the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 16 Oct 2023
Applying global warming levels of emergence to highlight the increasing population exposure to temperature and precipitation extremes
Abstract. Global temperatures exceeded pre-industrial conditions by 1.1 °C during the decade 2011–2020 and further warming is projected by climate models. An increasing number of climate variables exhibit significant changes compared to the past decades, even beyond the noise of internal climate variability. To determine the year when climate change signals can be detected, the concept of time of emergence (ToE) is well established. Additionally, climate projections are communicated increasingly frequently through global warming levels (GWLs) rather than time horizons. Yet, ToE and GWL have barely been combined so far. Here, we apply five Single Model Initial-condition Large Ensembles (SMILEs) to derive global warming levels of emergence (GWLoE) of four temperature and precipitation indices. We show that the concept of GWLoE is particularly promising to constrain temperature projections and proves a viable tool to communicate scientific results. We find that >75 % of the global population is exposed to emerged signals for nighttime temperatures at a GWL of 1.5 °C, increasing to >95 % at 2.0 °C. Daily maximum temperature follows a similar, yet less pronounced path. Emerged signals for mean and extreme precipitation start appearing at current GWLs and increase steadily with further warming (~20 % population exposed at 2.0 °C). Related probability ratios for the occurrence of extremes indicate a strong increase where temperature extremes reach widespread saturation (extremes occur every year) particularly in (sub)tropical regions below 2.5 °C warming. These results indicate that current times are a critical period for climate action as every fraction of additional warming substantially increases the adverse effects on human wellbeing.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(0 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2126', Anonymous Referee #1, 12 Nov 2023
Review of "Applying global warming levels of emergence to highlight the increasing population exposure to temperature and precipitation extremes"
This study uses large single model ensembles to explore projections of climate extreme indices at global warming levels using an emergence-based methodology. The authors not only look at emergence of climate extremes, but also exposure to emergence and examine effects of methodological choices (order of operations).
This is a robust analysis and a well-presented study. I'm confident that it will make a useful contribution to the literature. I do have three major comments for the authors to consider though:
Major comments:
- A benefit of SMILEs is that they can be used to explore sampling as well as structural uncertainties. I think with Figure 3 in particular it would be useful to show a range of area of emergence as a function of GWL for each SMILE. This could be derived from bootstrapping the simulations and computing a confidence interval based on the resampled ensembles.
- Some SSP population projections aren't particularly compatible with some emissions pathways. As such, SSP1 population projection is unlikely to be compatible with having a high GWL. I would suggest that SSP5 populations are used in Figure 4 and that uncertainty estimation through bootstrapping (as discussed in the previous comment) is shown instead.
- The results shown are applicable to the climate under a very high rate of global warming, but it should be noted that they aren't applicable to slower warming or stabilised climate states (e.g. King et al. 2020 (https://www.nature.com/articles/s41558-019-0658-7) and Mitchell et al. 2016 (https://www.nature.com/articles/nclimate3055)).
Minor comments:
Figure 1: Could you check that the lines are plotted correctly. For some models the range appears considerably smaller than due to interannual variability alone in the observations. In Maher et al. 2019 (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001639), the MPI range looks larger than is plotted here.
L190: There's a strange space that should be removed.
L200: "less" should be "fewer"
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC1 -
AC1: 'Reply on RC1', David Gampe, 16 Nov 2023
Thank you for the valuable inputs and comments and for your time and effort to critically review our manuscript. We will prepare a revised version once the discussion period is over. Yet, we want to stimulate the discussion by replying to your points already at this stage.
- Bootstrapping: We highly appreciate your suggestion to use bootstrapping to derive an uncertainty estimation of sampling the GWLs. We will consider this and will update Figure 3 (and related Figs in the SI) accordingly. We will apply bootstrapping to sample the n members of each SMILE from the available ensemble members. While we aim for a large number of samples the final decision on how many samples we will draw is yet to be determined. We will then update the Figures plotting the median/mean and the 5-95 percentile (or inter-quartile) range to indicate uncertainties.
- Compatibility of SSPs: we agree that not all SSPs are compatible with the selected SSP585 scenario. Nevertheless, we argue that one advantage of the application of GWLs is constraining scenario uncertainties. While the selection of the appropriate SSP is certainly a prerequisite we decided to include different spatially explicit estimates of future population distribution here to indicate the related uncertainties. We considered this a more appropriate assessment of uncertainties for these figures. However, we will explore also including the bootstrapped GWLs as suggested.
- Literature suggestions: Thank you for highlighting the aspect of slower warming and/or stabilized climate. We will incorporate the suggested publications and expand the related sections in the discussion.
- GWL range Figure 1: Thank you for critically reviewing this Figure. The GWLs for each ensemble were calculated using a 20-year moving window. This ultimately leads to a smoothening for individual years thus reducing the overall range considerably. As we then apply the GWLs based on the 20-yr window for further calculation of our results we decided to only show these in Fig.1. Nevertheless, we will mention this explicitly in the caption to improve this aspect.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC1
-
AC1: 'Reply on RC1', David Gampe, 16 Nov 2023
-
RC2: 'Comment on egusphere-2023-2126', Anonymous Referee #2, 29 Nov 2023
General comments:
This study presents the global warming level of emergence (GWLoE) of four temperature and precipitation indices, and the related exposure of population and land area based on the joint emergence of five SMILEs. While the finding is interesting, the manuscript needs much improvement regarding its writing.
Major Comments:
- The abstract needs to be reformatted. In the current version, the authors introduce the concept of ToE and GWLs, and mention that ‘ToE and GWL have barely been combined so far’. Yet, the scientific question of the research is not clear. The authors should clearly crystalize the specific problem/knowledge gap the study is aiming to address, i.e., the shortcomings of current research and why combining ToE and GWLs is significant.
- Again, there is a need to revise the Introduction to make it more accessible, particularly the current research progress and the knowledge gap. After reading the introduction, I am not clear about the current research progress and still have doubts about the importance of the paper. The authors only mention that “Recently, first studies combined GWL and ToE to provide global warming levels of emergence (GWLoE) instead of ToE (Abatzoglou et al. 2019, Kirchmeier-Young et al. 2019, Raymond et al. 2020). Yet, GWLoE remains a rarely applied concept in general as well as in the context of using SMILEs in particular”. The readers should be aware of the knowledge gap and be convinced that it is important to fill the knowledge gap after finishing reading the Introduction.
- As for the climate indices, why choose these four of different types? The temperature indices (TXx and TNx) are absolute indices; precipitation index PRCPtot is one of the amount indices, and R1Xday is the intensity index. Why not choose the indices of the same group for both temperature and precipitation?
- Why resample to the coarsest resolution (2.8°, CanESM5)? To what extent the selection of the resample resolution affects the exposure of land area and population? This should be clarified.
- Figure 1, the colored lines are very smooth, indicating almost no inter-annual variability (IAV). Has the time series been smoothed? The author should explain that.
- Figure 2, considering that there is almost no visible spatial difference in Figure b, the color bar can use unequal spacing instead of equal spacing.
Specific Comments:
- Many sentences lack commas, making them difficult to read. For example, Line 54, a comma is needed between “To disentangle a robust climate change signal from the background noise of internal climate variability” and “Single Model Initial-condition Large Ensembles (SMILEs) are widely used (e.g., Deser et al. 2020, Maher et al. 2021)”.
- Line 45. ‘W/m2’ ->’ W/m2’.
- Line 422. ‘proofs’ ->’ proves’.
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC2 -
AC2: 'Reply on RC2', David Gampe, 07 Dec 2023
We highly appreciate the constructive comments to improve our manuscript and acknowledge the time and effort to review our submission. To stimulate further discussion, we reply already at this stage prior to revising the manuscript.
- Highlighting knowledge gap/scientific merit: Thank you for critically screening our abstract and the introduction and pointing out the need to sharpen the corresponding sections. We agree that the knowledge gaps we address in this paper are not completely coming through at this stage. We will carefully revise the sections to highlight the novelty of and the need for our study. Certainly, we will emphasize the increased assessment of uncertainties using multiple SMILEs to derive the joint emergence in conjunction with increased sampling of internal variability through bootstrapping GWL/ToE/GWLoE. Both, the abstract and the introduction will be streamlined and carefully revised accordingly. We will also revise the entire manuscript to improve the language and readability of the paper.
- Selection of climate indices: we agree with the reviewer that the selected indices differ in their type and representation. However, we rather see this variety as a strength then a shortcoming as they cover absolute (extremes), quantities and intensity of temperature and precipitation. It was not necessarily the intention of our study to compare the indices with one another and rather showcase the potential of GWLoE for them. Further, we chose these indices as they are extremely frequently applied in other studies (also in the context of GWLs). In our opinion, it is thus of particular interest to address the aspects of uncertainties around GWLs and the merits of GWLoE for these indices.
- Resampling to CanESM5 resolution: Thank you for this valid and important comment. In our study we aim to address two major sources of uncertainty: internal variability (hence the use of SMILEs) and structural uncertainty (multiple SMILEs). One of the foci of our study is the joint emergence using multiple SMILEs to increase the robustness of our results (considering also structural uncertainty) on a grid scale. This, however, requires to harmonize the SMILEs to a common grid as otherwise coarser models would be penalized for their resolution, likely skewing the results. While this step is in our opinion crucial, we acknowledge and highlight the effect of resampling order for the calculation of the indices in the discussion and the supplement already. However, we agree that our reasoning to resample is partly lacking in our manuscript at the current stage and will revise the corresponding methods section accordingly.
- GWL range Figure 1: Thank you for critically reviewing this Figure. The GWLs for each ensemble were calculated using a 20-year moving window. This ultimately leads to a smoothing for individual years thus reducing the overall range considerably. As we then apply the GWLs based on the 20-yr window for further calculation of our results we decided to only show these in Fig.1. Nevertheless, we will mention this explicitly in the figure caption to improve this aspect.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC2
-
RC3: 'Comment on egusphere-2023-2126', Anonymous Referee #3, 30 Nov 2023
The study “Applying global warming levels of emergence to highlight the increasing population exposure to temperature and precipitation extremes” by Gampe et al assesses the emergence of changes in selected climate indices when temperatures are increasing, and also looks at how large a fraction of land areas and population experiences changes in these climate indices. An interesting novelty of this work is the use of Global Warming Levels (GWL) instead of time of emergence to disentangle the scenario uncertainty and the differences in ECS between different models from the response of the climate system to global warming. The main findings from the paper are that temperature related indices have already changed significantly at 1 deg warming relative to pre-industrial levels over large areas and fractions of the population, while changes in precipitation related indices are emerging slower and will affect large portions of land/population only with stronger warming. The paper is clear and concisely written, and the GWLoE form an interesting method to present climate change information in a novel way. There are however a few points where the authors could try to further improve the paper.
What is the motivation to use an ensemble of SMILEs? The reason for using a SMILE is that it eliminates uncertainty that stems from using different models yet retains the internal variability (as the authors mention several times in the paper). However, combining several SMILEs again introduces uncertainty from differences between the different model, and you give away the main advantage of a SMILE. Could you comment on that?
The paper hasn’t convinced me that an ensemble of SMILEs is essential for this type of analysis. How different would the results have been if an ensemble had been created from all CMIP6 models instead, say 1 member from each model that contributed to CMIP6? Would this change the finding? Or the robustness?
The paper lacks a proper assessment of the robustness of the results. How robust are the estimates for the fraction of land or fraction of population that experience a change in climate indicators? Ensembles in general are very useful for quantifying confidence intervals or estimating uncertainty, so an ensemble of large ensembles should be even better to assess uncertainties, or?
Sec 2.2 describes how ToE and GWL are computed. ToE and GWL are first computed individually for each ensemble member. Then the ToE for each SMILE is defined as the temperature at which 90% of the members show a significant change (BTW: how sensitive are results to this threshold?), while the GWL for each SMILE is the ensemble mean of the GWL from the members. So far I agree with the authors (and admit they describe the method much better than what I did here) but then this ensemble mean GWL is used to define GWLoE. However, isn’t this last step removing a large portion of the sensitivity to the global warming as the temperature variability across the ensemble is completely removed. An alternative could be to compute GWLoE for each ensemble member Individually and then average these individual GWLoE over the ensemble instead. Apart from better accounting for the temperature variability across the ensemble members this would also make it rather straightforward to form large ensembles by combining realisations from different models (or combining the SMILEs into one large superensemble) and use bootstrapping for estimating uncertainties.
What is the motivation for defining ratios of probabilities, trends and saturation in the ratios in Sec 2.4? Sure, probabilities for passing thresholds will change with increasing temperatures, but it is unclear to me what the added value of probability ratios and their level of saturation really is. Maybe the authors could better motivate the choice of the chosen method?
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC3 -
AC3: 'Reply on RC3', David Gampe, 07 Dec 2023
Thank you for your time and effort to review and comment on our manuscript. We will revise the manuscript accordingly in the revision stage but want to stimulate further discussion by replying to your points in advance.
- Motivation to use SMILEs for this study over ensemble of opportunity that is CMIP6: The main motivation of our study is to determine when a clear climate change signal emerges from the “noise”. It is thus essential to separate the climate signal from internal climate variability. We calculate the emergence of such signal in each member of each SMILE using a KS-test to compare the distribution of the desired time window with the pre-industrial state. Once statistical significance is reached and maintained until the end of the time series (2100) we conclude an emergence of the signal. This would be possible also with a multi-model CMIP6 ensemble. However, with one single member or respective CMIP6 model realization, the forced response cannot be cleanly separated from the internal variability of the climate. It rather just represents just one of many possible realizations. As the main focus of SMILEs is to allow for a robust representation and thus quantification of internal variability, it makes their application essential for our study purposes. We thus apply a 90% threshold for each SMILE where at least 90% of the members have to indicate emerged signals. This additional step ensures the consideration of internal variability and increases the robustness of our results.
- Use of multiple SMILEs: The main advantage of applying the multi-model CMIP6 ensemble is the excessive sampling of structural, i.e., model uncertainty. As the reviewer states, this is not reflected if using a single SMILE. However, the importance of structural uncertainty (in particular higher GWLs / towards the end of the century) is well established in the literature. We thus argue, that it is essential to also consider this uncertainty source for our study purposes. Therefore, we present the exposure of population / area for multiple SMILEs. Our results demonstrate that this is particularly relevant for the precipitation indices, where the related exposures are more sensitive to SMILE selection (compared to temperature indices). For the analysis of the probability ratio, we argue that the application of a joint emergence (using multiple SMILEs) thus further increases the robustness of the presented results and deem the application of more than one SMILE essential. However, we agree that this aspect could be more precisely highlighted in the manuscript. We will emphasize this more specifically in the revised version.
- Uncertainty assessment: we agree that a more rigorous assessment and presentation of uncertainties is required. We thus follow the suggestion by reviewer 1 to conduct bootstrapping for the sampling of GWL and ToE. The resulting updated exposure figures will thus include the 95% confidence intervals for each SMILE.
- Calculation of ToE/GWLoE: Thank you for bringing this to our attention. Our original intention was to apply GWL as the forced response of each ensemble (that is, the respective ensemble mean). However, we agree with the reviewer that the calculation of GWLoE based on each ensemble member is much more straight forward. We will thus update the methodology accordingly. Regarding the 90% threshold: this was selected (also following established Literature) so that the majority of members have to yield an emerged climate signal. A higher threshold, e.g., 95% would, however, put too much emphasis on the rather extreme members: for the MPI SMILE the current approach requires 27 of 30 members to show emergence. Raising the threshold to 95 would then consider 29 of 30 already. However, we agree that this is a rather arbitrary threshold. In the revised version of the manuscript, we will, following reviewer 1, implement a bootstrapping procedure to provide an estimate of uncertainties. This will ensure that all members are considered in the calculation of ToE and increase the robustness of the study.
- Application of probability ratio (PR): we agree that the motivation on the selection of this method in the current version of the manuscript is not obvious. In general, we focus on the assessment of the emergence of climate signals in relation to GWLs throughout the manuscript. However, we consider shifts in the entire distribution (KS-test) only, i.e., changes in magnitude/intensity. In this section, we thus aim to address potential changes in the tails of the distribution as well as changes in frequency of extremes. We agree that multiple other options for this would be available. We selected changes in PR as most appropriate to connect with the exposure of population in the previous section. The exposure to extremes is of particular interest for various reasons (public health, urban planning, general adaptation measures etc.). Further, the PR ratio is a common framework in detection and attribution which is easy to understand and communicate. The assessment of PR and related saturation in our opinion complements the other sections in a unique way. This analysis clearly highlights the need to include incremental changes compared to fixed warming levels, due to the shown non-linear response as well as the sensitivity of small incremental changes in GWL. We will, however, revise the methods section to highlight the motivation to select PRs accordingly.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC3
-
AC3: 'Reply on RC3', David Gampe, 07 Dec 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-2126', Anonymous Referee #1, 12 Nov 2023
Review of "Applying global warming levels of emergence to highlight the increasing population exposure to temperature and precipitation extremes"
This study uses large single model ensembles to explore projections of climate extreme indices at global warming levels using an emergence-based methodology. The authors not only look at emergence of climate extremes, but also exposure to emergence and examine effects of methodological choices (order of operations).
This is a robust analysis and a well-presented study. I'm confident that it will make a useful contribution to the literature. I do have three major comments for the authors to consider though:
Major comments:
- A benefit of SMILEs is that they can be used to explore sampling as well as structural uncertainties. I think with Figure 3 in particular it would be useful to show a range of area of emergence as a function of GWL for each SMILE. This could be derived from bootstrapping the simulations and computing a confidence interval based on the resampled ensembles.
- Some SSP population projections aren't particularly compatible with some emissions pathways. As such, SSP1 population projection is unlikely to be compatible with having a high GWL. I would suggest that SSP5 populations are used in Figure 4 and that uncertainty estimation through bootstrapping (as discussed in the previous comment) is shown instead.
- The results shown are applicable to the climate under a very high rate of global warming, but it should be noted that they aren't applicable to slower warming or stabilised climate states (e.g. King et al. 2020 (https://www.nature.com/articles/s41558-019-0658-7) and Mitchell et al. 2016 (https://www.nature.com/articles/nclimate3055)).
Minor comments:
Figure 1: Could you check that the lines are plotted correctly. For some models the range appears considerably smaller than due to interannual variability alone in the observations. In Maher et al. 2019 (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019MS001639), the MPI range looks larger than is plotted here.
L190: There's a strange space that should be removed.
L200: "less" should be "fewer"
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC1 -
AC1: 'Reply on RC1', David Gampe, 16 Nov 2023
Thank you for the valuable inputs and comments and for your time and effort to critically review our manuscript. We will prepare a revised version once the discussion period is over. Yet, we want to stimulate the discussion by replying to your points already at this stage.
- Bootstrapping: We highly appreciate your suggestion to use bootstrapping to derive an uncertainty estimation of sampling the GWLs. We will consider this and will update Figure 3 (and related Figs in the SI) accordingly. We will apply bootstrapping to sample the n members of each SMILE from the available ensemble members. While we aim for a large number of samples the final decision on how many samples we will draw is yet to be determined. We will then update the Figures plotting the median/mean and the 5-95 percentile (or inter-quartile) range to indicate uncertainties.
- Compatibility of SSPs: we agree that not all SSPs are compatible with the selected SSP585 scenario. Nevertheless, we argue that one advantage of the application of GWLs is constraining scenario uncertainties. While the selection of the appropriate SSP is certainly a prerequisite we decided to include different spatially explicit estimates of future population distribution here to indicate the related uncertainties. We considered this a more appropriate assessment of uncertainties for these figures. However, we will explore also including the bootstrapped GWLs as suggested.
- Literature suggestions: Thank you for highlighting the aspect of slower warming and/or stabilized climate. We will incorporate the suggested publications and expand the related sections in the discussion.
- GWL range Figure 1: Thank you for critically reviewing this Figure. The GWLs for each ensemble were calculated using a 20-year moving window. This ultimately leads to a smoothening for individual years thus reducing the overall range considerably. As we then apply the GWLs based on the 20-yr window for further calculation of our results we decided to only show these in Fig.1. Nevertheless, we will mention this explicitly in the caption to improve this aspect.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC1
-
AC1: 'Reply on RC1', David Gampe, 16 Nov 2023
-
RC2: 'Comment on egusphere-2023-2126', Anonymous Referee #2, 29 Nov 2023
General comments:
This study presents the global warming level of emergence (GWLoE) of four temperature and precipitation indices, and the related exposure of population and land area based on the joint emergence of five SMILEs. While the finding is interesting, the manuscript needs much improvement regarding its writing.
Major Comments:
- The abstract needs to be reformatted. In the current version, the authors introduce the concept of ToE and GWLs, and mention that ‘ToE and GWL have barely been combined so far’. Yet, the scientific question of the research is not clear. The authors should clearly crystalize the specific problem/knowledge gap the study is aiming to address, i.e., the shortcomings of current research and why combining ToE and GWLs is significant.
- Again, there is a need to revise the Introduction to make it more accessible, particularly the current research progress and the knowledge gap. After reading the introduction, I am not clear about the current research progress and still have doubts about the importance of the paper. The authors only mention that “Recently, first studies combined GWL and ToE to provide global warming levels of emergence (GWLoE) instead of ToE (Abatzoglou et al. 2019, Kirchmeier-Young et al. 2019, Raymond et al. 2020). Yet, GWLoE remains a rarely applied concept in general as well as in the context of using SMILEs in particular”. The readers should be aware of the knowledge gap and be convinced that it is important to fill the knowledge gap after finishing reading the Introduction.
- As for the climate indices, why choose these four of different types? The temperature indices (TXx and TNx) are absolute indices; precipitation index PRCPtot is one of the amount indices, and R1Xday is the intensity index. Why not choose the indices of the same group for both temperature and precipitation?
- Why resample to the coarsest resolution (2.8°, CanESM5)? To what extent the selection of the resample resolution affects the exposure of land area and population? This should be clarified.
- Figure 1, the colored lines are very smooth, indicating almost no inter-annual variability (IAV). Has the time series been smoothed? The author should explain that.
- Figure 2, considering that there is almost no visible spatial difference in Figure b, the color bar can use unequal spacing instead of equal spacing.
Specific Comments:
- Many sentences lack commas, making them difficult to read. For example, Line 54, a comma is needed between “To disentangle a robust climate change signal from the background noise of internal climate variability” and “Single Model Initial-condition Large Ensembles (SMILEs) are widely used (e.g., Deser et al. 2020, Maher et al. 2021)”.
- Line 45. ‘W/m2’ ->’ W/m2’.
- Line 422. ‘proofs’ ->’ proves’.
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC2 -
AC2: 'Reply on RC2', David Gampe, 07 Dec 2023
We highly appreciate the constructive comments to improve our manuscript and acknowledge the time and effort to review our submission. To stimulate further discussion, we reply already at this stage prior to revising the manuscript.
- Highlighting knowledge gap/scientific merit: Thank you for critically screening our abstract and the introduction and pointing out the need to sharpen the corresponding sections. We agree that the knowledge gaps we address in this paper are not completely coming through at this stage. We will carefully revise the sections to highlight the novelty of and the need for our study. Certainly, we will emphasize the increased assessment of uncertainties using multiple SMILEs to derive the joint emergence in conjunction with increased sampling of internal variability through bootstrapping GWL/ToE/GWLoE. Both, the abstract and the introduction will be streamlined and carefully revised accordingly. We will also revise the entire manuscript to improve the language and readability of the paper.
- Selection of climate indices: we agree with the reviewer that the selected indices differ in their type and representation. However, we rather see this variety as a strength then a shortcoming as they cover absolute (extremes), quantities and intensity of temperature and precipitation. It was not necessarily the intention of our study to compare the indices with one another and rather showcase the potential of GWLoE for them. Further, we chose these indices as they are extremely frequently applied in other studies (also in the context of GWLs). In our opinion, it is thus of particular interest to address the aspects of uncertainties around GWLs and the merits of GWLoE for these indices.
- Resampling to CanESM5 resolution: Thank you for this valid and important comment. In our study we aim to address two major sources of uncertainty: internal variability (hence the use of SMILEs) and structural uncertainty (multiple SMILEs). One of the foci of our study is the joint emergence using multiple SMILEs to increase the robustness of our results (considering also structural uncertainty) on a grid scale. This, however, requires to harmonize the SMILEs to a common grid as otherwise coarser models would be penalized for their resolution, likely skewing the results. While this step is in our opinion crucial, we acknowledge and highlight the effect of resampling order for the calculation of the indices in the discussion and the supplement already. However, we agree that our reasoning to resample is partly lacking in our manuscript at the current stage and will revise the corresponding methods section accordingly.
- GWL range Figure 1: Thank you for critically reviewing this Figure. The GWLs for each ensemble were calculated using a 20-year moving window. This ultimately leads to a smoothing for individual years thus reducing the overall range considerably. As we then apply the GWLs based on the 20-yr window for further calculation of our results we decided to only show these in Fig.1. Nevertheless, we will mention this explicitly in the figure caption to improve this aspect.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC2
-
RC3: 'Comment on egusphere-2023-2126', Anonymous Referee #3, 30 Nov 2023
The study “Applying global warming levels of emergence to highlight the increasing population exposure to temperature and precipitation extremes” by Gampe et al assesses the emergence of changes in selected climate indices when temperatures are increasing, and also looks at how large a fraction of land areas and population experiences changes in these climate indices. An interesting novelty of this work is the use of Global Warming Levels (GWL) instead of time of emergence to disentangle the scenario uncertainty and the differences in ECS between different models from the response of the climate system to global warming. The main findings from the paper are that temperature related indices have already changed significantly at 1 deg warming relative to pre-industrial levels over large areas and fractions of the population, while changes in precipitation related indices are emerging slower and will affect large portions of land/population only with stronger warming. The paper is clear and concisely written, and the GWLoE form an interesting method to present climate change information in a novel way. There are however a few points where the authors could try to further improve the paper.
What is the motivation to use an ensemble of SMILEs? The reason for using a SMILE is that it eliminates uncertainty that stems from using different models yet retains the internal variability (as the authors mention several times in the paper). However, combining several SMILEs again introduces uncertainty from differences between the different model, and you give away the main advantage of a SMILE. Could you comment on that?
The paper hasn’t convinced me that an ensemble of SMILEs is essential for this type of analysis. How different would the results have been if an ensemble had been created from all CMIP6 models instead, say 1 member from each model that contributed to CMIP6? Would this change the finding? Or the robustness?
The paper lacks a proper assessment of the robustness of the results. How robust are the estimates for the fraction of land or fraction of population that experience a change in climate indicators? Ensembles in general are very useful for quantifying confidence intervals or estimating uncertainty, so an ensemble of large ensembles should be even better to assess uncertainties, or?
Sec 2.2 describes how ToE and GWL are computed. ToE and GWL are first computed individually for each ensemble member. Then the ToE for each SMILE is defined as the temperature at which 90% of the members show a significant change (BTW: how sensitive are results to this threshold?), while the GWL for each SMILE is the ensemble mean of the GWL from the members. So far I agree with the authors (and admit they describe the method much better than what I did here) but then this ensemble mean GWL is used to define GWLoE. However, isn’t this last step removing a large portion of the sensitivity to the global warming as the temperature variability across the ensemble is completely removed. An alternative could be to compute GWLoE for each ensemble member Individually and then average these individual GWLoE over the ensemble instead. Apart from better accounting for the temperature variability across the ensemble members this would also make it rather straightforward to form large ensembles by combining realisations from different models (or combining the SMILEs into one large superensemble) and use bootstrapping for estimating uncertainties.
What is the motivation for defining ratios of probabilities, trends and saturation in the ratios in Sec 2.4? Sure, probabilities for passing thresholds will change with increasing temperatures, but it is unclear to me what the added value of probability ratios and their level of saturation really is. Maybe the authors could better motivate the choice of the chosen method?
Citation: https://doi.org/10.5194/egusphere-2023-2126-RC3 -
AC3: 'Reply on RC3', David Gampe, 07 Dec 2023
Thank you for your time and effort to review and comment on our manuscript. We will revise the manuscript accordingly in the revision stage but want to stimulate further discussion by replying to your points in advance.
- Motivation to use SMILEs for this study over ensemble of opportunity that is CMIP6: The main motivation of our study is to determine when a clear climate change signal emerges from the “noise”. It is thus essential to separate the climate signal from internal climate variability. We calculate the emergence of such signal in each member of each SMILE using a KS-test to compare the distribution of the desired time window with the pre-industrial state. Once statistical significance is reached and maintained until the end of the time series (2100) we conclude an emergence of the signal. This would be possible also with a multi-model CMIP6 ensemble. However, with one single member or respective CMIP6 model realization, the forced response cannot be cleanly separated from the internal variability of the climate. It rather just represents just one of many possible realizations. As the main focus of SMILEs is to allow for a robust representation and thus quantification of internal variability, it makes their application essential for our study purposes. We thus apply a 90% threshold for each SMILE where at least 90% of the members have to indicate emerged signals. This additional step ensures the consideration of internal variability and increases the robustness of our results.
- Use of multiple SMILEs: The main advantage of applying the multi-model CMIP6 ensemble is the excessive sampling of structural, i.e., model uncertainty. As the reviewer states, this is not reflected if using a single SMILE. However, the importance of structural uncertainty (in particular higher GWLs / towards the end of the century) is well established in the literature. We thus argue, that it is essential to also consider this uncertainty source for our study purposes. Therefore, we present the exposure of population / area for multiple SMILEs. Our results demonstrate that this is particularly relevant for the precipitation indices, where the related exposures are more sensitive to SMILE selection (compared to temperature indices). For the analysis of the probability ratio, we argue that the application of a joint emergence (using multiple SMILEs) thus further increases the robustness of the presented results and deem the application of more than one SMILE essential. However, we agree that this aspect could be more precisely highlighted in the manuscript. We will emphasize this more specifically in the revised version.
- Uncertainty assessment: we agree that a more rigorous assessment and presentation of uncertainties is required. We thus follow the suggestion by reviewer 1 to conduct bootstrapping for the sampling of GWL and ToE. The resulting updated exposure figures will thus include the 95% confidence intervals for each SMILE.
- Calculation of ToE/GWLoE: Thank you for bringing this to our attention. Our original intention was to apply GWL as the forced response of each ensemble (that is, the respective ensemble mean). However, we agree with the reviewer that the calculation of GWLoE based on each ensemble member is much more straight forward. We will thus update the methodology accordingly. Regarding the 90% threshold: this was selected (also following established Literature) so that the majority of members have to yield an emerged climate signal. A higher threshold, e.g., 95% would, however, put too much emphasis on the rather extreme members: for the MPI SMILE the current approach requires 27 of 30 members to show emergence. Raising the threshold to 95 would then consider 29 of 30 already. However, we agree that this is a rather arbitrary threshold. In the revised version of the manuscript, we will, following reviewer 1, implement a bootstrapping procedure to provide an estimate of uncertainties. This will ensure that all members are considered in the calculation of ToE and increase the robustness of the study.
- Application of probability ratio (PR): we agree that the motivation on the selection of this method in the current version of the manuscript is not obvious. In general, we focus on the assessment of the emergence of climate signals in relation to GWLs throughout the manuscript. However, we consider shifts in the entire distribution (KS-test) only, i.e., changes in magnitude/intensity. In this section, we thus aim to address potential changes in the tails of the distribution as well as changes in frequency of extremes. We agree that multiple other options for this would be available. We selected changes in PR as most appropriate to connect with the exposure of population in the previous section. The exposure to extremes is of particular interest for various reasons (public health, urban planning, general adaptation measures etc.). Further, the PR ratio is a common framework in detection and attribution which is easy to understand and communicate. The assessment of PR and related saturation in our opinion complements the other sections in a unique way. This analysis clearly highlights the need to include incremental changes compared to fixed warming levels, due to the shown non-linear response as well as the sensitivity of small incremental changes in GWL. We will, however, revise the methods section to highlight the motivation to select PRs accordingly.
Citation: https://doi.org/10.5194/egusphere-2023-2126-AC3
-
AC3: 'Reply on RC3', David Gampe, 07 Dec 2023
Peer review completion
Journal article(s) based on this preprint
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 242 | 0 | 0 | 242 | 0 | 0 |
- HTML: 242
- PDF: 0
- XML: 0
- Total: 242
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Clemens Schwingshackl
Andrea Böhnisch
Magdalena Mittermeier
Marit Sandstad
Raul R. Wood
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.