Separate domain for satellite image and nowcast output

[KristianHMoller]

Currently, the nowcast is provided on the same domain as the satellite image. This results in odd effects where the NaN values from outside the domain of the satellite image are blown into the domain of interest.

This PR allows for separation of the domains for satellite images and nowcast output, so that a larger domain can be used for the satellite image.

Based on back-of-the-envelope calculations assuming maximum (relevant) cloud velocities of 100 km/h, the domain NW_EUROPE_SATELLITE has been introduced as an example of a domain which can be used for producing nowcasts on the NW_EUROPE domain. The new domain extends 10 degrees outside of NW_EUROPE to all sides.

This PR does not change the default behavior, so prior to validation and active choice of using a different domain for the satellite image, this does not change the operational output. This simply allows for the option of changing.

[KristianHMoller]

@irenelivia you are our cloud expert - does the additional extent of the NW_EUROPE_SATELLITE domain sound reasonable?

[KristianHMoller]

Comparison showing the output on the NW_EUROPE domain after 5 hours of forecast time. Left uses the larger NW_EUROPE_SATELLITE domain for the satellite image and right uses the same domain for the satellite image and nowcast.

[irenelivia]

@KristianHMoller great work! My only consideration is that the nowcast ultimately changes if we use a larger domain (or a smaller domain, for that matter). You can see it in the case you showed - the advected clouds don't have the same exact shape in the two nowcasts run with different domains. So, we are ultimately providing a new model that we have not validated, and before making this kind of change, we should ideally "prove" that it improves the scores (or, at least, does not worsen them) over Denmark.

[KristianHMoller]

Will need updating to be compatible with this (much needed) functionality: #11

[KristianHMoller]

Without explicit specification of --domain_satellite or --custom_domain_satellite, the output is identical to the current version of the code, both when downloading and reading from files. This PR simply provides the option to have these be different.

[KristianHMoller]

CHANGELOG needs updating

[KristianHMoller]

@elbdmi do you think that it would make sense for you to review this from a technical point of view? Whether or not we should activate this operationally will require separate validation, possibly prior to next tagging

[elbdmi]

@elbdmi do you think that it would make sense for you to review this from a technical point of view? Whether or not we should activate this operationally will require separate validation, possibly prior to next tagging

I agree about the validation. Yes I think it makes sense to take a look from the technical point of view.

[elbdmi]

Looks good to me from a technical point of view. I only have a few comments/ questions. The small typo in the README should be fixed. The other comments are just for me to understand the functionality.

[elbdmi]

LGTM now. I have no further comments from my side.