2023/09/08 Guangdong - Hong Kong Floods

Heavy precipitation in Guangdong-Hong Kong mostly strengthened by human-driven climate change

Press Summary (First published 2023/10/24, Updated 2023/11/22)

Event Description

In September 8, 2023, Guangdong and Hong Kong regions experienced severe flooding due to torrential rains, with some areas receiving up to 800 mm of rainfall. This extreme weather, associated with the remnants of Typhoon Haikui, led to widespread disruption and significant damage. Tragically, at least two fatalities were reported as a result of the flooding.

The Hong Kong Observatory, responsible for monitoring weather conditions, indicated that the Pearl River Estuary was affected by a trough of low pressure, resulting from cyclone Haiku, and causing persistent heavy rain and thunderstorms. Rainfall exceeded 600 mm at the Observatory, and the Eastern and Southern Districts of Hong Kong Island saw rainfall totals higher than 800 mm, far exceeding the monthly average for September, which is typically 321.4 mm. The Observatory Headquarters recorded a record-breaking one-hour rainfall of 158.1 mm, the highest ever recorded since records began in 1884. This deluge caused streets to transform into rivers, with floodwaters over 1.5 meters. The consequences were severe, with disrupted road traffic, public transport, and subway stations. Schools, the Hong Kong Stock Exchange, and some offices were forced to close, and only essential travel was advised by the Labor Department.

The flooding also caused extensive damage to residential and commercial buildings, including a popular shopping mall. The Hong Kong Fire Department received numerous distress calls and evacuated 110 individuals, with 58 people trapped in elevators. The Home Affairs Department swiftly opened 14 temporary shelters to accommodate those displaced by the disaster. Tragically, Hong Kong police discovered the bodies of two individuals in different flooded areas of the city, and approximately 114 people were injured. The impact of the flooding extended beyond Hong Kong, affecting nearby Guangdong province, particularly areas in Shenzhen and Meizhou, where approximately 11,000 people had to be evacuated. Shenzhen also recorded an unprecedented 469 mm of rain in a 24-hour period, suggesting another record-breaking total in this devastating weather event. 

Surface Pressure Anomalies show a weak negative pressure anomaly over the western part of the domain analysed and associated with Precipitation Data over 100 mm/day.

Climate and Data Background for the Analysis

Changes in extreme precipitation patterns in Hong Kong in the 21st century have been an active research  field. According to the Intergovernmental Panel on Climate Change (IPCC), Hong Kong has already been experiencing noticeable changes in its precipitation patterns, particularly those exceeding 100 mm of rainfall in a single day. These changes are marked by a projected decrease in the annual number of rain days, alongside with a shift in the seasonality of rainfall, with more intense and prolonged wet seasons. However, it is crucial to acknowledge that future projections exhibit substantial variability among different model and emission scenarios, underscoring the significant uncertainties that persist in predicting future extreme rainfall events in the region, as well as, their effects on the low-lying areas and densely populated coastaline.

Our analysis approach rests on looking for weather situations similar to those of the event of interest having been observed in the past. For the Guangdong and Hong Kong floods we have low confidence in the robustness of our approach given the available climate data, as the event is largely unique in the data record.

ClimaMeter Analysis

We analyse here (see Methodology for more details) how events similar to the Guangdong - Hong Kong floods have changed in the present (2001–2022) compared to what they would have looked like if they had occurred in the past (1979–2000) in the region [108°E 120°E 18°N 25°N]. The Surface Pressure Changes show that cyclones leading to Guangdong - Hong Kong floods are slightly more intense (1 hPa) in the present than they were in the past. Precipitation Changes show that similar events produce larger (up to 5 mm/day) amounts of precipitation in the Western part of the region. However, they also tend to produce less precipitation in other areas, such as parts of the Eastern region of Hong Kong. Considering the affected urban areas, Dongguan, in the nearby Guangdong province, experienced an increase in precipitation in the present (up to 3 mm/day). We also find that Similar Past Events have become more frequent in September, and less common in October, suggesting a shift in the seasonality of these type of events. 

Finally, we find that sources of natural climate variability, notably the Atlantic Multidecadal Oscillation, may have partly influenced the event. This suggests that the changes we see in the event compared to the past may be mostly due to human driven climate change, with a secondary contribution from natural variability.


Based on the above, we conclude that cyclones leading to the Guangdong - Hong Kong floods similar to that observed in September 2023 are up to 5 mm/day wetter in the western Guangdong province than they would have been in the past and up to 3 mm/day dryer in the coastal part. We interpret the Guangdong - Hong Kong floods as an event whose characteristics can mostly be ascribed to human driven climate change.

Contact Authors

Davide Faranda, IPSL-CNRS, France  📨davide.faranda@lsce.ipsl.fr  🗣️French, Italian, English

Tommaso Alberti, INGV, Italy  📨tommaso.alberti@ingv.it  🗣️ Italian, English

Additional Information : Complete Output of the Analysis

The figure shows the average of surface pressure anomaly (msl) (a), average 2-meter temperatures anomalies (t2m) (e), cumulated total precipitation (tp) (i),  and average wind-speed (wspd) in the period of the event. Average of the surface pressure analogs found in the counterfactual [1979-2000] (b) and factual periods [2001-2022] (c), along with corresponding 2-meter temperatures (f, g),  cumulated precipitation (j, k), and wind speed (n, o).  Changes between present and past analogues are presented for surface pressure ∆slp (d),  2 meter temperatures ∆t2m (h), total precipitation ∆tp (i), and windspeed ∆wspd (p): color-filled areas indicate significant anomalies with respect to the bootstrap procedure. Violin plots for past (blue) and present (orange) periods for Quality Q analogs (q), Predictability Index D (r), Persistence Index Θ (s), and distribution of analogs in each month (t). Violin plots for past (blue) and present (orange) periods for ENSO (u), AMO (v) and PDO (w).  Number of the Analogues occurring in each subperiod (blue) and linear trend (black).  Values for the peak day of the extreme event are marked by a blue dot. Horizontal bars in panels (q,r,s,u,v,w) correspond to the mean (black) and median (red) of the distributions.