EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Seasonal weather regimes in the North Atlantic region: towards new seasonality?

Florentin Breton1, Mathieu Vrac1, Yiou Pascal1, Pradeebane Vaittinada Ayar2, and Aglaé Jézéquel3
Florentin Breton et al.
  • 1Laboratoire des Sciences du Climat et de l’Environnement, UMR8212 CEA – CNRS – UVSQ, Université Paris-Saclay & IPSL, Orme des Merisiers, Gif-sur-Yvette, France (
  • 2Institut National de la Recherche Scientifique | INRS · Eau Terre Environnement Centre, Québec, Canada
  • 3LMD/IPSL, Ecole Normale Superieure, PSL research University, Paris, France

European climate variability is shaped by atmospheric dynamics and local physical processes over the North Atlantic region. Both have strong seasonal features. So, a better understanding of their future seasonality is essential to anticipate changes in weather conditions for human and natural systems. We revisit the notion of seasons over the North Atlantic region through the concept of seasonal weather regimes (SWRs), by classifying daily fields of geopotential height at 500 hPa (Z500) without a priori separation of seasons. We use data from the ERA-Interim reanalysis, and from 12 climate models of the fifth phase of the Coupled Model Intercomparison Project. The spatial and temporal variability of SWR structures is investigated, as well as associated patterns of surface air temperatures. Although the climate models have biases, they reproduce structures and evolutions of SWRs similar to the reanalysis over 1979-2017: decreasing frequency of winter conditions, which start later and end earlier, and increasing frequency of summer conditions starting earlier and ending later in the year. These changes are stronger over 1979-2100 than over 1979-2017. By the end of the 21st century, the typical past winter conditions (e.g. 1979-2017) have almost disappeared and correspond to future extreme cold conditions. A new cluster related to summer that was almost absent in 1979-2017 (corresponding to past extreme warm conditions in the past) becomes dominant. To understand whether these changes are linked to uniform Z500 increase or changes in Z500 spatial patterns, we detrend the data (but impose a stationary seasonality) by removing the trend in the seasonal Z500 regional average to define detrended seasonal weather regimes (d-SWRs). The temporal properties of d-SWRs appear almost constant, whereas spatial patterns show evolution. Our results indicate that the evolutions of the SWR temporal features are caused by the regional Z500 trend and that changing spatial patterns in d-SWRs account for the heterogeneity of this trend. Previous research has shown that this large-scale Z500 trend is linked to human influence, suggesting that it drives the changes in seasonality that we find.

How to cite: Breton, F., Vrac, M., Pascal, Y., Vaittinada Ayar, P., and Jézéquel, A.: Seasonal weather regimes in the North Atlantic region: towards new seasonality?, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-5449,, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 2 – uploaded on 17 May 2020
Details added
  • CC1: Reply to AC2, Swinda Falkena, 18 May 2020

    Thanks! Good to know the first PC indeed captures most of the seasonal signal.

Presentation version 1 – uploaded on 04 May 2020
  • CC1: Comment on EGU2020-5449, Pedro M. Sousa, 05 May 2020

    Dear Florentin,

    Very nice presentation. My question is related to detrended / not-detrended analysis.
    Since you detrended the Z500series, I think you are "automatically" removing the signal related to global warming, since on the overall Z500 increases with tropospheric warming (thermal expansion). In this sense, the way your analysis is designed, it shows the resulting dynamical change. Correct.
    My question was if you looked or are thinking about looking (or doing a similar analysis) without detrending? If so, you could make some interesting comparisons. For example, besides seeing the overall change due to dynamical+thermodynamical changes, I guess you could "subtract" your current changes to total changes to "estimate" which of those are more relevant to surface impacts (dynamical or thermodynamical). Of course it's not that straigthforward, but there are several works doing this type of disentanglement for several variables.

    Best regards,
    Pedro Sousa

    • AC1: Reply to CC1, Florentin Breton, 05 May 2020

      Dear Pedro,

      Thank you for the feedback, that’s an interesting question. 

      We want to understand whether the evolution of the SWRs is linked to uniform Z500 increase (i.e. uniform warming), or to changes in Z500 spatial patterns (i.e. changes in circulation patterns). SWR7 corresponds to the total effects (dynamical and thermodynamical). Since we remove the calendar Z500 regional trend to define d-SWRs, we remove the effect from a uniform thermal expansion (we verified that this thermal expansion is uniform but it is not shown in the presentation). The resulting patterns in d-SWRs thus depend from two factors: non-uniform thermal expansion (unlikely) and dynamical changes. The dynamical changes are shown on slide 8 and are minor. 

      Nevertheless, as you suggest, it would indeed be interesting to use disentanglement techniques to further detail the specific contributions of dynamical and thermodynamical factors. Do you have examples of works in mind? (we can also discuss this directly via email).

      Best regards,

  • CC2: Comment on EGU2020-5449, Swinda Falkena, 15 May 2020

    Hi Florentin,

    Nice work (again)! I had another look at your presentation and if I understand correctly you are only using the first PC in clustering the data. Have you looked at using also the second PC? And do you think using more PCs would change your results?

    Best wishes,


    • AC2: Reply to CC2, Florentin Breton, 15 May 2020

      Hi Swinda,

      Thank you, that’s a very relevant question!

      The short answer is that using more PCs brings little more seasonality (our signal of interest here), a lot more noise (i.e. non-seasonal elements), and complicates the physical interpretation (in terms of seasonality).

      The long answer is that before clustering, we did a spectral analysis of the principal components (PCs) in order to assess how much information they capture regarding the variability (variance) and seasonality (signal at 1/365 of frequency) of the Z500 fields over the North Atlantic. Indeed, the first PCs correspond to most of the low-frequency signal and PC1 captures, depending on the dataset, between 49% and 60% of the total variance and between 95% and 99% of the total seasonal cycle. The totals correspond to the spectra power from the sum of all PCs.

      We tried using more PCs, and with a few PCs, the resulting SWRs are similar (weather patterns, seasonal cycle) to only using PC1 (seasonality is the dominating signal). However, the more PCs you use and the more you include signal of higher frequency (subseasonal: monthly, weekly) which is not desired in our study since the focus is on seasonality.   

      All the best,