Obtaining Useful Answers from Groundwater Models

NWRA 8:00 am – 12:00 pm, Tuesday, February 3, 2026

01_GWmodels_INTRO+Basics+Real.v2.pptx

Flow models provide consistent stories that can be compared to data
Lateral and vertical extents should be enough to solve problem without going too far
Hydraulic properties, stresses, and boundaries must be defined for entire model
Parameter estimation formalizes model calibration, where measured and simulated quantities are compared in an objective function
Parameters are changed so that objective function is reduced
Parameters typically define arrays of hydraulic properties with zones or pilot points
Parameter-to-array example and tube-model workbooks in SupportingMaterial\02_Interpolate+TubeMODEL subfolder

Geologic frameworks distribute hydraulic properties through groundwater-flow models
Groundwater controlled by hydraulically connected fractures
Detailed, complex frameworks do not improve hydraulic-property estimates
Compare transmissivity, not hydraulic conductivity, to rock types

Calibration should be evaluated with more than statistics, such as RMS error
Visual comparison of model results required to evaluate model plausibility
Maps and plots provide better visualization, compared to tables
Looking necessary to cross bar of “Not obviously wrong”
Broad visualization of simulated results uncovers undefined badness,
which leads to better models
Simplifying model features and integrating results for mapping eases comprehension
Cheap tools such as Model Viewer and Google Earth encourage looking by more users
Transmissivity distributions simplify presenting model results and aids comprehension
Water-level profiles are processed quickly and function
like hydrographs for steady state models

Geologic frameworks are not data, Distinguish between interpretation and reality
Geologic frameworks are foundations for distributing hydraulic properties
Revise geologic framework based on hydrologic data
Truncate geologic frameworks based on groundwater model purpose
Detailed, complex frameworks are not always better than simplified frameworks

Calibration & uncertainty analysis depend on judgement
- Model developers’ judgement explicitly embedded in objective function and parameters
  - Objective function—Data exclusion, data inclusion, and weighting of included data
  - Parameters—Spatial extent of faults or zones and pilot-point density,
    Ranges of defined parameters, where textbook ranges allow too much
Uncertainty analysis is not free and can cost measuring useful data

Results not overly sensitive to detail of model grid
Increasing discretization increases model run time, which weakens calibration and
post-processing of results
Models flexible—Answer many questions and are improved by clear purpose
Data and questions should drive model complexity
- Additional data dispels ignorance, not more models
- Show added details improve answers
Clear ties to measurements increase confidence
A figure of integrated model results is easier to comprehend than multiple figures and tables that need to be integrated mentally.
Require complete reporting
- Interior of Earth can be mysterious,
- No aspect of a model should be mysterious
Workbook 01_DD_CoarseFineGRID_Compare.xlsm and supporting files for creating, executing, and viewing coarse and fine models in SupportingMaterial\07_Apply+Grid+Report subfolder