Get longitude and latitude of grid points in cube

[krikru]

I have a netCDF file, A, in the Lambert conformal conic projection, so naturally it doesn't have any dimensions with the coordinates longitude and latitude. However, I need to extract the ranges for the longitude and latitude coordinates of the grid points, because I want to crop another netCDF file, B, which is stored on a regular grid, to a rectangular region (still on a regular grid) that is as small as possible while still covering the grid points in A completely.

Is there any way I can get the longitudes and latitudes for all points in A after loading it into an Iris cube? Or get the minimum and maximum values for the longitude and latitude coordinates of the points?

[krikru]

Considering that Iris can convert between different coordinate systems when regridding a cube, and that I can get the latitude and longitude when hovering the mouse cursor over a point on the map when plotting a slice of the cube using iris.quickplot.pcolormesh, surely, doing what I request here should be no problem, right?

[trexfeathers]

Hi @kriku, this is doable but actually quite complex. You'll need to use cartopy.crs.CRS.transform_points(), which means first getting the Cartopy CRS from the Cube's coordinate system. Have you looked into this route yet?

[krikru]

@trexfeathers Hm, I seem to be running into problems. I get

AttributeError: 'PlateCarree' object has no attribute 'as_cartopy_crs'

So, do you have any idea how to get the CRS from PlateCarree() or the CRS from a cube that doesn't have any coordinate system?

[krikru]

I see that the Projection class has a method called source_crs; I will try that.

[trexfeathers]

@trexfeathers Hm, I seem to be running into problems. I get

AttributeError: 'PlateCarree' object has no attribute 'as_cartopy_crs'

So, do you have any idea how to get the CRS from PlateCarree() or the CRS from a cube that doesn't have any coordinate system?

The PlateCarree you have imported already is a Cartopy CRS, so you do not need any 'conversion'.

[krikru]

Thanks! I think I have succeeded now to extract the latitudes and longitudes for all grid points in the cube that is on a Lambert conformal conic projection.

However, I see that PlateCaree() have the coordinates Easting, Northing and Ellipsoidal height (so that's what I get out of transform_points), while the cube that is on a regular grid, for which coord_system() returns None, has the coordinates latitude and longitude, which firstly don't match when considering the number of coordinates, and secondly have opposite orders of the horizontal coordinates (latitude and longitude corresponds to Northing and Easting, respectively, and not to Easting and Northing), but it still plots correctly when calling iris.quickplot.pcolormesh. Do you know how that works? I need to map latitude and longitude to Northing and Easting, respectively, but am a bit reluctant to hardcoding this mapping into my code. It seems like Iris already must do this mapping somehow.

[trexfeathers]

Do you know how that works?

Unfortunately my knowledge is limited when it comes to different coordinate systems and projections, sorry.

You could have a look at the full range of Iris coordinate systems and see whether any of them seem appropriate to you.

Alternatively look directly at the range of Cartopy CRS's. Full detail on this is available in the PROJ documentation.

Hopefully these resources help you find what you need.

[trexfeathers]

#5032 (reply in thread)

I am not keen for Iris to be making more decisions about the 'correct' coordinate system to use when one is missing; it is important that the user knows what is being done, and therefore it is important for the user to make the decision about the coordinate system. That will be why it is not auto populated when a Cube is loaded. Indeed I'm a little uncomfortable that Iris makes that default decision when plotting, but I don't know the history behind this choice.

Of course that's my opinion. @SciTools/iris-devs might think differently.