Cloud-Resolving Ensemble Prediction of the Extreme Rainfall Event on 2 June 2017 in Northern Taiwan

Abstract

An extreme Mei-yu rainfall event occurred on 2 June 2017 along the northern coast of Taiwan, where a peak amount of 635 mm fell in about 10 hours from 0100-1100 LST. Using ensemble-based sensivity analysis, an earlier study (Wang et al., 2021) showed that the areal-averaged rainfall in the wider region surrounding northern Taiwan in this event was influenced by frontal moving speed, frontal intensity (contrast), and location and timing of frontal disturbance in experiments using grid sizes of 2.5-5 km, but peak amounts in such runs were no more than 360 mm, suggesting low predictivity for its occurrence. Using 1-km grid size, predictions in a later study (Wang et al., 2023), on the other hand, showed that a peak amount of up to 541 mm can be achieved with a finer grid, but the detailed rainfall pattern and maximum amount are strongly dictated by the presence of a frontal low-pressure disturbance that affected the location and persistence of the surface convergence zone, thus again suggesting high nonlinearity and low predictability. However, in some initial times about 2.5-3 days before the event, the evolutions were more favorable for larger and more concentrated rainfall. In this study, the above extreme rainfall event is further examined, using ensemble forecasts both at 1 km and larger grid sizes, including the experiments carried out in the Taiwan Area Heavy-rainfall Prediction Experiment (TAHPEX). Special attention is given on the predictability of this extreme event and the strategy to produce useful results more effectively with longer lead times for early warning and preparation.