A Comparison of Response Surface Methodology and a One-Factor-At-A-Time Approach as Calibration Techniques for the Bioplume-II Simulation Model of Contaminant Biodegradation
Abstract
This thesis compared the Response Surface Methodology (RSM) approach to the one-factor-at-a-time approach for calibrating the Bioplume II finite-difference simulation model of groundwater flow, contaminant transport and biodegradation. The MADE-2 data set of hydrocarbon injection into pristine groundwater at Columbus Air Force Base, Mississippi was used in this research. Because the simulation includes both groundwater flow and contaminant transport, each calibration included both phases. The one-factor-at-a-time approach reduced the root-mean-squared (RMS) error criterion for the flow to 0.921225 feet in a total of 36 runs of Bioplume. The RSM approach reduced the error criterion to 0.918875 feet in a total of 47 runs. The one-factor-at-a-time approach was unable to reduce the RMS error criterion for the transport calibration below an initial value of 67.1831 parts per billion (ppb) benzene after 21 runs which spanned the feasible range of each of the parameters. The RSM approach was able to reduce the response to 67.0327 ppb after 47 runs of Bioplume. The RSM approach allows the modeler to identify parametric regions of improved response in a systematic way that would be extremely difficult to find using the one-factor-at-a-time approach. For this reason it may be very useful for calibration of Bioplume models to be used for research or long term monitoring of a contaminated site, where extra prediction accuracy may be needed. The major limitations of this work were the use of inefficient full factorial designs for the Response Surface Methodology approach and the limited improvement possible on the response surfaces possibly due to the assumption of homogeneous parameter values.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1995
- Accession Number
- ADA305927
Entities
People
- Benjamin Shuman
Organizations
- Air Force Institute of Technology