A stochastic framework for rainfall intensity–timescale–return period relationships. Part ΙΙ: Point modelling and regionalization over Greece

Abstract

In this work, we formulate a regionalization framework for rainfall intensity–timescale–return period relationships which is applied over the Greek territory. The methodology for single-site estimation is based on a stochastic framework for multi-scale modelling of rainfall intensity which is outlined in the companion paper. Five parameters are first fitted independently for each site and the resulting parameter variability is assessed. Following a systematic investigation of uncertainty and variability patterns, two parameters, i.e. the tail-index and a timescale parameter, are identified as constant in space and estimated using data pooling techniques. The other three parameters are regionalized over Greece by means of spatial interpolation and smoothing techniques that are assessed through cross-validation in a multi-model framework. The regionalization scheme is implemented in a sequential order that allows exploiting rainfall information both from rainfall stations with sub-daily resolution as well as from the higher reliability network of daily raingauges.