A colleague and I have been talking about when it is proper to use permanent ineffective areas. I generally use the normal ineffective areas when modeling bridges, on both the upstream and downstream sides (this is for steady flow).
My colleague says that for the upstream side of the bridge, permanent ineffective areas should be used since the embankment will always be blocking a part of the cross section.
Also, I use the lowest top-of-road for the upstream ineffective area elevation and the low chord for the downstream ineffective area elevation, unless there are special conditions at the site which would dictate otherwise.
My colleague says that he always uses low top-of-road for both sides.
The application of ineffective flow areas around bridges always has been and always will be very subjective. Every modeler has their own preferance. What you are trying to do, more than anything else, is maintain flow distribution consistency. In other words, if on the upstream side of the bridge you have 10% of the flow in the left overbank, 75% of the flow in the main channel, and 15% of the flow in the right overbank, then you want to have about 10%, 75%, and 15% on the downstream side of the bridge. In unsteady flow modeling, numerical stability becomes and issue as well. Using permanent ineffective flow areas can help to minimize the shock of a sudden change of storage volume to conveyance volume. How do you justify permanent? If the roadway embankment slope is shallow enough, I usually use non-permanent. If it is abrupt, or even near vertical, you could justify permanent.
Vertical position of the ineffective flow triggers is also very subjective and up to the preferances of individual modelers. Again, make sure you have flow consistency (if the upstream triggers turn off, the downstream triggers should trun off at the same time/profile). Personally, I like to put my upstream triggers about 0.5 ft above the high chord on the upstream side, and about 0.5 ft below the high chord on the downstream side. Then I check flow consistency and adjust from there.
I attended a HEC-RAS class given by the National Highway Institute, and I asked the instructors about this permanent ineffective areas business. They said that it was not applicable to steady-state modeling, and they would never recommend using it in a steady-state model. "Absolutely not" was one instructor's reply.
The reason that permanent ineffective areas are included in RAS is for use in unsteady flow applications, when a model is crashing or otherwise acting squirrelly and you need to 'put your thumb on it' in order to make it resolve.