PNNL

Abstract

This study presents a detailed intercomparison and evaluation of 19 gridded global soil moisture datasets, focusing particularly on the Contiguous United States (CONUS). These datasets include a collection of land surface models, remote sensing, reanalysis, and machine learning products. All datasets are harmonized to a 0.25-degree spatial resolution, covering various temporal spans and providing a comprehensive view of soil moisture dynamics. The analysis leverages the Koppen-Geiger Climate Classification to explore soil moisture’s spatiotemporal variability across different climatic zones. Our results highlight distinct patterns, with arid regions showing lower moisture variabilities and temperate areas exhibiting higher values. Remote sensing data sets tend to indicate drier conditions, while reanalysis products often present wetter estimates. In-situ soil moisture observations from the International Soil Moisture Network serve as the basis for wavelet power spectrum analyses for deeper exploration of discrepancies with respect to temporal scales across the 19 datasets. Despite challenges posed by data heterogeneity and resolution discrepancies, our approach provides a robust framework that can be used for recommendations such as irrigation scheduling, flood risk mitigation, and drought response planning based on soil moisture assessment.