WSP08834 - Laserfiche WebLink

Home Browse Search

OD13J4 35 calibration parameters of the conceptual models, than to the many calibration parameters of the physically-based models. In addition to these three predictive capability tests, the published results were subjected to a correlation test. The writers postulated that the predictive capability of a particular model might be some function of certain basin or event characteristics. For example, the storm water management model (SWMM) might predict the 5 year event in a 1 square mile basin that is 40% developed perfectly; but its predictive ability might decrease for other events, other basin sizes, or different percentages of development. In the correlation test, the writers evaluate the correlation of the peak discharge ratio (QpRATIO) to the basin area, the percent imperviousness, and the recurrence interval of the storm event. The highest correlation coefficient is less than 0.5 as shown in Table 1I-7. This low correlation indicates that either the limited amounts of data preclude a strong showing of correlation, or no sig- nifica~t correlation exists between the prediction measure and the three independent variables chosen. Summary of test results - The results of these comparison tests produce two significant findings: first, the physically-based models (as a group) do not provide significantly better runoff response pre. dictions; and second, the predictive capability of conceptual models is less sensitive to the model user than is the predictive capability of the physically-based models. These findings suggest that, at the present time, the cost effective rainfall-runoff models for local governments are the conceptual models. Their predictive ability is as good as the physically-based models, yet they are generally less