Back to News
InnovationAI Understanding briefing

Hong Kong deploys AI‑driven flood alerts and new stormwater tank to curb rainstorm flooding

Hong Kong’s Drainage Services Department has teamed with HKUST to launch an AI system that scans over 300 CCTV feeds for flood signs, while a new 2,000‑cubic‑metre stormwater storage tank in Ping Che aims to divert excess runoff during heavy storms.

4 min readRead the original reporting
Source-provided image accompanying Hong Kong deploys AI‑driven flood alerts and new stormwater tank to curb rainstorm flooding
Attributed reportingSource recorded
Publisher
scmp.com
Source type
Reporting by a news outlet — not a first-party document.

What we could not confirm independently: This claim is attributed to the named outlet. We did not verify it against a first-party document. (scmp.com)

ContextUnderstand this in 60 seconds

Key terms

Algorithm
A defined set of rules or steps that a computer follows to solve a problem or complete a task.
Latency
The time between sending a request and receiving the model's output.

What happened

The Drainage Services Department (DSD) announced two concrete flood‑mitigation measures: a 2,000‑cubic‑metre stormwater storage tank built in Ping Che and an AI‑powered monitoring platform that analyses live video from more than 300 CCTV cameras. The AI system, co‑developed with the Hong Kong University of Science and Technology (HKUST), automatically detects flooding severity and issues alerts to emergency teams. The tank, equipped with an intercepting channel and four pumps, can retain water equivalent to a standard swimming pool, diverting runoff from the Ping Yuen River during high‑water events.

In late October 2026, DSD completed construction of a stormwater storage tank in Ping Che, New Territories. The tank holds 2,000 cubic metres of water and is linked to an intercepting channel that diverts excess flow from tributaries of the Ping Yuen River. Four drainage pumps can release stored water back into the river system once levels recede. Engineers cited recent heavy rainstorms in May and June that caused flash floods in nearby villages as the catalyst for the project.

Concurrently, DSD and HKUST have co‑developed an AI monitoring platform that taps into the police force’s SmartView system, which aggregates video from more than 300 CCTV cameras across the city. The AI analyses real‑time footage to identify water accumulation, assess flood severity, and trigger alerts for emergency responders. Engineers Chan Pui‑ching and Joe Lee Ho‑ting said the system reduces the limitations of manual video monitoring and speeds up flood detection and response.

The AI system is integrated with DSD’s existing emergency dispatch workflow. When the flags a flooding event, alerts are sent to designated response teams, who can be dispatched to the affected site within minutes. The department also employed robots for immediate dredging and widening works on the Ping Che River, further supporting rapid mitigation efforts.

Source details: scmp.com

Why it matters

Hong Kong faces escalating flood risk as extreme rainstorms become more frequent, straining an already dense drainage network. Traditional manual monitoring of CCTV feeds is labor‑intensive and slow, limiting rapid response. By automating detection, the AI system promises faster dispatch of rescue crews, potentially reducing property damage and loss of life. The storage tank adds physical capacity to absorb sudden runoff, addressing the root cause of flash flooding in low‑lying New Territories villages. Together, the digital and infrastructural upgrades illustrate how city governments can blend AI with civil‑engineering solutions to improve climate‑resilience, a model other flood‑prone megacities may study.

Hong Kong recorded 72 confirmed flooding incidents by August 2026, up from 114 in 2025, despite a slight decline in total cases, indicating a trend toward more severe events. July 2026 saw 790.3 mm of rainfall, the second‑highest monthly total since records began in 1884. These statistics underscore the urgency of enhancing flood‑response capabilities.

AI‑driven video analysis can process far more visual data than human operators, reducing the lag between flood onset and emergency action. Faster alerts can translate into quicker road closures, evacuation orders, and deployment of pumps, potentially saving lives and reducing economic losses.

The storage tank adds a tangible buffer to the drainage system, directly addressing the volume of runoff that overwhelms existing channels during intense storms. By retaining water at the source, downstream flooding can be mitigated, protecting infrastructure and low‑lying communities.

The combined approach demonstrates a scalable model for other densely populated, flood‑prone cities. It highlights how public agencies can leverage academic partnerships to develop AI tools tailored to local infrastructure challenges.

Interactive Mechanism

Interactive Mechanism: How It Actually Works

Explore the underlying technology behind this development interactively.

Agent Lifecycle Stage:
1
User Intent & Planning: "Audit customer refund request #4092 and settle payment."
2
Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').
3
Guardrail & Verification:Paused: High-value action requires human operator sign-off.
4
Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.
Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.
Interactive Concept Check+10 Points
AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

What to watch next

Key indicators to monitor include the system’s detection accuracy, false‑positive rate, and response times during the upcoming rainy season. Officials have not disclosed the AI model’s technical specifications or any performance benchmarks, so independent validation will be essential. Watch for any policy updates on data privacy, given the system ingests footage from public CCTV networks. Additionally, the operational reliability of the Ping Che tank—maintenance schedules, pump performance, and sediment buildup—will affect its long‑term efficacy.

Performance metrics: Accuracy of flood detection, false‑positive/negative rates, and average alert during the next monsoon season.

Data governance: How DSD handles privacy concerns related to continuous CCTV monitoring, especially in residential areas.

Operational reliability: Maintenance schedules for the tank’s pumps and sediment removal, and any reported mechanical failures.

Policy replication: Whether other Hong Kong districts or neighboring jurisdictions adopt similar AI‑driven flood monitoring systems.

Related guides & quizzes

Found this useful?