Copula-based Statistical Modelling of Compound Flooding and the Effect of Climate Variability: A Case Study in the Netherlands

Abstract:

The complex interplay between (not necessarily extreme) precipitation and storm surge can drive extreme flooding events in coastal areas, threatening lives, assets, and the environment. In this study, we examine the suitability of copula-based models to replicate the physical interaction between surge and precipitation resulting in extreme inland water level at a gauge near Lauwersmeer (the Netherlands) observed in van den Hurk et al. (2015). A novel and challenging aspect lies in the high degree of human management that strongly affects the dynamics of the hydrological system, which needs to be translated into a realistic “impact function” that successfully describes the relationship between the impact (inland water level) and driving variables. We produce an impact-based multivariate statistical framework optimally designed to reproduce inland water level extremes. Upon event sampling, an impact function relating precipitation and surge processes was created by utilizing a bin-sampling approach that gives higher weights to the most extreme events. Copulas preserve the dependence structure of flooding drivers, simulating paired precipitation and surge synthetic events that are transformed to inland water levels via the impact function. The proposed statistical framework reproduces extreme inland water levels accurately, even when reducing the complexity of the dependence model. Moreover, we present an uncertainty assessment of the statistical framework accounting for the internal climate variability by assessing the impact of using a subset of 50 years of data (similar to what is typically available from observations) from the 800-year dataset. Most of the uncertainty is produced by the marginal analysis, whereas training the impact function with shorter records leads to an underestimation of the return levels. Results demonstrate that relatively short records are insufficient for a robust estimation of extreme water levels resulting from the interaction of both precipitation and surge, although this conclusion may be highly site-specific and dependent on the system management.

van den Hurk, B., van Meijgaard, E., de Valk, P., van Heeringen, K. J., & Gooijer, J. (2015). Analysis of a compounding surge and precipitation event in the Netherlands. Environmental Research Letters, 10(3), 035001.