Copula-based assessment of flood susceptibility in the island of Cyprus via stochastic multicriteria decision analysis

This study implements a probabilistic framework that a) involves the generation of realizations of the criteria weights by sampling methods that model the structural dependence of multicriteria correlated data, b) evaluates the similarity of the simulated data with respect to the original dataset, and c) uses these realizations to estimate the spatial distribution of the statistical metrics of flood susceptibility, and quantifies the uncertainties that are propagated from the simulated weights to the susceptibility levels. Seven criteria have been selected according to the literature, namely, terrain elevation, slope, flow accumulation, rainfall intensity, distance from the drainage network, land use/land cover and soil. The results reveal that copula-based samples exhibit a superior ability to capture the dependence among strongly correlated criteria, whereas the simulated data based on principal component analysis (PCA) achieve better agreement in terms of the standard deviation. Consequently, copula-based samples of weighting coefficients are used to assess flood susceptibility in Cyprus via multicriteria decision analysis (MCDA). The results revealed that 50 % of the study area is classified as either highly susceptible or very highly susceptible to flooding, with the majority of these regions being present in the southern and southeastern parts of the study area. With respect to the validation dataset, the known flood-prone areas are classified as having either “Very High” (12–14 %) or “High” (> 7880 %) probabilities of flood occurrence. The proposed probabilistic framework can be directly applied in other biogeographical regions and geological contexts, depending on the characteristics of the study area.