A historical record of four black rainstorm episodes in Hong Kong, China in July to August 2025

TL;DR

This paper reviews four record-breaking black rainstorm episodes in Hong Kong in 2025, analyzing their predictability using radar data and AI models, and discusses key atmospheric mechanisms involved.

Contribution

It introduces a novel analysis combining radar-derived wind fields and AI models to assess the predictability of extreme rainstorms in Hong Kong.

Findings

AI models predicted two major rainstorms a week in advance

Radar analysis identified surface convergence and jet streams as key factors

Predictability varied across the four rainstorm episodes

Abstract

Four episodes of black rainstorms, the highest tier of heavy rain according to the rainstorm warning system in Hong Kong, occurred within eight days from 29 July to 5 August 2025, breaking the record for torrential rain in the territory. Providing early alerts to the public about these black rainstorms has considerable application value and has a substantial impact on their lives. This initial review of the four black rainstorm cases focuses on the predictability of the events from the synoptic/mesoscale to the micro/storm scale. A newly available tool for three-dimensional wind fields retrieved from weather radars in the region was used to analyse the triggering factors for the development and maintenance of intense convection. Given the current level of technology, the four episodes of heavy rain had different degrees of predictability. Global artificial intelligence (AI) models were found to be skilful in capturing two major rainstorms, including one week before they occurred. Surface convergence, mid-tropospheric waves, and jets were important mechanisms in heavy rainfall occurrence. Future research directions on these black rainstorms have been briefly discussed.