Recommendation of forcing indicator for ozone forcing v2.0

[mo-abodas]

Given the recent update of the historical ozone forcing (v2.0) following #400, I wonder if the CMIP panel recommends the use of a different forcing indicator for historical and amip simulations with respect to the one used for other simulations that will be using v1.2 (e.g. piControl, abrupt-4xCO2). For example, piControl with ensemble indicator r1i1p1f1 using v1.2 and historical simulation with ensemble indicator r1i1p1f2 using v2.0 of the ozone forcing (instead of f1 for both simulations). Thanks!

[znichollscr]

Great question. There isn't a settled answer yet. However, I think that's ok, no one is publishing data yet so we have a bit of time to think.

Here are some thoughts for discussion

Your question makes a good point: the forcing is actually experiment specific. For example, the forcing used for an amip simulation is by definition different to the forcing used for a historical simulation (one needs SSTs, the other doesn't). So, first proposal: each 'f' number only applies to the simulation it is used for. I suggest we don't try and make 'f' numbers unique across simulations because that would be too big a change from CMIP6 (where groups more or less just always started from f1 for all simulations, even if the forcings used weren't actually identical) and we'd end up with too many 'f' numbers (maybe?).

I suggest we also don't try and identify an 'f' number for each collection of forcings. This was the idea in CMIP6, but I don't think it works because a) people don't check what collection they have actually used and what number they should accordingly use (at least that is the evidence from CMIP6) and b) if you make a mistake and apply the wrong 'f' number, the only way to fix it is to republish all your files, which is a crazy price to pay and c) some groups use their own custom forcings and these would never have an f number assigned.

I would suggest that, as step one, we ask modelling centres to record the source id of all the forcings they have used for each simulation internally using whatever system they want. Then, if two simulations of the same experiment use different forcing, then they should have different f numbers. However, the actual f number doesn't matter: groups can start at f1, then f2 then f3, or start at f10, then f15, then f20, whatever. As step two, the infrastructure team will set up a way for groups to record the mapping between f number and the forcings that were used for each simulation in a machine readable way. Then modelling teams can submit their information, before finally downstream users can use this machine readable information to sort, filter and group simulations however they want (only using latest forcings, only using v2 of ozone etc etc).

The benefit of this system are: a) modelling centres can start writing data now if they want (just pick an f number, record what it means and off you go) b) if you need to update the mapping between f number and forcings (because you realise you made a mistake the first time), you can do that without having to rewrite or republish any data and c) groups can use internal forcings without breaking things.

So, in your concrete case:

• do picontrol simulations. Pick an f number. Write down that you used v1.2 of ozone (and version X of GHGs and Y of emissions etc)
• do historical simulations. Pick an f number (can be same as picontrol or not). Write down that, for this f number, you used v2 of ozone (and whatever versions of other forcings you used too)
• if you do historical simulations with v1.2 of ozone too, use a different f number (from the one used for the step above e.g. if you used f1 for the simulations that used v2 of ozone, label these ones something other than f1 e.g. f2). Write down that, for this f number, you used v1.2 of ozone (and whatever versions of other forcings too)
• when the infrastructure team comes and asks you for your mapping between f number and forcings versions, don't ignore us :)

[mo-abodas]

@znichollscr thanks for your reply, that makes sense, and it coincides with our preferred approach.

[taylor13]

I fully endorse @znichollscr proposed method for assigning the "f" numbers. Other options I can think of will more likely lead to mistakes that aren't easily corrected. I think the model documentation task team is starting to consider "experiment" documentation (having mostly finished the EMD). Hopefully, they will find a way to collect and record the modeling groups' reported definitions of (what forcings they used for) each "f" value for each experiment.