FLOOD01798 - Laserfiche WebLink

Home Browse Search

UKAt"1 U tiaal( VI - t:stlmatlan OJ Large to t:.xtreme t"looas Table 17 Areal rainfall estimates for AEPs rarer than 1 in 2000 6 )12 72 132 164 190 233 268 322 146 181 211 258 297 357 3.057 3.057 3.057 3.057 3.057 3.057 3.057 3.057 0.091 0.087 0.082 0.075 0.071 0.064 0.058 0.049 0.077 0.073 0.069 0.063 0.063 0.061 0.059 0.059 0.229 0.229 0.229 0.228 0.229 0.229 0.229 0.229 1.056 1.053 1.051 1.046 1.045 1.043 1.041 1.040 115 146 187 231 268 328 375 451 192 246 299 346 418 473 559 218 278 337 388 465 523 609 247 316 380 436 518 577 662 295 376 449 511 598 656 734 i;:.i!oo 510 600 670 76.0 c.eto "~"~ ,~',,; . 86(l'~.1~,,: Table 18 Comparison of regression-estimate and Bureau of Meteorology's estimates of PMP. I I I I I I I I I I I I I I I I I I I 6.3.1 Selection of Design Parameter Values (a) Based on calibration to historic events and reconciliation with flood frequency estimates In the following, it is assumed that a RORB model has been fitted to four large observed floods. For each flood hydrograph it was found that a good fit could be achieved using parameter values of m and k, of 0.8 and 27, respectively. The loss values were found to vary appreciably, and the median values of initial and continuing loss values obtained were respectively 10 mm and 1.5 mmlhr. In addition, a flood frequency anaiysis was undertaken using 30 years of annual maxima, the results of which are shown in Figure 12. Given that the same set of routing parameters provided a good fit to all calibration hydrographs, the routing values obtained from calibration are adopted for reconciliation with flood frequency estimates. Design flood estimates are obtained using the median loss values obtained from calibration of the runoff-routing model to the four historic events (i.e. an IL of 10 mm and CL of 1.5 mm/hr). These loss values are used in conjunction with the areal design rainfalls for the 1 in 50 AEP and 1 in 100 AEP events. Design floods are then derived for a number of design burst durations to determine the critical duration, and in this case the 24 hour burst is found to be eritical. Tile /'esulls ferUle 1 in 50 AEP and 1 in 100 AEP events are shown in Figure 12 as filled diamond symbols. It is clearly seen that the results obtained using loss rates obtained from calibration are appreciably higher than the design flood estimates obtained from flood frequency analysi... For illustration purposes, results are also shown for design floods obtained using the design losses in conjunction with design storms (further discussion regarding application of the temporal patterns to obtain design bursts and storms is provided the Sections 6.3.3 and 6.3.4). The results obtained for the design burst approach are shown in the main graph area, and the results obtained using the design storm approach are shown in the inset panel. In practice, design storms should be used if they are available in lieu of design bursts, but both results are shown in Figure 12 for comparison purposes. If the flood frequency estimates were derived using at- site data alone, then the probability of indifference (Book III, Section 2) is 1 in 20. Thus, while there is an appreciable difference between the flood frequency and rainfall-based resulls, the flood frequency estimates must be considered unreliable; any adjustment of the loss values to derive rainfall-based results that are in better agreement would therefore need to take into account the wide confidence limits surrounding the flood frequency estimates. However, for this example it may be assumed that some regional information was used to fit the flood frequency distribution, and it is judged that the flood frequency results are likely to be more accurate than the runoff-routing results for events more frequent than 1 in 100 AEP. Accordingly, it is considered appropriate to increase the losses obtained from calibration to achieve a better match between the runoff-routing and flood frequency estimates. By trial and error, it is found that the design burst losses required to achieve better agreement between the runoff- routing and flood frequency estimates are an ll, of 10 mm and a CL of 3.5 mm/hr. If design storms are used instead of design bursts, then the loss values obtained are IL, of 20 mm and a CL of 3.6 mm/hr. The results obtained using these fitted loss values are shown in Figure 12 as filled, circular symbols. It is seen that the ratio between the 1 in tOO AEfl and 1 in 50 AE/> rUfloff-routing results is larger than that for the corresponding flood frequency estimates, however it is similar to the gradient between the two highest recorded annual maxima. The work of Hill et al. (1996')