Re: regridding faster with rcm2rgrid?

The 'search algorithm' used in rcm2rgrid is brute force.
It works well enough for smallish arrays but its performance is like
an N^2 sort. It slows significantly as the number of points increases.

---
If you have a "very specific satellite curvilinear grid
to a rectilinear grid" issue, I would suggest using the ESMF
regrid method.
[1] Generate the weight file once.
     This could be slow BUT it is done only once.
[2] Then regrid via ' ESMF_regrid_with_weights'. This is very fast.
[3] http://www.ncl.ucar.edu/Applications/ESMF.shtml
     See examples: 5, 8, 15
Regards
On 4/16/14, 6:04 PM, Paula Doubrawa wrote:
> Hello, everyone-
>
> I am using *rcm2rgrid* successfully to go from a very specific satellite
> curvilinear grid to a rectilinear grid.
>
> I am looking for some advice on how to make it work faster.
>
> The original data has 1539 x 830 points. When I try to go to 400 x 300
> points, it takes 6 minutes to regrid one variable. Is this normal?! I was
> hoping to go to 400 x 900.
>
>>From the original data, I am only interested in a portion of it that falls
> within a 2 degree x 6 degree box. Even within this box, several points are
> just missing data (as my valid data is only on water).
>
> Is rcm2rgrid still looking at every single point of the original data,
> outside or inside my box (my box meaning what I defined my lat1d and lon1d
> to be), and also looking at each missing value which could be using a lot
> of time? Is there a way to make it ignore the missing values so it runs
> faster? Or do you have any other suggestions?
>
> Additionally, is it possible to see the code that is doing this work so I
> can understand it better and try to plan accordingly to maximize efficiency?
>
> Any advice is appreciated!
> Thank you very much,
> Paula
>
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk