3-D stress

[winterbellgirl]

Hi,

I wanted to look at the vertical structure of stresses, and so I computed rho*AKv*du/dz (or dv/dz). The vertical gradient is computed using the finite difference method, not splines. I compared the values at the surface and bottom boundaries to sustr/bustr, which are output into history files, and found that at bottom the computed value is in opposite direction of bustr (at surface the direction is consistent with sustr). Although I knew that in ROMS the boundary condition of du/dz = bustr/Akv is not applied, and so the computed value will not be the same as real bottom stress, but at least the direction should be the same. Also, the computed value is not located exactly at the bottom, but half level above the bottom, and I don't think this offset can lead to a reverse. Any ideas about this problem? Thanks!