Testing the applicability of ecosystem services mapping methods for peri-urban contexts: A case study for Paris

Through their semi-natural and agricultural areas, peri-urban regions are pivotal in providing ecosystem services (ES) to city dwellers. To quantify the ES provided by these areas, it is possible to use ES mapping methods: many ES mapping methods rely on land cover maps, but most maps are coarse compared to the peri-urban scale. Nevertheless, readily-available land use data and methods are often used to map ES at such scales, without contextualisation. As a result, such methods may not be able to capture the diversity that is present in the peri-urban vegetation, which could have consequences for their accuracy and furthermore for urban planning policies. To increase our understanding of the applicability of ES mapping methods in peri-urban regions, we assessed to what degree sites with similar plant composition in the green belt of Paris, France, were also projected to have similar ES bundles. We considered two commonly used ES model types: proxy-based models (here: look-up tables) and phenomenological models. We used 252 sites for which botanical survey data were available and applied the ES models to seven ES relevant in the peri-urban context. A cluster analysis was used to group sites, hence facilitating analyse of the spatial congruence between types of vegetation and bundles of ES. Clustering sites based on plant composition revealed six distinct clusters. Clustering sites based on ES bundles as estimated by phenomenological models and proxy-based models, resulted in four and two clusters, respectively. The proxy-based clustering only highlighted broad-leaved forests as an important ES supply source. The phenomenological model estimates of ES allowed a more nuanced clustering of sites into four different groups. The level of congruence between the different sets of clusters based on plant composition and estimated ES bundles was low. Except for forests, the commonly used ES models tested here were not able to represent the same level of heterogeneity in the peri-urban landscape as was found in the vegetation. Our results demonstrate the need to integrate finer scale approaches and primary data in ES assessments of peri-urban areas