Return period of high-dimensional compound events. Part II: Analysis of spatially-variable precipitation

Abstract. This study introduces a comprehensive framework for modeling compound precipitation events, with a focus on handling zero intermittency in rainfall data. By expanding the existing methodologies to a five-dimensional approach and applying the joint return period (JRP) concept using both Gaussian copulas and R-vines with Gaussian, extreme value, and t-Student copulas, we offer a more accurate understanding of these complex events. A key contribution of this study is the proposal of a model that calculates the multivariate return period in five dimensions, surpassing the commonly used bivariate approach, and considers the dependence of precipitation events across multiple sites, accounting for both lower and upper tail dependencies. The comparison of dependency structures in the generated samples shows that the R-vine structure with extreme value copulas in a multivariate mixed model is particularly effective at capturing the spatial dependence in the data. Our findings emphasize that an inappropriate choice of copulas can lead to either overestimation or underestimation of design events with defined return periods, with significant implications for risk management.