SITE INFORMATION AND CORRESPONDENCE

Home Browse Search

Modeling Guidelines Page 2 7. What are the critical conditions that you are trying to model?Flow?Temperature?pH? Critical conditions may be combinations of conditions for several parameters that occur at the same time, e.g. simultaneous low flow and low hardness. Explain why these are critical conditions. Where parameters are not independent, explain and justify the assumed relations. 8. What is the model's performance objective or target? An example is: "Reproduce total nitrogen and total phosph orus concentrations within t 25% for 90%of observed data within the calibration and validation periods." [Surface Water Quality Modeling: An Introduction, May 2002,pg 15.] 9. How will model performance be demonstrated, and what efforts will be taken to improve the model's accuracy? Model performance may be improved through calibration and validation to reduce model uncertainty, and a sensitivity analysis can provide insight into the model uncertainty. a. How will the model be calibrated and validated? Calibration is the process of selecting model parameters to"fit"the model to the physical world that it simulates. Validation is the process of testing the model with the selected parameters to show that it simulates an independent data set. Calibration and validation information that should be provided include: (1) identification of the empirical data used, (2) discussion of the physical constants or other model variables that were modified or adjusted and why, (3)the procedures that were used to improve model accuracy, and(4)the results of calibration and validation. b. A sensitivity analysis evaluates whether a change in a model parameter results in either a relatively small change in a water quality response or a relatively large change in water quality response. It is typically performed by varying one calibration parameter or one model input at a time,usually by a fixed percentage,through an accepted range of values. A sensitivity analysis can identify the parameters that may be significant contributors to uncertainty in the model predictions. These parameters may then be further developed to reduce the uncertainty. The extent of sensitivity analysis that will be completed, including the parameters that will be analyzed, and the basis for selecting those parameters should be agreed upon prior to conducting the analysis. C. How will the model uncertainty be quantified? Model uncertainty describes how close the model results are to reality,and can be expressed as bias,precision, error,or as the margin of safety in the results. There should be agreement with Regional Board staff regarding the acceptable level of uncertainty in the model results. The quantification of uncertainty typically comes from the calibration, validation,and sensitivity analyses. There should be a goal to minimize the model uncertainty to a certain value if the model is being used to predict exact water quality constituent concentrations. However, if the model is being used to rank best and worst alternatives,the accuracy of the model may not be as important if the error in the model results is the same for each alternative.