Spatially-explicit integrated uncertainty and sensitivity analysis of criteria weights in multicriteria land suitability evaluation

We employ spatially-explicit uncertainty and sensitivity analysis to examine the robustness of land suitability evaluation. We use Monte Carlo simulation to sweep through criteria weight space, where weights are expressed using probability distributions. Multiple output suitability maps are generated and summarized using: an average suitability map, a standard deviation uncertainty map, and a number of sensitivity maps. We demonstrate how these surfaces help detect critical regions of suitability on the example of habitat suitability evaluation for a wetland plant. Areas of high average suitability and low uncertainty signify robust suitability sites, whereas high average suitability and high uncertainty characterize candidate areas. These candidate areas are potentially suitable but need further examination with variance-based sensitivity analysis, in which the variability of land suitability is decomposed and attributed to individual criteria weights. The resulting sensitivity maps delineate regions of weight dominance, where a particular weight greatly influences the uncertainty of suitability scores.