the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 May 2026
Intermodel differences in seasonal and regional CMIP6 divergent atmospheric heat transport
Abstract. The tropical rainband's location is closely tied to divergent atmospheric heat transport (AHT). Recent work decomposed annual and zonal-mean AHT into radiative fluxes, evaporative fluxes and sensible heat, finding that the latitudinal structure of divergent AHT strongly resembled that associated with the evaporative fluxes, and that imposed changes to evaporation in model simulations altered total divergent AHT.
Here, we generalise this decomposition to explore regional and seasonal intermodel differences in CMIP6 simulations. In historical climate, we find that the spatial structure of the total JJA and DJF AHT most resembles that linked to evaporative and radiative fluxes. Intermodel differences in divergent AHT predominantly relate to east-west rather than north-south rainband shifts. In future climate, in JJA most models show enhanced southward interhemispheric energy transport, while in DJF models instead undergo a zonal change towards more energy export from a warmer eastern Pacific.
We identify groups of models from different families that show similar decompositions of their total AHT response to climate change into the individual flux terms. This suggests that distinct storylines may exist through which warming affects the energy budget and associated tropical rainfall.
- Preprint
(32768 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 08 Jul 2026)
- RC1: 'Comment on egusphere-2026-2715', Anonymous Referee #1, 18 Jun 2026 reply
-
RC2: 'Comment on egusphere-2026-2715', Anonymous Referee #2, 28 Jun 2026
reply
Comments on “Intermodel differences in seasonal and regional CMIP6 divergent atmospheric heat transport” by Geen et al.
Geen et al. document the energy flux potential simulated by CMIP6 models and attribute it to the component caused by radiative, latent heat, and sensible heat fluxes. They provide a comprehensive diagnosis of the historical energy flux potential and its future changes under the SSP5-8.5 scenario, analyzing both mode means and inter-model spreads across the JJA and DJF seasons. Through diagnostics in the PC space, the authors find that models with similar historical distributions of energy flux potential can be driven by different underlying mechanisms. When the anomalies in the SSP5-8.5 simulations are projected onto the historical EOF modes, similar energetic pathways of change emerge across different model families.
Understanding the spatial distribution of energy budget is fundamental to the energetic framework and energy-circulation coupling studies; the diagnostics provided in this study will be helpful to future research in this field. In addition, the findings that different model families can share similar energetic pathways under the SSP5-8.5 scenario provide valuable insights into future climate projections. The manuscript is well-written and properly structured. A few points are listed below, which should be addressed prior to publication.
General comments:
- The clarity of the methodology needs to be enhanced. For instance, the definition of 'pr-lat' in Fig. 3 could not be found. Furthermore, the specific procedures and details regarding the EOF analysis are not clearly explained. I believe that the model-mean signal was subtracted prior to the EOF analysis (otherwise, EOF1 would simply capture the JJA/DJF model-mean climatology), but it is not mentioned explicitly in the text. Additionally, the procedures for projecting the SSP5-8.5 and ERA5 data onto these EOFs need to be clarified, particularly regarding whether any mean values were subtracted prior to projection. I suggest adding a subsection on the EOF procedures in the Methodology section
-
- At times, the paragraph breaks seem a bit too frequent to me, making the text feel slightly fragmented. Consolidating these into cohesive paragraphs would enhance readability. For instance, in section 3.2, consider merging the paragraphs starting at line 186 and 191, as well as those at line 195 and 199.Specific comments:
- Fig. 3:
I was surprised to find no correlation between annual-mean pr-lat and AHTeq, given that the correlations are generally high for annual-mean anomalies in response to external forcing (e.g. Donohoe et al. 2013, Schneider et al. 2014). Is this attributable to the definition of pr-lat, or does the link between pr-lat and AHTeq simply fail to explain the climatology model spread? Specifically, I wonder if the correlation would improve if alternative metrics were used instead, such as the precipitation centroid (Frierson and Hwang 2012, DOI: https://doi.org/10.1175/JCLI-D-11-00116.1) or the precipitation asymmetry (i.e., precipitation in the NH tropics minus the SH tropics).- Line 358-359: “As would be expected, models from the same family tend to be located near to one another in the PC space in both JJA and DJF.”
Can you specifically point out how do we get this conclusion in Fig. 8? Because it is not very obvious to me.- Line 403: “Hierarchical clustering was performed across the first 3 PCs of χ and latent heat, sensible heat and radiative input terms.”
Are you applying hierarchical clustering to the projections of SSP5-8.5 anomalies onto the first 3 EOF modes? So that you are grouping models with similar impacts of climate change to the EOF modes in the historical seasonal cycle?
Also, what is the pattern of the third EOF mode? It should at least be demonstrated in Supplementary if this is important for the results of hierarchical clustering.- Fig. 12
I feel the methodology of projecting the anomalies in SSP5-8.5 onto the EOF modes requires careful interpretation, as it may mislead readers into thinking that the model spread in climate change responses can be largely explained by the strengthening or weakening of seasonal EOFs; this requires thorough clarification.
The first two EOF modes represent the north-south asymmetry and east-west dipole, respectively, and I suspect that the third EOF simply represents another east-west dipole with a 90-degree phase shift compared to EOF2. In this case, these EOF modes collectively provide a spatial basis that can describe any wavenumber-1 structure. As the energy flux potential usually exhibits a spatial structure of wavenumber 1 or 2, it is not surprising that the first three EOF modes can adequately describe its overall pattern and thus be used for model selection in hierarchical clustering. However, replacing these three EOF modes with any arbitrary spatial basis that captures a wavenumber-1 structure might yield similar results. Therefore, this mathematical alignment does not necessarily support a physical linkage between the seasonal model spread and future climate change responses.Comments to figures:
- Figs. 9, 11:
The red, orange, and magenta lines are hard to distinguish. Please consider change the colors, and also make the lines thicker so that the readers can recognize each line easily.
The colors and shapes of the symbols are a bit distracting. As the model name is already on the top of each panel, a black dot should be enough.
Consider add a legend below the figure, showing the corresponding components of each line.Citation: https://doi.org/10.5194/egusphere-2026-2715-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 32 | 8 | 5 | 45 | 6 | 6 |
- HTML: 32
- PDF: 8
- XML: 5
- Total: 45
- BibTeX: 6
- EndNote: 6
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Comments on “Intermodel differences in seasonal and regional CMIP6 divergent atmospheric heat transport” by Geen et al.
This study evaluates the regional and seasonal atmospheric energetic simulation in CMIP6 historical climate (good to know that the simulations resemble ERA5!) and deconstructs its NEI contributions; and then follows on to evaluate intermodel differences using EOF analysis and understanding the change from historical to SSP5-8.5 referencing the EOF space. The latter reveals groups of models with different characteristic changes, suggesting that different storylines exist in interpretation of future changes
This is a straightforward and useful paper, adequately researched and well written. This study is exploratory rather than advancing a hypothesis, so not much can go wrong here. I thought the result showing that the intermodel variation in EFP can be summarized to two EOF representing land-ocean contrast and the interhemispheric gradient to be useful. An obvious question that is not really addressed in the manuscript is in understanding why the intermodel variation in the EFP are like this; but I understand that this is probably better left to a future study.
I only have relatively minor comments, listed below.
“Major” comments
Tuckman, P.J., Smyth, J., Lutsko, N.J. and Marshall, J., 2024. The zonal seasonal cycle of tropical precipitation: Introducing the Indo-Pacific monsoonal mode. Journal of Climate, 37(14), pp.3807-3824.
Specific comments
Figure 1 – check the labelling for SH and RAD, there’s an inconsistency.
Line 180 – its worth stating in the beginning that the LH and RAD contributions gives the structure to the overall EFP
Line 186 “In DJF, the zonal structure is dominated by the latent heat component (Fig. 2f)”. This is largely also true for JJA? There are quantitative differences to be sure between JJA and DJF for the zonal component between RAD and LH, but not as pronounced as what the description makes it out to be.
Figure 4 and paragraph starting line 269 – I interpret the EOF1 intermodal difference as reflecting more the strength of the convection over the warm pool and maritime continent, tied to the underlying SST. The warmer SSTs in the eastern equatorial Pacific is a dynamical consequence of the reduced trades in the western and central equatorial Pacific, through thermocline adjustment. Could you comment?
Line 277 “The region of cooler near-surface air temperature (Fig. 4j) overlaps with a known region of cold bias in CMIP models in the Western North Pacific, which has persisted across CMIP generations.” Since this is a regression on EOF1, presumably some models will be warmer and some will be cooler relative to the multimodel mean. Presumably it is the multimodel mean temperature that has the cold bias, but this isn’t associated with EOF1 per se.
Line 313-314 “The Southern Ocean is another known region of SST bias, with CMIP models on average having a too-warm Southern Ocean”. Again, unclear why this is of relevance.
Figure 6 and paragraph starting line 323. Presumably the shift in the EFP dipole (relative to JJA) is because of the eastward shift in the mean EPFM location in DJF (relative to JJA)? Tuckman et al. 2024 is worth taking a look at here.
Figure 8 caption. For readability, it would be helpful to add the physical description of the PC1 and 2 axes (as you did in the main text around line 355) to the caption.
Around line 360. You add the ERA5 PCs on figure 8. Can you specify how this is done? For the model EOF, you subtract out the multimodel mean prior to the EOF. For ERA5, do you also subtract out the multimodel mean from it and then project the difference on EOF1/2?