A composite approach to produce reference datasets for extratropical cyclone tracks: application to Mediterranean cyclones

Many cyclone detection and tracking methods (CDTMs) have been developed in the past to study the climatology of extratropical cyclones. However, all CDTMs have different approaches in defining and tracking cyclone centers. This naturally leads to cyclone track climatologies with inconsistent physical characteristics. More than that, it is typical for CDTMs to produce a non-negligible number of tracks of weak atmospheric features, which do not correspond to large-scale or mesoscale vortices and can differ significantly between CDTMs. Lack of consensus in CDTM outputs and the inclusion of significant numbers of uncertain tracks therein have long prohibited the production of a commonly accepted reference dataset of extratropical cyclone tracks. Such a dataset could allow comparable results on the analysis of storm track climatologies and could also contribute to the evaluation and improvement of CDTMs. To cover this gap, we present a new methodological approach that combines overlapping tracks from different CDTMs and produces composite tracks that concentrate the agreement of more than one CDTM. In this study we apply this methodology to the outputs of 10 well-established CDTMs which were originally applied to ERA5 reanalysis in the 42-year period of 1979-2020. We tested the sensitivity of our results to the spatiotemporal criteria that identify overlapping cyclone tracks, and for benchmarking reasons, we produced five reference datasets of subjectively tracked cyclones. Results show that climatological numbers of composite tracks are substantially lower than the ones of individual CDTMs, while benchmarking scores remain high (i.e., counting the number of subjectively tracked cyclones captured by the composite tracks). Our results show that composite tracks tend to describe more intense and longer-lasting cyclones with more distinguished early, mature and decay stages than the cyclone tracks produced by individual CDTMs. Ranking the composite tracks according to their confidence level (defined by the number of contributing CDTMs), it is shown that the higher the confidence level, the more intense and long-lasting cyclones are produced. Given the advantage of our methodology in producing cyclone tracks with physically meaningful and distinctive life stages, we propose composite tracks as reference datasets for climatological research in the Mediterranean. The Supplement provides the composite Mediterranean tracks for all confidence levels, and in the conclusion we discuss their adequate use for scientific research and applications.