09:15 〜 09:30
[AAS05-02] An Objective Detection of Separation Scenario in Tropical Cyclone Trajectories Based on Ensemble Weather Forecast Data
Tropical cyclones (TCs) are a major disaster, generating intense winds and heavy rains, causing the loss of human lives, important damages to the environment (erosion, mass movement...), social disruption (evacuation, public services...), and economic costs (suspended production in factories, discontinued economic fluxes...). Because the location of TCs plays a key role in the location of heavy rainfall and strong winds over lands, knowing their evolution in time is practically important to support disaster prevention. In ensemble numerical weather prediction (ENWP), TCs tracks sometimes group together into trajectories parting away from each other. Detecting different path (or separation) scenarios in an ENWP may help to activate (or not) mitigation measures. Also, detecting separation scenarios in ENWP is a necessary step toward understanding controllability. The existence of spatial separations in predicted TC tracks could be the indicator of a region where an energy-efficient input, performed at an appropriate timing, would help to manipulate the atmosphere, and reduce the effects of the storm over lands.
From this perspective, the Japan Meteorological Agency (JMA) has been developing a mesoscale ensemble prediction system (MEPS) that can provide probabilistic information and multi-scenarios for risk management. This is achieved with the help of 21 ensemble members (1 unperturbed member and 20 perturbed by paired singular vectors) and the 14 forecast horizons (from T+0 to T+39, every 3 hours), initialized every 6 hours (at T = 0000, 0600, 1200 and 1800 UTC).
This study proposes an objective detection method of separation scenario, using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm as a clustering approach. Taking advantage of the independence of the DBSCAN algorithm to the prior choice of the number of clusters, we first describe an objective way to calculate the aggregation distance, by searching the most frequent Euclidean distance between all the tracks. Here, this clustering approach has an advantage in clustering TCs without any parameter tunings. The clustering is then applied to all the 21 tracks forecasted by JMA-MEPS, at each of the initialization times. Separation scenarios exist when the number of clusters is greater than one. Three case studies illustrate this detection method: TCs “Dolphin” (2020), “Nepartak” (2021) and “Meari” (2022).
From this perspective, the Japan Meteorological Agency (JMA) has been developing a mesoscale ensemble prediction system (MEPS) that can provide probabilistic information and multi-scenarios for risk management. This is achieved with the help of 21 ensemble members (1 unperturbed member and 20 perturbed by paired singular vectors) and the 14 forecast horizons (from T+0 to T+39, every 3 hours), initialized every 6 hours (at T = 0000, 0600, 1200 and 1800 UTC).
This study proposes an objective detection method of separation scenario, using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm as a clustering approach. Taking advantage of the independence of the DBSCAN algorithm to the prior choice of the number of clusters, we first describe an objective way to calculate the aggregation distance, by searching the most frequent Euclidean distance between all the tracks. Here, this clustering approach has an advantage in clustering TCs without any parameter tunings. The clustering is then applied to all the 21 tracks forecasted by JMA-MEPS, at each of the initialization times. Separation scenarios exist when the number of clusters is greater than one. Three case studies illustrate this detection method: TCs “Dolphin” (2020), “Nepartak” (2021) and “Meari” (2022).