Strange behavior with data from grib file

[JFMass]

Hi

I've been using matplotlib and basemaps with pygrib to plot meteorological data from grib files.

However, when using a particular grib, I get some strange behavior where a large horizontal strip of value 80 appears on the map. Digging into the problem I can tell the problem is not related to the data inside the grib file because grads plots perfectly. I could can also confirm that the problem doesn't seem to come from pyplot either because checking individual grid values show normal value where the map shows 80. I could identify that the plotted value is near 80 by playing with levels.

The only thing left would be with the contour and contourf functions.

The data I am using 10km polar steregraphic projection rainfall data. Interestingly, 15km resolution files plot perfectly without any strip into the image.

Anyone can tell what is wrong?

#Retrieve gridded data from pygrib
grbs = pygrib.open('CMC_RDPA_APCP-006-final_SFC_0_ps10km_2015061906_000.grib2')
handle = grbs[1]
pltvar = handle.values
lat, lon = handle.latlons()

#troubleshooting to confirm data in grib is good where map show max values
print lat[365,765], lon[365,765], pltvar[365,765]
#46.0384291906 -70.7820353751 7.7229
#just as expected

#plot them with mpl/basemap
fig = plt.figure(figsize=(4.5,2.5),dpi=1024,frameon=False)
m=Basemap(epsg=3857,llcrnrlon=-80.15625,llcrnrlat=42.032974332441405,urcrnrlon=-54.84375,urcrnrlat=51.6180165487737,resolution='i')
x, y = m(lon, lat)
m.drawcountries()
m.drawcoastlines()
cs = m.contourf(x,y,pltvar,levels=(2,5,10,15,20,30,40,50,75,100))
plt.colorbar()
plt.savefig('preciptest.png',dpi=1024,transparent=True)
plt.close()

[tacaswell]

Do the axes limits matter at all? There are some known bugs with some projections.

attn @pelson @WeatherGod @ianthomas23

[pelson]

It's hard to see with the map we have, but this is an exceedingly common problem with plotting geometries which touch/cross the dateline and/or poles.

Are you getting any warnings with the contourf call? I'm expecting you to get something like a non-monotonic warning?

The general solution to this is not to work in lat/lons - doing so can result in a loss of coordinate topology (due to the modulus 360). In practice, I don't know how well basemap deals with that. @WeatherGod might be able to help with that.

This kind of problem is the bread and butter of a package called cartopy (which I originally wrote), which deals with the geometry/contour topology problems that basemap's approach cannot. It might be also worth looking at that as an option.

Either way, this is not really a matplotlib issue, more of a https://github.com/matplotlib/basemap issue.

[efiring]

@JFMass, Try plotting without basemap: just make a simple call to contourf with lon and lat as your x and y. Check min and max for lon and lat; are the lons all negative, consistent with your choice of negative corner longitudes? Are the lons and lats monotonic? Have you made a simple plot of lon versus index?
In any case, I agree with @pelson that this is not a matplotlib problem, and I suspect it is fundamentally not a basemap problem either; there is something about your data that doesn't match your expectations, or basemap's requirements. It is almost certainly related to the longitude grid.

[JFMass]

Thanks for the help, couln'd solve the bug but I could work around by regridding with wgrib and then opening new grib.

Problem solved that way.