Estimating root zone soil moisture across diverse land cover types by integrating in-situ and remotely sensed data

Abstract

Many soil moisture networks monitor only one land cover type, typically grassland, and the availability of in-situ soil moisture data in other land cover types is severely limited. Satellite-based radiometers lack adequate resolution to match the spatial variability in land cover, which often occurs at the sub-kilometer scale. Thus, spatial and temporal dynamics of root zone soil moisture in regions with heterogeneous land cover types remain poorly understood. Our objective was to determine how effectively root-zone soil moisture for diverse land cover types can be estimated using a water balance model driven by normalized high-resolution, remotely sensed vegetation indices (VI) data and in-situ meteorological data. Root zone soil moisture dynamics under four different land cover types were estimated using normalized VI data as a proxy for the basal crop coefficient. Correlation coefficients (r) between measured and modeled soil moisture ranged from 0.50–0.92, mean absolute error (MAE) ranged from 0.03–0.06 m3 m−3, and mean bias error (MBE) ranged from -0.05–0.02 m3 m−3 across tallgrass prairie, cropland, mixed hardwood forest, and loblolly pine plantation sites. Model-estimated soil moisture under each land cover type was more accurate than both measured data from the nearest long-term grassland monitoring site and data from the NASA-USDA Enhanced Soil Moisture Active-Passive (SMAP) soil moisture product, providing evidence that in-situ meteorological data and remotely sensed VI data may be integrated into a simple water balance model to better estimate root zone soil moisture across diverse land cover types.