The uniform risk engineering practices that are increasingly being adopted for structural design require estimates of the extreme wind loads with very low annual probabilities of exceedance, corresponding to return periods of up to 3000-years in some cases. These estimates are necessarily based on observational wind data that typically spans only a few decades. The estimates are therefore affected by both large sampling uncertainty and, potentially, non-negligible biases. Design practices that aim to meet mandated structural reliability criteria take the sampling uncertainty of long period wind speed or wind pressure estimates into account, but reliability could be compromised if estimates are also biased. In many circumstances, estimates are obtained by fitting an extreme value distribution to annual maximum wind speed observed over a few decades. A key assumption implicit in doing so is that wind speed annual maxima are max-stable. Departures from max-stability can exacerbate the uncertainty of long-period return level estimates by inducing systematic estimation bias as well. Observational records, however, are generally too short to assess max-stability. We therefore use wind speed data from a large (50-member) ensemble of CanRCM4 historical simulations over North America to assess whether wind speed annual maxima are max-stable. While results are generally reassuring at the continental scale, disquieting evidence of a lack of max-stability is often found in the central and southern parts of the continent. Results show that when annual maximum wind speeds are not max-stable, long period return level extreme wind speeds tend to be underestimated, which would compromise reliability if used to design infrastructure such as tall buildings and towers.

Abstract The recurring devastation caused by extreme events underscores the need for reliable estimates of their intensity and frequency. Operational frequency and intensity estimates are very often obtained from generalized extreme value (GEV) distributions fitted to samples of annual maxima. GEV distributed random variables are “max-stable,” meaning that the maximum of a sample of several values drawn from a given GEV distribution is again GEV distributed with the same shape parameter. Long-period return value estimation relies on this property of the distribution. The data to which the models are fitted may not, however, be max-stable. Observational records are generally too short to assess whether max-stability holds in the upper tail of the observations. Large ensemble climate simulations, from which we can obtain very large samples of annual extremes, provide an opportunity to assess whether max-stability holds in a model-simulated climate and to quantify the impact of the lack of max-stability on very long period return-level estimates. We use a recent large ensemble simulation of the North American climate for this purpose. We find that the annual maxima of short-duration precipitation extremes tend not to be max-stable in the simulated climate, as indicated by systematic variation in the estimated shape parameter as block length is increased from 1 to 20 years. We explore how the lack of max-stability affects the estimation of very long period return levels and discuss reasons why short-duration precipitation extremes may not be max-stable.

M.A. Ben Alaya was supported by the Climate Related Precipitation Extremes project of the Global Water Futures program.

Abstract Recently dam managers have begun to use data produced by regional climate models to estimate how probable maximum precipitation (PMP) might evolve in the future. Before accomplishing such a task, it is essential to assess PMP estimates derived from regional climate models (RCMs). In the current study PMP over North America estimated from two Canadian RCMs, CanRCM4 and CRCM5, is compared with estimates derived from three reanalysis products: ERA-Interim, NARR, and CFSR. An additional hybrid dataset (MSWEP-ERA) produced by combining precipitation from the Multi-Source Weighted-Ensemble Precipitation (MSWEP) dataset and precipitable water (PW) from ERA-Interim is also considered to derive PMP estimates that can serve as a reference. A recently developed approach using a statistical bivariate extreme values distribution is used to provide a probabilistic description of the PMP estimates using the moisture maximization method. Such a probabilistic description naturally allows an assessment of PMP estimates that includes quantification of their uncertainty. While PMP estimates based on the two RCMs exhibit spatial patterns similar to those of MSWEP-ERA and the three sets of reanalyses on the continental scale over North America, CanRCM4 has a tendency for overestimation while CRCM5 has a tendency for modest underestimation. Generally, CRCM5 shows good agreement with ERA-Interim, while CanRCM4 is more comparable to CFSR. Overall, the good ability of the two RCMs to reproduce the major characteristics of the different components involved in the estimation of PMP suggests that they may be useful tools for PMP estimation that could serve as a basis for flood studies at the basin scale.

In the context of climate change and projected increase in global temperature, the atmosphere’s water holding capacity is expected to increase at the Clausius-Clapeyron (C-C) rate by about 7% per 1 °C warming. Such an increase may lead to more intense extreme precipitation events and thus directly affect the probable maximum precipitation (PMP), a parameter that is often used for dam safety and civil engineering purposes. We therefore use a statistically motivated approach that quantifies uncertainty and accounts for nonstationarity, which allows us to determine the rate of change of PMP per 1 °C warming. This approach, which is based on a bivariate extreme value model of precipitable water (PW) and precipitation efficiency (PE), provides interpretation of how PW and PE may evolve in a warming climate. Nonstationarity is accounted for in this approach by including temperature as a covariate in the bivariate extreme value model. The approach is demonstrated by evaluating and comparing projected changes to 6-hourly PMP from two Canadian regional climate models (RCMs), CanRCM4 and CRCM5, over North America. The main results suggest that, on the continental scale, PMP increases in these models at a rate of approximately 4% per 1 °C warming, which is somewhat lower than the C-C rate. At the continental scale, PW extremes increase on average at the rate of 5% per 1 °C near surface warming for both RCMs. Most of the PMP increase is caused by the increase in PW extremes with only a minor contribution from changes in PE extremes. Nevertheless, substantial deviations from the average rate of change in PMP rates occur in some areas, and these are mostly caused by sensitivity of PE extremes to near surface warming in these regions.

Abstract Probable maximum precipitation (PMP) is the key parameter used to estimate the probable maximum flood (PMF), both of which are important for dam safety and civil engineering purposes. The usual operational procedure for obtaining PMP values, which is based on a moisture maximization approach, produces a single PMP value without an estimate of its uncertainty. We therefore propose a probabilistic framework based on a bivariate extreme value distribution to aid in the interpretation of these PMP values. This 1) allows us to evaluate estimates from the operational procedure relative to an estimate of a plausible distribution of PMP values, 2) enables an evaluation of the uncertainty of these values, and 3) provides clarification of the impact of the assumption that a PMP event occurs under conditions of maximum moisture availability. Results based on a 50-yr Canadian Centre for Climate Modelling and Analysis Regional Climate Model (CanRCM4) simulation over North America reveal that operational PMP estimates are highly uncertain and suggest that the assumption that PMP events have maximum moisture availability may not be valid. Specifically, in the climate simulated by CanRCM4, the operational approach applied to 50-yr data records produces a value that is similar to the value that is obtained in our approach when assuming complete dependence between extreme precipitation efficiency and extreme precipitable water. In contrast, our results suggest weaker than complete dependence. Estimates from the operational approach are 15% larger on average over North America than those obtained when accounting for the dependence between precipitation efficiency and precipitable water extremes realistically. A difference of this magnitude may have serious implications in engineering design.