Abstract
Urban green spaces (UGS) provide important ecological, social, and climatic functions, but their effectiveness depends not only on their extent but also on their vertical vegetation structure. To this end, we present a high-resolution urban green space typology (UGST) for four Swiss cities (Basel, Bern, Geneva, and Zurich) derived from remotely sensed thematic (NDVI) and structural (vegetation height and VCI) data, comprising 33 classes. The structural heterogeneity of the vegetation cover was consistently bimodal, with low-complexity vegetation (grass) or tall, high-complexity wooded areas dominating, while intermediate structures were underrepresented. The UGST was used to assess species richness across five taxonomic groups and a multidiversity score (mean normalized richness across all groups) using GAMs under varying thematic resolutions. Results indicated taxon-specific, generally non-linear associations between normalized recorded species richness or multidiversity and the individual component scores or percentage cover of several UGST classes. The fine-grained joint models had higher in-sample deviance and explained more variance compared to highly aggregated-class and total-green-cover formulations, although the formulations differed in complexity and were not evaluated out of sample.