Evolution of Distinct Air Streams during the Extratropical Transition of Tropical Cyclones

During extratropical transition (ET), tropical cyclones (TCs) undergo distinct structural changes: The generic primary and secondary circulations of TCs become increasingly more asymmetric and eventually develop into the conveyor belt like air streams of an extratropical cyclone. These changes in coherent air stream significantly modify the supply with air masses that constitute the fuel of the TC’s heat engine. This study takes a Lagrangian approach to identify and describe these topological changes as well as the associated thermodynamic properties of air masses. We investigate convection-permitting simulations with the operational, local-area model of the German weather service (COSMO). Trajectories are seeded in the vicinity of the TC (radius of 200 km) for 6h forward and backward in time using data with 5min temporal resolution. Seeding dates range from the onset to the completion of ET. We use a relatively high spatial resolution of the seeding points to represent well the statistics of trajectories that comprise the inner-core convection. Each seeding date thus features several 100.000 trajectories. Our aim is to identify coherent air stream by trajectory clustering. For computationally efficiency, the large number of trajectories dictates a reduction of dimensionality for such an analysis. We employ a variant of multidimensional scaling that uses a subset of trajectories only to approximate the embedding into lower dimensional space (so-called landmark multidimensional scaling). Subsequently, standard cluster algorithms can be used to identify distinct air streams, e.g. ascent through inner-core convection and air streams that swirl more passively around the TC. Examination of thermodynamic properties of different clusters further illustrates the distinctness of the air streams. The evolution of two contrasting ET cases in the North Atlantic (Karl (2016) and Katia (2011)) is examined from the perspective of such distinct air streams. Major structural changes, e.g. the transition from upright convection to warm-conveyor belt like ascent, become clearly evident in the clusters. In addition, challenges in the visualization of cluster key characteristics, such as the median location and the cluster-internal variability of trajectories in 3D are discussed.