Quantifying uncertainties in earth observation-based ecosystem service assessments

Abstract

Ecosystem service (ES) assessments are widely promoted as a tool to support decision-makers in ecosystem management, and the mapping of ES is increasingly supported by the spatial data on ecosystem properties provided by Earth Observation (EO). However, ES assessments are often associated with high levels of uncertainty, which affects their credibility. We demonstrate how different types of information on ES (including EO data, process models, and expert knowledge) can be integrated in a Bayesian Network, where the associated uncertainties are quantified. The probabilistic approach is used to map the provision and demand of avalanche protection, an important regulating service in mountain regions, and to identify the key sources of uncertainty. The model outputs show high uncertainties, mainly due to uncertainties in process modelling. Our results demonstrate that the potential of EO to improve the accuracy of ES assessments cannot be fully utilized without an improved understanding of ecosystem processes.