Mapping functional urban green types using hyperspectral remote sensing

The quantification of the ecosystem services provided by urban green is a challenging task, as the specific nature and magnitude of these services highly depend on the type, properties and context of the urban green element under consideration. Past remote sensing based mapping approaches of urban green have mainly focused on broad vegetation classes (tree - scrub - grass) or species identification, failing to deliver the detailed functional information required by urban ecologists, planners and managers. The end goal of our research is to develop a detailed functional urban green typology in combination with a multi-stage classification algorithm based on airborne hyperspectral data and LiDAR. In this paper, we conduct a spectral separability analysis and subsequent sub-pixel classification of common urban green types in Brussels, Belgium, to explore the potential of hyperspectral data for our purposes. Except for extensive green roofs and, to a lesser extent, coniferous trees, the distinction between different urban green types based solely on the amplitude and shape of the spectral signal proved to be difficult. In future research, we will therefore explore additional classification strategies, including spectral feature selection, integration with height information from LiDAR and object-based classification.status: publishe