Re: [ferret_users] Vertical velocity in Soda3.15.2 monthly mean data

[William Kessler - NOAA Federal <william.s.kessler@xxxxxxxx>]

x

1. Use exactly the same discrete continuity equation that the model uses.  Different discretization will often result in large error.

Yes, Ryo is correct. Models often have difference schemes that are more complicated than @DDC. For instance the differencing can extend across multiple grid cells in specific patterns. These are generally developed so that mass, properties and momentum balance over regions or the whole model.

 It might seem that a simpler differencing like @DDC should approximately work but I have seen it produce significantly wrong results, as Ryo says. w is sensitive to this because it is a vertical integral of horizontal derivatives; errors accumulate. To get exact results, there is no alternative to reading the model documentation and writing a script that does derivatives as in the model. If you don't have the original grid cell velocity (as it appears to be the case in SODA), you might very well get errors as you describe and not be able to calculate w.

Does the result improve if you average over time intervals or spatial regions? That might be a fallback.

Billy

On Fri, Sep 1, 2023 at 10:53 AM Satheesh Shenoi <shenoi1958@xxxxxxxxx> wrote: