Featured Product: Evaluating the Source of Water to Wells Using Metamodeling
Numerical groundwater modeling is a strength of the USGS, from software development (likeMODFLOW) to applications, such as the regional groundwater availability studies highlighted in the December 2015 News and Highlights.
Numerical models are our most powerful tools for decision making. But these models require expertise to operate and require significant computational resources. Natural resource managers need reliable models to predict surface water impacts dues to installation of new groundwater wells and may not have the expertise or computer resources to run the necessary simulations. An alternative is to emulate the numerical process model with a statistical “metamodel”, or surrogate model.
This new paper illustrates the use of a metamodel in the aquifers within the Lake Michigan basin, a temperate region where shallow groundwater is closely connected with lakes, wetlands, and streams. A groundwater flow model was emulated using three statistical techniques -- a Bayesian Network, an Artificial neural network, and a Gradient Boosted Regression Tree. A numerical MODFLOW-USG model was used for the evaluation.