RANKING SINKHOLES FOR PROTECTION BY WATERSHED DELINEATION

Karst features were ranked for protection by extracting their watershed divide and dividing by the area covered by thin soils in order to index them as a source of groundwater pollution. The study was conducted on the Onondaga escarpment in western New York in an area that has suffered several well contamination events in the past. In this presentation, we discuss the performance of two automated watershed delineation algorithms for performing this analysis. The approaches were tested against the “true” watershed divide mapped in the field to determine if these automated approaches can be successfully employed in this landscape, which consists of a heterogenous mixture of thinly soiled karst, sinkholes, and glacial features such as end moraines, ground moraines, channels and other till deposits. The two automated approaches can be considered to be at opposite ends of the spectrum in their treatment of internally-drained and zero slope regions. The Jensen and Domingue algorithm removes depressions to insure flow continuity. Its catchment delineation is thus inclusive, however, it has been known to develop unrealistic watershed divides in areas of flat or low-sloped topography. The PCSA does not remove depressions and assumes that all breaks in flow are hydrologically realistic. Its delineation tends to be more conservative and excludes areas that are zero-sloped and are isolated by depressions. It allows multiple flow directions however, which can sometime create hydrologic continuity to upslope areas outside of the catchment. The results suggest that both approaches had issues which occasionally caused inaccurate watershed delineation. The Jensen and Domingue algorithm performed well in areas dominated by convergant flow, however it was difficult to apply to linear karst features where flow was down straight hillslopes. Both approaches had issues with road berms, ditches and culverts which led to inaccurate watershed delineation. The former features commonly reduced the effective size of the catchment area. Catchment / Effective areas indices ranged from 2 to 237 with smaller solution sinkholes having the highest indices. Despite their issues, both automated methods performed well in ranking the sites.