Stochastic modelling of high-resolution rainfall time series

Two approaches to generate artificial high-resolution rain series for use as input to simulation of urban drainage systems have been tested, both based on waiting times between consecutive tips of tipping bucket gauges calibrated to sample rain in a 0.2 mm depth resolution. ARIMA-models give a reasonable description of data but they have found limited practical use due to difficulties with identification, estimation and simulation of individual extreme rain events. Markov chain models including a state variable representing accumulated rain depth are able to extract the statistical properties of the data series and may be used to generate artificial rain series that resemble the original data structure. The perspective is to couple a stochastic time series model with a regional model for extreme point rainfall in order to make inference about extreme rainfall at ungauged locations.