Using a high-density rain gauge network to estimate extreme rainfall frequencies in Minnesota

Abstract

This paper estimates rainfall frequencies over a portion of Minnesota’s high-density rain gauge network. Daily rainfall data from a grid of 10 km by 10 km cells centered on Minneapolis, MN are used to explore the spatial component of extreme rainfall frequencies. A suite of statistical and graphical techniques, performed on each cell’s time series of annual maximum daily precipitation for the period 1958–2002, revealed that the mean annual maxima were highly dependent on the number of observers within each cell. Considerable spatial variability in return-period estimates also resulted from the observer number disparities, as well as the fact that a small proportion of the cells experienced outstandingly large rainfalls. The estimates for 24-h, 100-yr rainfall in this study are larger—often by a factor of two-- than “official” published values. The highly urbanized study area may, therefore, be more susceptible to flooding and damage than its designers had anticipated. Many of the differences between these estimates and published values result from using multiple points to construct a synthetic, single-point time series for each cell; earlier studies have constructed each series from singular fixed points. While traditional point-based approaches to design values may be valid in some instances, it is argued that extreme rainfall is an inherently areal phenomenon, and that a small-area-based approach may be more appropriate when designing for features and processes that area also areal in nature.