Soil state and flux characterization using a stochastic data assimilation approach

Wastewater reuse via irrigation provides the potential for increases in water use efficiency, but excessive solutes from wastewater can pollute the groundwater beneath the irrigated lands. To avoid this adverse impact and provide a mechanism for informing optimal management practices, the first study develops a monitoring and modeling system by assimilating embedded sensor network measurements into a hydrologic model to provide real-time soil state and flux estimates. Results from synthetic observing system simulation experiments (OSSEs) show that state (i.e. soil moisture) estimation in isolation can lead to significant errors if flux estimates are a primary objective of the estimation framework and parameters are not well characterized. However, the OSSEs indicate that with sufficient measurement information, the system is capable of providing an accurate characterization of real-time soil state, model parameter, and flux estimates. The second study applied the proposed monitoring and modeling system to a real operational system of wastewater reuse. Results show that under an erroneous soil structure, the proposed filtering scheme is able to describe the spatial variability of moisture and salinity in soils. Also, the spatial data withholding experiment demonstrates that state predictions at unobserved areas can be improved via the assimilation, and the temporal data withholding experiment indicates that a high update frequently can maintain the reliability of state estimation. In the third study, large-scale moisture retrieval by assimilating microwave radiobrightness measurements is considered. A better model characterization is argued to further improve moisture estimates. Results show that for a synthetic homogeneous land surface (relative to the remote sensing footprint), the filtering scheme with parameter estimates can further improve estimates in moisture and associated energy flux. Within a heterogeneous measurement pixel, this proposed filtering scheme can also have better estimates of land surface variables at measurement scales compared to the conventional EnKF. Based on these studies, the real-time posterior estimates including moisture and associated fluxes through the proposed filtering scheme could be used to optimize irrigation management and/or describe the land surface processes associated with moisture dynamics.