Spectral Characterization of Nine Urban Tree Species in Southern Wisconsin

Urban trees provide essential environmental, health, social, and economic benefits. Consequently, researchers and stakeholders devote considerable effort to characterizing, mapping, and monitoring urban tree species. Traditional identification methods that rely on field surveys are labor-intensive and time-consuming. This study evaluated the potential of field hyperspectral spectroscopy to classify nine common urban tree species at the leaf level. Seven random forest classifiers, each using different combinations of spectral features, were compared for classification accuracy. The model that incorporated both first derivatives of spectral reflectance and vegetation indices achieved the highest overall accuracy (80.4%), whereas the model combining spectral reflectance and vegetation indices had the lowest predictive performance (70.1%). The most influential predictors were spectral bands and first derivatives in the red-edge and SWIR 1 regions; and the vegetation indices Red-edge Vegetation Stress Index (RVSI), Plant Senescence Reflectance Index (PSRI), and Blue Ratio (BR). These results support the use of hyperspectral remote sensing for identifying and classifying urban tree species.