Laserfiche WebLink
demands and projects that have changed over time and therefore filling of data based on non- <br />stationary data sets of gaged data results in miss-estimation of the gaged flows. <br />A test of the three methods for the San Juan Basin indicated that Method 2 did create data that <br />did not provide good estimates of base flows for the basin due primarily to the non-stationarity of <br />the gaged flows on several stations. The use of Method 3 closely matched the results of the <br />results of the original approach. <br />Table 2 shows a comparison of Method 1 and Method 3 for gages in the San Juan Basin. The <br />table shows the means of filled monthly data for method 1 and method 3. For comparison <br />purposes, the table also shows the mean monthly natural flow for the period of gaged record. <br />The comparison of the methods showed that the use of Method 3 does result in reasonable <br />estimates of filled data when compared to both Method 1 results and historic data. <br />Results and Analysis of Fill <br />The results of the data fill can be reviewed in the files that are generated as part of the data fill <br />process. The program generates three output files for each basin, the filled data files (*.xbf), the <br />summary file (*.sum) and the statistics files (*.sts). If the user selects that option, the *.xbf file <br />will contain the average monthly data for data that is not filled by the MSM. <br />When the data was initially filled, using the original method, there were 8 months in one station <br />in which data could not be filled (225 gaging sites and 1056 months). This was because there <br />were no stations that had correlations that satisfied the confidence intervals. The confidence <br />interval was relaxed from 95% to 80% and only added three points to the fill. Upon discussions <br />with the State, that rather than relaxing the data fill criteria for adding 8 points to the data set that <br />monthly averages should be used to fill those few points. As a result the option was given to fill <br />missing data with monthly average values. If the mixed station model does not complete filling <br />in the missing data and the user selects the average fill option then the remaining data will be <br />filled with average monthly values and the *.xbf file will be overwritten. The file will indicate <br />which stations and which years were filled with average data. <br />When studying the sensitivity of the data fill to the confidence interval it was noticed that when <br />changing the confidence interval the model sometimes selected a different independent station <br />for filling the data. This is because the model selects the lowest SEP that meets the confidence <br />interval. The confidence interval check utilizes the correlation and the number of data points <br />used in the correlation. If the confidence interval is relaxed then the selected independent station <br />may change because the number of possible independent variables that are acceptable may <br />change. <br />In order to get an idea of the confidence in the results we reviewed the amount of data filled in <br />each basin and which stations were used for the fill (using the original fill approach). Fi u~res 4 <br />through 8 show the percent of data filled and amount of fill that was obtained from each of the <br />five basins. These figures also show the data that was missing after the fill. We ran the mixed <br />station model to fill data for three periods, 1909-1996, 1929-1996, and 1949-1996. The results <br />of each run are shown in the chart. In the Colorado and Gunnison basins, the results for the <br />1949-1996 show significant data that was not filled. This is because there were several stations <br />Appendix E E-116 <br />