Using High Resolution Remote Sensing Data To Characterize Urban Forest Tree Species Composition

California=s urban forests provide habitat for over 90% of the state=s population and their impact on human and environmental health is a significant concern. Urban forests contain a diverse mix of tree species (over 200 species in most cities) arranged in sometimes heterogeneous patterns. Characterizing structure and species composition is fundamental to estimating energy and water balances, BVOC emissions, and other impacts of urban forests on air quality. However, the costs associated with methods relying on aerial photography and field sampling have been high. New technologies provide the capability to acquire and process multispectral data at very high resolution. With these technologies it may be possible to provide more accurate estimates of tree cover and urban forest structure at lower cost. Also, they may make it possible to detect differences among certain types of species and their leaf areas. In this study, we use remote sensing data from the Airborne Visible InfraRed Imaging Spectrometer (AVIRIS) and Geographic Information System (GIS) techniques to identify tree species in the City of Modesto, California. The unmixing method was used for analysis of the high spectral remote sensing data. Index and edge ratio methods were used for determining tree crown size. The results show that we can accurately distinguish different types of trees in urban forests (e.g., deciduous, broadleaf evergreen, conifer). At the individual tree species level, 5 of 18 tree species were accurately classified greater than 80% of the time, and another 6 of 18 were accurately classified greater than 60% of the time. The procedures to conduct this analysis are not site dependent and consequently, can be applied to other locations. These study results can be directly used for forest management and urban ecosystem research, as well as for air quality, energy, and hydrology analyses.