Odd DCCP variables in Eyr ? (DR 1.00.16)

[jypeter]

I'm wandering and losing myself in the DR, while checking PMIP related stuff...

I have come across some variables requested only by DCCP that look odd. There is a bunch of variables in Eyr, from treeFrac to residualFrac:

• that don't have time as their last dimension: e.g. dimensions for treeFrac is longitude latitude time typetree
• where the extra dimension is of type character, which is not always very convenient to use, and the allowed/expected values are not listed in the DR. e.g, nothing in the requested part of typetree

Is this correct??? @durack1 ?

I have had a very quick look at the DCPP paper (@taylor13 is a co-author) but I have not found obvious extra details. I have also asked LSCE vegetation people but they did not know about these variables. The closest thing they new about was gppLut that has a landUse dimension with 4 requested character values (primary_and_secondary_land pastures crops urban)

[martinjuckes]

In the data request database record defining the variable structure Temporal mean, Global field (single level) [XY-na] {:typetree} it is clear that typetree is a coordinate rather than a dimension, and in the MIP tables follow the CMIP5 convention in having coordinates listed after the time dimension. Every variable with standard name area_fraction must, according to the CF convention have a coordinate attribute which points to a variable identifying the area type . See the Data Request page on area coverage for more discussion.

[taylor13]

Actually, I think area_fraction has no special requirements (at least I couldn't find any mention of it). A variable with standard name "area_type", on the other hand, does require a coordinate defining the "type".

In CMIP5 there was no scalar coordinate dimension defined for variables like treeFrac, and I don't think it is essential, since it is obvious from the variable name and long_name. If it is included, then this will expand the list of coordinates that need to be defined (one for each area_type that is singled out ... order 10, I think)

[martinjuckes]

The CF Standard name description for area_fraction states To specify which area is quantified by a variable of "area_fraction", provide a coordinate variable or scalar coordinate variable of "area_type", which could be read as an instruction or guidance. Ticket https://cf-trac.llnl.gov/trac/ticket/153 tries to sort out the difference between requirements and suggestions, and currently identifies the use of an area_type as required ... but is hanging in limbo. I have, however, used the current status of this ticket to systematically analyse the compliance of the request with requirements which relate to specific standard names. In the CMIP6 request the majority of area_fraction variables do make use of an area_type coordinate.

[taylor13]

Thanks for pointing to proposal 153. If what you propose can be implemented in a way that avoids making CMIP5 files non-compliant (with respect to this issue), then I support it. I can see the value of defining a coordinate dimension to identify "type" using controlled vocabularies.

Does the current area_type table ( http://cfconventions.org/Data/area-type-table/6/build/area-type-table.html) contain all the options needed for CMIP6?

[jypeter]

OK. Thanks for the details!

[martinjuckes]

For information on non-conformant variables see here: https://www.earthsystemcog.org/projects/wip/area_coverage

[martinjuckes]

For information on non-conformant variables see here: https://www.earthsystemcog.org/projects/wip/area_coverage