the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Jul 2024
The 2023 global warming spike was driven by El Niño/Southern Oscillation
Abstract. Global-mean surface temperature rapidly increased 0.27 ± 0.05 K from 2022 to 2023. Such an interannual global warming spike is not unprecedented in the observational record with previous instances occurring in 1956–57 and 1976–77. However, why global warming spikes occur is unknown and the rapid global warming of 2023 has led to concerns that it could have been externally driven. Here we show that climate models that are subject only to internal variability can generate such spikes, but they are an uncommon occurrence (𝑝 = 2.6 ± 0.1 %). However, when a prolonged La Niña immediately precedes an El Niño in the simulations, as occurred in nature in 1956–57, 1976–77, 2022–23, such spikes become much more common (𝑝 = 16.5 ± 0.6 %). Furthermore, we find that nearly all simulated spikes (94 %) are associated with El Niño occurring that year. Thus, our results underscore the importance of El Niño/Southern Oscillation in driving the occurrence of global warming spikes such as the one in 2023, without needing to invoke anthropogenic forcing, such as changes in atmospheric concentrations of greenhouse gases or aerosols, as an explanation.
-
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
(1164 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1164 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1937', Anonymous Referee #1, 08 Jul 2024
This is a timely study exploring an important null hypothesis for explaining last year's extreme global warming. Although the approach is sound and the results solid I think that a bit more detail in the analysis will go a long way to rule out other explanations for the record-breaking GMTS in 2023. I really like the idea of quantifying spikes relative to a persistent La Nina - this idea was floated my Michael Mann on twitter so some credit should be give to him in the Acknowledgements.
Here are my suggestions for adding more depth or dimensions to the analysis:
In the end the El Nino of 2023 was not as strong as expected and therefore its effect on GMTS should not be compared with the events of 1997 and 2015 or with all events in models. I wonder if there is a sensitivity of the results to the magnitude of simulated ENSO. Perhaps figure 1A could be expanded to include PDFs for events of different magnitude. I suggest using other metrics of ENSO amplitude such as the SOI to avoid issues related to the warming trend on the definition of El Nino - see: https://iopscience.iop.org/article/10.1088/1748-9326/abe9ed
Also, the largest influence of El Nino on GMTS occurs on the year after the peak, not before. That should be 2024 not 2023. I wonder if a.more nuanced analysis should be performed to isolate the months when El Nino (and La Nina) have the largest influence on GMTS. I have performed this analysis using observations via lag-correlation of GMTS and Nino-34 and found that ENSO has the largest imnpact from Nov of year 0 to June of year +1. Perhaps GMTS should be computed for an "ENSO" year starting on Sep of year 0 and ending on August of year 1 for a more clear isolation of their correlation.
Otherwise a very important study that I hope gets published soon.
Citation: https://doi.org/10.5194/egusphere-2024-1937-RC1 -
RC2: 'Comment on egusphere-2024-1937', Mika Rantanen, 16 Jul 2024
Review of Atmospheric Chemistry and Physics manuscript egusphere-2024-1937 “The 2023 global warming spike was driven by El Niño/Southern Oscillation” by Raghuraman et al.
This manuscript tackles a topical issue by attempting to explain the 2023 global warmth by El Nino/Southern Oscillation. The authors employ a large set of climate model ensembles and calculate the probabilities of global warming “spikes” during different phases of ENSO and also after long La Nina events. Although the spike in global temperature in 2023 was an uncommon occurrence, its probability is multiplied when an El Nino event follows a prolonged La Nina. I think the manuscript is well written and adds well to the ever-growing literature explaining the causes of the 2023 warmth. Therefore, I recommend the publication of the study after my minor comments have been addressed.
- I downloaded the NASA GISTEMP table of global mean temperatures from here: https://data.giss.nasa.gov/gistemp/, and for me the annual mean data (column J-D) is 1.17°C for 2023 and 0.89°C for 2022 (accessed 16 July 2024). That gives an increase of 0.28°C. In the paper, the authors say 0.27°C. Can you explain where you got this value? It’s not a big deal, but since you say in the abstract that the increase is not unprecedented, even small differences are important. Also, 0.28°C would be the largest difference in the GISTEMP data, tied with 1977.
- That leads me to my 2nd comment, which is probably the most important. How sensitive are your results for the choice of observational dataset and its uncertainty? I quickly looked at how large was the year-to-year difference in other global temperature datasets from here: https://climate.metoffice.cloud/temperature.html. In 1950-2023, the year-to-year difference in 2023 is ranked 1-2 in HadCRUT5, Berkeley Earth, Gistemp, NOAA and ERA5 with values ranging between 0.28-0.30°C. For example, the difference between 2022 and 2023 in ERA5 and HadCRUT5 was 0.30°C. If the spike is defined by this number, the probability of it occurring would obviously be greatly reduced (see comment 8).
- Only the number of the 2023 spike was mentioned in the paper. I was also missing the numbers of the 1977 and 1957 spikes to put the 2023 value into context. They are in Fig. A1, but the numerical values seem not to be listed anywhere.
- L11. I think it’s not unknown that ENSO modulates global temperature and causes “spikes” in global temperature, such as 2016 or 1998. Rather, I think it is a well known fact. I’d suggest rephrasing this part of the sentence.
- L21-23. Relating to the comment 2, I find this a bit confusing. NASA GISTEMP is only one of the several observational datasets. The statement that GMST increased 0.27 +- 0.05 K is GISTEMP-specific, and other datasets might give slightly different numbers. I think you should not be so tied to GISTEMP in this context.
- L96. I think Rantanen and Laaksonen (2024) stress in their paper that internal variability alone cannot explain the September 2023 record margin with high likelihood. This does not exclude the possibility that internal variability still explains most of the spike. To me, saying that internal variability has little power sounds like Rantanen and Laaksonen (2024) is attributing less than half of the spike to internal variability, when this is clearly not the case.
- L110. I’d like to emphasize that your probabilities for Feb 1995 (0.13%) and May 1977 (0.1%) are still an order of magnitude higher than for September 2023 (0.01%). So calling them similar examples and 2023 being not unique sounds like a slight understatement. Also, El Nino tends to peak in NH winter, and thus I’d imagine that the global warming spikes tend to be stronger NH winter/spring, i.e. during or soon after the El Nino event. This was not the case in Sep 2023.
- L126. Again, I find it a bit odd to define the spike using the lower bound of the GISTEMP uncertainty interval. Why not define the spike using the average of multiple observational datasets (HadCRUT5, GISTEMP, Berkeley, NOAA...)? I guess the main finding, i.e. the probability of the spike (p=2.6 %) is greatly dependent on how you define the spike. If the spikes were calculated as averages over several datasets, using the best estimate from each dataset (rather than the lower bound), the 2023 spike would probably be estimated to be around 0.29 K (this is the number that most people consider to be the difference between 2022 and 2023). In this case, the probability would be much lower than the 2.6% reported in your paper?
- L138. “The ONI is defined as the sea surface temperature change”. Change from which? Do you mean sea surface temperature anomaly from some certain baseline?
Citation: https://doi.org/10.5194/egusphere-2024-1937-RC2 -
CC1: 'Comment on egusphere-2024-1937', Ales Kuchar, 19 Jul 2024
I recommend using violin plots in Fig. 1c in addition to dots for individual models. Violin plots are similar to box plots, except that they also show the probability density of the data at different values, usually smoothed by a kernel density estimator. While a box plot shows summary statistics such as mean/median and interquartile ranges, the violin plot shows the full distribution of the data. This would be useful to see how the individual models are distributed.
Citation: https://doi.org/10.5194/egusphere-2024-1937-CC1 - AC1: 'Comment on egusphere-2024-1937', Shiv Priyam Raghuraman, 04 Sep 2024
- AC2: 'Comment on egusphere-2024-1937', Shiv Priyam Raghuraman, 04 Sep 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-1937', Anonymous Referee #1, 08 Jul 2024
This is a timely study exploring an important null hypothesis for explaining last year's extreme global warming. Although the approach is sound and the results solid I think that a bit more detail in the analysis will go a long way to rule out other explanations for the record-breaking GMTS in 2023. I really like the idea of quantifying spikes relative to a persistent La Nina - this idea was floated my Michael Mann on twitter so some credit should be give to him in the Acknowledgements.
Here are my suggestions for adding more depth or dimensions to the analysis:
In the end the El Nino of 2023 was not as strong as expected and therefore its effect on GMTS should not be compared with the events of 1997 and 2015 or with all events in models. I wonder if there is a sensitivity of the results to the magnitude of simulated ENSO. Perhaps figure 1A could be expanded to include PDFs for events of different magnitude. I suggest using other metrics of ENSO amplitude such as the SOI to avoid issues related to the warming trend on the definition of El Nino - see: https://iopscience.iop.org/article/10.1088/1748-9326/abe9ed
Also, the largest influence of El Nino on GMTS occurs on the year after the peak, not before. That should be 2024 not 2023. I wonder if a.more nuanced analysis should be performed to isolate the months when El Nino (and La Nina) have the largest influence on GMTS. I have performed this analysis using observations via lag-correlation of GMTS and Nino-34 and found that ENSO has the largest imnpact from Nov of year 0 to June of year +1. Perhaps GMTS should be computed for an "ENSO" year starting on Sep of year 0 and ending on August of year 1 for a more clear isolation of their correlation.
Otherwise a very important study that I hope gets published soon.
Citation: https://doi.org/10.5194/egusphere-2024-1937-RC1 -
RC2: 'Comment on egusphere-2024-1937', Mika Rantanen, 16 Jul 2024
Review of Atmospheric Chemistry and Physics manuscript egusphere-2024-1937 “The 2023 global warming spike was driven by El Niño/Southern Oscillation” by Raghuraman et al.
This manuscript tackles a topical issue by attempting to explain the 2023 global warmth by El Nino/Southern Oscillation. The authors employ a large set of climate model ensembles and calculate the probabilities of global warming “spikes” during different phases of ENSO and also after long La Nina events. Although the spike in global temperature in 2023 was an uncommon occurrence, its probability is multiplied when an El Nino event follows a prolonged La Nina. I think the manuscript is well written and adds well to the ever-growing literature explaining the causes of the 2023 warmth. Therefore, I recommend the publication of the study after my minor comments have been addressed.
- I downloaded the NASA GISTEMP table of global mean temperatures from here: https://data.giss.nasa.gov/gistemp/, and for me the annual mean data (column J-D) is 1.17°C for 2023 and 0.89°C for 2022 (accessed 16 July 2024). That gives an increase of 0.28°C. In the paper, the authors say 0.27°C. Can you explain where you got this value? It’s not a big deal, but since you say in the abstract that the increase is not unprecedented, even small differences are important. Also, 0.28°C would be the largest difference in the GISTEMP data, tied with 1977.
- That leads me to my 2nd comment, which is probably the most important. How sensitive are your results for the choice of observational dataset and its uncertainty? I quickly looked at how large was the year-to-year difference in other global temperature datasets from here: https://climate.metoffice.cloud/temperature.html. In 1950-2023, the year-to-year difference in 2023 is ranked 1-2 in HadCRUT5, Berkeley Earth, Gistemp, NOAA and ERA5 with values ranging between 0.28-0.30°C. For example, the difference between 2022 and 2023 in ERA5 and HadCRUT5 was 0.30°C. If the spike is defined by this number, the probability of it occurring would obviously be greatly reduced (see comment 8).
- Only the number of the 2023 spike was mentioned in the paper. I was also missing the numbers of the 1977 and 1957 spikes to put the 2023 value into context. They are in Fig. A1, but the numerical values seem not to be listed anywhere.
- L11. I think it’s not unknown that ENSO modulates global temperature and causes “spikes” in global temperature, such as 2016 or 1998. Rather, I think it is a well known fact. I’d suggest rephrasing this part of the sentence.
- L21-23. Relating to the comment 2, I find this a bit confusing. NASA GISTEMP is only one of the several observational datasets. The statement that GMST increased 0.27 +- 0.05 K is GISTEMP-specific, and other datasets might give slightly different numbers. I think you should not be so tied to GISTEMP in this context.
- L96. I think Rantanen and Laaksonen (2024) stress in their paper that internal variability alone cannot explain the September 2023 record margin with high likelihood. This does not exclude the possibility that internal variability still explains most of the spike. To me, saying that internal variability has little power sounds like Rantanen and Laaksonen (2024) is attributing less than half of the spike to internal variability, when this is clearly not the case.
- L110. I’d like to emphasize that your probabilities for Feb 1995 (0.13%) and May 1977 (0.1%) are still an order of magnitude higher than for September 2023 (0.01%). So calling them similar examples and 2023 being not unique sounds like a slight understatement. Also, El Nino tends to peak in NH winter, and thus I’d imagine that the global warming spikes tend to be stronger NH winter/spring, i.e. during or soon after the El Nino event. This was not the case in Sep 2023.
- L126. Again, I find it a bit odd to define the spike using the lower bound of the GISTEMP uncertainty interval. Why not define the spike using the average of multiple observational datasets (HadCRUT5, GISTEMP, Berkeley, NOAA...)? I guess the main finding, i.e. the probability of the spike (p=2.6 %) is greatly dependent on how you define the spike. If the spikes were calculated as averages over several datasets, using the best estimate from each dataset (rather than the lower bound), the 2023 spike would probably be estimated to be around 0.29 K (this is the number that most people consider to be the difference between 2022 and 2023). In this case, the probability would be much lower than the 2.6% reported in your paper?
- L138. “The ONI is defined as the sea surface temperature change”. Change from which? Do you mean sea surface temperature anomaly from some certain baseline?
Citation: https://doi.org/10.5194/egusphere-2024-1937-RC2 -
CC1: 'Comment on egusphere-2024-1937', Ales Kuchar, 19 Jul 2024
I recommend using violin plots in Fig. 1c in addition to dots for individual models. Violin plots are similar to box plots, except that they also show the probability density of the data at different values, usually smoothed by a kernel density estimator. While a box plot shows summary statistics such as mean/median and interquartile ranges, the violin plot shows the full distribution of the data. This would be useful to see how the individual models are distributed.
Citation: https://doi.org/10.5194/egusphere-2024-1937-CC1 - AC1: 'Comment on egusphere-2024-1937', Shiv Priyam Raghuraman, 04 Sep 2024
- AC2: 'Comment on egusphere-2024-1937', Shiv Priyam Raghuraman, 04 Sep 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 798 | 300 | 112 | 1,210 | 18 | 18 |
- HTML: 798
- PDF: 300
- XML: 112
- Total: 1,210
- BibTeX: 18
- EndNote: 18
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 574 | 47 |
| Germany | 2 | 59 | 4 |
| China | 3 | 50 | 4 |
| Australia | 4 | 47 | 3 |
| United Kingdom | 5 | 39 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 574
Cited
1 citations as recorded by crossref.
Shiv Priyam Raghuraman
Brian Soden
Amy Clement
Gabriel Vecchi
Sofia Menemenlis
Wenchang Yang
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1164 KB) - Metadata XML