Title: Climate Modes evaluation datasets from CMIP6 pre-industrial control simulations and observations

Authors: Mohapatra et al.

This paper analyzes eight commonly examined climate modes derived from CMIP6 pre-industrial control simulations. The study aims to provide a foundation for evaluating climate mode interactions, model performance, attribution studies, and internal variability diagnostics. The topic is valuable, and the constructed dataset could be useful to the climate community. However, several clarifications and revisions are needed before the manuscript can be considered for publication. My detailed comments follow.

Line 58: please add citations distinguishing CP and EP ENSO types.

Line 98: ENSO teleconnection biases remain in CMIP6. Please see Fang et al. (2024)

Line 143: the manuscript compares only with ERSSTv5. Why not validate using multiple observational SST datasets? Please justify or add datasets to quantify observational uncertainty.

Line 170: clarify whether 1°×1° interpolation is optimal. Are results sensitive to interpolation resolution? A brief justification or sensitivity comment would strengthen confidence.

Line 203: “10-year”

Table captions should appear at the top of tables according to journal guidelines.

Line 335 / Fig. 4f: IPO period varies across models. Please explain possible causes.

Line 410: explain the mechanism responsible for cooling over the North Atlantic. Possible related factors include AMOC weakening, aerosol forcing representation, internal variability, or model drift.

Line 473: “50 years.”

Reference

Fang, Y., Screen, J. A., Hu, X., Lin, S., Williams, N. C., & Yang, S. (2024). CMIP6 Models Underestimate ENSO Teleconnections in the Southern Hemisphere. Geophysical Research Letters, 51(18), e2024GL110738. https://doi.org/https://doi.org/10.1029/2024GL110738

General assessment 

This manuscript presents a dataset of climate variability indices derived from pre-industrial control (piControl) simulations of multiple CMIP6 models. The authors compute a suite of well-known modes of climate variability (including ENSO, AMO, IPO, NAO, SAM, IOD, and SDM) using established methodologies, and provide these indices as a consolidated archive intended to facilitate studies of internal climate variability. Additionally, the authors compare the spatial structure and temporal variability of these modes with observations, aiming to showcase consistency or biases with respect to observations. 

The work is technically competent and clearly reflects substantial effort. Additionally, the manuscript is generally well written. However, after reviewing both the manuscript and the dataset itself in light of ESSD’s evaluation criteria, I conclude that the submission does not align with the journal’s requirements for significance, uniqueness, and added value to the scientific community. More specifically, I consider that the produced data substantially overlaps with already existing, broadly distributed, and thoroughly documented data. Besides, I consider that those aspects of the data presented in the manuscript that do not exactly overlap with previous literature do not represent enough novelty to merit publication. Therefore, my recommendation is to reject this manuscript from ESSD. Below, I expand on these points and present potential alternatives for future fruitful publication.

The two major problems: 

• Not enough novelty

The authors provide 8 indices. From those, four (Atlantic Multidecadal Oscillation, AMO; North Atlantic Oscillation, NAO; Southern Annular Mode, SAM; Interdecadal Pacific Oscillation, IPO) are publicly available and operationally maintained by the Climate Variability Diagnostics Package (CVDP; Phillips et al., 2014; Maher et al., 2025—please see supplementary material) community effort using the same methodology that the authors provide. Apart from the time period that the authors cover, CVDP also provides these indices for other CMIP versions, as well as for the broader suite of experiments that they encompass beyond the pre-industrial simulations.

The authors also provide two ENSO indices based on Empirical Orthogonal Functions for the Central Pacific and Eastern Pacific flavors of ENSO. Although these differ from the more broadly used ENSO indices provided, for example, in the CVDP package, the manuscript does not convincingly demonstrate that the chosen EOF-based definitions are more robust across models, superior to existing alternatives, or uniquely enabled by the dataset. The authors justify using these indices by stating that EOF-based indices are more robust to model biases, but, under my consideration, a deeper review of the literature and quantitative analyses are needed to justify that statement. Just as an example, I recommend reviewing the recently developed RONI index (L’Hereux et al., 2025). Furthermore, beyond the mere indices, community tools already provide composites, spectra, and regressions for piControl simulations (link). This same point applies to the Indian Ocean Dipole index.

As for the remaining mode, the Subsurface Dipole Mode, this component appears less covered by existing diagnostics frameworks and could represent a meaningful contribution if foregrounded, more thoroughly documented, and clearly positioned as the primary data product. If the authors decide to pursue this mode as the primary contribution of the paper, then a more thorough explanation should be provided, and the index should also be computed for the whole suite of CMIP experiments beyond pre-industrial simulations in order to make the paper stand out as a significant contribution useful to the broader community. The first author already published a paper on this mode of variability (CITE), which could serve as a foundation for a more complete data contribution to the community.

In summary, from the eight provided modes, four are fully covered by past literature, three are partially covered and deeper demonstration of superiority with respect to existing tools needs to be provided to make this paper stand out, and one is a novel contribution but not highlighted or explained in enough detail, and not computed for a broader set of experiments in order to have a significant impact.

• Unfair comparison between models and observations 

Although the exploration of the differences between the pre-industrial simulations and observations is interesting, its usefulness to the climate science community is somewhat limited. On one hand, it is methodologically inappropriate to rank model skill by directly comparing spatial patterns from pre-industrial control simulations with the historical observational record. It has been demonstrated that human emissions deeply altered the characteristics of modes of variability (e.g., Klavans et al., 2025), and substantial differences between pre-industrial and historical climates are therefore expected. It is not possible to know with certainty, solely based on the presented analysis, which version of the pre-industrial climate is more or less wrong. On the other hand, a deep and thorough comparison of modes of variability in models vs. observations is already available (e.g., Fasullo et al., 2020; Maher et al., 2025). Furthermore, these papers also take into account observational uncertainty and the fact that we only have one observed realization of internal variability (for the historical forcings - not pre-industrial) vs. the multiple realizations available in the CMIP archives. If assessing the correctness of the representation of modes of variability in Earth system models is the goal, authors should focus on aspects that have not been thoroughly explored before, and in colloquial words, be mindful of comparing apples with apples. Furthermore, if this is the main scope of the paper, I would recommend submitting to a journal with more emphasis on Earth system model evaluation rather than scientific data publication.

Conclusion

Based on the considerations above, I see two viable paths for successful publication of this work:

1. If the focus on a data publication is to be kept, a deeper justification of the methodological decisions (that must be different from the already available community resources) should be provided, showcasing and quantifying robustly why these would be preferred over the “status quo” of publicly available data. The novelty and usefulness of the data beyond what already exists should be explicitly proven.
2. If the authors would like to maintain and strengthen the models vs. observations comparison, this should focus on aspects that have not been covered before, which should be accompanied by a broad and thorough review of the existing literature. Additionally, care should be taken about the validity of the comparison in light of different forcings and the presence of internal variability and observational uncertainty.

References

- Phillips et al., 2014. Evaluating Modes of Variability in Climate Models. Eos.

- Maher et al., 2025. The updated Multi-Model Large Ensemble Archive and the Climate Variability Diagnostics Package: new tools for the study of climate variability and change. Geosci. Model Dev.

- L’Heureux et al., 2025. A relative sea surface temperature index for classifying ENSO events in a changing climate.

- Klavans et al., 2025. Human emissions drive recent trends in North Pacific climate variations.

- Fasullo et al., 2020. Evaluation of Leading Modes of Climate Variability in the CMIP Archives. Journal of Climate.