How might the occurrence frequency and intensity of extreme rainfall events in Ireland change in projected future climates?

Observed data of extreme rainfall occurrence in Ireland are available from approx. 30 stations over periods ranging from about 30 to 70 years.  They are provided as tables of occurrence dates and magnitudes of 9 different metrics, from “r4in15” (4 mm in 15 minutes) to “r30in24h” (30 mm in 24 hours).  These events are too few and too sparse to be treated with non-parametric statistics.  However, they are all “peak over threshold” (POT) events, and so the well-developed theories of POT statistics applies to them.  In particular, the number of such events in any given time period follow a Poisson distribution (which needs only one parameter, the mean), while the exceedances (i.e., magnitudes above threshold) follow a generalized Pareto distribution (which formally needs 3 parameters, but in practise only 2 need to be determined). 

As climate is projected to change in the future, our basic assumption is that the number of POT events (Poisson distribution) will change, while the exceedances (Pareto distributions) will remain the same, only represented by different numbers of samples.  Since the Poisson distribution is determined by just one parameter (the mean), the only quantity needed to fully characterize the nature of future intense rainfall events are projections of the mean of such events. 

Meanwhile, future climate projections (e.g., from CORDEX) are typically available on daily timescales, so the problem reduces to finding a robust relationship, whether physical  (based e.g., on a Clausius-Clapeyron scaling) or statistical (based, e.g., on the tails of frequency distributions)  between each of the 9 intense rainfall metrics and the daily values of standard surface variables under future climate projections.  Different options are explored, and results are shown based on future projections from the TRANSLATE project, as functions of different future emission scenarios. Typically, large increases in the frequency of all 9 metrics are projected regardless of the exact relationship assumed between the intense rainfall events and the frequency distributions of daily precipitation (or temperature).  Results are best shown as frequency distributions rather than single number percentage changes.