Reply to Referee 1. We thank the referee for their constructive feedback. Their comments are shown in bold with our responses in italics.

Overall evaluation:

The paper makes an argument for extending zero emissions simulations to millennial timescales. I strongly agree with the authors that having such simulations would be very useful for our science. However, millennial length simulations of zero emissions do exist for intermediate complexity Earth system models (EMICs), and were in fact part of the original ZECMIP (MacDougall et al, 2020). The authors need to better articulate why simulations with full ESMs are needed at millennial timescales to supplement the results from the EMICs.

Thanks. Yes, we previously hadn’t articulated well the specific need for long net zero ESM simulations relative to those from EMICs. We propose to edit the first paragraph (L17-23) in the main text to note the benefits of specifically using ESMs for long net zero simulations. We also propose to edit the third paragraph (L30-36) in the main text to note the prior use of EMICs in ZECMIP. In the sixth paragraph (L49-54) we would also add a note on prior use of EMICs and the analysis that has been conducted with them.

General Comments:

(1) The reason for using ESMs is given at line 70 of the paper "This is particularly problematic for understanding changes in interannual-to- multidecadal climate variability for which changes are only likely to be detectable over long periods. The same is true for multi-year extremes, such as multi-year droughts (Falster et al., 2024), for which limited information may be gained from shorter simulations." EMICs are unable to quantify such metrics, thus justifying the expense of using ESMs. I suggest articulating this point much earlier in the paper.

Indeed- we propose to edit the very first paragraph in the main text (L17-23) to highlight specific benefits of long ESM simulations, including analysis of extremes such as drought.

(2) Early in the paper you should acknowledge the millennial length simulations of zero emissions done with EMICs, summarize what they show then highlight the limitations of such EMIC simulations, thus highlighting the need for longer ESM simulations.

Yes, in the third paragraph (L30-36) that discusses ZECMIP we would add a sentence on prior use of EMICs as well as discussion in the sixth paragraph (L49-54) on previous findings using EMICs.

(3) It is also important to acknowledge that ESMs often lack processes the EMICs include, which become more important on long timescales. For millennial length simulations processes that are particularly important are: i) dynamic vegetation, ii) ice sheets, iii) permafrost carbon iv) ocean floor carbonate dissolution dynamics. Combined these feedbacks will strongly affect global CO2 concentration, ocean circulation, and regional climates. Also important to note is that many ESMs do not conserve mass and energy to machine precision and thus are not intended for millennial length simulations. Therefore only ESMs with little to no drift in their zero emissions pi-control simulation should be extended to millennial length (hopefully this will be less of a problem for CMIP7 models).

We agree these are important points to note. We propose to include an additional couple of sentences at the start of the seventh paragraph (L56) discussing ESM shortfalls and noting the need for further progress in Earth System Modelling. We also propose to add a note in the ninth paragraph (L76-82) that ESMs should be examined for drift prior to extended runs. In general, we plan to emphasise that ESMs and EMICs complement each other and both have a role.

Specific comments:

Line 18: "Humanity must achieve net zero emissions to slow down climate change" is not really correct. Reducing the rate of emissions should slow down climate change, since warming is roughly proportional to cumulative emissions. To stop global average climate change ZECMIP showed the near-zero emissions are needed. With the ZEC range implying that slightly positive to slightly negative emissions are compatible with zero global temperature change.

Yes, we see what you mean and agree this wasn’t correct. We propose to edit this from “Humanity must achieve net zero emissions to slow down climate change and to have any hope of meeting the Paris Agreement goals of limiting global warming to well below 2°C.” to “Humanity must achieve net zero emissions to have any hope of meeting the Paris Agreement goals of limiting global warming to well below 2°C.” As we are adding text elsewhere this is also useful in limiting the word count.

References:

MacDougall AH, Frölicher TL, Jones CD, Rogelj J, Matthews HD, Zickfeld K, Arora VK, Barrett NJ, Brovkin V, Burger FA, Eby M. Is there warming in the pipeline? A multi-model analysis of the Zero Emissions Commitment from CO 2. Biogeosciences. 2020 Jun 15;17(11):2987-3016.

Reply to Referee 2. We thank the referee for their constructive feedback. Their comments are shown in bold with our responses in italics.

Overall evaluation

The authors present an argument for extending ESM simulations under zero emissions to better understand the long term global and regional climate response. Though I agree this would generate useful information, the authors need to be more specific about what they are expecting to gain from model runs of this length, and why ESMs in particular are required. To support the paper’s title, the text would also benefit from an example of how millennial-scale ESM runs could impact decision making today.

We agree that the motivation for extended runs could be clearer and specifically why ESMs are needed for this purpose (as opposed to only EMICs; see also reviewer 1’s comments and our responses). We propose changes to discuss the additional information that may be gained from using ESMs (to be added in the first paragraph of the main text). We also propose to expand on our example of multi-year drought (briefly mentioned in Lines 72-74 in the initial submission) with discussion of decisions being made in response to worsening drought and increasing water demand in southern Australia, and the need to understand drought changes where there is a suggestion of mean precipitation recovery (at least in a single model; King et al. 2024). Multiple model simulations are needed for understanding changes in hydroclimatic extremes robustly. We believe that analysis of extremes based on extended model simulations could feed into climate policymaking.

General comments

The authors mention the significance of ESM runs for regional predictions. The proposal would benefit from a short explanation as to how millennial-length ESM simulations would enable this. For example, what are we expecting to gain from these runs vs multi-century simulations, and why are ESMs required over EMICs? Providing examples of which processes are resolved in ESMs but not EMICs, and what impacts these might reveal on millennial timescales would make the argument stronger.

Indeed- this wasn’t well enough articulated in the initial submission (also noted by Reviewer 1). A key example is for understanding changes in extremes where EMICs can’t be used. Returning to the drought case, there is a need for long-term planning to determine if augmentation of existing water supply is needed. We propose expansion on this beyond Lines 72-74, and also additional motivation for ESMs provided at the outset in the first paragraph (L17-23).

The case for informing decision making could also benefit from an example. What is it about millennial scale impacts that would change decisions made today?

This is in part about highlighting that there are long-term consequences to emissions delay and also about ensuring adequate sample sizes for analysis of rare events. We propose expansion of the bullet points to clarify this.

The authors need to be careful to differentiate between time, warming and emissions. In Figure 1 they show the change in impacts, largely to the ocean and cryosphere, in a model that stops emissions in the year 2030 and one that stops in 2060. The authors need to clarify whether it is the time it takes to reach net zero that changes these millennial scale outcomes, or the higher cumulative emissions and warming that has resulted during the delay. This also applies to the text in lines 87-90.

In reality, it’s hard to separate the effects of additional warming and cumulative emissions because they’re intrinsically tied, but we propose edits to clarify this point and to better highlight that there are different millennial-scale changes projected between simulations due to different levels of cumulative emissions/warming. These edits would be to the seventh paragraph (Lines 56-63).

Specific comments

Line 14: ‘New millennial-length Earth System Model simulations are required to better understand these committed changes and their dependence on delays in reaching net zero emissions.’ The authors should specify what we are expecting to change over that time frame compared to the medium term.

We would suggest retaining the current sentence as is because the Abstract text is supposed to be brief and the previous sentence lists some of these changes, although admittedly in general terms. We suggest that we instead expand the discussion of changes in the seventh paragraph (L56-63) when describing findings from the ACCESS-ESM-1.5 model study.

Line 17: ‘Humanity must achieve net zero emissions to slow down climate change’. Reducing emissions will slow the rate of climate change, and net zero emissions may bring us to a point of temperature stabilisation.

Thanks. Yes, reviewer 1 also pointed out this error and this needs to be corrected. We propose to edit this sentence from “Humanity must achieve net zero emissions to slow down climate change and to have any hope of meeting the Paris Agreement goals of limiting global warming to well below 2°C.” to “Humanity must achieve net zero emissions to have any hope of meeting the Paris Agreement goals of limiting global warming to well below 2°C.”

Lines 19/21: ‘regional and local climates will continue to evolve for many centuries. There is surprisingly little known about these longer timescale climate changes despite their policy relevance’. An example of a regional change would be helpful here. Global climate may also change over these timeframes, bearing in mind the uncertainty in the ZEC assessment.

Yes, we agree. We propose to use the example of substantial warming in many Southern Hemisphere land areas found using the 1000-year ACCESS-ESM-1.5 simulations.

Figure 1: small suggestion to change the colour scheme to avoid the automatic association between red and warming e.g. Arctic sea ice extent in March is actually increased for a net zero that’s imposed later, but is coloured red. It would also be useful to add in the figure caption what the emissions scenario was for these runs.

Yes, we see what you mean. We will make changes to the Figure colour scheme to avoid this issue. We will also provide more detail of the emissions pathway in the caption. Both runs branch from SSP5-8.5 with zero emissions imposed from 2030 onwards or 2060 onwards.

In Figure 1, it would also be useful to include the overall changes as well as the difference between the two time periods. For example, I suspect the very small reduction in Arctic sea ice for the 2060 run does not reflect less sea ice loss overall, and is instead the result of front-loading the impacts into the first 300 years in the 2060 scenario.

The Arctic sea ice extent changes are minimal in the 2060 run mainly due to summer sea ice loss during the 21^st century in SSP5-8.5 and a lack of recovery in the net zero simulations (see Figure 3f in King et al. 2024). Again, we can see the benefit in showing previous or overall changes for context, but equally we don’t want to make the Figure too busy. We propose to explicitly note the issue around Arctic sea ice extent in the text instead in the seventh paragraph (L56-63). 

Line 76 is an important point and would benefit from coming much earlier, possibly in the abstract. ‘The lack of extended net zero simulations risks understating the long-lasting impacts of increased global temperatures relative to pre-industrial levels’.

We agree. We propose to make an edit to the first paragraph (L17-23) in the main text to note this.

Line 78: The authors need to explain why extending these runs will offer the same or better utility than the alternative of extending the 1pct and bell emissions pathways to fall in line with the ZECMIP experiments.

Yes, we agree the ZECMIP runs could also be extended. Given the setup of ZECMIP (with branching from a concentration-driven run) and the shift towards emissions-driven model setups we proposed extensions from either flat10MIP or TipMIP with the view that these setups may be more likely to continue in future rounds of CMIP. We propose to make an edit noting the expansion of emissions-driven model frameworks and that leveraging this for extended net zero simulations would be beneficial.