HR: 1340h
AN: H33F-0516 [Abstracts]
TI: Evaluating Least Absolute Deviation Regression As An Inverse Model In Groundwater
Calibration
AU: * Huddleston, J M
EM: jhuddleston@itc.nrcs.usda.gov
AF: John Huddleston, USDA NRCS, 2150A Centre Ave, Fort Collins, CO 80536
United States
AB:
A new Least Absolute Deviation and Expectation Maximization (LAD-EM) FORTRAN procedure replaced the MODFLOW groundwater model
Least Squares (LSG) parameter estimation inverse procedure. The LSG and LAD-EM MODFLOW inverse models were applied to a
multi-layer 20,000 hectare watershed in Iowa. The hydraulic conductivity parameters computed with the LAD-EM MODFLOW inverse
model agree with published data. The LSG MODFLOW model predicted a confined aquifer system while the LAD-EM MODFLOW model
predicted an unconfined system. Literature documents the area as an unconfined system. The hydraulic conductivity
parameters computed by both inverse models were used to create two forward basin models of a larger encompassing 600,000
hectare river basin in Iowa. Application of the LAD-EM MODFLOW model hydraulic conductivities to the river basin produced
model heads that were within two meters of observed heads. Application of the LSG MODFLOW model hydraulic conductivities
produced model heads that were 15 meters below observed heads. The LAD-EM MODFLOW model is an excellent tool for modeling
multi-layer groundwater aquifers.
DE: 1829 Groundwater hydrology
SC: Hydrology [H]
MN: 2004 AGU Fall Meeting