AUTOMATED SINKHOLE MAPPING WITH ARCGIS USING SHAPE FACTORS IN HARRISONBURG, VA

Sinkholes are common features in karstic environments that pose threats to infrastructure and serve as conduits for potential groundwater pollution. Mapping sinkhole occurrence and potential formation is critical for susceptibility assessment. The presence of high-resolution digital elevation modems (DEM) allow simple visual identification of closed depressions that may be sinkholes. However, visually identifying sinkholes is ineffective and inefficient due to the large number of such features in karst areas. The objective of this study is to test the effectiveness of automated sinkhole mapping and develop models for further implementation. An approximately 15000 m x 15000 m study area encompassing Harrisonburg, VA was chosen for this study, with control sinkhole data sourced from VA Energy and 1 m LiDAR from the USGS. The high-resolution DEMs generated from LiDAR in ArcGIS make processing closed depressions in a landscape simple, but returns too much noise. Using a methodology adapted from Admassu & Woodruff (2021), DEMs were processed to isolate depression polygons and generate shape factors. To separate sinkhole depressions from artificial closed depressions, we used linear discriminant models based on shape factors such as circularity, sphericity, and curvature. The result showed that the models can identify sinkhole-caused depressions with accuracy as high as 71%. Maximum curvature was found to be the most efficient variable separating sinkholes from non-sinkholes. Additionally, all models identified sinkholes not originally mapped in the control. Future work should be concentrated on the potential difference in effectiveness in automated sinkhole mapping between urban, suburban, and rural landscapes.