Spatial Uncertainty Propagation in Ecosystem Service Assessment Modelling under change of scale

Ecosystem Service models are increasingly getting attention in current and future natural resource management and evaluation. Due to combinations of different submodels from various disciplines they belong to the family of integrated environmental assessment models. These models are often complex modelling chains with submodels developed and operating at different scales. Uncertainties associated to model predictions due to change of scales or support play a major role when different scenarios are tested for decision and policy making. This paper studies the impacts of land use and land cover change under spatial uncertainty on ecosystem services such as pollination, focussing on urban expansion in a study area of Luxembourg. We use spatial stochastic simulation to model the spatial distribution of sensitive model parameters and propagate these with a Monte Carlo approach throughout the entire modelling chain across the different spatial scales. To account for change of support we use geostastical up- and downscaling methodologies, such as Area-to-point kriging, to account for the correct spatial support (scale) of the different submodels. Results show that the final ecosystem service predictions are represented more realistically and can be derived as probability distributions distributed in space. This helps decision makers to better analyse not only overall uncertainties associated to model outputs but also get a more precise picture of uncertainties distributed spatially.