Calibration parameters and their range in EFDC model

EFDC is not a self calibrated model. There are several parameters that can be tuned in EFDC model to calibrate the parameters. Calibration is a process of comparing the modeled results with the observed data. If the modeled data and observed data match close then it is said that the model is calibrated. There are several parameters that are used to quantify the calibration process. They are average error, relative error, average absolute error, RMS error, Relative RMS error, Nash-Sutcliffe Efficiency Coefficient. 

Advertisements

Calibration in EFDC

EFDC is a short form of Environmental Fluid Dynamics Code. It is originally developed by John H. Hamrick and this is now curated by Tetratech Inc.

The model is quite complex in itself. The process of creating grid , selecting appropriate boundary conditions and finding the input necessary for the model is quite burdensome at times.

I have surpassed those hurdles and now prepared for the calibration of the model. I have been through several papers on EFDC and have read what they have done for the calibration. The common parameters the authors have used for quite a bit are as follows :

a) bottom roughness height

b) Wind drag coefficient

c) Solar radiation (increasing or decreasing)

d) WInd sheltering.

e) Adjusting Bathymetry

f) Appropriately using wind series.

Well, if you have developed your model properly, then adjusting these parameters may direct you to the good shape. Sometimes you may encounter the errors such as maximum iterations exceeded etc. Since the model doesn’t exactly tell you what was the error. The probable solution of such kind would be to change the iterations number to some bigger number for eg. say 2000. You may also try using small time step and rerun the model.

After calibration, we need to do some statistical analysis to access whether our model is working properly or not.

The parameters such as NES (Nash Sutcliffe Efficiency) , R square, MBE and RMSE can be used to see whether our model is working properly or not.

I would recommend to use EFDC Explorer since it is visual and you can easily figure out some minor mistakes.