Characterizing Garden Greenspace in a Medieval European City: Added Values of Spatial Resolution and Multi-Temporal Stereo Imagery

Abstract

Domestic gardens provide residents with immediate access to landscape amenities and numerous ecological provisions. These ecological provisions have been proven to be largely determined by greenspace composition and landscape, but the fragmentation and heterogeneity of garden environments present challenges to greenspace mapping. Here, we first developed a recognition method to create a garden parcel data set in the medieval Leuven city of Belgium, based on the land use layers and agricultural land parcels. Then, we applied multi-sourced satellite imagery to evaluate the added value of spatial resolution, plant phenology and 3D structure in identifying four vegetation types. Finally, we characterized the greenspace landscapes in garden parcels. Compared with single ALOS-2 imagery, SPOT-7 imagery and Pleiades-1A imagery increased the overall accuracy by 4% and 8%, respectively. The accuracy improvement (21%) produced from multi-temporal stereo Pleiades-1A imagery strongly verified the significance of plant phenology and 3D structure in garden mapping. The average greenspace cover in garden parcels was 71% but varied from 56% in urban gardens to 82% in rural gardens. The garden greenspace landscape is fragmented by the artificial structures in urban areas but has a more aggregated size and less complex shapes in rural areas. This study calls for greater attention to be paid to gardens, and for multi-disciplinary studies conducted in collaboration with urban ecologists and landscape designers to maximize the benefits to residents of both immediate landscape amenities and ecological provisions, in the face of global environmental changes and public health risks.