FloodSentinel is a two-in-one urban flood monitoring and intelligence system. It protects lives with real-time alerts — and gives cities the data they need to fix flooding at its source. Continuously valuable, not just in rainy season.
Communities across Nairobi receive no advance notice when conditions turn dangerous — whether from rivers overflowing or drainage channels backing up under heavy rain. Floods arrive in minutes.
Over 200,000 Nairobi residents live in areas where both river flooding and overwhelmed drainage systems cause regular inundation. Displacement and property loss recur every rainy season.
Much of Nairobi's drainage network is aging, under-maintained, and unmapped in terms of real performance. No system currently measures how quickly drainage channels fill, back up, or fail — so problems only get discovered after a flood.
When water rises upstream, FloodSentinel detects it immediately and pushes warnings to communities before conditions become dangerous.
The same sensors that trigger alerts are also building a long-term picture of how flooding behaves across the city — every day, flood season or not.
"FloodSentinel doesn't just warn people about floods. It helps the city understand where flooding happens, how severe it is, and where drainage systems are failing — so the right areas can be fixed."
FloodSentinel uses a distributed network of low-power sensor nodes placed at strategic river points upstream. Each node operates independently and communicates via GSM — no Wi-Fi or mains power required.
Central processing unit per node. Handles sensor reads, threshold logic, and communication triggers.
Non-contact water level measurement. Mounted above river surface, accurate to ±3mm.
Sends SMS alerts to registered residents, ward administrators, and emergency services automatically.
Fully off-grid power. Nodes operate continuously with no dependency on grid infrastructure.
Three-tier alert system: Watch → Warning → Critical. Rise rate detection prevents false alarms.
Historical level data sent via MQTT to a lightweight web dashboard for county water authorities.
How a flood event flows through the system
HC-SR04 fires ultrasonic pulse. ESP32 calculates distance-to-surface and converts to absolute river depth using baseline calibration.
Not just absolute level — rise rate (cm/hr) is calculated. A slow rise at 0.9m may be safe; a rapid rise at 0.6m triggers Watch status.
Each tier escalates the response. Watch notifies the ward office. Warning triggers resident SMS. Critical triggers all channels simultaneously.
Pre-registered numbers receive plain-language alerts in English and Swahili. No app, no internet, no smartphone required on the resident's end.
All readings pushed to MQTT broker → lightweight dashboard. County authorities get historical trends, node health status, and event logs.
The concern with flood systems is always: "what does it do when it's not raining?" FloodSentinel's answer is: it keeps working. Even on a dry day, the network is collecting baseline data, monitoring drainage channel behaviour, and refining its picture of how water moves across the city.
What gets built over time
Kenya Met and Google Maps provide general forecasts and regional data. FloodSentinel provides hyper-local, real-time detection at street and drainage level — especially in informal settlements and poorly mapped urban areas that existing systems ignore entirely.
Rain interference, debris, and mounting conditions affect any sensor. FloodSentinel handles this through per-node baseline calibration, emphasis on rise-rate rather than absolute level, and multi-threshold logic. The system tracks trends, not just snapshots.
GSM requires no smartphone, no Wi-Fi, no app. A plain SMS reaches anyone with any mobile phone. This is a deliberate accessibility choice — the communities most exposed to flooding are often the least connected.
Alerts require absolute water level + rise-rate + multi-stage threshold confirmation. A slow rise at 0.9m stays at Watch. A rapid rise at 0.6m triggers Warning. This prevents normal fluctuations from generating unnecessary alerts and eroding community trust.
Nodes are designed for easy replacement of individual components. Each node reports its own battery level and GSM signal status, enabling proactive maintenance. At ~$85 per node, replacement is affordable at ward or community level without central oversight.
Each node operates independently — no mesh networking, no central hub required. Adding a new location means installing one node. The system scales across an entire city with no increase in coordination complexity and no dependency on fixed internet infrastructure.
Component procurement, sensor calibration, ESP32 environment setup
Firmware: level reading, threshold logic, rise-rate algorithm
GSM alert integration, SMS dispatch, field enclosure & solar wiring
Dashboard, field testing at Nairobi River, demo & documentation