Laserfiche WebLink
<br />0J135S <br /> <br />27 <br /> <br />. . <br /> <br />This coefficient is always less than or equal to one and equals unity <br />when the simulated hydrograph perfectly corresponds to the observed <br />hydrograph. Sarma's ratings for this measure are given in Table 1I-3. <br />These three "goodness of fit" measures were computed for the <br />simulated and observed storm events reported in the eight papers. A <br />listing is presented in Appendix B. The tables in Appendix Bare <br />separa ted by model category. Tabl e B-1 cons i sts of the "goodness of <br />fit" measures for physically-based models and Table B-2 consists of <br />the measures for the conceptual models. These "goodness of fit" <br />measures provide useful comparative information. They illustrate pre- <br />cisely how well each model predicts all aspects of the observed runoff <br />response -- the peak discharge and the time distribution of storm water <br />runoff. <br />Comparison tests and results - It is difficult to directly compare <br />the tabulated "goodness of fit" measures; there are just too many. <br />For this reason, composite measures -- the mean and standard deviations <br />-- were computed for various groupings of the data as described below. <br />These composite "goodness of fit" measures were then used to compare <br />the predictive capability of the two model categories. The statistical <br />validity of using these composite measures rests upon the assumption <br />that the data in each grouping is a representative sample of the total <br />population of "goodness of fit" measures. While this may not be en- <br />tirely true for all the tests described below, the composite measures <br />allow a precise comparison of an otherwise unwieldy amount of data. <br />The predictive capability of the two model categories was compared <br />in three tests -- the one basin test, the overall test, and the con- <br />sistency test. <br />