Review of cpd-11-4003-2015

I feel this manuscript provides strong rationale and very helpful description of the PlioMIP2 experiment. I recommend that it is published subject to some revisions. I look forward to the actual experiment and hope that some interesting science will emerge from it. Below I’m suggesting some big revisions to the ensemble of simulations requested. I’m happy with the authors directly about whether these revisions truly represent better value for resources.

The vast majority of the simulations are required solely for the forcing factorisation. You might want to consider just important you feel this component of the research is. I worry that the amount of simulations required really justify the extra effort. They need 6x as much computation as just doing the PlioMIP2 entry card, but surely gain nothing like as much as six times the information (considering the fact all CMIP6 models must do the DECK, I’m not counting the preindustrial run). You may want to think of the factorisation as a sub-experiment, otherwise PlioMIP2 appears really daunting.


I did have one question about the scope of the manuscript. It wasn’t clear to whether it aims to serve to just as an experimental description, or will also act as the full description of the boundary condition datasets. I know that the previous experimental design (Haywood 2011) was complemented by a data description paper (Dowsett 2010). Whilst I think that most of datasets are adequately discussed in this manuscript (or prior publications), the topography feels under described. I hope that a separate manuscript is planned to describe all the underlying assumptions for this dataset and highlight the important changes. I would certainly like to see more discussion of the uncertainty inherent in the topography reconstruction. For example, a major change from PRISM3 is the closing of the Bering Strait. I remember seeing a poster at AGU 2014 by Dick Peltier presenting an alternate topography with it closed - this also included the novel scientific components described here. Whilst I’m happy with the reconstruction you present here, I don’t feel there is any acknowledgment that it may have uncertainties.

It isn’t clear to me how the ice sheets and topography are actually separated in the factorisation approach. I think you need to provide guidance in the manuscript. Does imposing ice-sheets also contain the topographic element associated with the ice-sheet or does that count in the topography? At its simplistic this could be ice-sheets could be thought of as white mountains, so the ‘i’ component only relates to the land surface specification. That doesn’t make much sense intuitively. The problem is however much more complicated than that, as you’ve taken account of the glacial isostatic adjustment in the your topographic reconstruction. At this point, I’m not sure it’s that important (it would need to be discussed seriously in the factorisation results). Here you do need to provide instructions to allow the runs to be performed.

I was also confused by the discussion of the standard experiment. Surely altering the Bering Strait is a change in the Land-Sea mask. If modelling groups have the ability to do this change (and in my experience making new land is more awkward than new ocean), then shouldn’t they be doing the other experiment. Incidentally, I would anticipate that this change is important for the AMOC, so well worth including if possible.


My other major worry relates to the quantity of simulations requested by the full experimental design. Whilst, I understand the justification of most of them from a rigorous scientific standpoint, I wonder if they will be tackled by sufficient groups to allow a model intercomparison. I also feel that most of the justification is for the Pliocene factoralisation experiments, and insufficient discussion is made of the CMIP DECK.

It is never mentioned that E\(^{280}\) (the preindustrial control run) must be performed as part of the DECK for CMIP6. As an outsider, I may read this paper and be really afraid of joining PlioMIP2 because of the quantity of simulations required. I suggest it is worth emphasizing that you can join (and be an important member of PlioMIP2) just be performing a single run (Eoi\(^{400}\)). I don’t know how many groups have signed up, but I’d expect most of them to only tackle that core simulation.

I was rather confused by the quantity of simulations requested for the Pliocene for Future side of Figure 2. There doesn’t seem to be much joined up thinking between this element and the wider picture of CMIP6. The purpose of the Tier 2 experiments E\(^{560}\) and Eoi\(^{560}\) appears to be to allow the Charney climate sensitivity to be calculated (Section 3.1.1). The DECK already involves a simulation specifically to calculate this metric - an abrupt quadrupling of CO\(_2\) - using a technique devised by (Gregory 2004). This approach only needs 150 years of computation rather than the minimum of 500 years proposed here. The DECK also involves standard transient forcing simulation where CO\(_2\) concentrations increase by 1% per year. Current approaches infer knowledge from past climates to constrain future projections involve either subsetting (i.e. discounting bad models) or using emergent constraints. Both of these approaches work based on just the CORE simulation and the simulations performed in the DECK. I therefore question whether any of the Pliocene for Future runs can be justified from that perspective.

The question of state dependence of climate sensitivity is an rather interesting one. The most efficient way to tackle it would be perform the Abrupt 4xCO2 experiment feature in the DECK, but using a Pliocene base state (perhaps denoted as Eio\(^{400}_{4xCO_2}\)). This would effectively replace Eio\(^{560}\), whilst E\(^{560}\) gains little over and above the E\(^{280}_{4xCO_2}\) simulation already in the DECK.

I don’t know whether there is any benefit to replicating the 1% simulation from a Pliocene base state, as I’m not sure people have previously investigated whether the ocean heat uptake efficiency is state dependent. It’s an interesting experiment to consider for some explorative science, but you’d probably need to do it in one model first to justify its inclusion.

There is an alternative potential justification for both E\(^{400}\) and E\(^{560}\), which is that they version of a stabilisation scenario. I didn’t immediately spot any runs in ScenarioMIP looking at determining what our target CO\(_2\) should be on the longer term. Nonetheless, I’d hestitate to include them in PlioMIP2 with that solely justification and feel they would sit better elsewhere. It would appear that you need E\(^{400}\) for the factorisation however. If you include that run, please talk to some folks involved in ScenarioMIP and try to get it used from both perspectives.

Despite all these negative comments about the scenario choices, you’re surely along the correct lines. You may want to instead have Eoi\(^{350}\) and Eoi\(^{450}\) as a Tier 1, that spans both the past and future elements, as I feel the CO\(_2\) uncertainty is pretty important to examine for both. This would mean that the Pliocene for Future section consists solely of Eio\(^{400}_{4xCO_2}\).

My recommended groupings would look like:

Expt Name Status
E\(^{280}\) CMIP6 DECK
E\(^{280}_{4xCO_2}\) CMIP6 DECK
Eoi\(^{400}\) CORE
Eoi\(^{350}\) Tier 1
Eoi\(^{450}\) Tier 1
Eoi\(^{400}_{4xCO_2}\) Future
E\(^{400}\) Linear Fact.
Eo\(^{400}\) Linear Fact.
Eo\(^{280}\) Non-linear Fact.
Eoi\(^{280}\) Non-linear Fact.
Ei\(^{400}\) Non-linear Fact.
Ei\(^{280}\) Non-linear Fact.
Eio\(^{280}\) Non-linear Fact.