Are there “rules of thumb” for an acceptable level of model error?
When calibrating to heads the distribution of the heads may be more meaningful than a root mean square error (RMSE). Look for about as many heads that are too high as that are too low. You do not want to see trends in the errors (all high heads in one area) and one should only look for errors in the near field (area of interest). What errors in the head can be accepted? Quite frankly, that is not easily answered since we are not modeling to predict heads! The modeling objective, in the context of source water assessment, is to predict the capture zone for one or more wells. Strictly speaking, therefore, you should calibrate against observed capture zones, but those don’t exist (this is why you are modeling in the first place). On the other hand, a good match of the heads (whatever that means) does not in any way guarantee a good capture zone. For instance, the aquifer porosity has no effect on your steady state solution in terms of heads or flows. However, the length of the “time of tr
Related Questions
- Would like to know about how the simple Keynesian cross model may be solved for a unique equilibrium level of national income. And What impact a fall in interest rate would have on the equilibrium?
- What is the level of uncertainty: percent clay, point model (topsoil and subsoil)?
- What is the quality level of Bat-Attitudes Pro Model Gloves?
