Laserfiche WebLink
the there is no relationship between the dependent and independent data sets. With MSM, the <br />program automatically selects the combination of independent gages that result in the lowest <br />standard error of prediction (SEP). Just as the "standard deviation" is a measure of how <br />widely data are dispersed from the average value (the mean), the SEP is a measure of how <br />widely the data are dispersed around the regression line (i.e. the amount of error in the prediction <br />of y for an individual x). If there is a wide spread in the data around the regression line, the SEP <br />is large and if the data are grouped tightly, the SEP is small. When the regression provides a <br />perfect fit (r = r~ = 1), the SEP will equal zero. Although the program uses the SEP to select <br />independent gages, the MSM summary report also provides correlation coefficients. <br />To further investigate the process of using TSTool and MSM to fill missing streamflow data, <br />three South Platte River streamflow gages believed to provide a representative sample were <br />filled using both models: Big Dry Creek at Mouth Near Fort Lupton (06720990), Big Thompson <br />River at Mouth, Near La Salle (06744000), and Laramie River Near Glendevey (06657500). <br />The following sections provide details on the specific functions utilized within each model. <br />Regression with TSTool <br />TSTool only allows a dependent gage to be filled using one independent gage at a time. <br />Regression relationships are developed using auser-specified "Analysis Period" and are applied <br />to auser-specified "Fill Period". If the independent gage has missing data within the dependent <br />gage missing period, the overlapping missing period will not be filled. Because TSTool does not <br />automatically select the best independent gage, one or more independent gages must be selected <br />and tested by the user. If, in the end, the user wants TSTool to fill a missing gage with two or <br />more independent gages, caution must be used to identify whether regressed (filled) data is used <br />by a subsequent regression. <br />TSTool can perform simple least squares regression using either monthly (twelve regression <br />equations) or annual (one regression equation) correlations with either linear or logarithmic data <br />transformations. This produces four potential combinations between any dependent and <br />independent gage: monthly correlations with linear data transformations, monthly with <br />logarithmic, annual with linear, and annual with logarithmic. Only one of the four combinations <br />can be tested at a time. Previous DSS experience shows that the combination producing the best <br />results varies on a case-by-case basis and that all four combinations should be tested. <br />The following steps describe a comprehensive approach to filling missing streamflow data using <br />the regression options within TSTool. <br />Where to find more information <br />^ The User Manual provided with TSTool describes the data filling methods <br />available in TSTool in more detail. <br />1. Select an independent gage for each dependent gage based on the following: <br />a. Period of record -sufficient overlap with the dependent gage. <br />Task2.doc 2 of 10 <br />