CR 63:31-41 (2015)  -  DOI: https://doi.org/10.3354/cr01280

ABSTRACT: For daily precipitation, previous studies have mainly identified extremes over fixed durations. The goal of our study was to identify over which multiday period a rainfall event can best be described as an extreme. Specifically, we extrapolated the starting time and duration within a rainfall episode to best describe an event with that starting time and duration as an extreme, compared to events with other starting times and durations. The principle is that the precipitation intensity averaged over this period is comparably (relative to duration) the strongest among all the events that have different starting times and durations. For this purpose, the ‘extreme’ intensity-duration (EID) relation is established through mathematical modeling based upon our understanding of the issue of identifying the extremes. The constraints in the model between ‘extreme’ intensity and duration require that the single parameter contained in the EID relation be between, but not too close to, 0 and 1. Tests show that extremes can be well identified with the EID approach by simply assigning a moderate value to the parameter, and the identification of extremes is not sensitive to which value is chosen for this parameter. The estimation with multiyear data and a regression indicates that the parameter is truly moderate between 0 and 1, but the value relies on the threshold used for determining the initial extreme intensities. It is therefore suggested that a fixed moderate value be given to the parameter in the operational identification of the extremes. As a real application of the method, we determined the starting time and duration of the extreme in a recent heavy rainfall that occurred over Beijing.

KEY WORDS: Climate extremes · Identifying extremes · Monitoring and detections · EID constraints · Mathematical modeling