AUTOMATED SINKHOLE MAPPING IN MIFFLIN COUNTY, PENNSYLVANIA: A TEST CASE FOR AN UPDATED, STATEWIDE KARST HAZARD-POTENTIAL MAP

Sinkhole mapping is vital for geohazard mitigation, groundwater protection, project planning and siting, and emergency management. Increased rainfall and storm intensity due to climate change, combined with development of communities in areas vulnerable to karst hazards, may result in more sinkhole incidents. Sinkholes pose a risk to human lives and infrastructure and are a conduit for groundwater contamination. Karst feature datasets and hazard maps are used by developers, planners, citizens, consultants, karst scientists, hydrologists, and the caving community. The current datasets and maps for Pennsylvania are based on visual examination of historical aerial photographs and reported sinkholes before 2000.

Mifflin County was selected as a test case area for automated sinkhole mapping. It is largely rural, reducing difficulties posed by urban areas, and is unglaciated, reducing challenges in identifying closed depressions in mantled karst. The county is underlain by Ordovician, Silurian, and Devonian carbonates, which are representative of many of the carbonates in Pennsylvania.

Lidar digital elevation models (DEM) with one-meter cells were modified according to steps outlined by Doctor and Wall (2018, 15th Sinkhole Conference, NCKRI, 54 p). Artificial dams created by roads were removed from the DEM. The modified DEM was used to identify closed depressions greater than nine square meters. The resulting 13,340 polygons were screened to remove false sinks including places the model failed to cross roads, alluvial features, quarries and mine collapses, ponds, and anthropogenic features. Field inspection checked questionable closed depressions such as solifluction lobes, vernal pools, and urban artifacts. The resulting 3,500 sinkholes were used to create a map of karst hazard potential. A comparison was made to the historical dataset to determine overlap and strengths of each method. The combined datasets will form a more accurate karst hazard-potential map and the continuing county assessments will build a better statewide map.