Downscaling soil moisture in the southern Great Plains through a calibrated multifractal model for land surface modeling applications

Accounting for small-scale spatial heterogeneity of soil moisture (θ) is required to enhance the predictive skill of land surface models. In this paper, we present the results of the development, calibration, and performance evaluation of a downscaling model based on multifractal theory using aircraft-based (800 m) θ estimates collected during the southern Great Plains experiment in 1997 (SGP97). We first demonstrate the presence of scale invariance and multifractality in θ fields of nine square domains of size 25.6 × 25.6 km², approximately a satellite footprint. Then, we estimate the downscaling model parameters and evaluate the model performance using a set of different calibration approaches. Results reveal that small-scale θ distributions are adequately reproduced across the entire region when coarse predictors include a dynamic component (i.e., the spatial mean soil moisture θ) and a stationary contribution accounting for static features (i.e., topography, soil texture, vegetation). For wet conditions, we found similar multifractal properties of soil moisture across all domains, which we ascribe to the signature of rainfall spatial variability. For drier states, the θ fields in the northern domains are more intermittent than in southern domains, likely because of differences in the distribution of vegetation coverage. Through our analyses, we propose a regional downscaling relation for coarse, satellite-based soil moisture estimates, based on ancillary information (static and dynamic landscape features), which can be used in the study area to characterize statistical properties of small-scale θ distribution required by land surface models and data assimilation systems.