Accounting for Spatial Dependencies in Flood Hazard Assessment

Flood hazard assessments commonly employ either scenario-based or event-based approaches, assuming uniform return periods for peak flows throughout the entire watershed. However, due to the spatial and temporal heterogeneity inherent in flood events, relying on uniform scenarios can yield inaccurate outcomes. While employing uniform return periods may be advantageous at the local level, it proves to be inadequate for large-scale assessments. For example, assuming a 100-year return period for the entire river network is unrealistic, given the rarity of a singular event of such magnitude occurring across the entire watershed. This approach may lead to an overestimation of flood risk, as it fails to accurately capture the intricacies of floods on a broader scale.Regardless of the chosen methodology to account for these dependencies and variations, synthesizing flood hazard emerges as a primary need to provide decision-makers with adequate knowledge. When no uniform return period is used, this synthesis may be realized using probabilistic maps, which, unlike deterministic approaches, may represent the inherent complexity of the considered hydrogeological system. This would enable the stochastic nature of flood hazard to be captured, ensuring a more reliable analysis.Several methods are available for generating flood events accounting for the spatial dependence and the variability of occurrence probability at the watershed scale. A comprehensive review of existing studies that have proposed alternative methods for generating flood events, considering the spatial dependence, will be presented. Noteworthy methodologies include approaches based on the conditional exceedance model by Heffernan and Tawn, the application of a multivariate multisite weather generator coupled with continuous hydrological simulation, the use of simplified stochastic rainfall generators, and models grounded on copulas.Furthermore, a sample application is presented of inferring a dependency structure for an area northern to Milan, considering spatial dependencies for contributing sub-basins, and prospects for stochastic hazard mapping are discussed.