Weight of evidence and artificial neural networks for potential groundwater spring mapping: an application to the Mt. Modino area (Northern Apennines, Italy)

Locations of potential groundwater springs were mapped in an area of 68 km(2) in the Northern Apennines of Italy based on Weight of Evidence (WofE) and Radial Basis Function Link Net (RBFLN). A map of more than 200 springs and maps of five causal factors were uploaded to ArcGIS with Spatial Data Modelling extensions. The WofE and RBFLN potential groundwater spring maps had similar prediction rates, allowing about 50% of the training and validation springs to be predicted in about 15 to 20% of the study area. The two maps were merged using a heuristic combination matrix in order to produce two hybrid maps: one representing susceptible areas in both the WofE and RBFLN maps (type A), while the other representing susceptible areas at least in one of the two maps (type B). For small cumulated areas, the success rate of both hybrid maps was higher than that of the parent maps, while for large cumulated areas, only the type B hybrid map performed similarly to the parent maps. This conclusion suggests different applications of these maps to water management purposes. (C) 2009 Elsevier B.V. All rights reserved.