Why can't I use cartopy to plot certain time averages of the same dataset?

Question

I have a 3-dimensional xarray DataArray of changes in surface temperature with coordinates of time, lat and lon. I am visualizing the data using Cartopy. You can find the 125 MB file here.

While producing plots of time-averages over different periods, I've found that I'm unable to produce orthographic projections when including certain time steps, such as the 132nd (index 131) time. Here is a plot of the time average from 0 to 130:

But this happens when I instead perform the time average from 0 to 131:

Here is the code I used to produce the plots:

# import statements

import cartopy.crs as ccrs
import xarray as xr
import numpy as np
import matplotlib.pyplot as plt
from cartopy.util import add_cyclic_point
%matplotlib inline
%config InlineBackend.figure_format = "jpg"

# read in data
ens_mean = xr.open_dataarray('temp_changes_ens_mean.nc')

# time average subset of data
to_plot = ens_mean.isel(time=slice(None,131)).mean(dim='time') # change 130 to 131 to break cartopy

# add cyclic point to avoid white lines
data = to_plot
lon = to_plot.coords['lon']

lon_idx = data.dims.index('lon')
wrap_data, wrap_lon = add_cyclic_point(data.values, coord=lon, axis=lon_idx)


# make an orthographic plot centered on north pole

fig = plt.figure(figsize=(4.5,3.5))
ax = fig.add_subplot(1, 1, 1, projection=ccrs.Orthographic(0, 90))

ax.coastlines()

im = ax.contourf(wrap_lon, to_plot.lat, wrap_data,
                transform=ccrs.PlateCarree())

# add colorbar
cb = fig.colorbar(im,orientation='horizontal',shrink=0.5,pad=0.05)

cb.ax.tick_params(labelsize=8)
cb.set_label('ΔSAT (K)',fontsize=8)

plt.tight_layout(w_pad=0.05)
plt.show()

This occurs whether I add a cyclic point or not. I am able to make quick plots of the data using matplotlib or xarray's built-in plotting without error. I've already checked for NaN values in the data. Lastly, if I remove the transform argument in the contourf line, it is able to produce a coherent plot, which leads me to think it is the transformation step that produces this odd plot.

Thanks for the help!

Answer 1

You can use ax.set_global() method to reset the coordinate limits:

#!/usr/bin/env ipython
# --------------------------------------------
import cartopy.crs as ccrs
import xarray as xr
import numpy as np
import matplotlib.pyplot as plt
from cartopy.util import add_cyclic_point
# --------------------------------------------------------------------------------------
#%matplotlib inline
#%config InlineBackend.figure_format = "jpg"

# read in data
ens_mean = xr.open_dataarray('temp_changes_ens_mean.nc')

# time average subset of data
to_plot = ens_mean.isel(time=slice(None,131)).mean(dim='time') # change 130 to 131 to break cartopy

# add cyclic point to avoid white lines
data = to_plot
lon = to_plot.coords['lon']

lon_idx = data.dims.index('lon')
wrap_data, wrap_lon = add_cyclic_point(data.values, coord=lon, axis=lon_idx)
# ------------------------------------------------------------------
# this is not working:
xlims = (np.min(ens_mean['lon']),np.max(ens_mean['lon'])); 
ylims = (np.min(ens_mean['lat']),np.max(ens_mean['lat']));
# ------------------------------------------------------------------
lon = to_plot.coords['lon']
# ====================================================================================
# make an orthographic plot centered on north pole
# Let us make a working/satisfying plot:

fig = plt.figure(figsize=(4.5,3.5))
ax = fig.add_subplot(1, 1, 1, projection=ccrs.Orthographic(0, 90))

ax.coastlines()

im = ax.contourf(wrap_lon, to_plot.lat, wrap_data,
                transform=ccrs.PlateCarree())

# -----------------------------------------------------------
# add colorbar
cb = fig.colorbar(im,orientation='horizontal',shrink=0.5,pad=0.05)

cb.ax.tick_params(labelsize=8)
cb.set_label('ΔSAT (K)',fontsize=8)

plt.tight_layout(w_pad=0.05)

ax.set_global();
#ax.set_xlim(xlims);
#ax.set_ylim(ylims);

plt.show()