Automatic detection of urban vacant land: An open-source approach for sustainable cities

Abstract

With urban vacant land including brownfield land attracting increasing attention; spatial information on vacant land is crucial for city planners and decision-makers. Yet vacant land mapping faces the challenges of confusing terminology, reliance on inefficient registration systems, and the cost of data collection. Automatic site detection has been attempted using classification of remote sensing images. However, because the morphology of vacant land can cover such diverse aspects as derelict structures, bare soil, and vegetation or a mix thereof, even when commercial high-resolution images are used, it remains difficult to achieve this goal. This study starts by defining an urban vacant land typology suitable for automatic site detection. It covers four categories: transportation-associated land; natural sites with unfavorable conditions such as wetlands, steep slopes, or riverbanks; unattended areas or reserve parcels as ``leftover spaces'' within the urban fabric; and brownfield sites previously used but now abandoned. The study goes on to describe the rule-based data fusion method applied for site detection. The fusion procedure combines remote sensing images, GIS layers and citizen science data. For each type of vacant land, a separate data processing flow was developed. The method was tested in 63 urban and rural districts in Germany, identifying a large number of vacant sites. The study used an open-source approach. Open-source spatial data was collected for analysis, aerial photos provided by open-data projects were used for quality checks, and free software was used for data processing. This means that, information on vacant land is retrievable by local administrations irrespective of their financial circumstances. Depending on local urban development strategies, the detected sites can be used to support different sustainability goals.