the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
GWL-REA: A highly optimized numerical method for classifying European Grosswetterlagen based on circulation and air mass considerations
Abstract. Grosswetterlagen (GWL) are a well-known classification of synoptic weather types that connect regional weather impacts over Central Europe to the large-scale circulation over the wider European and N.E. Atlantic regions. GWL-REA is a new algorithmic method for classifying the GWL from reanalysis, climate and/or forecast model data. The method is based initially on pattern correlations using multiple target patterns for each GWL. Further optimisations ensure an equitable classification covering three key aspects of these synoptic types: consistency of the large-scale circulation, regional circulation constraints introduced by each GWL’s nomenclature and regional areal-mean precipitation. For the latter, a neural network is deployed to improve discrimination quality between respective anticyclonic and cyclonic GWL types. Temporal filtering is applied to remove noise on short timescales to make it easier for a person studying the output to assimilate the key aspects. A comprehensive verification of the GWL-REA outputs shows that the new method, when used with ERA5 reanalyses, is of a much higher quality than the original manual Hess-Brezowsky GWL series and to other comparable weather type classifications, in terms of both large-scale circulation consistency and in its applicability for regional weather impacts.
- Preprint
(4144 KB) - Metadata XML
-
Supplement
(28982 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3911', Anonymous Referee #1, 30 Oct 2026
-
RC2: 'Comment on egusphere-2026-3911', Anonymous Referee #2, 30 Oct 2026
Review of manuscript egusphere-2026-3911 entitled "GWL-REA: A highly optimized numerical method for classifying
European Grosswetterlagen based on circulation and air mass considerations" by Paul M. James, Jennifer Ostermöller---------
Summary
---------
The paper introduces a revised classification of Grosswetterlagen (GWL) for central Europe. It builds on the empiric GWL approach of Hess-Brezowsky but brings it on a more objective basis. The paper starts with a longer introduction of the specificities of GWLs, caveats in current definitions and its history. It continues with a description of input data the pattern correlation method used and how nomenclature is found for the different weather types. Further methodology such as an AI model used to enhance the correlation of GWLs to precipitation and temporal filtering to further sharpen the GWLs as well as the names and patterns are introduced by a talble and Figure 5. Section 3 explains a grouping of the 30 weather types in 12 GWL groups according to shared characteristics. Section 4 tackles the problem of the often empiric and subjective nature of GWL classifications by introduces a four-axis verification score that verify certain aspects of circulation pattern classifications: meaningful nomenclature, a meaningful circulation, precipitation discrimination, and "readability". The new GWL-REA scheme outperforms earlier classifications in most of the aspects. Finally the measure "quantile entropy reduction" is introduced in order to verify in space how well a given weather type classification allows to discriminate or explain a surface weather variable. The validity of GWL-REA beyond central Europe for explaining precipitation anomalies is demonstrated and shown to improve compared to earlier GWL classifications. Finally the discussion chapter compares GWL-REA to a newly established classification of seven year-round Atlantic-European weather regimes again showing that GWL-REA is tailored towards explaining precipitation patterns. Finally the (limited) applicability to other weather parameters is demonstrated and some caveats discussed in an outlook section.----------------
Recommendation
----------------
The authors present a tedious and thorough revisit of Central European GWL and a major update to classical weather typing approaches used at DWD. As far as I can understand they provide a new basis to operational forecasting and climate monitoring at DWD and are therefore an important contribution to the field of weather types / circulation patterns / weather regimes.However, I have a couple of major concerns which must be tackled prior to consider the paper for publication in WCD. My main concerns are two-fold:
1) Fit in scope of journal: WCD requires overall a step ahead in the understanding of dynamical processes. I see the discussion of the novel circulation patterns and how much they explain precipitation variability not sufficient in that regard. Questions about trends in or dynamical drivers of the GWLs would shed more light on GWL. WCD require some more answers on the question "why GWLs establish" or "how they vary across decades".
2) Although there is a lot of textual description about the thoughts behind choices in defining the GWL classification, key aspects which would ensure reproducibility remain unclear. Furthermore code or data to reproduce the results are explicitly not given.
Regarding the descriptive nature of the paper I wonder if it would fit better in Geoscientific Model Developments or alike journals.
Furthermore there are more issues regarding e.g. the discussion of previous work which in my view need to be tackled, too, prior to publication. As I think that addressing these issues is difficult in one round of major revisions I recommend the authors to withdraw the manuscript and resubmit a revised version at a later time.Despite my critics I want to emphasise that I acknowledge the important and tedious work put into defining GWL-REA and providing the update, that I find it important having these GWLs documented and provided to the research and operational community, and I want to thank the authors for the huge effort put into implementing the new GWL-REA classification.
I leave the ultimate decision how to proceed with this manuscript considered for WCD to the authors and editor.
---------------
Main Concerns
---------------1) Fit in scope of journal: the paper lacks a sufficient discussion of novel insight in dynamical process understanding required by WCD
2) Lack of reproducibility: the methodological and data description are not sufficient to ensure reproducibility of the results without access to original code and input data. It makes the impression that still a lot of subjective choices have been made.
3) Introduction of terminology: terminology is partly differently used than in other established literature. This leaves it confusing which specific scales are considered. I recommend a clearer introduction and distinction of terms like large-scale circulation (pattern), weather pattern, weather types, (Gross)Wetterlagen, Grosswettertypen, weather regimes. In recent literature (Neal et al 2016, 2024, Grams 2026) the distinction between synoptic weather types (O(1d), O(1000km)) and continental-scale multi-day weather regimes (O(1week), O(10000km)) seems to establish.
4) the introduction does not provide a full picture, in particular of recent developments in weather type identification see detailed comments below.
5) Structure: In section 2 the GWL names are slowly used, prior to once clearly introducing all GWLs. This makes it difficult to follow for the non-expert reader and requires quite some pre-knowledge to access the text. Also in other sections it becomes difficult to follow due to long text passages without reference to figures and/or order of thoughts
6) Unclear why the continental scale North-Atlantic European 7 regimes are compared against GWL-REA. Both are designed for different purposes and address different scales.
7) Table 1 / Figure 5: How stable are these GWLs over the decades? Dorrington and Strommen (2020) argue that decadal variability of regime structure might obscure regime behaviour. Have you used detrended data to remove the recent trend in Z500 due to global warming?
----------------
Detailed comments
----------------
- l19: the abstract speaks of weather types and GWL, here in addition the term large-scale circulation pattern is introduced. The nomenclature needs unambiguous introduction, see main concern 3). The discussion that follows mixes different concepts.-l51: exceptions are Santos et al. 2016 and Grams et al. 2017 who at least discuss the patterns with meaningful identifiers
-l69: Schiemann and Frei 2010 provide a thorough investigation of how different weather type classifications explain rainfall patterns. In this paragraph also newer developments of pattern identification like Spuler et al. 2024 or Falkena et al. 2020
-85: unclear why the focus on precipitation has been chosen. Can the methodology also be applied to other impact variables?
-95: unclear why the WR7 have been picked. This definition must be better introduced and their purpose to characterise large-scale flow variability of the upper-level circulation must be stated.
-l103: suggest to use the more common abbreviation TPW instead of PPW
- Section 2.2: Unclear how it is accounted for different amplitude of MSLP, Z500, ZTH with the seasonal cycle. Does the method account for that or is it biased towards the winter season? Despite the textual description, methodologically this will be difficult to reproduce.
- Section 2.3: It is not easy to understand what has been done. For reproducibility more details on the calculation of the used Lamb types is needed. Admittedly, a lot of subjective experience seems to enter the methodology. In Fig 3 also the actual extent should be shown.
- l180-185: the statements here seem subjective and need more corroboration
- paragraph up to l215: a lot of pre-knowledge is needed on GWLs e.g. about TrM, TrW before this has even be introduced.
- Figure 4 / section 2.4: If there is already such a regional focus, why is ERA5 data used and not a regional / national precipitation data set?
- l220: Is the ERA5 data also at 2x2° resolution? this would be way to coarse to resolve regional rainfall variability.- Section 2.5. Also this section is descriptive, but lacks enough information for reproducibility.
- l301-307: too much pre-knowledge needed before the patterns even have been introduced.
- l319: such trends for all GWLs would be interesting, but appear here in a quite condensed way and only for one GWL. It appears too much of a mixing of results here.
- l325: It is a pity that the GWLs are only introduced via the Table and Fig. 5 without actually discussing their characteristics and relationships.
-l231: more details on the type of experimentation would be good in order to ensure reproducibility
- Figure 5/6: It would help to include the GWG labels in the GWL figures in figure 5, too, to see the grouping and similarity of GWLs in the same GWG.
- page 28, here is too much text with too little reference to the figures and a bit of subjective statements, which make it really difficult to follow for non-experts in the field of GWLs.
- l431, eq 1: The concept of penalty points needs a more thorough introduction.
- l448, eq 2: Likewise the Sp score needs more introduction. It is not understandable where exactly this score comes from.- Section 4, QER method: An interesting method to quantify the "resolution" of a weather type classification wrt. to explained precipitation variability. How does it compare to other methods, e.g. Schiemann and Frei (2010). This needs at least a bit of discussion.
-l592: the AI nature of the precipitation estimation was not clear to me. It needs to be clearer stated and explained in Section 2.
- Section 5: More dynamic content/results/explanations are needed for WCD: E.g. trend analysis, decadal variability. It would be good if self judgment was reduced and instead the quantitative analysis were repeated (the scores) e.g. summarise key results from the verification.
-l615: what about limiting the analysis to the about 70% of the actual regimes days, excluding no regime days of Grams et al. 2017?
-l639: Can you work this argument out more? Where does it come from?
- Figure 10 / Page 29: Not quite a fair comparison as this is not what the 7WR are made for. See Gerighausen et al. 2025 which discuss and provide a means on how to use the 7 WR in the context of surface weather.
- l654 / l690: here and in the introduction more discussion on recent work on targeted circulation types including the use of AI methods is needed. E.g. considering also newest developments such as Spuler et al. 2024.
- l702: Supplementary materials are mentioned only here but never referred to in the text.------------
References
------------
Dorrington, J., & Strommen, K. J. (2020). Jet speed variability obscures Euro-Atlantic regime structure. Geophysical Research Letters, 47, e2020GL087907. https://doi.org/10.1029/2020GL087907Falkena SK, de Wiljes J, Weisheimer A, Shepherd TG. Revisiting the identification of wintertime atmospheric circulation regimes in the Euro-Atlantic sector. Q J R Meteorol Soc. 2020; 146: 2801–2814. https://doi.org/10.1002/qj.3818
Gerighausen, J., J. Oldham-Dorrington, F. Mockert, M. Osman, and C. M. Grams. 2025. Understanding and Anticipating Anomalous Surface Impacts During Large-Scale Regimes. Meteorological Applications 32, no. 6: e70099. https://doi.org/10.1002/met.70099.
Neal, R., Fereday, D., Crocker, R. and Comer, R.E. (2016), A flexible approach to defining weather patterns and their application in weather forecasting over Europe. Met. Apps, 23: 389-400. https://doi.org/10.1002/met.1563
Neal, R., Robbins, J., Crocker, R., Cox, D., Fenwick, K., Millard, J., & Kelly, J. (2024). A seamless blended multi-model ensemble approach to probabilistic medium-range weather pattern forecasts over the UK. Meteorological Applications, 31(1), e2179. https://doi.org/10.1002/met.2179
Santos, J. A., M. Belo-Pereira, H. Fraga, and J. G. Pinto (2016), Understanding climate change projections for precipitation over western Europe with a weather typing approach, J. Geophys. Res. Atmos., 121, 1170–1189, doi:10.1002/2015JD024399.
Schiemann, R. and Frei, C., 2010: How to quantify the resolution of surface climate by circulation types: An example for Alpine precipitation. Physics and Chemistry of the Earth, Vol.35, 403-410.
Spuler FR, Kretschmer M, Kovalchuk Y, Balmaseda MA, Shepherd TG. Identifying probabilistic weather regimes targeted to a local-scale impact variable. Environmental Data Science. 2024;3:e25. doi:10.1017/eds.2024.29
Citation: https://doi.org/10.5194/egusphere-2026-3911-RC2 -
EC1: 'Comment on egusphere-2026-3911', Florian Pantillon, 06 Nov 2026
Dear authors,
Both referees have now returned their reports.
You will see that, while they recognize your effort at objectivizing the historical Hess-Brezowsky classification of “Grosswetterlagen”, the reports also point several issues. The main issue is that the paper is essentially descriptive and does not sufficiently discuss results and dynamical processes. Another issue is the lack of reproducibility of the algorithm that constitutes the core of the paper, which limits its interest to the community.
Please submit your final author comments on how, on a high level, you would address the major concerns of the referees. This may be beyond a typical round of revisions, thus, I also invite you to consider withdrawing the paper to prepare a substantially new version to submit to WCD or, alternatively, to a more technical or a data journal.
Best regards,
Florian PantillonCitation: https://doi.org/10.5194/egusphere-2026-3911-EC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 276 | 206 | 56 | 538 | 176 | 41 | 38 |
- HTML: 276
- PDF: 206
- XML: 56
- Total: 538
- Supplement: 176
- BibTeX: 41
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The paper describes the new version of objectivized classification of atmospheric circulation patterns after Hess-Brezowsky (HB) and provides its evaluation according to several metrics. The production of such a classification catalogue, with the advantage consisting, among others, in its homogeneity, is definitely a worthwhile exercise. I have, however, several substantial reservations regarding the paper that do not allow me to recommend it for publication.
Major comments
While a further step (the first step having been made by the first author twenty years ago) in the objectivization of the HB classification is certainly worthwhile, its relevance for the scientific community is diminished by the fact that this particular classification is not transferable from central Europe to anywhere else (unlike e.g. the Lamb / Jenkinson-Collison classification or any fully objective classification method). The international relevance of the updated HB classification should be explained and substantiated in much more detail, particularly by comparing with other classifications that are geographically transferable.
There is little scientific content in the paper; most of the text is devoted to the description of procedures how the updated classification is produced (Secs 2 and 3, extending on 17 pages out of the total of 30, not counting the reference list). Even the verification part contains a substantial amount of the description and explanation of verification criteria, and not of results. My opinion is that this paper would be much better suited (maybe with some rather minor modifications) to a data journal as a paper describing the new classification dataset.
The methodology of producing the updated classification is not explained enough. The procedure contains a number of choices that seem to be subjective and lack any support, either from data or from literature. Some examples follow (the list is not exhaustive): 1. “The MSPL field’s contribution is weighted to be around 27% higher” (l. 105): Why the exact number is not provided? And why 27%, and not let’s say 28 or 30? 2. Lamb weather types are calculated at six domains (l. 192, Fig. 3): why these domains, why just six? 3. How is the penalty weighting done? How are the correlation coefficients adjusted? (l. 200) 4. It is not clear how the AI feature set is constructed (l. 239) and how it is “converted into a weighted penalty” (l. 242). 5. “The window is shifted forward a few days” (l. 261): how many days exactly? 6. The description of how the individual types are grouped into GWG’s (Sec. 3) is vague; stating that “the grouping has been carefully developed” (l. 352) is not sufficient.
The assertion that there are several distinct circulation states (“several differing MSLP, Z500 and ZTH field combinations,” l. 130) grouped together in one circulation type, on which one part of the procedure is based, is questionable because atmospheric circulation forms a continuum, not discrete separate states.
The choice of verification scores is largely arbitrary. A more detailed justification for their use, including a support from literature, is needed. Why are the scores that have been commonly used for evaluation of classifications (e.g., Beck and Philipp 2010) not applied here? Also, the description of the scores is not detailed enough (e.g., what are the “penalty points” in Eq. (1), l. 432?) and bears considerable subjectivity (why square root in Eq. (3) and not another function?, why “the number of days in … two classes should be balanced” (l. 452)?, why the limit for the frequency bias is 0.8, and not 1.0 or any other reasonable number (l. 457)?).
Only a small set of classifications (in fact two) is used as a benchmark for the updated GWL-REA classification. However, whenever a new method is developed, its superiority must be demonstrated clearly, so that its existence is substantiated. For such a clear demonstration of superiority, the new method must be compared with a representative (and sufficiently large) set of classifications.
Only precipitation is used as a target variable for the updated classification. Why is this choice made; why e.g. temperature is entirely ignored in the development of the classification and in the verification (with the exception of one figure in the evaluation)?
Minor comments
The current version of the classification uses the same abbreviation (GWL-REA) as its predecessor from twenty years ago (James 2006, 2007), which is highly confusing.
The German description of the types in Tab. 1 is redundant.
The feature that several types are assigned to two different groups of types (GWG’s) (Fig. 6 and Tab. 2) is impractical.
References (only those not included in the manuscript are provided here)
Beck C, Jacobeit J, Jones PD 2007: Frequency and within-type variations of large-scale circulation types and their effects on low-frequency climate variability in Central Europe since 1780. Int J Climatol 27, 473-491
Beck C, Philipp A 2010: Evaluation and comparison of circulation type classifications for the European domain. Phys Chem Earth 35, 374-387