Tropical Cyclone Trends
Note: I updated this in 2024 to include 2023 data.
In a previous post, I covered the distinction between detection and attribution. These are normally considered sequentially. First you detect a signal above natural variability, then you examine if some portion of that signal can be attributed to human activity. The previous post shows that the two are statistically independent from each other, so there is no need to limit attribution studies to signals that are detectable above natural variability. This distinction and clarification I think is important in discussions of tropical cyclone trends. While trends are clear since 1980, the longer term record (since ~1850-1900) is less clear due to selection biases arising from our increasing ability to detect tropical cyclones. This makes detection difficult long term, but that doesn't necessarily mean that scientists can't also examine ways in which human activity is affecting tropical cyclone trends.
Detection in the Satellite Record (Since 1980)
The satellite record provides us with 42+ years of accurate counts of tropical cyclones globally as well as as the intensity of each of these storms. Over this time frame, we can have confidence in the accuracy of detected trends in hurricane frequency and intensity. NOAA maintains a dataset for TCs called the International Best Track Archive for Climate Stewardship (IBTrACS).[5] This dataset is pretty clear. There is no statistically significant trend in the frequency of hurricanes globally; if anything, they are becoming slightly less frequent. However, there is an increase in major (category 3+) storms.
%20Hurricanes%20and%20All%20Hurricanes%20(2).png)
This means necessarily that major hurricanes are increasing as a proportion of all hurricanes. I took the above data and plotted Cat 3+ hurricanes as a percentage of all hurricanes. Here's what I came up with.
%20Hurricanes%20as%20a%20Percentage%20of%20All%20Hurricanes%20(2).png)
What should be clear here is that we can detect a signal here that is consistent with the expectations of climate science. That is, warming is expected to correlate with a decrease in hurricane frequency and an increase in the proportion of major hurricanes, and that's what we observe. But this is only part of the equation. It's one thing to detect as signal; it's quite another to show that this signal is distinct from natural variability. This is where things get more complicated.
The problem with this, of course, is that much of the increase in hurricane count is due to our increasing ability to count hurricanes. In the 19th century, for instance, a hurricane could be counted only if it made landfall somewhere or crossed the path of a ship at sea which allowed the hurricane to be detected. So the above graph clearly suffers from a sample bias that will exaggerate any actual trends in hurricane frequency or intensity. Whatever long-term trend exists, it must be smaller than what is depicted above, if a long-term trend exists at all. To determine long-term trends that actually exist, therefore, we must find ways to eliminate this sample bias. There are several ways that you could attempt to do this: one by limiting recent data to match the limited data in the historical record, and two by estimating the storms that occurred but were not counted in the historical record.
Detection in the Historical Record (Since 1850)
A superficial count of hurricanes since 1850 shows that hurricanes have increased in frequency since the mid to late 19th century. This graph from NOAA below shows a clear trend in increasing hurricanes in the Atlantic basin.[6]
1. Count only U.S. Storms Making Landfall. One way to mitigate the sample bias is to limit current hurricane counts to what was historically detectable . Counting only U.S. landfalling hurricanes is one way to attempt this, since we have accurate counts of landfalling hurricanes since 1900. This approach to mitigating the sample bias shows no trend in U.S. landfalling hurricanes or major hurricanes since 1900.
The only problem with this approach is that it potentially introduces another sample bias. By limiting the count to only those hurricanes that happen to strike the US (about 2% of the globe), you do not count 1) hurricanes that make landfall outside of the US or 2) hurricanes that don't make landfall anywhere. This approach assumes that trends in the percentage of total TCs that make landfall in the US is constants, such that landfalling US hurricanes is representative of total hurricanes. This assumption can easily be (and likely is) wrong. The ratio of US hurricane strikes to Atlantic basin hurricane frequency is not constant, and the ratio of major hurricanes has actually been decreasing.[7] Climate changes affect where hurricanes are more likely to make landfall or trends in how many hurricanes fail to make landfall anywhere. It's also a bit ethnocentric to count only those hurricanes that would affect those doing the counting and ignoring all others.
Limiting current hurricane counts to be as bad as historical counts doesn't seem to be the best way to go. A better approach is to estimate how many TCs are missing from the historical record. There are two ways scientists can do this.
2. Estimate Missing TCs using Ship Data. Ships traversing the Atlantic and Pacific Oceans sometimes encounter hurricanes. This data alone cannot eliminate the sample bias of detected TCs, but this data can be used to estimate the number of missed hurricanes from observing practices in the late 19th century. Vecchi and Knutson[7] attempted to correct for sample bias in long-term homogenized dataset using this data and changes in the US hurricane strike fraction. This reconstruction shows that there has been no detectable increase in the proportion of hurricanes that become major hurricanes.
3. Estimate Missing TCs using Reanalyses. Another approach to recover missing hurricane data is to estimate missing hurricanes using reanalysis data - that is, a mix of empirical data and model simulations, to reconstruct likely trends in hurricane frequency and intensity since the mid 18th century. One such study (Emmanuel 2021)[8] simulated an increase in TCs in the Atlantic basin, but found no statistically significant trend globally in number of TCs since 1850. However, two of three reconstructions showed an upward trend in the number of major hurricanes. "Major hurricanes downscaled from the two NOAA reanalyses show statistically significant upward trends, but most of these trends occurred before 1920 and the major hurricanes downscaled from CERA-20C show no significant trend."
The above graph shows reanalysis data for TCs and Cat 3+ hurricanes with the IBTraCS data in black with a 7-year running mean applied. This result shows, at least superficially, that there has been an increase in TCs globally, but most of the increase in TCs comes from the reconstructions prior to 1920. Another study (Chan et al 2021)[9] reconstructed TCs in the Atlantic basin alone and found little to no trends at century scale in Atlantic TCs. Still another study (Chand et al 2022)[10] found a decreasing trend in TCs in the Atlantic basin and globally. Since reconstructions of Atlantic TCs show anywhere from increasing to decreasing trends in frequency on century time scales, it seems premature at this point to make any assessments of trends. There is insufficient evidence to rule out the null hypothesis and accept that TC frequency or intensity have increased above natural variability on century long time scales.
It may well be that we will never answer this question; the data simply doesn't have the resolution and accuracy to make any firm conclusions. I strongly suspect that improved reanalyses have the best shot of detecting a signal above the noise with confidence, but that analysis does not yet exist to my knowledge.
Attribution
As we've already seen, the fact that historical data is insufficient to determine with confidence that there have been any long term trends in TC frequency or intensity doesn't mean that AGW isn't having an impact. Afterall, the global warming signal can be detected above natural variability sometime in the 1980s. However, warming separates itself from expected trends without AGW as early as the 1920s[2]. Scientists can attribute global warming to anthropogenic causes before the signal is louder than the noise. The AGW global warming signal is complicated by aerosol pollution, which exerts a cooling influence that would otherwise be dominated by increases in GHGs. Aerosols caused a "pause" in warming following WWII that resumes in the late 1970s after we begin mitigating our aerosol pollution while at the same time accelerating our carbon emissions through fossil fuel use.
Studies have shown that several known factors contribute to TC frequency and intensity, and the combination of these lead climate scientists to expect that global warming will not increase the frequency of TCs (in fact, frequency may decrease) but those that form are more likely to become more intense, leading to an increase in the proportion of major (Cat 3+) hurricanes. This is precisely what we see in empirical data since 1980, though these trends are complicated by unreliable storm data from 1850 to 1980. However, the physics behind the expectations of climate science is pretty clear. Let's consider three factors that are known to contribute to hurricane frequency and intensity.
1. Increased vertical wind shear and increasing upper tropospheric temperature relative to the surface limits the formation of TCs.[4] Hurricanes form in part because of the temperature gradient between the upper troposphere and the surface. As the upper troposphere warms with respect to the surface (as is expected in all meteorological and climate models), we should see a decrease in that gradient and an increase in vertical wind shear. This trend should dampen the formation of TCs.
2. Increased SSTs fuel the formation of more TCs. There is a strong correlation between increasing sea surface temperature and the formation of hurricanes. The warmer water, the more hurricanes can form and the more intense they can become. This suggests that there are two competing forces at work here - one limits the formation of TCs and encourages the formation of TCs. But when they form, wind shear is likely low and TCs are able to become stronger storms.
The graph below shows Atlantic major hurricane number and its correlation of with the inverse of Atlantic vertical wind shear and sea surface temperatures. Storms are more likely to become major hurricanes with warm SSTs and low wind shear.
 |
| From Yann et al 2017 [11] |
3. Increasing aerosol pollution can limit the formation of TCs, while mitigating aerosol pollution removes that limitation.[12] Aerosols are somewhat of a wildcard introduced into the above trends. The North Atlantic experienced a decrease in hurricane frequency in the 1960s and 1970s, and the increase in aerosol pollution is the most likely candidate explaining that decrease in frequency. As efforts mitigate aerosol pollution halted the increase in aerosol pollution in the 1980s, hurricane frequency and intensity in the North Atlantic increased. This suggests that much of the increase in the proportion of Cat 3+ hurricanes following 1980 may be due to efforts to mitigate aerosol pollution. In much the same way that aerosols limited warming in the 1960s and 1970s, they may also have been responsible for the decrease in North Atlantic hurricane activity during that time frame. The IPCC says reduced aerosol forcings following the 1970s are likely responsible for a portion of the increase in TCs Atlantic hurricanes with medium confidence.
Murakami et al 2020[12] used models to simulate the frequency and spatial distribution of TCs using a combination of external forcings including GHGs, aerosols and volcanic activity. The model was able to produced simulations that were similar to observations for the formation of storms in the tropics from 1980 to 2018. This suggests that the scientific understanding of the forcings responsible for the formation of TCs are sufficient to detect an anthropogenic influence from GHGs and aerosol pollution.
Conclusion
While scientists can't detect a signal beyond the noise in trends beginning in the late 19th century, since the beginning of the satellite era, scientists do detect the expected trends in hurricane frequency and intensity - that is, slightly decreasing hurricane frequency with an increasing proportion of Cat 3+ storms. As best I can tell, the evidence even in the long-term data is consistent with what scientists expect with AGW. TC frequency is not increasing with AGW, but the proportion of major hurricanes does increase with AGW. Aerosol pollution works against warming from AGW, so the increase in aerosol pollution following WWII likely limited the formation of hurricanes, and when we began to mitigate aerosol pollution in the 70s and 80s, TC frequency, especially in the Atlantic Basin, increased again. This would suggest that the increase in the proportion of major storms since the beginning of the satellite era was due both to cleaning up aerosol pollution and increasing temperatures from AGW. Better stated, the masking effect of aerosols is being removed so that the increase in Cat 3+ storms is returning to what should be expected from AGW. Future studies and continuing data from storms will likely add more clarity to our understanding to the impact of AGW on tropical cyclones.
References:
[1] IPCC. "SR1.5 Glossary." https://www.ipcc.ch/sr15/chapter/glossary/
[2] Gavin Schmidt, "Watching the detections." RealClimate. September 25, 2022. https://www.realclimate.org/index.php/archives/2022/09/watching-the-detections
[3] Patrick T. Brown, PhD "Signal, Noise, and Global Warming's Influence on Weather." https://patricktbrown.org/2018/09/11/signal-noise-and-global-warmings-influence-on-weather/
[4] Tom Knutson. "Global Warming and Hurricanes: An Overview of Current Research Results." NOAA: Geophysical Fluid Dynamics Laboratory. October 3, 2022. https://www.gfdl.noaa.gov/global-warming-and-hurricanes/
[5] NOAA. International Best Track Archive for Climate Stewardship (IBTrACS). https://www.ncei.noaa.gov/products/international-best-track-archive.
[6] NOAA. Tropical Cyclone Climatology. https://www.nhc.noaa.gov/climo/
[7] Vecchi, G.A., Landsea, C., Zhang, W. et al. Changes in Atlantic major hurricane frequency since the late-19th century. Nat Commun 12, 4054 (2021). https://doi.org/10.1038/s41467-021-24268-5
[8] Emanuel, K. Atlantic tropical cyclones downscaled from climate reanalyses show increasing activity over past 150 years. Nat Commun 12, 7027 (2021). https://doi.org/10.1038/s41467-021-27364-8
[9] A Chan, Duo, Vecchi, Gabriel A, Yang, Wenchang, Huybers, Peter. Improved simulation of 19th- and 20th-century North Atlantic hurricane frequency after correcting historical sea surface temperatures. 2021. Science Advances. eabg6931 7 26. doi:10.1126/sciadv.abg6931. https://www.science.org/doi/abs/10.1126/sciadv.abg6931[10] Chand, S.S., Walsh, K.J.E., Camargo, S.J. et al. Declining tropical cyclone frequency under global warming. Nat. Clim. Chang. 12, 655–661 (2022). https://doi.org/10.1038/s41558-022-01388-4
[11] Yan, X., Zhang, R. & Knutson, T.R. The role of Atlantic overturning circulation in the recent decline of Atlantic major hurricane frequency. Nat Commun 8, 1695 (2017). https://doi.org/10.1038/s41467-017-01377-8
[12] Murakami, Hiroyuki. Delworth, Thomas L. Cooke, William F.. Zhao, Ming. Xiang, Baoqiang. Hsu, Pang-Chi. Detected climatic change in global distribution of tropical cyclones. Proceedings of the National Academy of Sciences. 117.20 (2020). 10706-10714. doi:10.1073/pnas.1922500117
[13] Ken Rice. "No, a cherry-picked analysis doesn’t demonstrate that we’re not in a climate crisis." And Then There's Physics.... https://andthentheresphysics.wordpress.com/2022/10/07/no-a-cherry-picked-analysis-doesnt-demonstrate-that-were-not-in-a-climate-crisis/
[14] Alimonti, G., Mariani, L., Prodi, F. et al. A critical assessment of extreme events trends in times of global warming. Eur. Phys. J. Plus 137, 112 (2022). https://doi.org/10.1140/epjp/s13360-021-02243-9 https://link.springer.com/article/10.1140/epjp/s13360-021-02243-9
[15] Klotzbach, P. J., Wood, K. M., Schreck, C. J., Bowen, S. G., Patricola, C. M., & Bell, M. M. (2022). Trends in global tropical cyclone activity: 1990–2021. Geophysical Research Letters, 49, e2021GL095774. https://doi.org/10.1029/2021GL095774
[16] Emanuel, Kerry. Evidence that hurricanes are getting stronger. (May 29, 2020). PNAS 117 (24) : 13194-13195.
https://www.pnas.org/doi/pdf/10.1073/pnas.2007742117
[17] J. P. Kossin, T. L. Olander, K. R. Knapp, Trend analysis with a new global record of tropical cyclone intensity. J. Clim. 26, 9960–9976 (2013).
Comments
Post a Comment