A Unified Framework for Extreme Subdaily Precipitation Frequency Analyses Based on Ordinary Events

Abstract

The metastatistical extreme value approach proved promising in the frequency analysis of daily precipitation from ordinary events, outperforming traditional methods based on sampled extremes. However, subdaily applications are currently restrained by two knowledge gaps: It is not known if ordinary events can be consistently examined over durations, and it is not clear to what extent their entire distributions represent extremes. We propose here a unified definition of ordinary events across durations and suggest the simplified metastatistical extreme value formulation for dealing with extremes emerging from the tail, rather than the entire distributions, of ordinary events. This unified framework provides robust estimates of extreme quantiles (<10% error on the 100 yr from a 26 yr long record) and allows representations in which ordinary and extreme events share the scaling exponent. Future applications could improve our knowledge of subdaily extreme precipitation and help investigate the impact of local factors and climatic forcing on their frequency. Plain Language Summary We propose here a unified methodology to quantify the intensity of extreme rainfall of short duration, such as events expected to occur on average once every 100 yr. As opposed to alternative methods in literature, we rely on the simultaneous analysis of all everyday rainfall events, which, being much larger in number than extremes, were shown to provide improved estimates for daily rainfall. We show that, under our approach, the hypothesis of everyday and extreme events being similar enough holds also for short-duration rainfall. Application of our method to 26 yr of data from an individual station reproduces analyses based on more than 150 yr of observations from multiple nearby stations, with less than 10% error on the estimation of rain intensities expected to occur on average once every 100 yr, which are not directly quantifiable from the 26 yr of observations. The proposed methodology could help improve our knowledge of short-duration rainfall extremes, with implications for water resources and risk management, and could help investigate the impact of climate change on extreme rainfall events. et al., 2016). The framework can include any class of distributions for F and allows to consider multiple ©2020. American Geophysical Union. All Rights Reserved.

Orographic Effect on Extreme Precipitation Statistics Peaks at Hourly Time Scales

“…The shape parameter of the ordinary events right-tail distribution (Figure 2a-2d) decreases with duration in the plains and, less markedly, in the hills, indicating heavier-tailed distributions at longer durations. This confirms what found in Marra et al (2020) in a coastal region at slightly higher latitude.…”

“…The shape parameter of the ordinary events right-tail distribution (Figure 2a-2d) decreases with duration in the plains and, less markedly, in the hills, indicating heavier-tailed distributions at longer durations. This confirms what found in Marra et al (2020) in a coastal region at slightly higher latitude. Conversely, a nonmonotonic behavior of the shape is observed in the mountains (Figure 2c), with increasing shape parameter, hence lighter tails, at longer durations, and a maximum around durations of 1 h. In the downwind rift the shape smoothly increases with duration, indicating heavier-tails at shorter durations (Figure 2d).…”

“…Following previous studies in similar climates (Restrepo-Posada & Eagleson, 1982;Marra et al, 2020), independent storms are defined as consecutive wet (≥0.1 mm/10 min) time intervals separated by 6-h dry hiatuses; short (<30 min) storms are removed to avoid noise due to individual tips of the rain gauge. Storms associated with Mediterranean cyclones are isolated according to a semi-objective synoptic classification as detailed in Table S1 (Alpert et al, 2004); as the classification is daily, the time in which each storm ends is used.…”

“…The subsequent analyses are based on Mediterranean cyclones only. Ordinary events are computed as the maximum intensities observed within each storm using moving windows of durations between 10 min and 6 h (10, 20, 30 min, 1, 2, 3, 6 h) (Marra et al, 2020). The ordinary events' right-tail is defined as the Weibull-identically-distributed portion of the events ( Figure S1); in our region this was identified as the largest 25% of the ordinary events (Marra et al, 2019(Marra et al, , 2020.…”

“…Ordinary events are computed as the maximum intensities observed within each storm using moving windows of durations between 10 min and 6 h (10, 20, 30 min, 1, 2, 3, 6 h) (Marra et al, 2020). The ordinary events' right-tail is defined as the Weibull-identically-distributed portion of the events ( Figure S1); in our region this was identified as the largest 25% of the ordinary events (Marra et al, 2019(Marra et al, , 2020. The scale and shape parameters of the Weibull distributions describing the ordinary events are computed left-censoring the remaining portion of the data, that is without considering the intensities but retaining the weight in probability, and using a least squares regression in Weibull-transformed coordinates (Marani & Ignaccolo, 2015).…”