Laserfiche WebLink
b. Location -give preference to gages located on the same tributary or on tributaries <br />with similar drainage area sizes and sub-basin characteristics (e.g. human <br />influences). <br />2. Complete regression analyses using all four combinations of monthly and annual <br />correlations with linear and logarithmic data transformations. Use the entire period of <br />overlapping data between the independent and dependent gage for the Analysis Period; <br />set the Fill Period to 1950 through the current year (the SPDSS study period). <br />3. Review the regression results using scatter plots, line graphs, and the statistical output. <br />Compare annual versus monthly and linear versus logarithmic regressions to determine <br />the combination that produces the best statistical results (correlation coefficient, r, and <br />square root of the mean of the sum of the square errors, RMSE) with reasonable results. <br />Examples of unreasonable results include regressions that produce extreme outliers or <br />winter baseflows that are significantly different than the historical data set. <br />4. Test multiple independent gages to find the best fit that passes the "reasonableness" test. <br />When the period of record, location, and sub-basin characteristic criteria as described above in <br />Step 1 are met, monthly regression equations generally produce better correlations than a single <br />annual equation. <br />Regression with MSM <br />Rather than performing regression on a single dependent gage at a time, MSM can fill several <br />dependent gages at the same time, using data from multiple independent gages. This allows a <br />number of independent gages with different periods of record to be used to fill or extend the <br />incomplete record for the dependent gage. Both dependent and independent gages are provided <br />in a single input file and each gage is treated as both an independent and dependent gage. Only <br />the original historical data are used to fill other gages (estimated values are never used to fill <br />missing data). <br />MSM can perform simple least squares regression using monthly and annual correlations with <br />logarithmic data transformations (a straight linear regression is not available). Because a <br />logarithmic transformation is always used, zero values within the dependent gage data set are not <br />used to develop correlations with independent gages. The MSM default setting uses the SEP to <br />select the best fit for each missing data point in the dependent gage from both monthly and <br />annual correlations with each of the independent gages. The user can override the default setting <br />and specify that only monthly (cyclic) or only annual (non-cyclic) correlations be used. Other <br />program settings include the minimum concurrent values (previous DSS efforts recommended <br />using a minimum value of 20) and the confidence interval (default is 95%). <br />Following is a stepwise description of how missing streamflow data can be filled using MSM. <br />Where to find more information <br />^ Appendix A of this memo provides more detailed MSM user instructions. <br />1. Develop MSM input files using TSTooI to extract streamflow data from HydroBase in <br />standard StateMod (*.stm) format. The first set of input files should include every <br />Task2.doc 3 of 10 <br />