Dear authors,

Thank you for submitting your responses to the reviewer comments on your paper. The reviewers raised many important concerns and your responses go some way towards addressing these. I am happy to allow the revision process to continue, with the view that a revised manuscript would undergo further review to assess whether the concerns have been adequately addressed.

In particular, there are a few points where I think that your responses do not go far enough in addressing the concerns raised by the two reviewers. These are:

The proxy network: a number of coral records were not used because coral d18O contains signals related to both SST and rainfall. This is true but it shouldn't automatically rule out these archives, which are among some of the most high-fidelity proxy records available. In some cases variability within a coral d18O record from a particular site can be shown to be controlled primarily by SST or rainfall, so could be assigned to one of your networks. Also, this dual environmental control that is known for corals, is also a factor in many other types of proxy records, but just may not be as explicit in other proxies (e.g. tree rings) as it is for coral d18O where the geochemical components can be identified. These factors need to be considered, as well as a more thorough review of archived data. As well as the NOAA Paleoclimate database, can you please also review the PAGES2k (including Ocean 2k HR and LR) database to cross check record selection.

30 year averaging: more work needs to be done to justify this method choice. In particular using 30 year averages means that IPO/PDO needs to be much more clearly brought into the study as at 30-year average time scales you are firmly in the regime of PDO-scale variability rather than the interannual variability associated with ENSO.

Calibration/validation: The proposal to use CMIP5 last millennium simulations for calibration and validation of the proxy datasets is a very shaky one. This would require confidence that these models accurately represent temperature and rainfall variability associated with ENSO and the PDO - both in terms of the spatial patterns of these anomalies and in their time evolution. I don't think that we can have confidence yet that this is the case, and even if models were able to accurately capture the forced climate responses of ENSO/IPO through time, there would be an inevitable mismatch between proxy records and model output related to unforced climate variability even in a perfect world. The long model experiments will be very useful for testing your methodology within a model world, using either pseudo-proxy techniques or even just straight temperature and rainfall sampling from the grid cells representing proxy locations. But it isn't realistic to expect that the proxy records can be calibrated/validated against model output. This problem for calibration/validation also feeds back into your method choice of using 30 year averages.

Please let me know if you have any questions about these comments, or the process for continuing to revise your work.
Sincerely,
Nerilie Abram

Dear Lilo Henke and co-authors,

Thank you for submitting your revised manuscript to Climate of the Past. The manuscript has been greatly improved by the additional work carried out in the revisions process.

As communicated earlier, because of the extensive revisions made to the manuscript I elected to undertake a thorough review of the manuscript myself. It has taken longer than I would have liked to complete this and I apologise for the delay.

Overall I think that the manuscript is suitable for publication in Climate of the Past with a few further minor revisions. In particular I think that the method development is a very useful addition to the research community, and even though firm conclusions can’t be reached about ENSO-like temperature-precipitation changes at this stage the results are a useful addition to discussions around the interpretation of palaeoclimate data.

Specific comments that I would like see further minor revisions around are:

1) I’m a bit uncomfortable about the choices around coral records to use for the temperature network. In particular, some coral d18O records are excluded because of the additional influence of precipitation/SSS on these signals. This is fine but it is a very grey line around which coral d18O records are included and which are not. All coral d18O records will have some influence from seawater d18O/rainfall, so perhaps a more objective decision would be to only use coral Sr/Ca records, which should more purely record SST only.

2) The PDO paragraph at line 681 feels a bit “tacked on” at the moment. Throughout the manuscript I kept on wondering about whether the method was reconstruction ENSO or IPO – the 30y windows are on the IPO rather than ENSO timescale. The temperature EOF in figure 6 of your paper looks strikingly similar to the IPO pattern shown in Figure 1 of Henley et al (2015) Climate Dynamics, DOI 10.1007/s00382-015-2525-1. I feel that this aspect hasn’t yet been clearly dealt with from the original reviews, even though I understand why the 30-y method has been employed.

3) How reliably do global/NH/SH temperature reconstruction ensembles document a robust temperature difference between the LIA and the MCA. (e.g. line 98, line 387)? In particular, I’m thinking of the Neukom et al 2014 Nature Climate Change SH reconstruction ensemble. Also a finding of the PAGES2k Consortium 2013 Nature Geoscience paper was that the LIA and MCA were not global events.

4) Could the lack of T-P relationship be related to P changes being dominated by ITCZ shifts rather than changes in ENSO mean state?

5) Would it be possible to include some statistics to demonstrate if 20CRv2c has a good representation of ENSO indicies, for the period where it overlaps with more widely used datasets.

Please let me know if you want further clarification around these points.

Sincerely,
Nerilie

Dear Lilo Henke and co-authors,

Thank you for submitting a revised version of your manuscript to Climate of the Past. I am very pleased to now be able to accepted this paper for publication.

Sincerely,
Nerilie Abram