The domain I am working with is quite near the equator -1S to -21S. This seems to be quite problematic for DA and especially since it seems that only a geostrophic balance is applied for updating the 3D UV momentum balance increments.
The ECMWF NEMOVAR system deal with this by applying a β-plane approximation combined with a modified pressure increment to enforce a symmetric pressure increment about the equator [http://apdrc.soest.hawaii.edu/doc/tm668.pdf].
For my research I am looking at improving ocean currents with DA so it's hard for me to grasp how much of an impact turning off the 3D UV momentum balance in my 4D-Var update is having on my results (the model blows up near the equator keeping this turned on).
I don't think I am confident enough to go into the balance codes and try and implement this myself so I'm adding it to the wish list.