A fully-coupled physically-based land surface hydrologic model, Flux-PIHM, is developed by incorporating a land-surface scheme into the Penn State Integrated Hydrologic Model (PIHM). The land-surface scheme is mainly adapted from the Noah LSM, which is widely used in mesoscale atmospheric models and has undergone extensive testing. Because PIHM is capable of simulating lateral water flow and deep groundwater, Flux-PIHM is able to represent both the link between groundwater and the surface energy balance, as well as some of the land surface heterogeneities caused by topography. Flux-PIHM has been implemented and manually calibrated at the Shale Hills watershed (0.08 km2) in central Pennsylvania. Model predictions of discharge, soil moisture, water table depth, sensible and latent heat fluxes, and soil temperature show good agreement with observations. The discharge prediction is significantly better than state-of-the-art conceptual models implemented at similar watersheds.
The ensemble Kalman filter (EnKF) provides a promising approach for physically-based land surface hydrologic model calibration. A Flux-PHIM data assimilation system is developed by incorporating EnKF into Flux-PIHM for model parameter and state estimation. This is the first parameter estimation using EnKF for a physically-based hydrologic model. Both synthetic and real data experiments are performed at the Shale Hills watershed to test the capability of EnKF in parameter estimation. Six model parameters selected from a model parameter sensitivity test are estimated. In the synthetic experiments, synthetic observations of discharge, water table depth, soil moisture, land surface temperature, sensible and latent heat fluxes, and transpiration are assimilated into the system. Observations are assimilated every 72 hours in wet periods, and every 144 hours in dry periods. Results show that EnKF is capable of accurately estimating model parameter values for Flux-PIHM. In the first set of experiments, different initial guesses are given to different test cases. The estimated parameter values from all test cases are very close to the true parameter values. For five out of the six parameters, the deviations of the mean parameter values are within onestandard deviation of the true values for more than 50% (up to 77%) of the calibration period. Synthetic experiments are also performed to test the impact of different observations on parameter estimation. It is found that discharge, soil moisture, and land surface temperature (or sensible and latent heat fluxes) are the most critical observations for Flux-PIHM calibration. In real data experiments, in situ observations of discharge, water table depth, soil moisture, and sensible and latent heat fluxes are assimilated. Results show that, for five out of the six parameters, the EnKFestimated
parameter values are very close to the manually-calibrated parameter values. The hydrologic predictions using EnKF-estimated parameters and manually-calibrated parameters are very similar. The results demonstrate that, given a limited number of site-specific observations, an automatic sequential calibration method (EnKF) can be used to optimize Flux-PIHM for watersheds like Shale Hills.
Shi, Y., Davis, K.J., Zhang, F., Duffy, C. (2012): Parameter Estimation of a Physically-Based Land Surface Hydrologic Model Using the Ensemble Kalman Filter. AGU Annual Fall Conference Proceedings.
This Paper/Book acknowledges NSF CZO grant support.