EVALUATING USER INTERPRETATION ASSOCIATED WITH MANUAL DIGITIZATION OF STONE WALLS AND RELICT CHARCOAL HEARTHS USING AIRBORNE LIDAR

Light Detection and Ranging (LiDAR) can be used in the northeast United States as a tool to uncover archeological features hidden beneath forest canopy. Stone walls and relict charcoal hearths (RCHs) have been the focus of ongoing research to map and document the legacy of 17th-19th century agricultural practices in the region. Stone walls and RCHs stand out clearly in derivative LiDAR digital elevation model (DEM) products such as slope and hillshade maps. To date, mapping has been mainly carried out by on-screen digitization, but the sensory gap associated with such mapping, particularly as it relates to user interpretation, has not been evaluated. Furthermore, as large regions have yet to be analyzed and manual digitizing being a time-consuming task, regional datasets are compiled from multiple users of varying expertise. One important issue associated with digitizing these relict features in LiDAR is that many natural and anthropogenic structures can be misinterpreted as stone walls or RCHs. This study examines user interpretation associated with digitizing stone walls and RCHs using airborne LiDAR for several distinct landscape types in southern New England. Results show that factors such as LiDAR quality and topographic roughness, due to natural and anthropogenic features, can have an impact on stone wall and RCH mapping using LiDAR products. Terrain with a high degree of local (1-10 m) variability (from low LIDAR point spacing or significant anthropogenic modification) proves challenging for GIS mappers when digitizing because the features of interest are small, with relevant dimensions in the 1-15 m range (height, width, length, diameter, etc.). The stone wall study indicates that individual mappers varied, but agreed with the consensus 70-94% of the time across five test areas, with an average of 82%. For the RCH study, mappers agreed with the consensus 50-97% of the time across seven test areas, with an average of 80%. Overall, stone walls were found to be the more difficult feature to map reliably - all digitizers in the study mapped the exact same feature only 59% of the time in ideal terrain, while in ideal terrain for RCHs, all digitizers in the group mapped the exact same feature 94% of the time.