An Integrated Spectral–Structural Workflow for Invasive Vegetation Mapping in an Arid Region Using Drones

Mapping invasive vegetation species in arid regions is a critical task for managing water resources and understanding threats to ecosystem services. Traditional remote sensing platforms, such as Landsat MODIS, are ill-suited distinguishing native non-native due their large pixels compared plant sizes. Unmanned aircraft systems, or UAS, offer the potential capture high spatial resolution imagery needed differentiate species. However, order extract most benefits from these there need develop more efficient effective workflows. This paper presents an integrated spectral–structural workflow classifying Lower Salt River region of Arizona, which has been site fires flooding, leading proliferation Visible (RGB) multispectral images were captured processed following typical structure motion workflow, derived datasets used inputs two machine learning classifications—one incorporating only spectral information one utilizing both data structural layers (e.g., digital terrain model (DTM) canopy height (CHM)). Results show that including classification improved overall accuracy 80% 93% spectral-only model. The important features CHM DTM, with blue band indices (normalized difference index (NDWI) normalized salinity (NDSI)) contributing models.