Hi Pearse, 3d interpolation is not simple - like any interpolation. You have to make assumptions on the rules how to find values between source grid points. May be "trilinear" interpolation. But how do you justify this? The ocean is highly anisotropic between horizontal and vertical. Stratification is formed by mostly vertical processes, mixing is isoneutral in many cases. Hence ocean dynamics is not following linear "interpolation rules". So why to invest much work in this direction. Also, in some areas the WOA is a climatology that is supported by only a few measurements anyway. So seeking accuracy from best interpolation onto the model grid may be hopeless. So I just use combined 2d-interpolation horizontally (bilinear with @ave) and 1d-interpolation vertically. The steps are: - extend the original field into land by fill_xy before interpolation to the model grid. This is a 2d nearest-neighbour averaging filler. Please consult the manual. The aim is to improve interpolation results for coastal points. Nevertheless, WOA is not well supported by data points near coasts in some cases. - map horizontally to the model grid by bilinear interpolation with @ave - apply your model land mask - interpolate vertically - fill missing model (mostly bottom and surface) cells by the 1d-nearest neighbour filler of ferret @fnr. This is an extrapolation. I am aware that your initial total nitrate content of the model may be incorrect this way. But the WOA itself has uncertaincy. Stratification from inclined isolines may be disturbed from this method. But this should be corrected during the models spin up by the model dynamics itself. At least the model should run into its own equilibrium during some time and should "forget" initial errors due to nitrate surfcace flux, denitrification, nitrification flux to sediments ... If you are interested I can send you my scripts, just let me know. I hope this helps a little bit, Martin Am 30.11.2015 09:48, schrieb Pearse Buchanan:
|