Ben, You will want to set it to permanent ineffective. There will be some conveyance happening below the berm elevation but not full depth.
You may want to use a combination of a high vertical n value (0.99) to essentially block conveyance below say a foot or two of the berm and ineffective set at the berm elevation so that as water rises on the “wet side” of the berm it doesn’t calculate conveyance in the dry side. Once the berm is topped then it can start using say the top 1 to 2 feet as conveyance with a lower n-value (0.02?).
I would test it out first by just setting permanent ineffective at the berm elevation and then a few feet below the berm and see if it makes much of a difference in computed WSE. If it doesn’t just leave it at the top of the berm. I assume this is a steady state model?