The model independent parameter estimation and predictive uncertainty analysis software PEST is used to support model calibration and quantification of model predictive uncertainty. PEST is unique in that no programming is required to interface an existing model with PEST because it communicates with a model through the model's own input and output files.
PEST is based on weighted least squares and a robust implementation of the Gauss-Marquardt-Levenburg method. It also includes methodologies to support regularized inversion and quantification of model predictive uncertainty. A suite of ancillary support software expedite the PEST / model interface and deployment process.
The Watershed Systems Group at CHL has guided and supported further development of PEST. The research efforts have principally been directed to the development of additional software / capabilities within the PEST software suite to support efficient accommodation of
- local minima in watershed model calibration
- highly parameterized watershed models
- model predictive uncertainty
The Watershed Systems Group also utilizes the PEST software suite to support practical watershed modeling studies, with clients including district offices of the U.S. Army Corps of Engineers, the U.S. Navy, U.S. Geological Survey, and agencies / departments within local and state governments.
The Watershed Systems Group can provide training with how to properly use PEST, and / or consultancy services in model calibration and predictive uncertainty analysis.
Print Page Content...