Simple scaling of extreme precipitation in North America

Abstract

Extreme precipitation is highly variable in space and time. It is therefore important to characterize precipitation intensity distributions on several temporal and spatial scales. This is a key issue in infrastructure design and risk analysis, for which intensity–duration–frequency (IDF) curves are the standard tools used for describing the relationships among extreme rainfall intensities, their frequencies, and their durations. Simple scaling (SS) models, characterizing the relationships among extreme probability distributions at several durations, represent a powerful means for improving IDF estimates. This study tested SS models for approximately 2700 stations in North America. Annual maximum series (AMS) over various duration intervals from 15 min to 7 days were considered. The range of validity, magnitude, and spatial variability of the estimated scaling exponents were investigated. Results provide additional guidance for the influence of both local geographical characteristics, such as topography, and regional climatic features on precipitation scaling. Generalized extreme-value (GEV) distributions based on SS models were also examined. Results demonstrate an improvement of GEV parameter estimates, especially for the shape parameter, when data from different durations were pooled under the SS hypothesis.