Characterizing the Effect of Spatial Heterogeneity and the Deployment of Sampled Plots on the Uncertainty of Ground “Truth” on a Coarse Grid Scale: Case Study for Near‐Infrared (NIR) Surface Reflectance

Abstract

AbstractThe quality of surface reflectance product is fundamentally important to ensure the information extracted from the downstream products can be trusted. However, ground validation of surface reflectance satellite products is challenging, because ground “truth” on a coarse grid scale based on sparse ground measurements is subject to uncertainty due to spatial heterogeneity. Based on a high‐resolution numerical simulation, this study quantifies the influence of spatial heterogeneity on the uncertainty of surface reflectance ground “truth” in different sampling cases (i.e., the number and positions of sampled plots). The effect of spatial heterogeneity is very small and even negligible when optimal locations are identified. By contrast, for the averaged condition without optimal sampling, spatial heterogeneity is almost the only source of uncertainty of ground “truth,” with determination coefficients of 0.94, 0.99, and 1 for 1, 2, and 3 sampled plots, respectively. The response coefficients between them are about 0.77, 0.55, and 0.47 for 1, 2, and 3 sampled plots, respectively. Optimal sampling plays a more important role in reducing the uncertainty of ground “truth” than increasing the number of sampled plots. Particularly, the latter is not helpful in reducing the uncertainty when the worst locations are adopted. Furthermore, the thresholds of spatial heterogeneity of surface reflectance to meet a predefined uncertainty of 0.01 (about 3.5% for surface reflectance over the study area) were given for different sampling cases over northern China. This study provides important guidance to aid in situ network design for the validation of 1 km satellite surface reflectance products.