Using dimension reduction PCA to identify ecosystem service bundles

The concept of ecosystem services (ES) has facilitated the identification, mapping and communication about the many non-marketable benefits of green infrastructure. These benefits are important to consider during a spatial planning process. For spatial prioritisation of sites, with a high societal importance, there is need to filter this information to insightful spatial indicators. The mapping of ES-hotspots and identification of ES-bundles have been put forward as promising methods for spatial prioritisation and the assessment of multifunctionality. While "hotspot mapping" and "ES-bundles" speak to the imagination of many, it is open to many different interpretations. In addition, there is a risk that the commonly applied hotspot mapping of single services and subsequent overlay analysis does not capture true hotspots of multifunctionality, where we expect multiple services to co-occur, but at lower intensities. Therefore, hotspot mapping should be applied on ES-bundles, rather than single ES. Yet, there are few methods to objectively identify and map such bundles of co-occurring services. In this research we propose dimension reduction principal component analysis (PCA), as a solution to identify and map bundles of ES. This technique is an established technique in remote sensing, where it is used to reduce unnecessary clutter in a data set. This research shows that if the methods for quantification and mapping of ES are sufficiently independent and biophysically sound, the PCA method can reveal multifunctionality between services and lead to (new) insights that can be used for better informed decisions on management and planning. The PCA graphs, ES-bundle maps and the integrated RGB-visualisation are objective and factual outputs of a statistical analysis that can be used for communication and discussion with stakeholders. It gives insight in co-occurrence of services and challenges to look for answers to why things are the way they are. Although scale effects did not play an important role in the results of this study, we advise to use this method on relatively small scales and repeat analysis rather than generalizing large scale results to the local scale or transfer findings between study sites as land-use patterns (and its interplay with abiotic conditions) are the result of many different socio-ecological developments throughout history, which can obviously differ from region to region.