Fusing Spatial Datasets and the Change of Support Problem

The fusing of spatial datasets is becoming more common as the availability of remote-sensing datasets grows. Data fusion combines different observations into a single, higher-quality representation of the same underlying phenomenon of interest, benefiting from the unique strengths of the inputs while mitigating their shortcomings. Fusing two or more spatial datasets has the potential to produce more accurate estimates of natural processes than modeling each dataset separately. However, different datasets often have different spatial supports (e.g., areal versus point data or multiple areal datasets with differing areal units), and the resulting differences in statistical properties require careful handling during modeling—a problem known as “change of support.” In this talk, we review the change of support problem as it relates to the fusing of two or more spatial datasets in climate modeling. We discuss why accounting for change of support matters, review and compare existing and newly-expanded tools for statistical modeling available in the R programming language, and discuss the strengths and weaknesses of the different approaches. Our study provides practitioners with an understanding of the problem, and a detailed primer on the tools available for appropriately quantifying uncertainty in spatial models from fused datasets. Sandia National Laboratories is a multimission laboratory managed and operated by National Technology & Engineering Solutions of Sandia, LLC, a wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.