Influence of Spatial Heterogeneity in Error Characterization Using Triple Collocation

Accurate error characterization is essential for validating satellite-based geophysical products. Triple Collocation (TC) estimates random error variances of three mutually independent datasets but assumes a common spatial scale—a condition rarely met in practice. Spatial heterogeneity in the ground truth and mismatches in spatial resolution introduces "spatial representativeness errors", whose influence on error variance estimates remains unexamined. In this study, we have analyzed the sensitivity of the triple collocation estimates using the synthetically generated soil moisture dataset under varying sample sizes and spatial heterogeneity. Our results indicate that sample size (N) affects the TC estimates, with % bias decreasing from ±15% to ±2% for N ranging from 100 to 1000. The study finds that % bias also varies with the degree of spatial heterogeneity across the area under consideration. Additionally, the TC framework exhibits an equal likelihood of overestimation and underestimation. These findings underscore the critical importance of addressing spatial heterogeneity to enhance the reliability and robustness of error characterization in geophysical measurement systems. The study provides valuable insights for improving the applicability of TC in satellite product validation and underscores the need for more advanced approaches to handling spatially diverse datasets.