Vertical interpolation

460 views
Skip to first unread message

Nick Savage

unread,
Jan 19, 2013, 11:31:00 AM1/19/13
to scitoo...@googlegroups.com
I need to be able to convert grib data on a hybrid pressure grid (ECMWF model data) to data on a hybrid height grid (Met Office UM model data).

While there are python tools that can easily be adapted to do horizontal regidding, vertical interpolation is a more difficult problem. I feel that it would be a useful addition to IRIS though.  Code from the UM reconfiguation code (written in Fortran) could be used as a source of algorithms for this.

Would anyone else find this capability useful?

Phil Elson

unread,
Jul 16, 2013, 9:25:15 AM7/16/13
to scitoo...@googlegroups.com
I've just been pointed to this post as I've recently been asked by Bernd Becker about vertical interpolation. Personally I think hooking into external sources of algorithms (such as vertical interpolation from the UM) is a capability which we would love to make easy to do, but I'm not sure it is the kind of capability that Iris should ship with by default (save that for an Iris-UM-bridge spin-off package maybe...) though I have to admit I haven't thought about the logistics too much here.

As it stands, Iris has the capability to interpolate between various vertical coordinate systems (I've tested with a combination of hybrid height vs iso-height levels) using linear interpolation. I'm guessing that linear is a pretty bad choice for interpolation of pressures though and doing some logarithmic interpolation would be more desirable. If logarithmic interpolation were available, is that something that you would find useful?

Nick Savage

unread,
Jul 22, 2013, 3:48:44 AM7/22/13
to scitoo...@googlegroups.com
I possibly was not aware of all the capabilities of Iris, or it has had additional capability added. Two questions:

1) does it do hybrid pressure (ECMWF) to hybrid height (MetUM) interpolation? Perhaps I can work with you to try and put a test example together?
2) can it cope with grids which do not completely overlap (as the Reconfiguration program can)?

I am not sure that logarithmic interpolation versus linear is the issue - more a question of what vertical co-ordinate you use for interpolation (i.e. height or pressure). A lot of information about what reconfiguration does is available in UMDP S1
"
Interpolation techniques and grid transformations"
Reply all
Reply to author
Forward
0 new messages