Northfold

[doos]

The "northfold" is not working for the project nemo with orca1. It worked fine before with orca12.

[joakimkjellsson]

Here's what I've found out so far:

There are three different treatments of the north fold in Tracmass. All are still there, but one is commented out in branch standard_configs. So what worked with ORCA12 must be one of those options.
The other two are available by using nperio=4 or 6.
I tried to follow the NEMO nomenclature, which is
nperio = 0, closed boundaries
nperio = 1, cyclic east-west
nperio = 2, symmetric around equator (we will probably never do this...)
nperio = 3, North-fold (T-pivot)
nperio = 4, North-fold (T-pivot) and cyclic E-W
nperio = 5, North-fold (F-pivot)
nperio = 6, North-fold (F-pivot) and cyclic E-W

All global ORCA runs either use nperio=4 or 6. The NEMO documentation says nperio=4 is for ORCA2, ORCA025 and ORCA0083. I run nperio=6 for ORCA05, but I'm not sure what folks to with ORCA1.

A quick check in the NEMO source code (https://forge.ipsl.jussieu.fr/nemo/browser/NEMO/releases/release-4.0/src/OCE/LBC/lbcnfd.F90), I find something like
do i=1,IMT
p(i,JMT,k) = sgn * p (IMT-i+2, JMT-2, k)
enddo
do i=IMT/2+1,IMT
p(i,JMT-1,k) = sgn * p (IMT-i+2, JMT-1, k)
enddo
for T-points on T pivots

So the northernmost rows are mirror images of each other with reversed east-west order, and the eastern half of j=JMT-1 is mirrored at the western half of j=JMT-1.
I think sgn=-1 for vectors in all cases, so that u > 0 becomes u < 0 over the fold.
There are also special cases for i=1 etc.

So it seems to me that an ORCA025 trajectory at j=JMT should be put at i=IMT-i+2, j=JMT-2. Shouldn't that be enough?
For F-pivots (ORCA05, ORCA1), the indices are shifted so that
do i=1,IMT
p(i,JMT,k) = sgn * p (IMT-i+1, JMT-1, k)
enddo

In tracmass, nperio = 4 triggers a special case where, if y=JMT-1, x1=IMT+3-x1, and same for nperio=6 but with x1=IMT+2-x1.
We seem to be off by one I-index from NEMO and also doing the shifting at j=JMT-1 and not j=JMT. Should this be the case?