mirin said:
As part of our double-checking of coupler mapping function, we plotted monthly averages of taux and tauy as seen by the ocean and atmosphere codes, and compared. I suppose just differences in when each quantity is sampled over the month could make a difference when trying to compare them. We were focusing on the Arctic region, particularly near Greenland, and for both T85 and 1x1.25 the plots were similar but in some places there were significant (over 50%) differences between ocean and atmosphere instantiations. This is work in progress.
Are these the exact same physical quantities in each code?
What kind of agreement should we exect between ocean and atmosphere instantiations?
Thanks.
When comparing atm/ocn taux & tauy on an atm grid vs. an ocn grid,
here are some things to keep in mind:
1) in the cpl and its history files, the basis vectors for wind
and stress are oriented east and north. Component models may use
different basis vectors, and if they do, they must rotate vector
fields appropriately when sending/receiving data to/from the cpl.
pop, for example, uses basis vectors aligned with its shifted
pole grid, and these basis vectors apply in pop history files.
2) atm/ocn wind stress is computed on the ocn grid and then mapped,
via area averaging, to the atm grid. Using area averaging for
mapping results in conservation of momentum at the expense of
accuracy at any given point.
3) The cpl/model interface assumes/requires all data is exchanged
on an A-grid. When using pop, this means pop receives surface
stress data on T-points, which means it must be relocated to
U-points, which I believe results in essentially a smoothing
operation on the wind stress. There's no evidence of this
T-to-U grid mapping in cpl history files as this is done in pop.
4) Wind stress sent to the atm component comes from one instantaneous
calculation, whereas wind stress sent to the ocn component is
typically time averaged over one day (typically 24 samples). This
is conservative but it generally results in a much smoother field.
A daily average of (10m wind speed squared) is also sent to the ocn
component (which is more useful for computing vertical mixing than
a daily average wind stress).
