Wednesday, April 2, 2008 : 1:00 p.m.

On Model Selection Criteria in Multimodel Analysis

Shlomo P. Neuman, University of Arizona, Ming Ye, Ph.D., Florida State University and Philip D. Meyer, Pacific Northwest National Laboratory

There has been a growing tendency to postulate several alternative hydrologic models for a site and use model selection criteria to (a) rank these models, (b) eliminate some of them and/or (c) weigh and average predictions and statistics generated by multiple models. This has led to some debate among hydrogeologists about the merits and demerits of common model selection criteria such as AIC, AICc, BIC, and KIC. In particular, whereas we have based our approach to multimodel hydrologic ranking and inference on the Bayesian criterion KIC (which reduces asymptotically to BIC), others have voiced a strong preference for the information-theoretic criterion AICc (which reduces asymptotically to AIC). We compare the theoretical underpinning of the four criteria, explain why KIC is the only criterion accounting validly for the likelihood of prior parameter estimates, elucidate the unique role that the Fisher information matrix plays in KIC, and demonstrate through an example that it imbues KIC with desirable model selection properties not shared by AIC, AICc or BIC. Our example appears to provide the first comprehensive test of how AIC, AICc, BIC and KIC weigh and rank alternative models in light of the models’ predictive performance under cross-validation with real hydrologic data.

Shlomo P. Neuman, University of Arizona Regents' Professor of Hydrology, B.Sc. (1963) Geology Hebrew University Jerusalem, M.S. (1966) and Ph.D. (1968) Engineering Science UC Berkeley, Member U.S. National Academy of Engineering, Fellow American Geophysical Union and Geological Society of America.

Ming Ye, Ph.D., Florida State University Dr. Ming Ye is an Assistant Professor in the School of Computational Science and Department of Geological Sciences of the Florida State University. Before joining the Florida State University, he was an Assistant Research Professor of the Desert Research Institute, and post-doc of the Pacific Northwest National Laboratory. He received his Ph.D. in hydrology from the University of Arizona in 2002, and a B.S. in geology from Nanjing University, China, in 1997. His research interests include groundwater modeling in saturated and unsaturated porous and fracture media, parameter estimation, applied geostatistics, and uncertainty analysis of groundwater modeling.

Philip D. Meyer, Pacific Northwest National Laboratory Philip D. Meyer is a Sr. Research Engineer at Pacific Northwest National Laboratory. He received a B.A. degree in Physics from Cornell University and M.S. and Ph.D. degrees in Civil Engineering from the University of Illinois at Urbana-Champaign. Dr. Meyer has 20 years experience in applying models of flow and transport through unsaturated and saturated porous media to the solution of engineering problems, including the estimation and interpretation of hydrologic uncertainties in dose/risk assessment, analysis of flow and transport in soil covers, engineered barriers, and the near-field environment at waste disposal facilities, and groundwater monitoring network design under uncertainty.


2008 Ground Water Summit