Detecting patches of invasive shrubs using high-density airborne LiDAR data and spectral imagery

Abstract

Urban forests are increasingly threatened by invasive shrub species in the understory. Accurate mapping of these species is crucial for effective management of their prevention and mitigation. This study focuses on the detection of invasive shrub species in forested areas of the Chicago metropolitan region using high-density airborne LiDAR data and high-resolution imagery. We analyzed the structural and spectral characteristics of forests with and without invasive shrub species. We tested whether forests with invasive shrub species have a simpler canopy structure than those not invaded, and whether high-density LiDAR based structural metrics are more effective in detecting the presence of invasive shrub species than spectral-based NDVI. Our binomial logistic model demonstrated a test accuracy rate of 93 % for detecting invasive shrub species. Our findings show that forest patches invaded by these species exhibit higher vegetation area density, lower height, and lower NDVI (full leaf on season) values than non-invaded patches. LiDAR-derived canopy structural metrics are shown to be more effective than NDVI for detecting invasive shrub species. To ensure the reliability of our predicted map, we tested correspondence between our invasion estimation map with the ground identified samples that were not used in model testing and training. These validation results demonstrated the predictive ability of our model, showing accuracy rates of 83 % and 77 % for tree survey (DBH > 5 cm) and shrub survey (Height > 1 m & DBH < 5 cm), respectively. Our study demonstrates that LiDAR-derived metrics along with spectral imagery can successfully map invasive shrubs in urban forests.