Laserfiche WebLink
<br />Memorandum To: File <br />Page 6 <br />September 28, 2001 <br />Several approaches were tested for disaggregating monthly demand to daily, for the simulation runs. <br />Ultimately, new Statemod code was developed to disaggregate monthly demands by estimating that the <br />monthly average demand rate occurs on the fifteenth of the month, and linearly interpolating between <br />these mid-month points. Since this calculation can result in the sum of daily data exceeding the monthly <br />total, daily values are scaled to insure the monthly value is preserved.. <br />Base name for the model was dlyelk3. <br />Results <br />Calibration models versus historical gage <br />Simulated gage flows from the three daily calibration runs were compared to the historical gage at <br />Clark to determine how well the models simulate historical conditions. Figures 2 through 4 show the <br />hydrographs for 1977, 1983, and 1988 - a dry year, a wet year, and an average year. The graphs <br />show that the Average Daily model (D1yElk1) flows lie at the mean monthly flow and do not change <br />through the month. <br />The Daily Pattern (D1yElk2) and Daily Input (D1yElk3) model results are very close to historical <br />flows. The significant deviations in both these models, relative to historical, can be traced to <br />differences in the amount of streamflow put to or taken from reservoir storage. Consistent with the <br />Yampa Phase IIIb model, there are no operational rights in the Elk River models to draw water out <br />of Steamboat Lake or Lester Creek Reservoir. They store only to replace evaporation losses. <br />Historically, though, there were periods where one or both reservoirs released water for unknown <br />reasons, and subsequently replaced that water in storage. These are identified for the selected water <br />years in Figures 2, 3, and 4. Another factor that contributes to small differences between simulated <br />and historical gage flows is Calculated agricultural demand that exceeds historical diversions. <br />Figures 5, 6, and 7 illustrate deviation from historical daily flows for each day of the simulation, for <br />the Average Daily, Daily Pattern, and Daily Input models, respectively. In a perfect model, all the <br />points would fall on the diagonal line representing Y = X, where Y is the simulated flow and X is the <br />historical flow. The point scatter for the Average Daily model is much wider than for either of the <br />other two daily models. Furthermore, daily flows above 5,000 of/day are never simulated by the <br />Average Daily model. Point scatter for the Daily Pattern and Daily Input models are much closer to <br />the Y = X line, and also very much alike. <br />These observations are quantified in Table 4. Mean error, or average difference between historical <br />and simulated values, is very low for all three models. However, as Figure 5 illustrates, errors for <br />the Average Daily model balance each other out, being both above and below observed values. The <br />standard deviation of the error terms is one expression of dispersion about the Y = X line. Thus the <br />standard deviation of error for the Average Daily model is much greater than for the other two <br />models. These statistics illustrate that the Daily Pattern and Daily Input models do a much better job <br />of predicting flows than the Average Daily model. Furthermore, there is not a lot of difference <br />between the Daily Pattern and Daily Input model results. <br />TaskMem2Final.doc g~~LE <br />