## Plotting comments
### Forecast animations
* Make individual frame PNGs same orientation and colourmap as animation
* Don't use blues_r colourmap - impossible to distinguish different bins which may be very important operationally
* Reduce fps - March lasts 1 second! Maybe fps=1?
* Increase figsize so grid cells are more visible upon zooming (would also reveal `forecast_date` in title of the pngs)
* Replace contourf with standard NaN land mask (with `set_bad` to grey). Try adding `ax.coastlines` to plots
* Might be finicky with the rotation we do - maybe `cartopy.crs.LambertAzimuthalEqualArea(central_longitude=-90, central_latitude=90)` would work?!

### Historical forecast animiation
* Don't use Reds colourmap - hard to compare - use same colourmap we settle on from above
* plt.subplots_adjust(bottom=0.2) to reveal the colourbar ticks that are cropped off

## Plotting ensemble spread (stddev)
* use `Greys` colourmap to represent uncertainty
* maybe keep just to use initially and send to Am/Jonathan if we think it looks good

## Gradients
* Encompassing error by implication
* Threshold plots would not encompass errors, so probably nice

## Uncertainty

## General
* Get the coastlines into the plots if possible
* Color maps and normalisation - CLI args