GC51A-0377:
Detection of Nonstationarity in Seasonality of Extreme Precipitation Using a New Statistical Approach

Friday, 19 December 2014
Nirajan Dhakal and Shaleen Jain, University of Maine, Orono, ME, United States
Abstract:
Changes in seasonality of extreme precipitation have important implications for public safety, stormwater infrastructure and, in general, adaptation strategies in a changing climate. In this context, an understanding of shifts in the extreme event seasons—emerging, weakening or intensification within seasonal windows is an important step. In this study, we applied a nonparametric circular method to assess the temporal changes in the seasonality of the extreme precipitation for 10 USHCN stations across the state of Maine. Two 30-year blocks (1951-1980 and 1981-2010) of daily annual maximum precipitation were used for the analysis. Extreme precipitation dates were used to compute the circular probability distribution. Important information regarding the multimodality in the seasonal distribution of extreme precipitation dates were obtained from the probabilistic assessment of seasonality using the kernel circular density estimates. Nonstationarity in seasonality was observed for most of the stations; some stations exhibit shifting of significant mode towards Spring season for the recent time period while some stations exhibit multimodality for both the time periods. Despite the limitation of being sensitive to the smoothing parameter, the kernel circular density estimates method is clearly superior and robust when dealing with diverse seasonal pattern of extreme rainfall comprising of multiple seasonal modes.