Reply to: Moon, I.-J. et al.; Lanzante, J. R.

The Original Article was published on 05 June 2019

The Original Article was published on 05 June 2019

replying to: I.-J. Moon et al. Nature (2019).

replying to: J. R. Lanzante Nature (2019).

I am grateful to Moon et al.1 and Lanzante2 for their interest in my original Letter3, the results of which can be summarized as follows. Using the available historical record of regional and global tropical cyclone tracks for the period 1949–2016, I found that the forward translation speed of tropical cyclones is generally slower in the later periods than in the earlier periods. The potential impacts of such a slowdown and a possible link to anthropogenic climate change were discussed. Such a link can be considered in terms of a growing number of previous studies that show a general weakening of general atmospheric circulation patterns in a warming climate4,5,6,7,8,9,10,11. In the broadest sense, it can be argued that Arctic amplification12,13 is expected to slow the general circulation through the reduction of meridional temperature and pressure gradients. Since tropical cyclones are largely carried passively within these circulation patterns, it is reasonable to say that the observed slowdown of tropical-cyclone translation speed is consistent, at least in sign, with expected circulation changes forced by anthropogenic factors. However, I was careful to point out that no direct claims of a causal mechanism were supported by the study3, stating that “The analyses presented here do not constitute a detection and attribution study because there are likely to be many factors, natural and anthropogenic, that control tropical-cyclone translation speed. For example, the best-track data exhibit a global 10% reduction in translation speed during a period in which global-mean surface temperatures increased by about 0.5 °C; however, this finding does not provide a true measure of the climate sensitivity of these related phenomena. To determine the true sensitivity (that is, the expected change in tropical-cyclone translation speed as a function of anthropogenic forcing), further analyses and numerical simulations are required.”

A shared concern of Moon et al.1 and Lanzante2 is that the trends shown in Kossin3 do not follow a monotonic decrease in speed over the full period of record but rather are due to step changes. From this observation they argue that the step changes are caused by data artefacts. As stated in Kossin3 and reiterated above, the observed changes in tropical-cyclone translation speed over the past 70 years or so are most probably caused by a combination of factors and so there should be less expectation of a steady linear trend, particularly in a metric that is linked to atmospheric circulation, which is highly stochastic on a broad range of timescales. For example, global mean surface temperature is directly linked to CO2 concentration, which exhibits a quasi-steady linear trend, but the trend in global mean surface temperature still exhibits change points owing to the combined influences of natural and anthropogenic forcing. The global warming ‘hiatus’ is a good example of such behaviour14. Any link between CO2 concentration and tropical-cyclone translation speed would be much less direct and there is less expectation for a clear one-to-one relationship or a steady monotonic trend. As I pointed out3, the explicit relationship between tropical-cyclone translation speed and anthropogenic forcing, when all other factors are removed, is not yet clear. Formal attribution studies that utilize carefully controlled numerical simulations are required to address this question.

I acknowledged3 that the historical tropical cyclone record contains heterogeneities, but that tropical-cyclone translation speed should be comparatively less sensitive to these issues. This was stated in the context of comparison with known heterogeneities in tropical cyclone frequency- and intensity-based metrics, which are highly heterogeneous15,16,17. A number of statements by Moon et al.1 and Lanzante2 refer to issues with frequency- and intensity-based metrics, and are less relevant to the results of ref. 3, which rely only on storm location data that are then averaged along each storm track. Nevertheless, I acknowledge here that data-related biases could affect the fidelity of the results of ref. 3, and Moon et al.1 and Lanzante2 provide useful insights into this.

The main issue raised by both Moon et al.1 and Lanzante2 is the potential for systematic temporal biases in the mean latitude of the tropical-cyclone tracks. Such biases would be expected to project onto tropical-cyclone translation speed because tropical-cyclone steering flow, which describes the ambient environmental wind that tropical cyclones are embedded in, is a function of latitude. This is a reasonable concern. There are two ways that such a bias could appear: (1) in a systematic meridional shift in the tracks and (2) in inter-basin frequency trends, which could affect the global trend because different basins have different mean-state steering flows. As argued by Moon et al.1, the former could be related to data-collection heterogeneities or through a true climate signal, while the latter can introduce a change in tropical-cyclone translation speed independently of steering-flow changes or meridional track shifts.

In order to better quantify the potential for inter-basin frequency variability to project onto the global slowing trend in tropical-cyclone translation speed, random sampling can be employed (see Methods). When the tropical-cyclone translation speeds from each basin are randomly sampled to remove inter-basin frequency variability and trends, the global trend is reduced from 10% to 7% while remaining highly statistically significant (Extended Data Fig. 1). Thus inter-basin frequency variability does project onto the global trend, but when accounted for, the effect is relatively small in terms of potential impacts.

The potential effect of meridional tropical-cyclone track shifts on tropical-cyclone translation speed in any given basin is more difficult to quantify because it must first be determined whether any such shifts are caused by climate factors or data issues (or both), as also noted by Moon et al.1 and Lanzante2. If the shifts are related to climate factors, then they merely represent another possible pathway for climate variability to affect tropical-cyclone translation speed and the associated impacts of a slowdown. The analyses of Moon et al.1 and Lanzante2 quantify these meridional shifts and indeed show changes in the latitudinal distribution that would be expected to increase the slowing trends. This provides evidence that the trends shown in ref. 3 have a latitudinal component, although it is not clear why Lanzante2 arbitrarily chose two different epochal periods to represent the introduction of satellite data in the Northern and Southern hemispheres. Polar orbiting satellites, which were introduced in the 1960s, view the entire globe and geostationary satellites, which were introduced in the 1970s and 1980s, are positioned above the Equator and view both hemispheres equally. The choice of periods used by Lanzante2 may maximize the latitudinal signal, but it is not clear that the data should be expected to partition in that way. Nevertheless, the evidence presented by Moon et al.1 and Lanzante2 that the regional trends have a latitudinal component, from a few per cent in some basins to substantially more in others, is compelling. It remains unclear whether the latitudinal shifts are entirely due to data artefacts caused by the introduction of satellite data or whether a climate signal may have a role, but this question certainly bears further scrutiny before we can have high confidence that the trends shown in ref. 3 reflect a true global slowdown of tropical-cyclone translation speed.

As noted, it is not generally obvious how to objectively discern and quantify the potential effects of data heterogeneities in the analyses presented in Moon et al.1, Lanzante2 and Kossin3. Some progress towards this can be made, however, by considering the historical record of landfall and over-land data for the continental USA. This record has been used in a number of previous studies and is often invoked as the best possible proxy for ground truth in century-scale Atlantic hurricane variability and trends17,18,19,20,21. These data begin in the mid-nineteenth century, but are typically constrained to begin in the year 1900, to represent the reliable part of the longer record because it is “the first year when nearly all landfalling tropical cyclones would probably have been monitored”17. The time series of tropical-cyclone translation speed over the continental USA (Extended Data Fig. 2) demonstrates a 17% slowdown over the 118-year period 1900–2017 and the linear trend is marginally insignificant at the 95% confidence level (P = 0.058).

The known drivers of climate and tropical-cyclone variability in the Atlantic are linked to the El Niño–Southern Oscillation, which operates on inter-annual timescales, and Atlantic multi-decadal variability (AMV), which most probably represents a convolution of internal and external regional variability22. The relationship between the El Niño–Southern Oscillation and tropical-cyclone translation speed over the continental USA is weak, but the decadal variability of the tropical-cyclone translation speed time series over the continental USA is apparently linked to observed AMV (Fig. 1). The increased tropical-cyclone translation speeds in the 1940s, which project positively onto the trends shown in Kossin3, appear to be related to a warm AMV phase, whereas the suppressed tropical-cyclone translation speeds in the 1970s and 1980s appear to be related to the subsequent cool phase. The transition to the present warm phase coincides with increasing tropical-cyclone translation speeds over the continental USA.

Fig. 1: Low-pass filtered time series of annual-mean tropical-cyclone translation speed over the continental USA and an AMV index.

Both time series are normalized for plotting purposes by removing their means and dividing by their standard deviations.

When AMV is regressed from the tropical-cyclone translation speed time series (Fig. 2), the magnitude and statistical significance of the slowdown increases and becomes more uniform over the 118-year period. That is, there is a statistically significant slowing trend over a reliable 118 year period that remains after accounting for the dominant drivers of regional variability. Additionally, the observed shifts or change-points in tropical-cyclone translation speed are apparently linked to AMV rather than being data artefacts as argued by Lanzante2. The slowdown is not related to any systematic equatorward shift in the mean latitude of the tropical cyclones (Extended Data Fig. 3). The annual number of data over the continental USA remains relatively constant after 1900 until the present period of heightened Atlantic tropical-cyclone activity (Extended Data Fig. 4). The increase in the number of data concurrent with the recent trend toward higher latitude would be expected to reduce the slowing trends in tropical-cyclone translation speed shown in ref. 3 rather than inflate them.

Fig. 2: Time series of residuals of the regression of annual-mean tropical-cyclone translation speed over the continental USA onto an AMV index.

The bold line shows the low-pass filtered time series. The trend is based on the 118-year period 1900–2017 and grey shading shows the 95% confidence bounds of the trend. The slope of the trend line is −0.05 km h–1 yr–1 with a 95% confidence interval of [−0.09, −0.01] and P value of 0.02.

A weakness of the analyses of Lanzante2 is the arbitrary choice to exclude over-land data analyses, with the claim that the over-land data are “unreliable indicators of the long-term behaviour of TCS” (where TCS is tropical-cyclone translation speed). This statement is contrary to present knowledge and the introduction of satellite data should have a minimal effect on the quality of tropical-cyclone position data over land compared to over open water. There is no obvious expectation that the location of tropical cyclones while over land should be highly dependent on the availability of satellite data. This is especially true for tropical cyclones travelling over land along the east coast of the USA since 190017,18,19,20,21, and here it is shown that these tropical cyclones have in fact slowed down substantially over the past 118 years. Furthermore, the observed century-scale variability of these data exhibits shifts or change-points that are apparently physical and linked to regional multi-decadal variability rather than data artefacts and discontinuities. Nevertheless, the concerns stated by Moon et al.1 and Lanzante2 are valid, and exploring the possible ways for data-related biases to project onto the trends shown in ref. 3 is an important exercise. Given the substantial impacts of a slowdown in tropical-cyclone translation speed, particularly over land, it is essential to understand whether the trend is real or an artefact of the data (or some combination of both) and whether the slowdown, if real, is related to climate change.

Unfortunately there are no other regions outside the continental USA that allow for an analysis of a similarly long and reliable record, so this result is clearly limited in addressing the more global questions posed by Moon et al.1 and Lanzante2. That is, it does not directly address concerns about the fidelity of the global trend, but rather supports the existence of a long-term physical slowing trend in a particular region, which provides at least some additional confidence in the fidelity of the observed and like-signed trends in other broader regions. In spite of uncertainties in the data, given the potential for very substantial impacts on lives and property, the evidence presented here and in Kossin3 point towards an important need for further study. For example, it is possible to perform “attribution without formal detection” studies using controlled numerical simulations22,23. One such related study24 has recently found that the tropical-cyclone translation speed of Atlantic hurricanes is reduced, on average, under global warming scenarios, which adds to the balance of evidence for a slowdown and a physical connection to human activities. Additional attribution studies are clearly warranted.


None of the trend significance tests required correction for serial correlation, as indicated by a Durbin–Watson test. In Fig. 1, low-pass filtering was performed using a 21-year sliding centred mean applied to the annual-mean tropical-cyclone translation speed over the continental USA and an annual-mean index of AMV. The AMV index is taken from the unsmoothed Atlantic Multidecadal Oscillation index available from the National Oceanic and Atmospheric Administration/Earth System Research Laboratory (NOAA/ESRL; In Extended Data Fig. 1, inter-basin frequency variability was removed as follows: for each year, the full sample of tropical-cyclone translation speeds from each basin was randomly sampled with replacement at a fixed rate while maintaining the observed proportion of data in that basin relative to the global data.

Data availability

All data used in the analyses are available from the author on request.

Code availability

All codes used to read, analyse and plot the data are available from the author on request.


  1. 1.

    Moon, I.-J., Kim, S.-H. & Chan, J. C. L. Climate change and tropical cyclone trend. Nature (2019).

  2. 2.

    Lanzante, J. R. Uncertainties in tropical-cyclone translation speed. Nature (2019).

  3. 3.

    Kossin, J. P. A global slowdown of tropical cyclone translation speed. Nature 558, 104–108 (2018); Author Correction Nature 564, E11–E16 (2018).

    CAS  ADS  Article  Google Scholar 

  4. 4.

    Mann, M. E. et al. Influence of anthropogenic climate change on planetary wave resonance and extreme weather events. Sci. Rep. 7, 19831 (2017).

    Google Scholar 

  5. 5.

    Held, I. M. & Soden, B. J. Robust responses of the hydrological cycle to global warming. J. Clim. 19, 5686–5699 (2006).

    ADS  Article  Google Scholar 

  6. 6.

    He, J. & Soden, B. J. Anthropogenic weakening of the tropical circulation: the relative roles of direct CO2 forcing and sea surface temperature change. J. Clim. 28, 8728–8742 (2015).

    ADS  Article  Google Scholar 

  7. 7.

    Vecchi, G. A. & Soden, B. J. Global warming and the weakening of the tropical circulation. J. Clim. 20, 4316–4340 (2007).

    ADS  Article  Google Scholar 

  8. 8.

    Vecchi, G. A. et al. Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing. Nature 441, 73–76 (2006).

    CAS  ADS  Article  Google Scholar 

  9. 9.

    Grise, K. M. & Polvani, L. M. Understanding the timescales of the tropospheric circulation response to abrupt CO2 forcing in the Southern Hemisphere: seasonality and the role of the stratosphere. J. Clim. 30, 8497–8515 (2017).

    ADS  Article  Google Scholar 

  10. 10.

    Plesca, E., Grützun, V. & Buehler, S. A. How robust is the weakening of the Pacific Walker circulation in CMIP5 idealized transient climate simulations? J. Clim. 31, 81–97 (2018).

    ADS  Article  Google Scholar 

  11. 11.

    Plesca, E., Buehler, S. A. & Grützun, V. The fast response of the tropical circulation to CO2 forcing. J. Clim. 31, 9903–9920 (2018).

    ADS  Article  Google Scholar 

  12. 12.

    Coumou, D., Di Capua, G., Vavrus, S., Wang, L. & Wang, S. The influence of Arctic amplification on mid-latitude summer circulation. Nat. Commun. 9, 2959 (2018).

    CAS  ADS  Article  Google Scholar 

  13. 13.

    Coumou, D., Lehmann, J. & Beckmann, J. The weakening summer circulation in the Northern Hemisphere mid-latitudes. Science 348, 324–327 (2015).

    CAS  ADS  Article  Google Scholar 

  14. 14.

    Kosaka, Y. & Xie, S.-P. Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature 501, 403–407 (2013).

    CAS  ADS  Article  Google Scholar 

  15. 15.

    Kossin, J. P., Olander, T. L. & Knapp, K. R. Trend analysis with a new global record of tropical cyclone intensity. J. Clim. 26, 9960–9976 (2013).

    ADS  Article  Google Scholar 

  16. 16.

    Vecchi, G. A. & Knutson, T. R. Estimating annual numbers of Atlantic hurricanes missing from the HURDAT database (1878–1965) using ship track density. J. Clim. 24, 1736–1746 (2011).

    ADS  Article  Google Scholar 

  17. 17.

    Landsea, C. W., Vecchi, G. A., Bengtsson, L. & Knutson, T. R. Impact of duration thresholds on Atlantic tropical cyclone counts. J. Clim. 23, 2508–2519 (2010).

    ADS  Article  Google Scholar 

  18. 18.

    Landsea, C. W. Counting atlantic tropical cyclones back to 1900. Eos 88, 197–208 (2007).

    ADS  Article  Google Scholar 

  19. 19.

    Klotzbach, P. J., Bowen, S. G., Pielke, R., Jr & Bell, M. Continental U.S. hurricane landfall frequency and associated damage: observations and future risks. Bull. Am. Meteorol. Soc. 99, 1359–1376 (2018).

    ADS  Article  Google Scholar 

  20. 20.

    Landsea, C. W. Comments on “Monitoring and understanding trends in extreme storms: state of knowledge.”. Bull. Am. Meteorol. Soc. 96, 1175–1176 (2015).

    Article  Google Scholar 

  21. 21.

    Landsea, C. W. Hurricanes and global warming. Nature 438, E11–E12 (2005).

    CAS  ADS  Article  Google Scholar 

  22. 22.

    Kossin, J. P. et al. in Climate Science Special Report: Fourth National Climate Assessment (eds Wuebbles, D. J. et al.) Vol. I, 257–276 (U.S. Global Change Research Program, 2017).

  23. 23.

    Murakami, H. et al. Investigating the influence of anthropogenic forcing and natural variability on the 2014 Hawaiian hurricane season. Bull. Am. Meteorol. Soc. 96, S115–S119 (2015).

    Article  Google Scholar 

  24. 24.

    Gutmann, E. D. et al. Changes in hurricanes from a 13-yr convection-permitting pseudo–global warming simulation. J. Clim. 31, 3643–3657 (2018).

    ADS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to James P. Kossin.

Ethics declarations

Competing interests

The author declares no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data figures and tables

Extended Data Fig. 1 Time series of global annual-mean tropical-cyclone translation speed with inter-basin frequency variability removed.

Bold lines show the low-pass filtered time series and trend. Grey shading shows the 95% confidence bounds of the trend. The slope of the trend line is –0.02 km h–1 yr–1 with a 95% confidence interval of [–0.03, –0.01] and P value of 0.0003. The change over the period represents a 7% slowdown.

Extended Data Fig. 2 Time series of annual-mean tropical-cyclone translation speed over the continental USA (CONUS).

The trend, which is constrained to the reliable period 1900–2017, has a slope of –0.04 km hr–1 yr–1 with a 95% confidence interval of [–0.07, 0.0008] and P value = 0.058. The change over the period represents a 17% slowdown. Grey shading shows the 95% confidence bounds of the trend.

Extended Data Fig. 3 Time series of annual-mean tropical-cyclone latitude over the continental USA.

The bold line shows the low-pass filtered time series. There is no trend in the time series.

Extended Data Fig. 4 Time series of the annual count of data over the continental USA.

The bold line shows the low-pass filtered time series. The increasing trend prior to 1900 is due to data collection changes. The increasing trend at the end of the time series is associated with the present active phase in Atlantic tropical-cyclone activity, which is associated with the present warm AMV phase.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kossin, J.P. Reply to: Moon, I.-J. et al.; Lanzante, J. R.. Nature 570, E16–E22 (2019).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing