projection dimension coordinates to decimal lat/long

[rajkumar8581]

Hello friends, I have an xarray dataset with projection dimension coordinates (meters in x and y directions from the projection origin in the grid mapping space). Could anyone help me with regridding to decimal lat/long.

(y)
float64
2.165e+06 2.195e+06 ... 4.325e+06
units :
m
standard_name :
projection_y_coordinate
axis :
Y
array([2164827.903951, 2194827.903951, 2224827.903951, 2254827.903951,
2284827.903951, 2314827.903951, 2344827.903951, 2374827.903951,
2404827.903951, 2434827.903951, 2464827.903951, 2494827.903951,
2524827.903951, 2554827.903951, 2584827.903951, 2614827.903951,

[aulemahal]

Hi @rajkumar8581,
We would be happy to help, but I think we need more information.

What's your question exactly? What are you trying to do?

xESMF works well for transforming data between different grids, but it is all based on decimal lat-lon information. If your dataset does not have any information about latitude and longitude, then you won't be able to use xESMF. You would first need to translate the x, y coordinates in lat(x, y) and lon(x,y) maps. Depending on what information you have, this could be done with pyproj.

[rajkumar8581]

Thanks for your response @aulemahal, I am basically working with WRF model output files (staggered grids) with coordinates as follows:-
Goal is to simply regrid with decimal lat/long and visualize the variables.

[aulemahal]

I see. So here the issue when using xESMF might be that there are multiple lat/lon coordinates, so you must help it find the correct one.

XLAT, XLONG, XLAT_U, XLONG_U, XLAT_V and XLONG_V are all decimal latitude and longitude coordinates.

Let's say your variable is called 'UVELand has dimensions(y, x_stag), this means the correct lat/lon coordinates to associate it to are XLAT_U, XLONG_U`. The simplest way I can see here, without knowing anything about your data, would be to do something like:

import xesmf as xe

 # This creates a dataset with only 1 data variable and only the coordinates that relate to it
ds_in = ds.UVEL.to_dataset()
# Rename the lat/lon coordinates to lat/lon, so that xESMF knows to use them.
ds_in = ds_in.rename(XLAT_U='lat', XLONG_U='lon')

regridder = xe.Regridder(ds_in, grid_out, method='bilinear')
ds_out = regridder(ds_in)

Here, grid_out is another dataset with the 1D lat and lon you want.

[rajkumar8581]

Hi again Pascal, thanks for the sample code. But I get a NameError for regridder.

regridder = xe.Regridder(ds_in, grid_out, method='bilinear')

NameError Traceback (most recent call last)
Cell In [12], line 1
----> 1 regridder = xe.Regridder(ds_in, grid_out, method='bilinear')
2 ds_out = regridder(ds_in)

NameError: name 'grid_out' is not defined

Secondly, would you be willing to take a look at my data here:-
https://actvet-my.sharepoint.com/:f:/g/personal/rajkumar_sivaprahasam_adpoly_ac_ae/EofPOyhq8UhPlq3VBrvqLJABVR0G8xj_y0gmKRgLW1wagA?e=4hfZuT

[aulemahal]

The NameError: name 'grid_out' is not defined might be because grid_out is not defined.
I think you could read the documentation for xESMF to learn how to use it. For example this page : https://pangeo-xesmf.readthedocs.io/en/latest/notebooks/Curvilinear_grid.html

Good luck!

[rajkumar8581]

Got it @aulemahal Thank you. May I request you to visit xWRF and collaborate if it interests you.

[raphaeldussin]

to convert between x,y and lon, lat you can use the pyproj package with the relevant "PROJ" string of your Coordinate Reference System. SInce this is not related to xesmf, can I close the topic?

[rajkumar8581]

@raphaeldussin thank you. I was referred to esmf for a possible solution. The dataset has decimal lat long as a staggered grid. I need to cull out 1D lat long from it. @aulemahal