Modeling groundwater flow on a regional scale has its own challenges because a regional model itself deals with refinement requirements, larger baselines, higher differences in elevations, complex geologies and areas without observation data.

The applied model of this tutorial covers the Angascancha Reservoir basin groundwater flow model in Ayacucho, Peru. This example have been modeled in several courses, tutorials and webinars with Modflow 2005, Flopy and Model Muse, however, this is the first time we port the model (succesfully) to Modflow 6 and Model Muse 4.

Great free and open groundwater flow modeling software are in constant development process, therefore year by year we have newer versions or even new software. This time we have uptated a basic and useful tutorial from 2016 to the new version of Modflow and Model Muse. The tutorial shows the complete procedure to download and install Model Muse 4 and configure Modflow 6.

We have explored the new (or not so new) options of Modflow 6 and Model Muse 4 to create local refinements on a regional scale based on hydraulic features. So far, we came up with a decent example of regional groundwater modeling on a andean basin with three boundary conditions and decreasing hydraulic conductivity with depth. The tutorial also explores the complexities to model basins with high difference on elevation and the use of text defined parameters to implement the Newton formulation in Modflow 6.

Finite difference method as well as any other discretization method allows the conceptualization of a geological media into cells or other volumes. Geological models come in diverse formats in binary or text format and need to be “translated” to the cell extension of a groundwater model.

This tutorial has a applied example of the implementation of a 3D geological model from a neural network into a groundwater model with determined horizontal discretization and layer thickness. The tutorial covers all the steps for model construction and hydrogeological unit determination with scripts in Python with Flopy and other libraries. Comparisons of the original and translated geological model were done as Matplotlib plots and Vtk files in Paraview.

We have done tutorial in Python and recent and powerful libraries as Scikit Learn to create a geological model based on lithology from drillings on the Treasure Valley (Idaho, USA). The tutorial generates a point cloud of drillings lithologies that are transformed and scaled for the neural network. The selected neural network classifier is Multi-layer Perceptron classifier implemented on the Scikit Learn library as sklearn.neural_network.MLPClassifier. An analysis of the confusion from the neural network is performed. The tutorial also includes a georeferenced 3D visualization from well lithology and interpolated geology as Vtk format in Paraview.