Differential ability of AMS and PDS to detect trends in occurrence of heavy rainfall events - a rainfall simulator approach

Understanding of regional distribution of extreme precipitation events as a response to global warming scenario is a subject of active research as suggested in IPCC-4. A number of authors have attempted to establish trends in these extreme events. It, however, has been recognized that the perceived trends may vary depending on the method chosen for characterizing the frequency of these events. In this work we comparatively examine the ability of Annual Maximum Series (AMS) and Partial Duration Series (PDS) in resolving trends in the heavy rainfall events via changes in parameter values in the Generalized Extreme Value (GEV) distribution. A spatial-temporal rainfall simulator was deployed to generate daily rainfall amounts over the state of North Carolina. Then the observation processes at COOP stations were simulated with randomized missing periods assigned. The artificial station observations were used to construct AMS and PDS whereby the GEV distribution was calculated for each decade and for each station. The efficacy of these two approaches is assessed and discussed.