Detecting intense hurricanes from low resolution datasets via dynamical indicators

Tropical cyclones are among the most extreme and costly weather events. In the United States, hurricane Katrina in 2005 alone caused damage equal to roughly 1% of the country’s gross domestic product. Identifying trends in the frequency or intensity of such cyclones is difficult, however, due to their rare occurrence, as well as the brevity of associated datasets. Many future climate projections predict more intense hurricanes in the North Atlantic, but only with medium confidence because of the difficulties in reproducing the dynamics of the most severe hurricanes even in the most advanced climate models.
Destructive tropical cyclones often experience rapid intensification, as a storm gains strength dramatically over a short period of time, enhancing its destructive potential and also making its trajectory less predictable. Researchers often detect intensification via a characteristic increment of sustained winds over a period of 24h, but it’s hard to detect this phenomenon in global climate models (GCMs), due to the coarse resolution of wind field data. In a new paper, LML External Fellow Davide Faranda and colleagues explore the possibility of identifying intense tropical cyclones instead from coarse-grained atmospheric data. They compute two metrics describing cyclones as states of a chaotic high-dimensional dynamical system, and consider the persistence and dimension (i.e. the number of active degrees of freedom) of such states. These metrics have recently provided insights for a number of geophysical phenomena, including transitions between transient metastable states of the mid-latitude atmosphere, palaeoclimate attractors and changes in mid-latitude atmospheric predictability under global warming. All these applications have taken a Eulerian approach, focused on a fixed spatiotemporal domain, rather than tracking the evolution of specific physical phenomena. Instead, Faranda and colleagues apply a Lagrangian perspective which is particularly convenient for studying the complex behaviour of convectively unstable flow systems, such as tropical convection.
As the authors show, the Lagrangian implementation enables a convenient characterisation of the phase-space structure of tropical cyclones in terms of their dimension and persistence, variables which help single out the most intense cyclones. Their computations reveal a strong relationship between the values of indicators estimated from coarse atmospheric fields and the maximum sustained winds taken from observations. In the most intense phase of the cyclones, large maximum sustained winds are associated to low values of both dimension d and persistence θ, especially when estimated in the horizontal velocity field. To explain this result, the authors note that distinct air parcels appear to behave almost identically, with their rotational degrees of freedom oriented along the global axis of the storm. This behaviour reflects highly persistent, low-dimensional states found at unstable fixed points of the dynamics. These results imply that intense cyclones are characterized by different dynamical properties (entropy, stability) from those not accompanied by large maximum sustained winds. As a result, d and θ give a very good degree of discrimination from other, less intense, tropical cyclones.
The paper is available at

Leave a Reply

Your email address will not be published. Required fields are marked *