the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Sep 2026
The moving target of simulating ENSO: A timeline of Community Earth System Model version 3 (CESM3) Development
Abstract. The process of developing Earth System Models (ESMs) varies across modeling centers globally, but the overarching goal is largely the same – to improve the representation of physically-based processes such that biases in the mean climate state and its variability are minimized. Developers face a number of scientific and technical hurdles to ensure the best use of limited computational resources, storage space, and time. These challenges are particularly pronounced for modes of climate variability that are characterized by high internal variability, like the El Niño Southern Oscillation (ENSO). In this study, we leverage the development of the Community Earth System Model version 3 (CESM3) as a case study to illustrate some of these difficulties as well as to highlight new model developments and their impacts on ENSO.
ENSO is a complicated indicator of model performance in that it can be characterized by a wide number of metrics to assess how well (or poorly) it is simulated. This is at odds with the model development process as a whole, which requires that a manageably small number of metrics be selected for model analysis as teams conduct hundreds of simulations to arrive at a single "best" model configuration. Selecting too many metrics runs the risk of dramatically slowing progress in developing a coupled ESM. As a result, we discuss a minimal set of ENSO metrics here that we consider indicative of overall model performance. We find that biases in the spatial extent of ENSO events persisted throughout the development cycle, with sea surface temperature anomalies (SSTa) that extend too far into the West Pacific for all simulations. Other metrics are more sensitive to model changes, including ENSO amplitude and duration. Such metrics are, however, prone to significant internal variability. This is confirmed by temporally sub-sampling long pre-industrial control simulations of previous model versions (CESM1 and CESM2), which also adds critical context to the evaluation of CESM3; any new model version should ideally not be markedly worse than past iterations. Ultimately, CESM3 produces a reasonable ENSO in comparison to previous model versions and relative to observations, but it remains difficult to attribute changes in its representation to individual model changes due to the significant internal variability.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Geoscientific Model Development.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(6672 KB) - Metadata XML
-
Supplement
(308 KB) - BibTeX
- EndNote
Status: open (until 04 Nov 2026)
- RC1: 'Comment on egusphere-2026-4558', Anonymous Referee #1, 29 Sep 2026 reply
-
RC2: 'Comment on egusphere-2026-4558', Anonymous Referee #2, 09 Oct 2026
reply
I think this is a great paper and provides unique insights on model development that were previously unknown to me and are often never published (probably because they’re intrinsically ‘messy’). The paper explains how even when a certain metric (e.g. western edge of 299K contour) looks good it could still be discarded because of other considerations. While I fully recognize that ENSO performance is only a part of how a model is chosen I still remain curious how the “best” model is selected (lines 298, 361) and would be happy if the authors offered other additional details on that aspect— what are the main overall metrics for an earth system model?
I also really enjoyed the authors acknowledgement of internal variability and their accounting for it when they compute these metrics. I like that they showed how the assessment of internal variability can change with model (CESM1 vs. CESM2).
The only major missing part I see in this paper is that the major meteorological centers around the world (BoM, ECWMF, NOAA, etc) are using relative Nino3.4 to track and forecast ENSO. I realize that is a recent development, which may not have been evident when this work was being done, but now that it is clear, I think at least a figure should be considered to evaluate relative Nino-3.4, especially as the focus is on ENSO and the goal of this paper is discuss a “minimal” set of ENSO metrics. I recognize that most of these runs weren’t done with this measure in mind but perhaps some discussion could be added and if a figure could be put together, even better.
I also realized that the linear trend was removed (line 130) which probably means the results may be fairly consistent with the relative anomalies, but it would still be nice to show this and would serve to reinforce the point that you’re not examining the typical SST anomaly calculation, which includes model trends.
After reading section 2.2 (Methods) I’m surprised shortwave fluxes are not mentioned. It feels like if TAUX is going to be analyzed some surface fluxes should be analyzed and particularly SW fluxes b/c they are associated with known biases in ENSO. Would consider adding such an indicator.
Less significant issues:
Line 30: I’m not aware of NCAR’s seasonal predictions other than in a research capacity. If possible I would also cite some of the national operational centers who regularly create seasonal predictions.
Line 33-34: There are some other very good papers that examine ENSO’s physical mechanisms in models and I think those should be cited and perhaps summarized a bit as well. Some suggestions:
Bellenger, H., Guilyardi, E., Leloup, J. et al. ENSO representation in climate models: from CMIP3 to CMIP5. Clim Dyn 42, 1999–2018 (2014).
Bayr, T., Wengel, C., Latif, M. et al. Error compensation of ENSO atmospheric feedbacks in climate models and its influence on simulated ENSO dynamics. Clim Dyn 53, 155–172 (2019).
Bayr, T., J. F. Lübbecke, J. Vialard, and M. Latif, 2024: Equatorial Pacific Cold Tongue Bias Degrades Simulation of ENSO Asymmetry due to Underestimation of Strong Eastern Pacific El Niños. J. Climate, 37, 6167–6182.
Line 39: The distinct shift in Nino-3.4 anomalies in HadISST after the satellite era is an interesting comment. Is this published anywhere and, if not, can you show it with a plot? Also, do other well established datasets have this shift such as Cobe and ERSSTV6 (the latter does not include satellite data)? If they do not then that’s a case for making sure comparisons are done with multiple observations.
Line 42-43: There is other research that establishes that there may be a temporary period of increases followed by a decrease. Suffice it to say the picture may be more complicated than Cai et al 2023 citation and would mention it.
Maher, N., Wills, R. C. J., DiNezio, P., Klavans, J., Milinski, S., Sanchez, S. C., et al. (2023). The future of the El Niño–Southern Oscillation: Using large ensembles to illuminate time-varying responses and inter-model differences. Earth System Dynamics, 14(2), 413–431.
Callahan, C. W., Chen, C., Rugenstein, M., Bloch-Johnson, J., Yang, S., and Moyer, E. J.: Robust decrease in El Niño/Southern Oscillation amplitude under long-term warming, Nat. Clim. Change, 11, 752–757
Line 181: Why coefficient of variation and not the more widely used signal-to-noise ratio? Here you want a high signal-to-noise ratio (instead of a low CV).
Lines 183 and 189: What does “more reliably computed” mean in this context?
Figure 4 (lines 233-251): I genuinely wonder what may account for the different biases— is there a particular tuning parameter that seems to most strongly affect it . If you were to adjust that parameter to better fit it would there be a tradeoff somewhere else? More discussion along these lines I suspect would interest the reader.
Line 268-269: If you were to scatterplot the variance against the mean bias is it just random scatter? Just want to be sure there is no connection. It would also help if some citations were provided for the comment on line 269 (even though one might anticipate there is a physical connection).
Line 273: “cross Pacific gradient” do you mean “zonal Pacific gradient?”
Fig S1: How are you computing the zonal gradient difference? It’s not clear to me which regions are being averaged. Also it would help to know whether more negative values of the SST difference are stronger gradients (I assume).
Line 322-323: I wonder if this might have been dominated by La Nina events which are known to be multi-year. Can you identify the sign of SST anomalies in the the single long ENSO event? Ah I see you followed up on this with Figure 7 which partially answers my question (though total SSTs).
Line 333: I’m not sure what is meant by La Nina events struggling to emerge. Is it possible that these cold values are reflecting long duration La Nina events? Can an anomaly be calculated?
Figure 9: I’m not sure what the y-axis is showing here. Should’t it be in correlations? (-1 to 1)
Line 378-380: I’d try to clarify this sentence (CESM3 primarily shows…) and provide a bit more physical insight. Predictability usually is tied to correlation.
Figure 10c and 10d need a title/label somewhere other than the tiny x-axis label indicating the units of what these panels are showing.
Line 390: I would try to explain physically what “denser tails” means. What does this PDF say about the characteristics of SSTa?
Line 391-392: The connection between stronger anomalies and more realistic patterns isn’t clear to me. Is this a signal-to-noise argument stemming from Pacific forcing?
Figure 10 and related text needs much more description between what these plots are showing and what the physical interpretation means. I’m not sure Figure 10d is discussed at all. Also, Figure 10e show the kurtosis of the CESM1 and CESM2 distributions as well? I’m having a difficult time interpreting what this means across model generations. From Figure 10d it looks like ERA5 has more density in the tails of the distribution so it surprises me to see that the kurtosis in Fig 10e is near zero. Is this linked more to the height of the PDF near zero anomaly values? If so, then it needs to be much clearer what this kurtosis metric is measuring because my understanding is kurtosis is often used to measure the data in the tails so more of the discussion should be baed on this aspect.
Citation: https://doi.org/10.5194/egusphere-2026-4558-RC2
Data sets
Data & scripts for "The moving target of simulating ENSO: A timeline of Community Earth System Model version 3 (CESM3) Development" Authors/Creators M. D. Fowler et al. https://doi.org/10.5281/zenodo.21658223
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 276 | 196 | 70 | 542 | 50 | 62 | 69 |
- HTML: 276
- PDF: 196
- XML: 70
- Total: 542
- Supplement: 50
- BibTeX: 62
- EndNote: 69
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of 'The moving target of simulating ENSO: A timeline of COmmunity Earth System MOdel version 3 (CESM3) development' by Fowler et al.
The manuscript examines the challenges of simulating ENSO during the development cycle of CESM3. The authors track ENSO performance across a range of developmental model configurations rather than relying solely on controlled sensitivity experiments. One important finding is that the high internal variability of ENSO requires relatively long simulations—often exceeding 100 years—to robustly assess model performance, which is consistent with the general understanding of ENSO variability.
I appreciate the authors’ considerable effort and exceptional transparency in documenting ENSO behavior throughout the model-development process. The framework and methodologies presented here could also be useful to other modeling centers. Overall, I am generally supportive of publication, and I hope the authors can consider addressing the following comments, which I believe could further strengthen the manuscript.
1) Broader Applicability
While the manuscript transparently documents the inherently iterative and sometimes messy reality of model development, I feel that it currently provides relatively limited actionable guidance for other modeling centers. Contrasting the “simulations of opportunity” approach with more systematic atmospheric-parameter optimization methods, such as those described by Yu et al. (A systematic atmospheric parameter optimization method to improve ENSO simulation in the ICON Earth system model), could help elevate the manuscript from a documentation of the CESM3 development history to a more broadly applicable methodological guide. I fully recognize that this may not be straightforward, since model development typically involves optimizing multiple metrics simultaneously and balancing realistic representations of individual physical processes against overall model performance. Nevertheless, along these lines, I encourage the authors to explore the underlying model physics in greater depth, such as the discussion of the treatmen of gustness and its impacts on ENSO in this paper.
2) Link to AMIP Simulations
ENSO is a highly air-sea coupled mode and is strongly impacted by both intrinsic atmospheric and oceanic model physics and the coupled mean climate state. For model development, it is common to build a reasonable atmospheric and oceanic models before coupling them. I think incorporating AMIP simulations would significantly strengthen the analysis by using prescribed SST forcing. It may help distinguish whether the identified ENSO-related biases originate primarily from intrinsic atmospheric model physics or emerge through interactions with the coupled mean state.
3) Other factors/metrics critical for ENSO simulations:
The analysis currently places substantial emphasis on atmospheric diagnostics. A more comprehensive evaluation of the oceanic and thermodynamic contributions would help complete the physical picture. In particular, I encourage the authors to consider expanding the analysis to include something like: thermocline depth/tilt, mean upwelling, thermodynamic damping. These factors all influence ENSO evolution.
The manuscript identifies a weakened Bjerknes feedback, particularly through a muted wind-stress response to SST anomalies. I think this is, at least partially, related to the model’s mean state: CESM3 exhibits a mean cold SST bias, which may make the threshold for deep convection more difficult to reach. Consequently, even for comparable SST anomalies, the associated convective response—and therefore the subsequent wind-stress response—could be systematically weaker. Given the mean SST bias, this may not be quite surprising. Again, if the AMIP model does show the underestimated wind stress response, it could be a big concern and should be related to convection parameterization and PBL processes.