Detecting intense hurricanes from low resolution datasets via dynamical indicators

Although the life-cycle of hurricanes is well understood, many of the underlying physical processes occur at scales below those resolved by globalclimate models (GCMs), so that projecting future changes in hurricance characteristicsremains challenging. We assess the capability of dynamical system metrics to identify intense cyclones even in coarse resolution datasets, where wind speed may be not accurately represented. We compute dynamical indicators, namely the persistence and number of active degrees of freedom, from the horizontal wind field of 146 tropical cyclones occurred between 2010 and 2018 using ERA5 reanalysis data at 0.25° horizontal resolution, and link these to the maximum sustained winds as detected from observational datasets. Our analysis provides a representation of cyclones in phase space and allows to: i) identify different stages of the cyclones’ life cycle as distinct regions of the phase space; and ii) locate regions of the phase space associated with intense cyclones (as detected from observations). Specifically, we find that the most intense cyclones are associated with a strong decrease of the instantaneous dimension and an increase in persistence of tropical cyclones. This relation could be used for detection of intense cyclones in comparativelycoarse resolution datasets, such as those issued from GCM simulations or century-long reanalyses.