Karst aquifers are difficult to model because the discrete, heterogeneous nature of groundwater flow through conduits, rather than through pore spaces, leads to high uncertainty. Existing models rely either on a detailed conduit map, or on effective parameters approximating a porous medium. Neither approach is adequate for most systems, where conduits are unmapped, yet flow patterns are fundamentally different from those in porous media.
We present pyKasso, a Python package for conduit evolution modeling, which stochastically generates numerous plausible conduit maps for a study site. The only inputs are the locations of the inlets and outlets to the system, the orientation and distribution of fracture families, and a geologic model of the system, based on existing geologic maps. Conduit evolution is represented by an anisotropic fast marching algorithm, making our approach extremely computationally efficient compared to chemistry-based models, and better able to handle complex geology and topography than previous attempts based on isotropic fast marching.
We include several statistical analysis and visualization tools for the resulting ensemble of networks, such as calculating topological and geometrical metrics, and visualizing a probability map indicating where conduits are most likely to occur. We use these tools to identify potential collapse and/or contamination hazard zones, and to identify the most informative locations for field measurements such as dye tracer tests. The key benefit of this approach is that it allows small, resource-limited communities dependent on karst groundwater to maximize the value of the information they can acquire on a limited budget.