Baldwin et al., 2015

Talk/Poster

Downscaling Satellite Data for Predicting Catchment-scale Root Zone Soil Moisture with Ground-based Sensors and an Ensemble Kalman Filter

Baldwin, Douglas C., Henry Lin and Erica Smithwick (2015)
H43H-1635 Remote Sensing and Modeling of the Terrestrial Water Cycle II Posters, presented at 2015 Fall Meeting, AGU, San Francisco, CA, 14-18 Dec.  

Abstract

Predicting root zone (0-100 cm) soil moisture (RZSM) content at a catchment-scale is essential for drought and flood predictions, irrigation planning, weather forecasting, and many other applications. Satellites, such as the NASA Soil Moisture Active Passive (SMAP), can estimate near-surface (0-5 cm) soil moisture content globally at coarse spatial resolutions. We develop a hierarchical Ensemble Kalman Filter (EnKF) data assimilation modeling system to downscale satellite-based near-surface soil moisture and to estimate RZSM content across the Shale Hills Critical Zone Observatory at a 1-m resolution in combination with ground-based soil moisture sensor data. In this example, a simple infiltration model within the EnKF-model has been parameterized for 6 soil-terrain units to forecast daily RZSM content in the catchment from 2009 – 2012 based on AMSRE. LiDAR-derived terrain variables define intra-unit RZSM variability using a novel covariance localization technique. This method also allows the mapping of uncertainty with our RZSM estimates for each time-step. A catchment-wide satellite-to-surface downscaling parameter, which nudges the satellite measurement closer to in situ near-surface data, is also calculated for each time-step. We find significant differences in predicted root zone moisture storage for different terrain units across the experimental time-period. Root mean square error from a cross-validation analysis of RZSM predictions using an independent dataset of catchment-wide in situ Time-Domain Reflectometry (TDR) measurements ranges from 0.060-0.096 cm3 cm-3, and the RZSM predictions are significantly (p < 0.05) correlated with TDR measurements [r = 0.47-0.68]. The predictive skill of this data assimilation system is similar to the Penn State Integrated Hydrologic Modeling (PIHM) system. Uncertainty estimates are significantly (p < 0.05) correlated to cross validation error during wet and dry conditions, but more so in dry summer seasons. Developing an EnKF-model system that downscales satellite data and predicts catchment-scale RZSM content is especially timely, given the anticipated release of SMAP surface moisture data in 2015.

Citation

Baldwin, Douglas C., Henry Lin and Erica Smithwick (2015): Downscaling Satellite Data for Predicting Catchment-scale Root Zone Soil Moisture with Ground-based Sensors and an Ensemble Kalman Filter. H43H-1635 Remote Sensing and Modeling of the Terrestrial Water Cycle II Posters, presented at 2015 Fall Meeting, AGU, San Francisco, CA, 14-18 Dec..

This Paper/Book acknowledges NSF CZO grant support.