Article ID: 2019-035
A novel method for detecting tropical cyclones in high-resolution climate model simulations is proposed herein and subjected to examination. The proposed method utilizes a two-dimensional scatterplot based on two quantities that represent the radial gradient and the tangential asymmetry of mid-to upper-level thickness around a simulated vortex. A comparison between the modeled and observed tropical cyclones using the non-hydrostatic regional climate model (NHRCM) with 20-km grid spacing under reanalysis-driven boundary conditions for one year revealed that no cyclones were missed and there was only one false alarm over a part of the western North Pacific near Japan. The simulated vortices were classified into two categories; tropical cyclones and extratropical cyclones. These two groups, having specific features, were also found in the results using present-day climate datasets, indicating that the tropical cyclones were reasonably distinguished from extratropical cyclones although a one-by-one comparison could not, in principle, be conducted. Comparison of the results obtained from datasets with 5-km and 20-km grid spacing demonstrated that the detection scheme was only weakly dependent on the horizontal resolution. This dependence was further reduced by using the radial gradient over the outer radii instead of near the center of the vortex. The resolution-independent feature in this method is due to a procedure in which the tangential asymmetry of mid-to upper-level thickness is utilized instead of the relative vorticity at 850 hPa, often used in conventional schemes. This procedure allows the method to identify tropical cyclones without the need to determine a grid-dependent threshold. The method proposed here provides a useful tool for detecting tropical cyclones in high-resolution climate simulations.