[SciPy-User] [SciPy-user] 2d interpolation, non-regular lat/lon grid - help with delauney/natgrid??

[Scott Sinclair]

> 2009/8/13 Robert Kern <robert.kern at gmail.com>:
> On Thu, Aug 13, 2009 at 08:47, John [H2O]<washakie at gmail.com> wrote:
>>
>> YOU should've read the (FM)?? Or I? ;)

Looks like everyone needs to :)

>> Actually, I have referred to both docs, but I'm just missing something. &-(
>>
>> Unfortunately, it seems now the problem is that m (basemap instance) expects
>> grid_lon, grid_lat to be of the same length. I tried to convert them into
>> meshgrid objects, but then I get the error I received in the prior message
>> about monotonically increasing axes...
>
> Ah, yes. griddata() only handles regular grids for some reason, not
> arbitrary interpolation points. You will have to use the underlying
> delaunay package to interpolate arbitrary points. Using your variable
> names:
>
>        # triangulate data
>        tri = delaunay.Triangulation(x,y)
>        # interpolate data
>        interp = tri.nn_interpolator(z)
>        Z0 = interp(gridx, gridy)

Cool, I didn't know about going directly to the delaunay package!

I think the original problem was with the large area of missing data
in the plot. John?

This is just a result of the transformation from a regular grid in
long/lat to the projection co-ordinates. The projected grid is no
longer a regular rectangular grid in the polar projections co-ordinate
system. Have a look at the attached script and set proj_grid to True
or False before running the script, to see the difference between
interpolating from scattered points to a grid that is regular and
rectangular in long/lat and one that is regular in a polar projection.

Cheers,
Scott
-------------- next part --------------
A non-text attachment was scrubbed...
Name: griddata_demo.py
Type: application/x-python
Size: 1481 bytes
Desc: not available
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20090813/52c9c495/attachment.bin>