Conceptual uncertainty is considered one of the major sources of uncertainty in groundwater flow modeling. Hypothesis testing is essential to increase system understanding by analyzing and refuting alternative conceptual models. We present a systematic approach to conceptual model testing aimed at finding an ensemble of conceptual understandings consistent with prior knowledge and observational data. This differs from the traditional approach of tuning the parameters of a single conceptual model to conform with the data through inversion. We apply this approach to a simplified hydrogeological characterization of the Wildman River area (Northern Territory, Australia) and evaluate the connectivity of sinkhole-type depressions to groundwater. Alternative models are developed representing the process structure (i.e., different fluxes representing interactions between surface water and groundwater) and physical structure (i.e., different lithologies underlying the depressions) of the conceptual model of the depressions. Remote sensing data are used to test the process structure, while geophysical data are used to test the physical structure. Both data types are used to remove inconsistent models from an ensemble of 16 models and to update the probability of the remaining alternative conceptual models. Three out of five depressions that are used as a test case are conditionally confirmed to act as conduits for recharge, while for the last two depressions, the data are inconclusive. Although the framework is not directly prediction oriented, the testing of plausible conceptual models will ultimately lead to increased confidence of any groundwater model based on accepted posterior conceptualizations.
Bibliographical noteFunding Information:
This research project was funded through CSIRO's Deep Earth Imaging Future Science Platform. The seismic datasets for this research are available at https://doi.org/10.17632/dx328m8cg2.1. The Planet Labs Education and Research Program are thanked for providing access to PlanetScope imagery. The authors would like to thank Chris Turnadge, Ursula Zaar, and Steve Tickell for their assistance in the field campaign and Pascal Castellazzi for help in downloading the PlanetScope data and offering insights into its interpretation. The authors would also like to thank Russell Crosbie, Tao Cui, and three anonymous reviewers for review and constructive comments on the manuscript.
©2020. The Authors.
Copyright 2020 Elsevier B.V., All rights reserved.
- conceptual model
- model rejection
- model testing
- multimodel framework