the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Dec 2022
A Weather Type classification based on the CESM-LE over the Middle Americas
Abstract. In this study, two classifications of 20 Weather Types (WTs) were used to identify large-scale and synoptic-scale patterns over the Middle Americas region (MAR) that comprises Mexico, intra-American seas, Central America, and northern South America. The Self-Organizing Maps (SOM) method was used to detect both classifications using standardized Mean Sea-Level Pressure (MSLP) anomalies from the ERA-Interim (ERA-I) reanalysis and the Community Earth System Model-Large Ensemble (CESM-LE) in the historical period and its future projection under an RCP8.5 scenario. The WTs obtained with the CESM-LE in the historical period were assigned to each day of the future projection. Averages of the days belonging to each WT of the historical period were compared between both classifications (ERA-I and CESM-LE) employing seasonal and monthly frequencies of occurrence, correspondence in days of occurrence, Pearson's spatial correlation, and position changes of high-pressure semi-permanent centers such as the North Atlantic Subtropical High (NASH) and the North Pacific High (NPH). From precipitation and MSLP, it was observed that WTs obtained with CESM-LE showed a marked seasonality in their temporal distribution, mainly in the wet period (May–October), similar to the ERA-I classification. Three characteristic phenomena of MAR were the North American Monsoon System (NAMS), the Mid-Summer Drought (MSD), and the Caribbean Low-Level Jet (CLLJ). The CESM-LE adequately represented these phenomena in the historical period compared with ERA-I. Regarding the future projection, the CESM-LE ensemble showed that the spatial patterns were very similar in the historical period. However, differences in precipitation between August and September decreased. To assess the effect of internal climate variability of the CESM-LE, we analyzed the spatial average of precipitation in two regions: NAMS and MSD, for the 34 members of the ensemble in the future projection. In the CLLJ region, differences between the historical and the future projection in terms of averages of the zonal-wind component and precipitation were less than 1 %. This analysis showed that the SOM technique detected the signal of climate change on a regional scale without being affected by the internal global variability of the model. Therefore, SOM emerged as a useful tool for the analysis of numerical experiments such as CESM-LE.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(6241 KB)
Interactive discussion
Status: closed
- RC1: 'Comment on egusphere-2022-1332', Anonymous Referee #1, 05 Jan 2023
-
RC2: 'Comment on egusphere-2022-1332', Tim Woollings, 20 Mar 2023
Review of ‘A Weather Type classification based on the CESM-LE over the Middle Americas’ by Cala-Pérez et al
This paper uses a Self-Organising Map approach to characterise the climate of the Central American region in a large ensemble of historical and future climate model simulations. The work presented shows that this method has potential to provide useful information on how atmospheric circulation patterns relate to climate impacts in the simulations and how these compare to reanalyses. For example, the section linking the weather types to features such as the Caribbean Low-Level Jet shows that the authors can usefully connect circulation features and impacts to the statistical analysis with this approach. The analysis shows quite robust changes in the Mid-Summer Drought in the model, which is also a useful result.
However, I have several concerns over the paper and I think that considerable effort will be required to deliver on the potential promised. As a result, I think that this manuscript is not suitable for publication in WCD. I have outlined my concerns below along with suggested ways to improve the research. I would encourage the authors that this avenue of study is worthwhile and I hope that the feedback is useful for developing a future manuscript for publication, either in WCD or elsewhere.
Major Points:
1. Appropriateness of the method
The SOM method has been used extensively, as described by the authors, but I think its application here could be more convincing. One of the main problems is that there are many Weather Types (WTs) and that lots of these seem to be very similar. This is evidenced by several of the maps and also the high spatial correlations presented in the paper. Choosing to use fewer WTs may help here, and I also wonder whether other aspects of the approach could be reconsidered. For example, MSLP is indeed commonly used in this type of analysis, but is that the best approach in this tropical/subtropical region? It seems that the higher latitude parts of the region may be dominating, where the Coriolis paramater is larger and so winds are more closely linked to MSLP.
2. Objectivity
The application of the method could be more objective in several regards. The choice of 20 WTs is a good example, as it is not clear that the subjective choice of 20 is working well here. The analysis often chooses pairs of WTs to focus on later, which again seems like a subjective choice. It is not clear how much of the relevant data variance is included or excluded in these analyses. It is hard to defend the choice of one pair of related WTs over others which are also highly correlated. How do we know that conclusions based on a choice of particular WTs are fully representative of the model?
3. Reproducability
I think that more details of the method need to be given so that the analysis of this paper could be repeated by others. I appreciate that more detail is given in some of the papers cited, so perhaps that just needs clarifying. But also some more information on how the method works would considerably improve the utility of the present paper. For example, what is the SOM algorithm trying to achieve and how does it handle uncertainty? What does the ranking of WTs mean, since the numbering of patterns does not seem to be random? Is number 1 more important than 2, 3, … etc?
4. Analysis
I thought that some other aspects of the analysis could also be improved. For example, the significance test applied seems quite weak. Hence, there are cases where some ensemble members show significant increases in a quantity while other members show a significant decrease in the same quantity; presumably both cannot be meaningful (fig 11). The ‘correspondence in days’ analysis also does not seem suitable to me - why should individual calendar days correspond between model and reanalysis? This seems at best to be an inefficient way to quantify the similarity of the model and reanalysis patterns.
5. Framing
Overall, I thought it might help to frame the article more on scientific aims and conclusions rather than methodological ones. It seems there are three distinct aims: 1) to evaluate the applicability of the SOM method, 2) to evaluate the performance of the model, and 3) to use the method to characterise the projected climate changes in this region. I would try to separate these out more in the paper, right from the introduction, and ensure that there are clear conclusions relating to each of these aims which are supported by quantitive analysis. Items 2 and 3 on this list would seem to be of most interest to the wider community, while at present I feel that most of the conclusions relate to item 1. This framing should help to ensure that the most useful conclusions are fully supported by the analysis. As one example, CESM is suggested to perform better in the wet period than the dry period; can statements like this be better quantified? Figure 11 shows some clearly interesting and robust changes in the mid-summer drought; are WTs really needed to extract these changes?
6. Literature
The paper does well in linking to other literature on SOMs, but I think it could go further in connecting to some broader topics, in particular 1) literature that includes relevant assesment of CESM in simulating tropical / subtropical climate and its variability, 2) literature on climate change in the local (and broader region). The latter of these is important to connect more strongly to the science aims, as above; what is our understanding of how climate change will affect this region, how does this depend on specific theories and models, and what are the key uncertainties? It is also important to more clearly separate this paper from Ochoa-Moya et al (2020, Atmosphere) which is similar in many aspects, albeit focused exclusively on reanalysis.
7. Introduction
The introduction could do with more structure, perhaps using the framing suggested above. It should review the relevant literature more broadly, and with more synthesis; at present it feels a bit like a list of relevant papers and what they did. A more structured and synthetic approach should help to set up the science questions better. One specific comment is that often the authors refer to ‘changes’ when more specific words would be better, such as bias, response, evolution, discrepency etc.
Citation: https://doi.org/10.5194/egusphere-2022-1332-RC2 -
EC1: 'Comment on egusphere-2022-1332', Tim Woollings, 20 Mar 2023
I’d like to thank the authors for their submisson to WCD. I apologise that this process has taken a little while; we have been let down by a reviewer. In order not to delay the process any further I have reviewed the paper myself. You should see this as a review comment in the discussion of the paper. I am sorry to say that my own assessment is quite consistent with that of Reviewer 1. As a result, and after discussion with the executive editors, I am afraid that I will likely have to reject the paper. You are, of course, entitled to post a final response in the online discussion and I will gladly consider any suggestions you have before proceeding.
Both the reviewer and I had serious concerns over the methodology and analysis in the paper, and hence that the conclusions were not supported by the evidence given. We have both provided extensive comments which I hope will be of use in taking this work further. We both see merit in your work and I believe there is potential for this to lead to a high quality and very interesting publication in the future. However, the nature and extent of the suggestions made imply that it would be quite a different paper to the current manuscript, and so I believe a rejection is appropriate. I hope you have the opportunity to act on this feedback and that this ultimately leads to a successful publication in the future, either in WCD or elsewhere.
Regards
Tim Woollings
Citation: https://doi.org/10.5194/egusphere-2022-1332-EC1
Interactive discussion
Status: closed
- RC1: 'Comment on egusphere-2022-1332', Anonymous Referee #1, 05 Jan 2023
-
RC2: 'Comment on egusphere-2022-1332', Tim Woollings, 20 Mar 2023
Review of ‘A Weather Type classification based on the CESM-LE over the Middle Americas’ by Cala-Pérez et al
This paper uses a Self-Organising Map approach to characterise the climate of the Central American region in a large ensemble of historical and future climate model simulations. The work presented shows that this method has potential to provide useful information on how atmospheric circulation patterns relate to climate impacts in the simulations and how these compare to reanalyses. For example, the section linking the weather types to features such as the Caribbean Low-Level Jet shows that the authors can usefully connect circulation features and impacts to the statistical analysis with this approach. The analysis shows quite robust changes in the Mid-Summer Drought in the model, which is also a useful result.
However, I have several concerns over the paper and I think that considerable effort will be required to deliver on the potential promised. As a result, I think that this manuscript is not suitable for publication in WCD. I have outlined my concerns below along with suggested ways to improve the research. I would encourage the authors that this avenue of study is worthwhile and I hope that the feedback is useful for developing a future manuscript for publication, either in WCD or elsewhere.
Major Points:
1. Appropriateness of the method
The SOM method has been used extensively, as described by the authors, but I think its application here could be more convincing. One of the main problems is that there are many Weather Types (WTs) and that lots of these seem to be very similar. This is evidenced by several of the maps and also the high spatial correlations presented in the paper. Choosing to use fewer WTs may help here, and I also wonder whether other aspects of the approach could be reconsidered. For example, MSLP is indeed commonly used in this type of analysis, but is that the best approach in this tropical/subtropical region? It seems that the higher latitude parts of the region may be dominating, where the Coriolis paramater is larger and so winds are more closely linked to MSLP.
2. Objectivity
The application of the method could be more objective in several regards. The choice of 20 WTs is a good example, as it is not clear that the subjective choice of 20 is working well here. The analysis often chooses pairs of WTs to focus on later, which again seems like a subjective choice. It is not clear how much of the relevant data variance is included or excluded in these analyses. It is hard to defend the choice of one pair of related WTs over others which are also highly correlated. How do we know that conclusions based on a choice of particular WTs are fully representative of the model?
3. Reproducability
I think that more details of the method need to be given so that the analysis of this paper could be repeated by others. I appreciate that more detail is given in some of the papers cited, so perhaps that just needs clarifying. But also some more information on how the method works would considerably improve the utility of the present paper. For example, what is the SOM algorithm trying to achieve and how does it handle uncertainty? What does the ranking of WTs mean, since the numbering of patterns does not seem to be random? Is number 1 more important than 2, 3, … etc?
4. Analysis
I thought that some other aspects of the analysis could also be improved. For example, the significance test applied seems quite weak. Hence, there are cases where some ensemble members show significant increases in a quantity while other members show a significant decrease in the same quantity; presumably both cannot be meaningful (fig 11). The ‘correspondence in days’ analysis also does not seem suitable to me - why should individual calendar days correspond between model and reanalysis? This seems at best to be an inefficient way to quantify the similarity of the model and reanalysis patterns.
5. Framing
Overall, I thought it might help to frame the article more on scientific aims and conclusions rather than methodological ones. It seems there are three distinct aims: 1) to evaluate the applicability of the SOM method, 2) to evaluate the performance of the model, and 3) to use the method to characterise the projected climate changes in this region. I would try to separate these out more in the paper, right from the introduction, and ensure that there are clear conclusions relating to each of these aims which are supported by quantitive analysis. Items 2 and 3 on this list would seem to be of most interest to the wider community, while at present I feel that most of the conclusions relate to item 1. This framing should help to ensure that the most useful conclusions are fully supported by the analysis. As one example, CESM is suggested to perform better in the wet period than the dry period; can statements like this be better quantified? Figure 11 shows some clearly interesting and robust changes in the mid-summer drought; are WTs really needed to extract these changes?
6. Literature
The paper does well in linking to other literature on SOMs, but I think it could go further in connecting to some broader topics, in particular 1) literature that includes relevant assesment of CESM in simulating tropical / subtropical climate and its variability, 2) literature on climate change in the local (and broader region). The latter of these is important to connect more strongly to the science aims, as above; what is our understanding of how climate change will affect this region, how does this depend on specific theories and models, and what are the key uncertainties? It is also important to more clearly separate this paper from Ochoa-Moya et al (2020, Atmosphere) which is similar in many aspects, albeit focused exclusively on reanalysis.
7. Introduction
The introduction could do with more structure, perhaps using the framing suggested above. It should review the relevant literature more broadly, and with more synthesis; at present it feels a bit like a list of relevant papers and what they did. A more structured and synthetic approach should help to set up the science questions better. One specific comment is that often the authors refer to ‘changes’ when more specific words would be better, such as bias, response, evolution, discrepency etc.
Citation: https://doi.org/10.5194/egusphere-2022-1332-RC2 -
EC1: 'Comment on egusphere-2022-1332', Tim Woollings, 20 Mar 2023
I’d like to thank the authors for their submisson to WCD. I apologise that this process has taken a little while; we have been let down by a reviewer. In order not to delay the process any further I have reviewed the paper myself. You should see this as a review comment in the discussion of the paper. I am sorry to say that my own assessment is quite consistent with that of Reviewer 1. As a result, and after discussion with the executive editors, I am afraid that I will likely have to reject the paper. You are, of course, entitled to post a final response in the online discussion and I will gladly consider any suggestions you have before proceeding.
Both the reviewer and I had serious concerns over the methodology and analysis in the paper, and hence that the conclusions were not supported by the evidence given. We have both provided extensive comments which I hope will be of use in taking this work further. We both see merit in your work and I believe there is potential for this to lead to a high quality and very interesting publication in the future. However, the nature and extent of the suggestions made imply that it would be quite a different paper to the current manuscript, and so I believe a rejection is appropriate. I hope you have the opportunity to act on this feedback and that this ultimately leads to a successful publication in the future, either in WCD or elsewhere.
Regards
Tim Woollings
Citation: https://doi.org/10.5194/egusphere-2022-1332-EC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 316 | 105 | 21 | 442 | 10 | 7 |
- HTML: 316
- PDF: 105
- XML: 21
- Total: 442
- BibTeX: 10
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1