How to Report Uncertainty Chapter

[kenkehoe]

Fixes: #320

See issue #320 for discussion of these changes.

Release checklist

• Authors updated in cf-conventions.adoc?
• Next version in cf-conventions.adoc up to date? Versioning inspired by SemVer.
• history.adoc up to date?
• Conformance document up-to-date?

For maintainers

After the merge remember to delete the source branch.
Tags are set at the conclusion of the annual meeting; until then master always is a draft for the next version.

[ngalbraith]

This looks very good, I appreciate all the work that went into it. I DO have a question (or 3) about how this could be applied to vectors (e.g. northward/eastward_sea_water_velocity or northward/eastward_wind), where there are uncertainties in both speed and direction, and the units for the directional uncertainty won't match the units for the vector. Unless I missed it, the proposal doesn't require that the units match, but I'd just like to double check that that wasn't assumed.

Also, can I include uncertainty as a percentage of a measurement?

The document doesn't spell out (again, maybe this is assumed) that a single set of uncertainty variables can be applied to more than one data variables, in this case both eastward and northward vectors. Is there any problem with both northward_ and eastward_sea_water_velocity sharing an ancillary variable (or a group of them) for this purpose?

Thanks!

[kenkehoe]

Great questions. Let me answer them in reverse order.

• There is no one-to-one assumption between variables. Multiple data variables can use a single uncertainty variable and multiple uncertainty variables can be linked to a single or multiple data variables.
• The units for the uncertainty do not need to match the data variables. If the units attribute is listed then the uncertainty is in those units, else if no units attribute then the units match the data variable. The data provider would need to ensure units are clearly understood.
• I think your question about velocity vs speed & direction is dependent on how the uncertainty is presented. If the uncertainty is a velocity then the units should (would?) match the data velocity. But if there is need to present the uncertainty of speed and direction independently that is possible. For those the units attribute must be defined.
• I see no reason why the uncertainty could not be a percentage of a measurement as long as it's clear how to use that uncertainty description. I would see providing the uncertainty as a percentage as requiring a formula to convert from percentage to data units. So there would need to be a formula attribute to describe how to convert the values to data units.

Does this help?

[ngalbraith]

Great questions. Let me answer them in reverse order.
...
Does this help?

Yes, thanks very much, this answers all of my querstions!

[kenkehoe]

It may be difficult to read the standard without running ASCIIdoc. To help reviewers here is the PDF of the pull request.
cf-conventions.pdf