithm was proposed. The correlation of land surface temperature and brightness temperature was explored from time series to establish the statistical model between them without ground cover classification, which effectively avoided the influence of underlying surface complex- ity on the reconstruction accuracy. The temperature values of large-scale area were reconstructed through traversing pixels. In addi[tion, the inversion error was controlled by using filter algorithm based on the pe- riodicity of land surface temperature time series to correct the results of statistical model. The results of MODIS land surface temperature reconstruction showed that the inversion accuracy was significantly im- proved, and this method could be used for all kinds of underlying surface.