MICROZONATION OF GROUNDWATER FLOW MODEL PARAMETERS

Tightly spaced and concentrated monitoring points at large sites with abrupt changes of potentiometric head and stratigraphy present a formidable challenge for groundwater flow model calibration. Honoring all of the geologic information presented from boring logs, geophysical logs, aquifer tests and seismic data makes the task even more challenging. State of the art computer tools such as optimization software reduce the time requirements, improve the modeler's ability to honor the data, and improve the models predictive reliability. At the initiation of the model development process, the horizontal hydraulic conductivity, vertical hydraulic conductivity, and recharge are zoned to match the range of values in the field data. During calibration the modeler can elect to reconfigure the (parent) zones by either changing zone boundaries or by subdividing the original zones into smaller (daughter) zones. The author has coined the term microzonation to describe this process of zone subdivision. Often a review of the early model simulation data shows that one group of wells in a particular parent zone has a positive residual error while another group of wells in the same parent zone has a negative residual error. Subdividing this problematic parent zone into two or more daughter zones will reduce the model error resulting in an improved model calibration. The daughter zones are iteratively reconfigured to honor the site geology and then optimized with the parameter estimating software. At several test sites in the United States the process has resulted in the root mean square error being less than 5% of the total dynamic head across the model domain. This 5% root mean square error rule of thumb has been shown to be adequate for most predictive groundwater models.