The MOM6 code style guide at https://github.com/mom-ocean/MOM6/wiki/Code-style-guide#array-syntax disallows this type of usage of array syntax. It seems like fix the identified problem, and adhere to the style guide, by keeping the original multiplication of m_leithy where it is and introducing i,j indices to the arrays.

Thanks! I will make that change.

That will also mean I don't need the "where" statements in the leithy taper function.

So it looks like the original code was multiplying the entire array m_leithy by leithy_taper_function many times, instead of just once. I think this is equivalent to multiplying by leithy_taper_function**N, where N is the number of grid cells in a block. So the effective tapering is almost like a step function. The bug was not just an efficiency problem, it was a correctness problem. Please be sure to examine the effect of correcting this on solutions.

Thanks again! I'm rushing this as I approach the weekend. In fact I'm writing this from the car.

I agree about the correctness. This does not affect the 090 series of G cases which I used to tune the taper, but it does affect coupled run 236.002. Maybe that's a partial explanation for why the MOC is different in 236.002 as compared to the G case runs.