the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Sep 2024
The Role of Weather Regimes for Subseasonal Forecast Skill of Cold-Wave Days in Central Europe
Abstract. Weather regimes (WRs) represent the large-scale tropospheric flow and therefore may contain useful information about the subseasonal predictability of cold waves, one of the most severe weather extremes in Central Europe. Firstly, we investigate in how far the succession of WRs during a forecast can be used to explain skill differences of forecasts initialized during different WRs. As an example, we use the skill differences of mean-bias-corrected 14-day reforecasts of the European Centre for Medium-Range Weather Forecasts for the occurrence of wintertime cold-wave days in Central Europe. Reforecasts initialized during the WR Greenland Blocking (GL; characterized by a high pressure system over Greenland) show the best Brier skill while those initialized during the WR Scandinavian Trough (ScTr; characterized by a low pressure system over Scandinavia) show the worst skill compared to a climatological ensemble for the winters 2000/2001–2019/2020. We find, that for forecasts initialized during GL, more often WR succession which follow typical climatological pattern are found during the 14 days of forecasts than for forecasts initialized during ScTr. We suggest that this is one of the main reasons for an increased forecast skill of predictions initialized during GL in contrast to predictions initialized during ScTr. Secondly, we analyze the WR succession for the best (worst) predicted days within the observed cold waves in the winters 2000/2001–2019/2020 independent from the WR present at initialization. We find, that forecast skill is significantly higher, when the European Blocking WR (characterized by a high pressure system over the British Isles and southern Scandinavia) is present a few days before the predicted cold-wave day. These results can be used to assess the reliability of cold-wave day predictions at the subseasonal lead time of 14 days.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(4773 KB)
-
Supplement
(918 KB)
-
This preprint has been withdrawn.
- Preprint
(4773 KB) - Metadata XML
-
Supplement
(918 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2955', Anonymous Referee #1, 10 Oct 2024
This work described in this manuscript extends earlier studies by the same authors, focusing on a thorough analysis of the connection of weather regimes (and their succession) with the predictability of cold-wave days in Central Europe. The analysis shows that more common ('climatological') WR successions tend to be more predictable than uncommon WR successions, while other factors like the number of regime transitions between forecast initialization and valid time did not show a clear association with forecast skill. The paper is interesting, but some clarifications and more evidence for the main conclusion is required as detailed below.
General comment:
The main conclusion of the manuscript is that among the different WR-related explanations of increased/decreased predictability the frequency of WR successions following climatological patterns plays an important role. This conclusion is primarily based on the observation that 61.6% vs. 53.9% of a subselected set of cases follows such climatological patterns. That difference is noticeable but not huge, and given the additional complication due to the subselection criterion (only the most frequent WR successions per WR at the target date are considered), which presumably has the effect of amplifying the observed difference, I feel that more evidence for this conclusion should be provided. Would it be possible, for example, to calculate the Brier score for forecasts with the GL/ScTr WR at initialization time separately for the cases where the WR successions do and do not follow a climatological pattern and test whether the score differences are statistically significant?
Specific comments:
- Section 2.2: Are the ECMWF reforecasts also temporally smoothed (like the observation data), or is that unnecessary due to the subsequent post-processing?
- 131: Aren't these just forecast errors of an ensemble mean forecast? I find it strange to call them biases, which to me is a systematic error, while without further aggregation the quantities calculated here contain (a substantial amount of) random forecast errors as well.
- Section 3.2, 2nd paragraph: More detail is required for this ERA5-based predictor. Is ERA5 data at the different hours from the day before initialization time used here? Can you briefly describe the preprocessing operations mentioned in 149?
- 189: I was very confused about this concept of 'hypothetical' forecasts when I read it here, and understood only later that it's not really a forecast, but that the weather regimes on these dates can still be analyzed. Maybe this can already be clarified here.
- 235-236: I don't understand what is meant by 'single actual WR successions', and found this sentence very confusing. This paragraph is generally hard to follow, but it becomes clear what is studied here in connection with Figure 4. The aforementioned sentence, however, could easily be removed without loss of information.
- 267-268: I don't understand what is meant by 'without taking persistence of the individual WRs per se into account'. What if the WR at initialization time persists for the 14 days lead time? Please rephrase and/or explain.
- 275: Perhaps clearer to say '..., the number of possible WR successions varies ...'
Typos and language:
72: -> their skill
130: Therefore -> To this end
139: Either "the ECMWF S2S reforecast ensemble" or "ECMWF's S2S reforecast ensemble"
159: Please check this reference, I have never seen a citation with a range of publication years beforeCitation: https://doi.org/10.5194/egusphere-2024-2955-RC1 - AC1: 'Reply on RC1', Selina Kiefer, 19 Dec 2024
-
RC2: 'Comment on egusphere-2024-2955', Anonymous Referee #2, 22 Nov 2024
Review of "The role of weather regimes for subseasonal forecast skill of cold-wave days in Central Europe"
This study attempts to relate Euro-Atlantic weather regimes (and their transitions) to the predictability of severe cold events in Central Europe in ECMWF forecasts. However, I have several major concerns with this study, which make it unsuitable for publication in its present form. In addition, I found the manuscript to be rather muddled and difficult to read.
Main comments:(1) The authors claim to investigate how the transition of weather regimes "during a forecast" can be used to explain differences in the predictability of extreme cold events from forecasts initialised during different regimes. From my perspective, this should include some assessment of the links between cold events and weather regimes within the forecasts themselves. At the most basic level, it would have been useful to show how the frequency of cold events is modulated by regimes in both observations and forecasts. For example, a prerequisite for the presented analysis is a demonstration that forecasts accurately capture the location and magnitude of (lagged) relationships between the regimes and surface impacts. Instead, the authors limit their analysis to the observed regime behaviour during the forecast period. This leads to conclusions such as "WR successions following typical climatological patterns might be easier to forecast and thus leading to an increased forecast skill". However, this conclusion remains speculation while there is no assessment of the regime behaviour within the associated forecasts. In particular, the is no demonstration that the forecasts with increased skill for the prediction of cold events have actually simulated the observed regime transition. Although it is physically plausible to link large scale regimes to extreme cold events, there could be other explanations for the success of a given forecast that are unrelated to the regimes considered.
(2) The introduction should include a clear and concise description of the hypotheses to be tested and the associated diagnostics. Some of this information is embedded in the results section and should be moved to the introduction to be structured something like: "We hypothesise that it is more difficult to forecast extreme cold events when Euro-Atlantic regimes exhibit the following characteristics within the forecast period: (i) XXX, (ii) XXX, (iii) XXX. We measure these characteristics using the following metrics...(i) XXX, (ii) XXX...". The authors also need to expand on their physical reasoning and explicitly discuss their interpretation of the links between regime transitions and forecast uncertainty. However, as stated in comment (1), some of these metrics should include information from the forecasts. It is not possible to conclude that cold event forecasts are good/bad because they have predicted a particular regime transition, without assessing this property of the forecasts.
(3) It is not clear why the random forecast calibration method is emphasised in the introduction and methods and then only mentioned in a single sentence of the results (line 377).
(4) The authors present a lot of descriptive statistics derived from very small samples (e.g. descriptions of figures 4 and 5). However, with eight classes (7 WR + Null regime) this means there are 64 possible transitions to consider and it is not clear if the derived "climatological" regime successions are statistically robust. For example, how do figures 4 and 5 compare when they are limited to the dates for which forecasts are available? The sampling uncertainty is something could be tested more easily using regimes derived from forecasts by subsetting from the available members, for example. Given the authors' conclude that "climatological successions" are easier to predict, I think they need to demonstrate that the climatology is robustly defined based on the available data.
Other comments:Title & throughout: The authors emphasise "subseasonal" forecasts, but the analysis is limited to a lead time of 14 days. I would call this "medium-range".
Abstract: "These results can be used to assess the reliability of cold-wave day predictions" - this has not been demonstrated.
Abstract and throughout: "we investigate in how far the succession of WRs during a forecast can be used to explain skill differences of forecasts initialized during different WRs" - I assume this refers to skill differences in the predictability of extreme cold events. However, this is ambiguous and I initially assumed it referred to the predictability of regimes themselves. This happens elsewhere in the paper where it would improve clarity to refer to the "skill of cold extreme predictions" rather than just "skill".
Line 24: "on that timescale" - ambiguous and unnecessary. Could delete "on that timescale, which comprises".
Line 48: The discussion of Greenland Blocking can be linked negative NAO.
Line 59 (and elsewhere): I find it confusing to have results from previous studies by the same author described in the present tense. In some cases this is done in the results section, which makes it difficult to distinguish what is new in the present study.
Line 65-66: At this point in the introduction it would be very useful to clearly describe the hypotheses to be tested and the associated diagnostics. See main comment (2).
Line 67 (and elsewhere): given the focus on conditional statistics and specific regime transitions, which dramatically reduces the available sample size, why limit analysis to specific winters?
Line 75: What is the motivation for using RF calibration? Also see main comment (3).
Section 2.1: This would benefit from separation into two sections containing (i) the description of the index to be predicted (i.e. methodology that is common to both forecasts and observations) and (ii) the construction of climatological ensemble.
Figure 1: The choice of colour scale means there is no distinction between ocean and land > 800m. Perhaps highlight the cold wave index region with a red contour or similar?
Line 96: Which IFS cycle is use?
Line 97: Are forecast data processed identically to the climatological ensemble? Is the same 7-day smoothing applied? If forecasts and reference data are not processed identically, this will systematically impact differences in Brier Score.
Section 2.3: This section could be shorter if it is based on the exact same methodology as Grams et al.
Section 3.1 This could be part of the description of the IFS data.
Line 130: "Reforecast ensemble [mean]"?
Line 131: I would rephrase this to "time series of errors" and use "bias" to describe the average/expectation of the error over many cases.
Section 3.2: I find it extremely odd to have a long section dedicated to the description of a calibration method that is then only mentioned in passing in the results section. See also main comment (3). The authors should either (i) provide a clear motivation of the use of the calibration method and how it helps understand the link between regimes and cold wave predictions and adequately discuss the results or (ii) remove this element of the paper.
Line 165: This notation of the Brier score is quite arcane. It is typically presented as the expectation of squared differences in forecast/observed probability for a specific event. https://en.wikipedia.org/wiki/Brier_score
Line 173: What is the motivation for use of BS differences rather than the more typical skill score form i.e. BSS = (BS_benchmark - BS_model)/BS_benchmark)?
Line 180: The results section begins with a statement of results from a previous study by the same author in the present tense. It is then extremely ambiguous as to whether the following sentence ("We find that the ..." is another result from the previous study or something new. It would be easier for the reader if previous results are kept in the past tense and initially described in the introduction, with reference as required in the results.
Lines 188-190 (and elsewhere, including figure captions): The description of observational data as "hypothetical forecasts" is extremely confusing. For example, Figure 3 is titled "Number of regime transitions during the forecast", which is confusing for two reasons: (1) this is not forecast data and (2) it is not limited to forecast dates. A more accurate description would be "Number of regime transitions in ERA5 within a 14 day moving window". Similarly, Figure 4 is title "Frequency of WRs in forecasts initialised during the GL regime". Again, this is not an accurate description. For this description (and associated "lead time" axis labels) to make sense, the authors would need to show the regime frequencies from the forecast model. It would then also be useful to compare with the observations limited to the dates of forecast initialisation.Line 195: Why GL and ScTr only?
Line 197: "GL is the dominant regime at the initialisation for forecast predicting the occurrence of (non-)cold-wave days" - what does this mean? GL is dominate regime for both cold-wave and non-cold-wave days?
Line 209: "GL regimes tend to be more persistent" - is this result robust for the limited sample size?
Lines 203-220 & 235-240: These paragraphs contain hypotheses that would be useful to include in the introduction, alongside additional discussion of how this links to forecast/skill uncertainty and description of the the proposed diagnostics.
Line 212: "number of active WRs regimes at each day" - what is the physical interpretation of multiple active regimes per day? If there is no clear distinction is there really a regime?
Line 223: "Regime changes might be more difficult to forecast than persistence" - again, this is speculation/hypothesis that should be introduced earlier rather than part of the results.
Figure 3: Plotting the median and interquartile range seems unnecessary for the small sample sizes. It would be more transparent to show every data point. For example, what does it mean in the case of GL->Zo and GL->GL that there is no interquartile range box? I assume there it means all contributing data points are equal but it would be clearer to just show the data.
Line 306: "Analogously as done for to the occurrence of (non-)cold-wave days, we investigate the WR characteristics during days within cold waves." I don't understand the distinction between (non-)cold-wave days and days without cold waves. Surely the set of non-cold-wave and cold-wave days includes all days?
Summary and conclusions: I don't think the authors can answer the proposed research question how is the "WR succession during a forecast linked to [...] forecast skill" without looking at weather regimes in forecasts.
Lines 443-449: The links between weather regimes and teleconnections are interesting but this topic should have been introduced earlier (e.g. in the introduction). I agree it may be interesting to investigate the modulation of weather regime transitions/successions by other modes of variability, but the sample sizes is a major challenge. The authors already implicitly stratify data into 64 possible cases (GL-Zo ...etc), which leaves very little data for further stratification by MJO/polar vortex phase, for example. Maybe the authors could consider what would be required to make progress in this area (e.g. reforecasts with very many start dates covering long periods?)
Citation: https://doi.org/10.5194/egusphere-2024-2955-RC2 - AC2: 'Reply on RC2', Selina Kiefer, 19 Dec 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-2955', Anonymous Referee #1, 10 Oct 2024
This work described in this manuscript extends earlier studies by the same authors, focusing on a thorough analysis of the connection of weather regimes (and their succession) with the predictability of cold-wave days in Central Europe. The analysis shows that more common ('climatological') WR successions tend to be more predictable than uncommon WR successions, while other factors like the number of regime transitions between forecast initialization and valid time did not show a clear association with forecast skill. The paper is interesting, but some clarifications and more evidence for the main conclusion is required as detailed below.
General comment:
The main conclusion of the manuscript is that among the different WR-related explanations of increased/decreased predictability the frequency of WR successions following climatological patterns plays an important role. This conclusion is primarily based on the observation that 61.6% vs. 53.9% of a subselected set of cases follows such climatological patterns. That difference is noticeable but not huge, and given the additional complication due to the subselection criterion (only the most frequent WR successions per WR at the target date are considered), which presumably has the effect of amplifying the observed difference, I feel that more evidence for this conclusion should be provided. Would it be possible, for example, to calculate the Brier score for forecasts with the GL/ScTr WR at initialization time separately for the cases where the WR successions do and do not follow a climatological pattern and test whether the score differences are statistically significant?
Specific comments:
- Section 2.2: Are the ECMWF reforecasts also temporally smoothed (like the observation data), or is that unnecessary due to the subsequent post-processing?
- 131: Aren't these just forecast errors of an ensemble mean forecast? I find it strange to call them biases, which to me is a systematic error, while without further aggregation the quantities calculated here contain (a substantial amount of) random forecast errors as well.
- Section 3.2, 2nd paragraph: More detail is required for this ERA5-based predictor. Is ERA5 data at the different hours from the day before initialization time used here? Can you briefly describe the preprocessing operations mentioned in 149?
- 189: I was very confused about this concept of 'hypothetical' forecasts when I read it here, and understood only later that it's not really a forecast, but that the weather regimes on these dates can still be analyzed. Maybe this can already be clarified here.
- 235-236: I don't understand what is meant by 'single actual WR successions', and found this sentence very confusing. This paragraph is generally hard to follow, but it becomes clear what is studied here in connection with Figure 4. The aforementioned sentence, however, could easily be removed without loss of information.
- 267-268: I don't understand what is meant by 'without taking persistence of the individual WRs per se into account'. What if the WR at initialization time persists for the 14 days lead time? Please rephrase and/or explain.
- 275: Perhaps clearer to say '..., the number of possible WR successions varies ...'
Typos and language:
72: -> their skill
130: Therefore -> To this end
139: Either "the ECMWF S2S reforecast ensemble" or "ECMWF's S2S reforecast ensemble"
159: Please check this reference, I have never seen a citation with a range of publication years beforeCitation: https://doi.org/10.5194/egusphere-2024-2955-RC1 - AC1: 'Reply on RC1', Selina Kiefer, 19 Dec 2024
-
RC2: 'Comment on egusphere-2024-2955', Anonymous Referee #2, 22 Nov 2024
Review of "The role of weather regimes for subseasonal forecast skill of cold-wave days in Central Europe"
This study attempts to relate Euro-Atlantic weather regimes (and their transitions) to the predictability of severe cold events in Central Europe in ECMWF forecasts. However, I have several major concerns with this study, which make it unsuitable for publication in its present form. In addition, I found the manuscript to be rather muddled and difficult to read.
Main comments:(1) The authors claim to investigate how the transition of weather regimes "during a forecast" can be used to explain differences in the predictability of extreme cold events from forecasts initialised during different regimes. From my perspective, this should include some assessment of the links between cold events and weather regimes within the forecasts themselves. At the most basic level, it would have been useful to show how the frequency of cold events is modulated by regimes in both observations and forecasts. For example, a prerequisite for the presented analysis is a demonstration that forecasts accurately capture the location and magnitude of (lagged) relationships between the regimes and surface impacts. Instead, the authors limit their analysis to the observed regime behaviour during the forecast period. This leads to conclusions such as "WR successions following typical climatological patterns might be easier to forecast and thus leading to an increased forecast skill". However, this conclusion remains speculation while there is no assessment of the regime behaviour within the associated forecasts. In particular, the is no demonstration that the forecasts with increased skill for the prediction of cold events have actually simulated the observed regime transition. Although it is physically plausible to link large scale regimes to extreme cold events, there could be other explanations for the success of a given forecast that are unrelated to the regimes considered.
(2) The introduction should include a clear and concise description of the hypotheses to be tested and the associated diagnostics. Some of this information is embedded in the results section and should be moved to the introduction to be structured something like: "We hypothesise that it is more difficult to forecast extreme cold events when Euro-Atlantic regimes exhibit the following characteristics within the forecast period: (i) XXX, (ii) XXX, (iii) XXX. We measure these characteristics using the following metrics...(i) XXX, (ii) XXX...". The authors also need to expand on their physical reasoning and explicitly discuss their interpretation of the links between regime transitions and forecast uncertainty. However, as stated in comment (1), some of these metrics should include information from the forecasts. It is not possible to conclude that cold event forecasts are good/bad because they have predicted a particular regime transition, without assessing this property of the forecasts.
(3) It is not clear why the random forecast calibration method is emphasised in the introduction and methods and then only mentioned in a single sentence of the results (line 377).
(4) The authors present a lot of descriptive statistics derived from very small samples (e.g. descriptions of figures 4 and 5). However, with eight classes (7 WR + Null regime) this means there are 64 possible transitions to consider and it is not clear if the derived "climatological" regime successions are statistically robust. For example, how do figures 4 and 5 compare when they are limited to the dates for which forecasts are available? The sampling uncertainty is something could be tested more easily using regimes derived from forecasts by subsetting from the available members, for example. Given the authors' conclude that "climatological successions" are easier to predict, I think they need to demonstrate that the climatology is robustly defined based on the available data.
Other comments:Title & throughout: The authors emphasise "subseasonal" forecasts, but the analysis is limited to a lead time of 14 days. I would call this "medium-range".
Abstract: "These results can be used to assess the reliability of cold-wave day predictions" - this has not been demonstrated.
Abstract and throughout: "we investigate in how far the succession of WRs during a forecast can be used to explain skill differences of forecasts initialized during different WRs" - I assume this refers to skill differences in the predictability of extreme cold events. However, this is ambiguous and I initially assumed it referred to the predictability of regimes themselves. This happens elsewhere in the paper where it would improve clarity to refer to the "skill of cold extreme predictions" rather than just "skill".
Line 24: "on that timescale" - ambiguous and unnecessary. Could delete "on that timescale, which comprises".
Line 48: The discussion of Greenland Blocking can be linked negative NAO.
Line 59 (and elsewhere): I find it confusing to have results from previous studies by the same author described in the present tense. In some cases this is done in the results section, which makes it difficult to distinguish what is new in the present study.
Line 65-66: At this point in the introduction it would be very useful to clearly describe the hypotheses to be tested and the associated diagnostics. See main comment (2).
Line 67 (and elsewhere): given the focus on conditional statistics and specific regime transitions, which dramatically reduces the available sample size, why limit analysis to specific winters?
Line 75: What is the motivation for using RF calibration? Also see main comment (3).
Section 2.1: This would benefit from separation into two sections containing (i) the description of the index to be predicted (i.e. methodology that is common to both forecasts and observations) and (ii) the construction of climatological ensemble.
Figure 1: The choice of colour scale means there is no distinction between ocean and land > 800m. Perhaps highlight the cold wave index region with a red contour or similar?
Line 96: Which IFS cycle is use?
Line 97: Are forecast data processed identically to the climatological ensemble? Is the same 7-day smoothing applied? If forecasts and reference data are not processed identically, this will systematically impact differences in Brier Score.
Section 2.3: This section could be shorter if it is based on the exact same methodology as Grams et al.
Section 3.1 This could be part of the description of the IFS data.
Line 130: "Reforecast ensemble [mean]"?
Line 131: I would rephrase this to "time series of errors" and use "bias" to describe the average/expectation of the error over many cases.
Section 3.2: I find it extremely odd to have a long section dedicated to the description of a calibration method that is then only mentioned in passing in the results section. See also main comment (3). The authors should either (i) provide a clear motivation of the use of the calibration method and how it helps understand the link between regimes and cold wave predictions and adequately discuss the results or (ii) remove this element of the paper.
Line 165: This notation of the Brier score is quite arcane. It is typically presented as the expectation of squared differences in forecast/observed probability for a specific event. https://en.wikipedia.org/wiki/Brier_score
Line 173: What is the motivation for use of BS differences rather than the more typical skill score form i.e. BSS = (BS_benchmark - BS_model)/BS_benchmark)?
Line 180: The results section begins with a statement of results from a previous study by the same author in the present tense. It is then extremely ambiguous as to whether the following sentence ("We find that the ..." is another result from the previous study or something new. It would be easier for the reader if previous results are kept in the past tense and initially described in the introduction, with reference as required in the results.
Lines 188-190 (and elsewhere, including figure captions): The description of observational data as "hypothetical forecasts" is extremely confusing. For example, Figure 3 is titled "Number of regime transitions during the forecast", which is confusing for two reasons: (1) this is not forecast data and (2) it is not limited to forecast dates. A more accurate description would be "Number of regime transitions in ERA5 within a 14 day moving window". Similarly, Figure 4 is title "Frequency of WRs in forecasts initialised during the GL regime". Again, this is not an accurate description. For this description (and associated "lead time" axis labels) to make sense, the authors would need to show the regime frequencies from the forecast model. It would then also be useful to compare with the observations limited to the dates of forecast initialisation.Line 195: Why GL and ScTr only?
Line 197: "GL is the dominant regime at the initialisation for forecast predicting the occurrence of (non-)cold-wave days" - what does this mean? GL is dominate regime for both cold-wave and non-cold-wave days?
Line 209: "GL regimes tend to be more persistent" - is this result robust for the limited sample size?
Lines 203-220 & 235-240: These paragraphs contain hypotheses that would be useful to include in the introduction, alongside additional discussion of how this links to forecast/skill uncertainty and description of the the proposed diagnostics.
Line 212: "number of active WRs regimes at each day" - what is the physical interpretation of multiple active regimes per day? If there is no clear distinction is there really a regime?
Line 223: "Regime changes might be more difficult to forecast than persistence" - again, this is speculation/hypothesis that should be introduced earlier rather than part of the results.
Figure 3: Plotting the median and interquartile range seems unnecessary for the small sample sizes. It would be more transparent to show every data point. For example, what does it mean in the case of GL->Zo and GL->GL that there is no interquartile range box? I assume there it means all contributing data points are equal but it would be clearer to just show the data.
Line 306: "Analogously as done for to the occurrence of (non-)cold-wave days, we investigate the WR characteristics during days within cold waves." I don't understand the distinction between (non-)cold-wave days and days without cold waves. Surely the set of non-cold-wave and cold-wave days includes all days?
Summary and conclusions: I don't think the authors can answer the proposed research question how is the "WR succession during a forecast linked to [...] forecast skill" without looking at weather regimes in forecasts.
Lines 443-449: The links between weather regimes and teleconnections are interesting but this topic should have been introduced earlier (e.g. in the introduction). I agree it may be interesting to investigate the modulation of weather regime transitions/successions by other modes of variability, but the sample sizes is a major challenge. The authors already implicitly stratify data into 64 possible cases (GL-Zo ...etc), which leaves very little data for further stratification by MJO/polar vortex phase, for example. Maybe the authors could consider what would be required to make progress in this area (e.g. reforecasts with very many start dates covering long periods?)
Citation: https://doi.org/10.5194/egusphere-2024-2955-RC2 - AC2: 'Reply on RC2', Selina Kiefer, 19 Dec 2024
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 165 | 55 | 116 | 336 | 18 | 4 | 4 |
- HTML: 165
- PDF: 55
- XML: 116
- Total: 336
- Supplement: 18
- BibTeX: 4
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Patrick Ludwig
Sebastian Lerch
Peter Knippertz
Joaquim G. Pinto
This preprint has been withdrawn.
- Preprint
(4773 KB) - Metadata XML
-
Supplement
(918 KB) - BibTeX
- EndNote