Evaluating the potential for site-specific modification of LiDAR DEM derivatives to improve environmental planning-scale wetland identification using Random Forest classification

摘要: Abstract Wetlands are important ecosystems that provide many ecological benefits, and their quality presence protected by federal regulations. These regulations require wetland delineations, which can be costly time-consuming to perform. Computer models assist in this process, but lack the accuracy necessary for environmental planning-scale identification. In study, potential improvement of identification through modification digital elevation model (DEM) derivatives, derived from high-resolution increasingly available light detection ranging (LiDAR) data, at a scale small-scale delineations is evaluated. A novel approach flow convergence modelling presented where Topographic Wetness Index (TWI), curvature, Cartographic Depth-to-Water index (DTW), modified better distinguish upland areas, combined with ancillary soil used Random Forest classification. This applied four study sites Virginia, implemented as an ArcGIS model. The resulted significant average compared commonly National Wetland Inventory (84.9% vs. 32.1%), expense moderately lower non-wetland (85.6% 98.0%) overall 92.0%). From this, we concluded modifying TWI, DTW provides more robust signatures improving rates classifications using original indices. resulting general tool able modify these local LiDAR DEM derivatives based on site characteristics identify wetlands high resolution.