Adding population_density as standard_name

[dompap]

Topic for discussion

Dear All,
in CMIP7 one of the forcing data variables is population density (https://input4mips-cvs.readthedocs.io/en/latest/dataset-overviews/population/) which does not feature in CF. Therefore, the CMIP Climate Forcings Task Team (https://wcrp-cmip.org/cmip7-task-teams/forcings/) has been discussing the value of having a standard_name for this variable (possibly "population_density"). This is a socio-economic rather than a physics variable, but is no less necessary for many Earth system models.

At the moment, we define this as follows:

float pop_dens(time, lat, lon)
pop_dens:long_name = "population_density_per_square_km" ;
pop_dens:units = "1" ;
pop_dens:description = "Estimated population (mid-year) per square kilometer of grid cell area. The value was calculated
by dividing population by the whole grid cell area, i.e. both land and water parts of the grid cell."

One point of discussion is also how to write the unit, which given that population density is per square kilometer, then I guess it should be "km-2" in connection to standard_name "population_density".

Thank you in advance for any feedback.

[JonathanGregory]

Dear Dominik @dompap

If this quantity is going to be included in CMIP7 it is likely that a standard name will be needed for it, since several models will require it. I don't think there is a conceptual difficulty about it. We already have a standard name for number_concentration_of_biological_taxon_in_sea_water in m-3, for example. Homo sapiens is also a biological taxon with geophysical influence, although mostly limited to two dimensions.

Since "density" usually means kg m-3, I think it would be better to use per_unit_area, which appears in many standard names, such as number_of_icebergs_per_unit_area. I suggest number_of_humans_per_unit_area. The canonical unit should be m-2 but it could be expressed in any equivalent unit, such as km-2. If you want to take it forward, it would be sensible to propose a standard name in a new vocabularies issue.

Best wishes

Jonathan

[efisher008]

Dear Dominik @dompap,

Many thanks to Jonathan @JonathanGregory for initiating the discussion. I agree with Jonathan's suggestion that this should be modelled on a biological taxon number concentration name, which makes logical sense. Density in CF terms does indeed imply a mass per volume, rather than a number per area, so number_of_humans_per_unit_area would be reasonable. In the CF Standard Name table, we have a phrase number_content which is taken to mean "a quantity [in this case, number of humans] per unit area". Perhaps this is more suitable for the use case, making the name number_content_of_humans_per_unit_area?

In terms of the unit, referring to the NVS P06 vocabulary which is used by CF, this would be either /m^2 or #/m^2. I am not certain which is more correct, although I would argue the second is more restrictive/specific of it being a "number" based quantity?

In any case, I think this would be well worth opening a new issue in the cf-vocabularies repo, which I will leave to you to do @dompap. We can carry on the discussion there, if necessary.

Best wishes,
Ellie

[dompap]

Dear Ellie and Jonathan @efisher008 @JonathanGregory
thank you for your feedback. I will open an issue in vocabularies on this.
"number_of_humans_per_unit_area" makes sense, with /km^2 as unit being preferred as it is the one everybody would expect.
"number_content_of_humans_per_unit_area" I think goes rather too far :)
Best regards,
Dominik

[efisher008]

@dompap I apologise if I've missed this or have misremembered, but did you end up opening an issue in the cf-vocabularies repo on this topic?

[dompap]

No, I didn't. Sorry, but the vacatation season intervened :)

[larsbarring]

If you want to follow the current CF Conventions the canonical units is "m-2" because it consists of SI units (without any prefixes), and is formatted following UDUNITS2 requirements:

$ udunits2 -A -H "m-2" -W ""
    m-2
$ udunits2 -A -H "# m-2" -W ""
udunits2: Don't recognize "# m-2"
$ udunits2 -A -H "/m2" -W ""
udunits2: Don't recognize "/m2"
$ udunits2 -A -H "#/m2" -W ""
udunits2: Don't recognize "#/m2"

For convenience here is the UDUNITS2 brief help:

$ udunits2 -h
Usage:
    udunits2 -h
    udunits2 [-A|-L|-U] [-r] [-H <have>] [-W <want>] [<XML_file>]

where:
    -A          Use ASCII encoding (default).
    -L          Use ISO-8859-1 (ISO Latin-1) encoding.
    -U          Use UTF-8 encoding.
    -h          Help.  Print this message.
    -r          Reveal any problems in the database.
    -H <have>   Use <have> unit for conversion. Default is reply to prompt.
    -W <want>   Use <want> unit for conversion. Empty string requests
                definition of <have> unit. Default is reply to prompt.
    <XML_file>  XML database file. Default is "/usr/share/xml/udunits/udunits2.xml".

The NVS P06 allows the "#" to more clearly indicate that it is a count (as does QUDT). In the "units definition communities" there is long-standing discussion on 'dimensionless units' or 'quantities of unit 1' to which count belongs, and CF has not yet adopted the NVS P06 and QUDT extensions. But that is for another time and place to discuss.

Finally, while I do understand that population density expressed as number of person per square meters seems less useful (if not outright ridiculous), CF uses the concept of canonical unit that allows users to choose for any equivalent unit, such as "square kilometers" in your case. But this should be handled by the CMIP7 data request process.