In ROMS, the vertical grid changes temporally at a fixed location and spatially at a given time. Suppose I want to take the time mean of some variable, say temperature, over several months. Should I first interpolate the temperature field at every time instant to a new vertical grid with fixed cell centers and dz (i.e., constant in time)? Otherwise, I will be averaging a field at a given (lon,lat) over different depths over the averaging duration, which makes it harder to interpret.
The same question also applies for a spatial horizontal mean. Suppose I want to average the temperature over the horizontal extent of the domain to produce a single vertical profile T(z). Without interpolating everything first onto a new vertical grid whose spacing does not change with (lon, lat), I will be averaging fields at different depths.
In summary, it looks like any time or horizontal averaging operation in ROMS requires interpolation onto a new vertical grid. Is my understanding correct?
[I have a related question regarding the use of pm, pn but I will post that in a separate thread.]