Leveraging Freely Available Remote Sensing and Ancillary Datasets for Semi-Automated Identification of Potential Wetland Areas Using a Geographic Information System (GIS)

Abstract

The purpose of this study was to develop a wetland identification tool that makes use of freely available geospatial datasets to identify potential wetland locations at a spatial scale relevant for transportation corridor assessments. The tool was developed to assist the Virginia Department of Transportation in wetland identification over large geographic regions. Wetland identification is an integral part of many construction projects performed by state departments of transportation. However, current methods for wetland identification in support of these activities are lacking in one or more of the following ways: inadequate use of ancillary data, little automation, failure to leverage freely available data, excessive computation times, high expense, or the requiring of software not typically available to state departments of transportation. This study addressed these limitations through development of a geographic information system (GIS)-based wetland screening tool with freely available data and automated geoprocessing workflows to assist in wetland identification over large geographic regions. The tool was designed as a screening tool able to identify potential wetland areas that would require further investigation by a trained wetland identification expert. Therefore, the tool was designed to minimize false negatives: cases where the tool incorrectly designates wetland as non-wetland. Application of the tool to a study region with detailed wetland delineations showed that the tool correctly identified wetlands nearly 70% of the time, produced false positives 24% of the time, and produced false negatives only 6% of the time. The tool allows decision makers to adjust the sensitivity of the wetland identification algorithm in order to decrease false negatives at the expense of increasing the fraction of the study area identified as potential wetland. The tool, therefore, allows decision makers to balance trade-offs between the amount of area requiring more detailed wetland identification and the frequency with which wetland areas are misidentified by the screening tool as false negatives. Although the wetland identification tool was shown to be effective, future studies will be required to calibrate and validate the tool further using a broader range of application areas. The study recommends that this be done by way of additional corridor analyses to facilitate further improvements to the tool.