Principal Investigator


At a Glance

Over the last year, the Vecchi lab continued their efforts to understand the mechanisms behind tropical cyclone (TC) activity changes on timescales of years to decades. Tropical cyclones impact society and ecosystems through extreme wind, rain and surge. The Vecchi lab uses climate and atmospheric models, combined with analyses of the observed record, to help distinguish the extent to which observed multi-decadal to centennial changes in TC activity have been driven by large scale factors, such as ocean temperature changes, greenhouse gases, volcanic eruptions, or El Niño, as opposed to random atmospheric fluctuations. A better understanding of TC changes over time is key to building strategies to mitigate their damages for the public and private sectors.

 


Research Highlight

Tropical cyclones (TCs) are of profound societal and economic significance. TC characteristics, such as their track, frequency, wind and rainfall intensity, exhibit variations on a range of timescales. Predicting these variations requires improved understanding of the character of and mechanisms behind these changes. Throughout 2023, Vecchi and his researchers worked to understand the climatic controls on TC frequency (Hsieh et al., 2023) and track (Kortum et al., 2023). They developed a novel machine learning-based method to reconstruct the three-dimensional structure of hurricane winds (Eusebi et al., 2024). This method could help improve weather-scale forecasts of hurricanes and explore the history of and mechanisms behind hurricane variations in the North Atlantic over the past millennium (Yang et al., 2024).

The relatively short historic record of observed hurricanes poses an important limitation to a better understanding of multi-decadal to centennial changes in hurricane activity. While ship-based observations of hurricanes in the Atlantic go back to the 1850s, satellite-based estimates only go as far as the 1960s. Vecchi and his team have been working with colleagues around the U.S. to explore the impact of climate changes over the past millennium on tropical cyclone activity using a reconstruction from sedimentary paleohurricane records and a statistical model of hurricane activity using sea surface temperatures (Yang et al., 2024). The goal is to betterunderstand the variations in hurricane activity. This work indicates that there may have been large centennial changes in Atlantic hurricane activity over the past 1,000 years, resulting largely from natural climate variations. The amplitude of these past changes may be comparable to model-projected changes in Atlantic hurricane activity projected for the coming century. This indicates that future changes arising from increasing greenhouse gases may be temporarily masked (for periods of many decades) or enhanced – providing a complex landscape for adaptation strategies, and complicating interpretation of the historical hurricane record. The researchers found that statistical and dynamical models of hurricane activity, when provided with estimates of ocean temperature changes over the past millennium, are able to recover multi-decadal to centennial scale variations in Atlantic hurricane activity estimated independently from sediment cores. A manuscript detailing this work is in press (Yang et al., 2024).

The Vecchi group continued their work to understand the mechanisms controlling tropical cyclone frequency, building on their novel paradigm of “seeds and probabilities.” This research provides a framework for interpreting the impact of large-scale environmental factors on tropical cyclone activity. It shows that one must consider the change in the number of pre-tropical cyclone vortices (or “TC seeds”), and the impact of large-scale environmental conditions on the probability of cyclone genesis from those seeds. Both are needed to accurately account for the sensitivity of tropical cyclone frequency to global forcing, such as the presence of atmospheric aerosols. Previous work, which had focused primarily or solely on the genesis probability, failed to accurately explain the response of TC frequency to forcing. Over the past year the researchers published a study demonstrating that the inter-model spread in TC seeds (and thus TC genesis) in response to warming could be explained by the large-scale radiative energy convergence in the atmosphere, providing a large-scale constraint on global TC frequency change (Hsieh et al., 2023). This work has shown that inter-model spread in TC genesis sensitivity to changing climate is largely driven by differences in the climate-inducedresponse of pre-TC synoptic disturbances, which can be connected to large-scale changes in vorticity and ascent – and these can be understood in terms of atmospheric energy flux convergence.

Given the risks posed by hurricanes, improved forecasts of the track, intensity and impact of hurricanes are of value. The researchers aim for more precise weather-scale predictions of storms to provide weather forecast models with an accurate representation of the three-dimensional structure of a hurricane at the start of the forecast. Although tropical cyclone winds are energetic and spatially variable, airplane-based observations of storms recover relatively sparse samples. Figure 9.1a shows the wind observations of Hurricane Ida while it was in the Gulf of Mexico on the 27th of August 2021. These observations are confined to the lower atmosphere (below 500hPa) and at discrete locations. The Vecchi group built a hybrid statistical-dynamical (i.e., a “Physics-informed Neural Network”, or PINN) method for reconstructing the three-dimensional wind field of tropical cyclones using machine learning/artificial intelligence techniques (Eusebi et al., 2024). As can be seen in Figure 9.1b, the method is able to generate a spatially complete reconstruction of the storm, even outside of the area with observations, and recover the asymmetries in the storm structure. These wind fields will be used in future experimental retrospective hurricane forecasts to explore the potential for improving weather-scale forecasts of hurricane track and intensity.

Figure 9.1 a-b.
PINN output trained with a combination of real hurricane hunter observations and SHiELD forecast data. (a) The locations and magnitudes of Ida flight level and dropsonde wind speed observations recorded roughly between 10z and 12z, Aug. 27. (b) PINN 3D reconstruction of Hurricane Ida on August 27th, 12z (Eusebi et al., 2024).

Vecchi and his team are using observations, advanced statistical methods, and climate model experiments (Kortum et al., 2023) to explore the mechanisms controlling decadal changes in Atlantic hurricane tracks. They have shown that the recent multi-decadal eastward shift in Atlantic hurricanes (from 1971 to 2020) was not principally a climate-driven signal. The shift included a dominant component associated with random weather fluctuations – highlighting a limited predictability for decadal hurricane track changes. The study further demonstrated that decadal changes in the location of TC genesis have had a larger role than changes in atmospheric steering winds in controlling these multi-decadal changes in hurricane track. This highlights the need to improve our predictions of TC genesis location in order to improve assessments of regional TC risk.

 


References

Eusebi, R., G.A. Vecchi, C.-Y. Lai, and M. Tong, 2024. Realistic tropical cyclone wind and pressure fields can be reconstructed from sparse data using deep learning. Nature Communications: Earth and Environment 5(8). (https://doi.org/10.1038/s43247-023-01144-2).

Hsieh, T.-L. et al., 2023. The influence of large-scale radiation anomalies on tropical cyclone frequency. Journal of Climate 36(16):5431–5441. (https://doi.org/10.1175/JCLI-D-22-0449.1).

Kortum, G., G. Vecchi, T.-L. Hsieh, and W. Yang, 2024. Influence of weather and climate on multidecadal trends in Atlantic hurricane genesis and track. Journal of Climate 37(5):1501-1522. (https://doi.org/10.1175/JCLI-D-23-0088.1).

Yang, W., E. Wallace, et al., 2024. Last millennium hurricane activity linked to endogenous climate variability. Nature Communications 15(816). (https://doi.org/10.1038/s41467-024-45112-6).