A Multiscale Nested Sampling Method for Representative Albedo Observations at Various Pixel Scales

Validation of satellite products is necessary to enable end-users to know the fitness and exploit their full potential in supporting the decision-making process. Representative in situ observations need to be obtained to make an independent, unified, standard, and traceable validation due to the spatial scale mismatch between satellite and in situ-based observations. Under the demand of validating various pixel scales of satellite products ranging from a few meters to kilometers, this article proposed a multiscale nested sampling method (MNSM), which is able to provide representative ground observations at various pixel scales under a unified accuracy requirement. The basic idea is to extract the spatiotemporal variation information of surface variables using long time sequences of high-resolution images. And the sample locations were optimized with a random combination method using an index of spatiotemporal representativeness. The optimal sampling was carried out from a coarse pixel scale to a high pixel scale with the top-down nested approach. The final “point” in situ observations can represent surface variables at different pixel scales with a unified and predetermined accuracy. It was found that the representativeness error of a single in situ measurement and the minimum number of required filed plots under a unified sampling accuracy requirement are partly determined by spatial heterogeneity and partly determined by the spatial scale mismatch between the plot size and the pixel size to be matched. The sampling method is applicable to many other surface variables (e.g., soil moisture, biomass, albedo, GPP, and LAI) that exhibiting spatial heterogeneity.