Laserfiche WebLink
<br />OJ13j6 <br /> <br />37 <br /> <br />expensive to initiate. Just as important. the conceptual models are <br />more likely to yield consistent runoff response predictions regardless <br />of the model user. This consistency is extremely important when the <br />chosen model will be accessed by the various model users in the community <br />such as municipal staff personnel and engineering consultants. <br />These findings must, of course, be tempered with the limitations <br />of this analysis. One must remember the following; <br />1) The statistical inferences are based on the assumption that <br />the predictive information used is representative of the entire "popu- <br />lation" of such information. <br />2) The storm events used were generally the more frequent <br />events; consequently, the mean values of the "goodness of fit" measures <br />represent the mean values for frequent storms and not the whole range <br />of storm events. Thus, models with perfect prediction for the 10-year <br />storm event may predict anyone of the runoff distributions shown in <br />Figure 11-4. The consequences of this distribution variability are <br />significant when calculating damage-frequency curves. <br />3) The published prediction results are from basins with <br />adequate verification and calibration data. Such data generally does <br />not exist in urbanizing basins. <br /> <br />Economic Sensitivity of Errors in Predicting Runoff Responses <br />One of the findings reported in the previous section states that <br />the predictive ability of physically-based rainfall-runoff models is <br />not significantly better than the predictive ability of conceptual <br />models. This condition may change with the availability of better <br />data and improved simulation algorithms. The writers note that the <br />high cost of collecting proper model data for every urbanizing basin <br />