Spatial Clustering of Extreme Rainfall Events in Greater New York Area Using Weather Radar Data

Extreme rainfall events, specifically in urban areas, have dramatic impacts on society and can lead to loss of lives and properties. Despite these hazards, little is known about the city-scale variability of heavy rain events. In the current study, 13 years of gridded Stage IV radar data, 2002-2014, is employed to investigate the statistical properties of the spatiotemporal variability of simultaneous rainfall exceedances in Greater New York Area. The 95th percentile of each gridpoint's annual max intensity is considered as a threshold for storms. Then, multivariate k-means clustering is applied on extreme rainfall events' intensity and area exposure for each rainfall duration and season of occurrence. Comparison of timing indicates most of extreme rainfall events (more than 40%) are occurring in summer. Clustering analysis results show that for short rainfall duration, most of the study area is hit by high intensity-large area storm in warm seasons while in cold seasons rainfall intensity is low and the areal exposure is also low. In contrast, long rainfall duration follows an opposite spatiotemporal pattern. Resultant maps geo-reference the probability of occurrence of high-intensity large-area exposure storm over the study area. These maps can become inputs for design of hydraulic systems with the spatial and temporal resolution of 4km X 4km and 1-hour respectively which corresponds to the input radar data.