masking and smoothing, who goes first

[feroda]

I just want to make sure that for the masking and smoothing, who should go first?
Here I would like to raise an example:

A point, B, in the model grid with the location index is (i,j).
On the assumption that hraw(i,j)=4 which is interpolated from topography data. Also, hmin=10.

If smoothing comes first, then the depth in point B may get a new value of B(i,j)=20.
After that, do masking will treat point B as a water point with the depth of 20.

In other case,

If masking comes first, point B will be a land point for the depth is less than hmin=10. Even if in the hereafter smoothing, the value of point B will not changed for we do NOT do smoothing at the land point.
So, in the ROMS, depth for B(i,j) is 10.

What's your opinion? Thank you!

[arango]

The smoothing is independent of masking. That is, smoothing should not affect land/sea masking. However, you can use the land/sea masking in local smoothing to improve computations. The coast is always at the same place. What the smoothing does in coastal areas is either to make it shallower or deeper. We never apply land/sea masking to the bathymetry (h) since we divide by it in several places in the code. Recall that division by zero is not mathematically possible. The minimum value (hmin) is assigned over land areas. We need to have a finite water column to define model levels.

[kate]

My smoother knows about the land mask. This is to preserve the fjordness of fjords, where you have deep water between shallow masked points. I didn't worry about it before moving to Alaska, so it depends on your application (like many things).