• Cave entrances represent the confluence of surface and subsurface morphology
  • Cave entrances locations can be predicted using a machine learning approach


Cave entrances directly connect the surface and subsurface geomorphology in karst landscapes. Understanding the spatial distribution of these features can help identify areas on the landscape that are critical to flow in the karst groundwater system. Sinkholes and springs are major locations of inflow and outflow from the groundwater system, respectively, however not all sinkholes and springs are equally connected to the main conduit system. Predicting where on the landscape zones of high connectivity exist is a challenge because cave entrances are difficult to detect and imperfectly documented. Wildlife research has a similar issue of understanding the complexities of where a given species is likely to exist on a landscape given incomplete information and presence-only data. Species distribution models can address some of these issues to create accurate predictions of species or event occurrence across the landscape. Here we apply a species distribution model, MaxEnt, to predict cave entrance locations in three geomorphic regions of Kentucky. We built the models with cave locations from the Kentucky Speleological Survey database and landscape predictor variables, including distance from sinkholes, distance from springs, distance from faults, elevation, lithology, slope, and aspect. All three regional models predict cave locations well with the most important variables for predicting cave entrance locations consistent between models. Throughout all three models, sinkholes and springs had the largest influence on the likelihood of cave entrance presence. This unique use of species distribution modeling techniques shows that they are potentially valuable tools to understand spatial patterns of other landscape features that are either ephemeral or difficult to identify using standard techniques.



Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-Share Alike 4.0 License.

Blitch etal.ris (1 kB)
Export RIS