Laserfiche WebLink
<br />The end result is a greater basin contribution in the design storm. The simulation of the 2-year <br />event is outside the 95 percent confidence limits, making the simulation invalid for this return <br />period. Since the calibration is based on a single event, the final parameters are subject to the <br />antecedent conditions of that event. The initial loss parameter is fairly low indicating a relatively <br />moist soil condition. The success of the simulation for less frequent storms suggests this is a <br />reasonable expected condition. Failure in simulating the 2-year flood suggests a larger degree of <br />variability in basin conditions for more frequent storms. Using more than one storm for <br />calibration could alleviate this problem. <br />....------ <br /> <br />The SCS CN model simulations are, overall, rair. The problems encountered with the 2-year <br />event do not appear to be an issue using this method. The model begins to over simulate <br />between the 2- and 25-year floods and begins to deviate below the ensemble beyond the 25-year <br />flood. The SCS CN approach assumes a constant maximum amount of surface retention once <br />the initial losses are satisfied. The result is a constant fraction loss rate. This is a significant <br />difference from the infiltration decay curve that is the basis of the Green and Ampt model. The <br />constant fraction loss results in greater infiltrated volume in larger storms. This basin appears to <br />be more sensitive to the shape of the actual infiltration decay. <br /> <br />4.2.2 Geary Creek tributary <br /> <br />Geary Creek tributary was only 25 percent censored with the paleoflood data. An estimate of the <br />1997 flooding was not made at this location. This level of censoring falls just within the <br />threshold recommended in Bulletin l7B. With this, the Bulletin l7B method could be used <br />without the multiple scenarios. For this site, both procedures were performed. The resulti!lg <br />frequency curve was slightly less than the meaJ} frequency curve of the ensemble. The ensemble <br />method should account for the censored data with less sampling error simply due to the number <br />of scenarios. The ensemble results are used in the remaining analysis. <br /> <br />Figure 12 illustrates the how the peak flow data fit the frequency curve. The goodness of fit test <br />indicated that the distribution could not be rejected. Storm data proved to be unreliable at this <br />site due to apparent clock errors in the equipment. The 5- and lO-year floods were used for <br />calibration. As seen in Figure 12, this is the narrowest portion of the ensemble results. Figure <br />13 illustrates the results of the simulation. <br /> <br />The simulations for both models are good with the exception of the 2-year flood. This point falls <br />just below the lower limit of the ensemble. The SCS CN model follows the mean of the <br />ensemble very closely to the I DO-year flood. <br /> <br />16 <br /> <br />v <br />