Author: Andrew Johnston
Urban forests are a central element of the urban environment. Improved observations of historic tree cover dynamics are required to better understand how the urban environment changes though time. In this study, satellite remote sensing techniques were applied to observe past variability in tree cover area within the District of Columbia using highly calibrated Landsat data. Validation was performed with data from field surveys and public geospatial data on standing tree cover. Testing of alternate methodologies demonstrated that an approach utilizing support vector regression produced results with greater accuracy across the city when compared to linear spectral mixture analysis. Per-pixel uncertainty remained high using both techniques. Spectral mixture analysis overestimated tree cover in low population density areas and underestimated tree cover in the urban core, while support vector regression provided consistent accuracy across land use types. The consistent reliability allowed results from support vector regression to be used for observing tree cover changes between different land use zones. This will make it possible to identify past tree cover changes in low density residential zones within the District of Columbia. These results provide useful background information for maintenance and resource management as part of efforts to monitor and expand urban tree cover. Further development of these methods may enable their application with archival moderate resolution satellite remote sensing data for other study areas.
You are reviewing the
Download.
Your machine is locate at 3.145.63.131.