Need suggestions on regridding

Hi Xin,

You can see an indepth explanation of our various regridding methods near the bottom of this section:
http://earthsystemmodeling.org/esmf_releases/non_public/ESMF_5_3_0/ESMF_refdoc/node5.html#SECTION050332400000000000000

It's not clear what kind of comparison that you're doing between the fields.

+ If you want to compare values point by point then I would suggest the bilinear or patch interpolation methods. The bilinear method does linear interpolation between the source and destination fields. Which means that the value of the destination point will vary smoothly between the values of the source points surrounding it. The patch method does second order interpolation. It has a wider stencil and results in smoother derivatives than the bilinear, but uses more time and memory.

+ If you want to conserve the integral of the values (e.g. total water content) then I would use the conserve method. It was built to preserve the integrals of fields across the interpolation, although because of this added constraint the interpolation of the values isn't quite as accurate as the bilinear or patch.

Regarding your question about the difference between these and a weighted average. In all of these methods the destination point is basically a weighted sum of various source points. With the weights calculated to give the various properties above. I guess the difference is that you don't need to take the time and effort to compute the weights yourself. :-)

Does that answer your question?

Thanks,

- Bob

On Nov 2, 2012, at 9:43 AM, Mary Haley wrote:

> Hi Xin,
>
> I'm CC ESMF support on this message, as they are experts in this department.
>
> --Mary
>
>
>
> On Oct 31, 2012, at 1:47 PM, Xin Xi wrote:
>
>> Hi there,
>>
>> I have followed recent discussions on regriding, and I happen to be dealing with this issue recently. I need to compare two soil moisture (unit: m^3/m^3) fields, one is on a coarse rectilinear grid (1x1 degree) and the other is on a curvilinear (from WRF model) grid at 40 km resolution. I could regrid from 1x1 grid to wrf domain, or from wrf domain to 1x1 grid. Which way and which regrid method can better suit my needs here?
>> To avoid interpolation errors, I guess it is better to go from fine to coarse resolutions. If so, I can do the interpolation myself by looping over the coarse grid, and assigning the average value (or weighted average, and other methods) of all fine pixels that fall within each coarse cell. What are the differences of doing this compared to the NCL regrid methods? Can someone explain this to me?
>>
>> Thanks a lot !
>> Xin
>> _______________________________________________
>> ncl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
> _______________________________________________
> Esmf_support mailing list
> Esmf_support@list.woc.noaa.gov
> https://list.woc.noaa.gov/cgi-bin/mailman/listinfo/esmf_support
Received on Fri Nov 2 13:54:28 2012