The current design storm estimation method used in Canada is based on single site frequency analysis and single site intensity-duration-frequency relationships and involves large uncertainties, especially at short-term record stations and ungauged sites. To overcome the shortcoming of the current approach, a new improved method based on regional frequency analysis and regional depth-duration-frequency equations is proposed. The L-moments are used in the three stages of regional frequency analysis, namely the delineation of homogenous regions, the identification of a regional parent distribution, and the estimation of distribution's parameters. Following a numerical analysis of short duration (5 minutes to 24 hours) rainfall extremes from 375 stations, it was found that Canada may be considered as one homogeneous region where L-skewness and L-kurtosis display no significant spatial variability. Also, based on mean annual precipitation (map), Canada may be subdivided into climatologically homogeneous sub-regions, wherein the L-coefficient of variation in virtually constant. The regional parent distribution was identified as the general extreme value (GEV), the parameters of which depend on the map and storm duration. These findings are different from the present method, where the extreme value type I (EVI) is used irrespective of storm duration. A hierarchical regional approach is proposed for fitting the identified GEV distribution, where the L-skewness, L-coefficient of variation, and mean are estimated on a regional, sub-regional, and at-site basis, respectively. Monte Carlo simulation studies indicate that the hierarchical regional GEV frequency approach is substantially more accurate than the single site frequency method. In particular, it is shown that three times as much data are required for the single site method to provide the same accuracy as the hierarchical regional approach. The depth-duration and depth-frequency ratios computed by the developed hierarchical regional GEV approach are used to assess the hypothesis that convective cells associated with short duration storms (i.e. less than 120 minutes) have common properties in different hydrologic regions. Depth-duration ratios (defined as the ratios of the t-min to the 60-min rainfall depth of the same return period) are found to be independent of return period and geographical location for any storm less than 60 minutes. However, for storms of longer durations, depth-duration ratios depend on both the return period and the geographical location indexed by the at-site map. Depth-frequency ratios (defined as the ratios of the T-yr to the 10-yr rainfall depths of the same storm duration) are also found to depend on the return period and geographical location. Hence, the assumption of geographically independent depth-frequency ratios used in previous studies is incorrect. Generalized expressions of depth-duration and depth-frequency ratios are combined to develop a set of regional depth-duration-frequency equations that are applicable in Canada. These equations are found to be more accurate than other regional equations developed in previous studies. Furthermore, a split sampling experiment has verified that the proposed equations reproduce the rainfall frequency data at long-term record stations in different hydrologic zones better than the existing single site AES equations. Finally, the proposed hierarchical regional GEV approach and depth-duration-frequency equations are combined to develop a new design storm estimation method at ungauged sites. This method is shown to be a viable alternative to the current arbitrary interpolation procedure from isoline maps.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/6831 |
| Date | January 1994 |
| Creators | Alila, Younes. |
| Contributors | Adamowski, Kazimierz, |
| Publisher | University of Ottawa (Canada) |
| Source Sets | Université d’Ottawa |
| Detected Language | English |
| Type | Thesis |
| Format | 288 p. |
Page generated in 0.0021 seconds