Paper No. 9
Presentation Time: 4:15 PM

SIMPLIFIED ACCESS TO THE POWER OF PEST THROUGH PEST++ AND KEYPEST


HUNT, Randall J., Wisconsin Water Science Center, U. S. Geological Survey, 8505 Research Way, Middleton, WI 53562, FIENEN, Michael N., Wisconsin Water Science Center, U.S. Geological Survey, 8505 Research Way, Middleton, WI 53562, WHITE, Jeremy, United States Geological Survey, Tampa, FL 33612, WELTER, David, Computational Water Resource Engineering, Stuart, FL 34995 and DOHERTY, John, National Centre for Groundwater Research and Training, Flinders University, GPO Box 2100, Adelaide, 5001, Australia, rjhunt@usgs.gov

PEST++, an object oriented adaptation of the parameter estimation code PEST, has been designed to lower the barriers of entry for both users and developers. Initial PEST++ (version 1.X) development focused on implementing the most popular features of PEST using more streamlined user input, and in a manner amenable for extension by other software developers. Two notable differences to PEST are that PEST++: 1) implements singular value decomposition in an efficient manner that retains the functionality of Marquardt lambda; and 2) has the ability to automatically switch between native and super parameters “on the fly”. PEST++ retains backward compatibility with PEST, such that the input files are interchangeable. However, backward compatibility was attained by retaining the original relatively complex PEST input structure, which creates a high learning curve for new users. To reduce input requirements, a “keyPEST” translator has been developed to move minimal keyword-oriented input to a full control file suitable for PEST++ and PEST. Input not explicitly provided by the user is supplied by developer-provided defaults that reflect settings suitable for most problems; alternatively, users have the option to override these defaults by simply specifying the associated keyword and value. It is hoped that these efforts will provide a user-friendly entry point that leads to more robust and efficient access to recently developed sophisticated calibration and uncertainty methods. This, in turn, should provide an accessible conduit to interject the insight and transparency afforded by parameter estimation to more environmental models in the future.