Standardise Metric Naming

open-AIMS / ADRIA.jl

ADRIA: Adaptive Dynamic Reef Intervention Algorithms. A multi-criteria decision support platform for informing reef restoration and adaptation interventions.

MIT License

14 stars 5 forks source link

Standardise Metric Naming #743

Open DanTanAtAims opened 5 months ago

DanTanAtAims commented 5 months ago

Coral Cover metric naming should be standard across ADRIA.

absolute_cover: coral cover in absolute units, e.g., m^2
relative_cover: coral cover relative to habitable area. Ideally in [0, 1]
total_cover: coral cover relative to total reef area including uninhabitable. Ideally in [0, 1]

Interfaces loading external data sets should make sure the naming is assigned to the correct data and that any unit conversions are applied.

ConnectedSystems commented 5 months ago

Note for everyone: this is associated with #465, and could be part of the refactor of all metrics to leverage YAXArrays/metadata (#524).

Zapiano commented 4 months ago

I think there might be a confusion when using "absolute" and "total". The word "relative" gives a hint that the cover will be a fraction of something, but it doesn't say "relative to what". Since we have cover relative to the habitable area and relative to the total area, what do you think about relative_habitable_cover and relative_total_cover? Too much?

arlowhite commented 4 months ago

I agree with including "habitable" in the name. relative_habitable_cover 👍 maybe "all" instead of "total"? so relative_all_cover or what about relative_reef_cover?

VHallerBull commented 2 weeks ago

I agree with Arlo, both would be considered relative just to a different reference.

The other part to consider is resolution, from my use of adria so far it seems to use total_cover when there is one value per scenario while the other two are used for one value per location (so hundreds of timeseries per scenario). So, the question here is, do we use the same names independent of the resolution of should we have differences i.e. loc_relative_cover and relative_cover would be the same metric but at the two different resolutions

ConnectedSystems commented 2 weeks ago

so far it seems to use total_cover when there is one value per scenario

You should be able to use any metric summarised by scenario or location. Less supported are the functional group-based metrics.

ConnectedSystems commented 2 weeks ago

Further to the above, the metrics are documented (although how it's presented needs work)

https://open-aims.github.io/ADRIA.jl/dev/API/#Metrics