The east/west boundaries are periodic, and the
north/south boundaries are solid walls, with a sponge
layer from 
S to 
S, and N to
N for the standard setting.
At the north-south boundaries, meridional velocity v
is set to zero. Thus, at the northern boundary, both v0(i,NY)
and v1(i,NY) are set to zero. To describe the southern boundary, an
extra point has been added at the southern-most extent of
arrays v0 and v1. Thus, the y-direction of
these two arrays (and arrays which depend on them, such as
psi0) are actually dimensioned 0:NY instead of
1:NY, as the other arrays (such as u0 and u1) are.
An exception is vort0, which one might think is
dimensioned 0:NY, but actually is dimensioned 1:NY.
Then, this southern-most point of both v0 and v1
(i.e. v0(i,0) and v1(i,0)) are set to zero. As a
reminder, v0(i,0) and v1(i,0) describe
and 
, respectively.
Note that though the east/west boundaries are periodic, numerically this is implemented without the use of ghostpoints.
Between V2.0 and V2.1, there has been a shift of grid points that offsets
which point occurs at the equator. This is to accommodate the pole-to-pole
option in V2.1 and makes the V2.1 grid more similar to the Gaussian grid
of GCMs. In V2.1 and higher versions, the equator has a v
(v0 and v1) point, with T and
u(u0 and u1) points half a grid north and south of the
equator. In V2.0 and lower, the equator has a T point.