THE REAL-WORLD COMPLEXITY OF KARST—LIMITATIONS AND CONSIDERATIONS RELATIVE TO APPLYING PREDICTIVE MODELS TO THESE AND SIMILAR AQUIFERS

Karst aquifers, by their very method of formation, are highly nonhomogeneous and anisotropic. Ground-water flow is concentrated along unpredictable, preferred flowpaths that are typically open and much more permeable than the surrounding rock matrix. Hydrogeologic modeling studies in areas underlain by limestone and other easily dissolvable karst rocks generally are undertaken at a regional scale, because at the larger scale, unknown and unpredictable flowpaths become more integrated, discrete conduits become less influential for specific hydrologic-budget components, flow more closely obeys Darcy's Law, and prediction thus becomes more tractable from an empirical and theoretical point of view. Predictive models work better at these larger scales.

Environmental and contaminant questions in karst, on the other hand, typically require site-specific watershed-scale answers that are commonly 3 to 5 orders of magnitude more refined than the 10 kilometer scale typical of regional studies. The real-world complexity of karst at this smaller scale does not lend itself to predictive modeling, and although many have tried to apply models to site-specific studies, the reality is that the predictions are constrained by the available data, and the models almost always are oversimplified.

Drawing on examples from the Savoy Experimental Watershed and from around the world, case studies show that modeling is a valuable tool for hypothesis testing, but an inappropriate tool for prediction. Predictive models at site-specific scales give the false appearance of hydrogeologic understanding, and their application to karst aquifers is at best, misleading, and at worst, unethical.