Application of object-oriented image classification to identify potential brownfield sites in Syracuse, New York

The United States Environmental Protection Agency estimates that there are between five hundred thousand and one million brownfield sites in the United States. Brownfields are underutilized properties, the redevelopment or reuse of which is complicated by the presence or potential presence of contamination. Brownfield redevelopment efforts are expected to expand as local governments take advantage of recent brownfield legislation to revitalize communities and enhance local economies. This work applies an object-oriented classification method to high spatial resolution aerial imagery of the city of Syracuse, NY to identify potential brownfield sites during the inventory stage of the redevelopment process. The results were most reliable among larger brownfield sites where abandonment was evident and thus more readily detected by the classification. Differences in the land cover structure within certain property classes also affected the classification. Though commission errors were common, from an inventory perspective, this outcome is preferable to having a high omission error rate that would preclude many sites from further investigation into redevelopment options. The results are a promising step forward for research into the automation of brownfield identification using remote sensing.; Keywords. remote sensing, object-oriented classification, brownfields, fuzzy rules, underutilized parcels...