Uncertainty propagation between web services - A case study using the eHabitat WPS to identify unique ecosystems

Schulz M., Skøien J., Gerharz L., Dubois G., Pebesma E.

Zusammenfassung

Protected areas (PAs) are designed to protect ecosystems and their associated species against anthropogenic threats. When assessing the importance of the current network of PAs, and when considering new areas which should be protected, one of the criteria is the uniqueness of the ecosystems found inside the existing or planned PA when compared to other parks. As a helping tool for park managers and potential funders, eHabitat has been designed using the Open Geospatial Consortiums (OGC) Web Processing Service (WPS) interface specification. It allows end-users to compute, using different data and models, the likelihood of finding ecosystems with similar properties, as well as the potential changes these areas are exposed to according to different climate change scenarios. The most important input parameters, typically thematic geospatial "indicator layers" characterizing the ecosystem, are provided to the WPS as references using standardised web services or catalogues. This allows for a virtually infinite number of combinations to describe these ecosystems. However, the layers used can range from geophysical data captured through remote sensing to socio-economical indicators. eHabitat is therefore exposed to a broad range of different types and levels of uncertainties. Assessing these uncertainties, and as an additional component further propagating them when potentially included in a chain of model services, is a key aspect in the context of the Model Web. The use of the Uncertainty Markup Language (UncertML) as developed within the UncertWeb project to promote interoperability between data and models with quantified uncertainty and different approaches for encoding and visualising uncertainty information will be presented. Retrieving feedback of intermediate processing results like input layer histograms or variability using supplied uncertainty information will be discussed.