Re: ESMF_regrid

I've already responded offline about this, but thought I'd post to ncl-talk as well.

You can try the "conserve" interpolation method. See attached script and image.

--Mary

On Jul 10, 2013, at 11:15 PM, Jiaxin Zhang wrote:

> Hi Mary,
>
> Thank you so much for clarifying. Now, I understand why we get this output after regridding.
>
> I was assuming that using a mask in the regridding would allow to get regridded values for all the grids in the mask. But maybe this is not correct, right?
>
> ESMF is the best option for this interpolation. Because the original file is 0.5x0.5 (360x720), and I need to interpolate temperature values to 1x1 (180x360), I wonder if there are any available options in ESMF that would allow to regrid from 0.5x0.5 to 1x1, but regridding temperature values for the land-water boundaries. Are there any available options to solve this problem in ESMF?
>
> In case there are no options in ESMF to directly interpolate temperature values to the land-water boundaries from 0.5x0.5 to 1x1, perhaps there may be some "alternative" solutions: for instance filling these grids (included in the mask but located in land-water boundaries) with the original temperature values at 0.5x0.5 instead of leaving the grids with empty values for temperature. Or perhaps doing something else?
>
> I really need to have the temperature data regridded at 1x1 and with temperature values for all land grids (including land-water boundaries). Your help and suggestions about how to succeed in this will be very appreciated.
>
> Thank you!
>
> Jiaxin
>
> 2013/7/10 Mary Haley <haley@ucar.edu>
> Hi Jiaxin,
>
> I believe the issue is simply that in the 1 degree grid, you don't have enough points along some of the boundary areas to get an interpolated value. I believe the ESMF bilinear and patch regridding requires a full four points (a square) of data in order to do an interpolation.
>
> In order to illustrate this better, I modified your script so that the lat/lon grid centers are drawn as red and blue dots (missing and non-missing locations). If you look at the plot with the "1deg" title, you will see that any place along the coast of Australia where you don't have a point near the coast, it's because you don't have four valid points for the interpolation.
>
> In the 0.5 degree grid, you have more resolution, and hence you do get more interpolated values by the coast.
>
> --Mary
>
>
>
>
> On Jul 8, 2013, at 11:00 PM, Jiaxin Zhang <jiaxinzhang3@gmail.com> wrote:
>
>> Dear NCL users,
>>
>> I am using ESMF_regrid to perform regridding of temperature. Thanks for your help Mary, now ESMF works.
>>
>> The main problem that I have is that after running ESMF_regrid to regrid temperature from 0.5x0.5 to 1x1, and trying two different methods (“bilinear” and “patch”), there are several grids in the land-water boundaries that have no interpolated temperature value. Does anyone know why this happens? What should I do to get regridded temperature values for the land-water boundaries?
>> I would need to have interpolated temperature values for all the land grids (including the ones in the boundaries with water).
>>
>> Attached are the original netcdf and the code. The attached code also creates a plot of the original file and the regridded file. If you compare the plot of the original file with the plot of the interpolated file (either using “bilinear” or “patch”) seems to have several grids without interpolated values, in contrast to the original file. I even tried to use regridding with weights but the problem seems to continue.
>>
>> I started using “linint2_Wrap” and have the same problem, so I was recommended to use ESMF instead, but I still do not know how to solve this problem. What do I need to modify in the attached NCL code? Is it better to use “Opt@SrcMask2D = where(.not.ismissing(temp),1,0)” or maybe a different file as a mask instead?
>>
>> I guess there must be an error somewhere in the code or a different way to solve this problem. If anyone could please let me know.
>>
>> Thank you!
>>
>> Jiaxin
>> <tair_2010_initial_1time.nc><test_regrid.ncl>_______________________________________________
>> 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

_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk