Precipitation Extremes: Linking Consistent Intensity-Duration-Frequency Relation to Large Scale Atmospheric Flow (IDF-AF)
Precipitation events can have very different characteristics, from long-lasting light drizzle to short but intense precipitation. For every precipitation event duration, extremes can be defined using extreme value statistics, resulting in intensity-duration-frequency (IDF) relationships. A popular approach to IDF curves is the estimation of extreme value models for individual durations separately and subsequently and, in a second separate step, modelling the duration dependence for fixed quantiles (or return-levels) individually. This leads necessarily to inconsistencies such as a crossing of IDF curves associated to different quantiles (return levels): i.e. a 0.9-quantile being larger than a 0.95-quantile. We use a single consistent statistical model based on a duration-dependent Generalized Extreme Value Distribution (d-GEV) to estimate the IDF relationships for all durations simultaneously (Koutsoyiannis, 1998) and compare different approaches to model the duration dependency of the GEV. We aim for a meaningful estimation of the IDF relationships across all relevant event durations and for ungauged sites through spatial covariates in a spatial model. To further investigate the parameter range, a Bayesian Hierarchical Model (BHM) will be set up, including a Gaussian Process for spatial modelling. Conditional on large-scale atmospheric flow indices, IDF relationships in a changing climate shall be obtained locally (i.e. at the gauge level) and results will be presented on an interactive web-interface for end-user's convenience.