An Optimal Sampling Design for Land Surface Temperature Validation with Spatial and Diurnal Variations

Abstract

The development of ground-based sampling strategies is vital to the validation of medium- or coarse-resolution satellite-derived land surface temperature (LST) products, extremely over heterogeneous ground with dramatic diurnal LST change. An optimal sampling strategy in support of LST validation across both spatial and diurnal scales (SDS) was proposed in this study. The SDS integrated prior knowledge of land-cover, multi-temporal feature information and spatial distribution of samples to improve the representativeness of the samples. The SDS were also compared with three sampling strategies including random, systematic, and land-cover base sampling. The results obtained by the remote sensing simulation data indicated that the SDS performed best with stable root mean square errors (RMSE) less than 0.1k when sample ratio was more than 2%, and the representativeness of samples selected by the SDS in both diurnal space and spatial space was superior to the current sampling strategies.