An Application of Reification for a Rainfall-Runoff Computer Model

• Activity Number: 593
• Type: Topic Contributed
• Date/Time: Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Leanna House*+ and Ian Vernon and Allan Seheult
• Companies: Virginia Polytechnic Institute and State University and University of Durham and University of Durham
• Keywords: computer model ; discrepancy ; reification ; Bayes linear ; subjective ; rainfall-runoff

Abstract

This is an application paper that exemplifies the need for subjective prior specification and uses a method called Reification (Goldstein and Rougier, 2008) to do so. The applied research goal is to assess the hydrological process by which rain becomes runoff in the Swiss Haute-Mentue research catchment (Iorgulescu et. al, 2005) given two types of data: real and computer-simulated data. To use both data types, an assessment of the discrepancy between the true hydrological process and the simulator is necessary, and several analysis methods exist. Most methods, unlike reification, do not rely on experts as a data source for model discrepancy, yet in our application, the model developer can identify specific model inadequacies. Thus, we use reification in our Bayes linear analysis of the Swiss Haute-Mentue research catchment.