Laserfiche WebLink
that had no data for that time period and thus there were no regressions possible. In this case <br />even average data was not able to fill missing values so the final *.xbf files contained -999 <br />values. <br />Generally the results show good correlation. Table 2 shows a comparison of means of filled data <br />to historic data for gage sites in the san juan basin. The model is robust enough to be able to fill <br />a high percentage of missing data at the 95% confidence interval. As was discussed in the task <br />10 memo, the reliability of the data is dependent upon the amount of data to be filled. As shown <br />in Fi ug res 4 through 8 the percentage of data filled increased as the period of analysis is <br />increased. For example in the Colorado river, the amount of data that was predicted for the 1909 <br />period is 58%. This means that almost 60 percent of the filled data set are generated from the <br />msm. Where-as for the 1949-96 period only 48 percent of the data are estimated. As the <br />percentage of data filled increased, the reliability of that data should be verified to assure that the <br />data generated is a reasonable estimate of the actual historic baseflows. <br />The user should note that, for the best results, using the data from 1909-1996 for the MSM will <br />utilize all potential concurrent periods for independent variables. Th.e user can then select the <br />more recent periods, i.e., 1949-1996, to have a data set that has the least percentage of estimated <br />(filled) data. <br />When the Method 3 approach was tried, the percentage of filled data was about the same but <br />generally the data was filled more from base flows within the basin. Fi ug res 9 through 13 show <br />the results for this method in terms of data filled from in-basin base flows and key gages flows. <br />Since the results using Method 3 appeared to be very close to the results obtained with the <br />original method, it was concluded that Method 3 would be used to generate the *.xbf files for <br />each basin. This would provide greater convenience when operational modifications are made in <br />the Statemod models for each basin. Specifically, although the MSM program would need to be <br />rerun for that basin, in other basins *.xbf files would not be affected. <br />This data will be used in task 11.12-16 to evaluate the estimation of baseflows by the MSM. <br />This will provide further validation of the results of the model. Until these simulations are <br />complete, the reliability of the MSM data filling won't be completely validated. <br />Comments and Concerns <br />The USGS Mixed Station Method was used to fill missing baseflow data. The PC version of the <br />program, written in 1989, and was adapted with a GUI for use in CRDSS. In addition, several <br />routines were written that provides summaries of the results. <br />The results show that the model is robust enough to be able to fill high percentage of missing <br />data at the 95% confidence interval. Some basins contained gaps in the data across the whole <br />basin and, therefore for some time periods, data cannot be filled. <br />Several approaches to data filling were tried and it was concluded that the best method, based on <br />modeling and maintenance issues, was to utilize the base flows within the basin along with key <br />gage flows from all basins to create the filled data set for the basin of interest. Comparison of <br />this method to other methods indicates favorable results. <br />Appendix E E-117 <br />