Be the first to know. Get instant alerts for floods, fires, crime and other dangers nearby — Sign up free / download mobile app and keep yourself and loved ones out of harm's way.Sign up free
The long version

A warning is a chain of people, and every link is on the map.

This is the whole thing, in order, in plain words: where the information comes from, who is allowed to say it is true, how it reaches your phone, and what happens in the days after.

Where a warning begins

A city floods from rain that fell an hour away, so the first link is upstream. Solar stations on the bridges and culverts read the water level every few minutes and send it over a radio mesh that keeps working when the phone masts do not.

For fire, crime, waste and weather there are no sensors. There are residents with phones, and there is the news. A report needs a photo, a place and a few plain words. A news story is read by the platform, placed on the map inside an honest ring of uncertainty, and held until somebody checks it.

5 minbetween station readings. A station that goes quiet is itself a signal, and the map says so.

In Accra every step of that chain has a name: the upper Odaw catchment, a gauge at the Avenor bridge, a gateway on a rooftop, and a model that has seen every rainy season since it was switched on.

Rainfallupper Odaw catchment
↓
Rain gauge + water-level sensorsradar · ultrasonic · rain · T/RH · pressure
↓
LoRa node → GatewayLoRaWAN mesh · LTE-M/Cat-M1 failover
↓
Cloud + AI predictionGBM + RNN ensemble · GMet radar fusion
↓
App · Push · Dashboard · NADMOtiered alerts: advisory → warning → evacuate
1 · SensingSolar-powered nodes with radar and ultrasonic water-level sensors, a rain gauge, temperature/humidity and atmospheric-pressure sensors — GPS-located, tamper-detecting, in flood-rated IP67 enclosures. Each node reports every 60–120 seconds and runs 5+ days on battery backup alone.
2 · ConnectivityA LoRaWAN mesh (10+ km per gateway) is the low-power backbone; LTE-M/Cat-M1 cellular takes over automatically if it fails — so the network stays up at the exact moment it matters most.
3 · IntelligenceA cloud AI engine fuses sensor streams with GMet radar and satellite rainfall data, producing street-level flood-risk scores refreshed every five minutes — and retraining after every event.
4 · ActionTiered alerts — advisory → warning → evacuate — go out through the app, push notifications, email and the public dashboard, in English and local languages.
See the stations →

The people who check

Nothing is drawn on the public map until it has been verified. The agency responsible for that service, inside its own area, sees every report for it first. Where no agency is onboarded yet the platform's own reviewers do the work, and the pin says so plainly rather than borrowing authority it does not have. Neighbours matter too: a few people pressing Confirm on a pin they can see move it faster than the same number of separate reports ever could.

Who answers is stated, never implied.

The model helps, and is labelled as help. It merges one event reported by three outlets, resolves a place name to a street, and estimates where the water goes next — always marked as an estimate, never as a fact.

Report an issue →

The map

The home page is the whole network at once. Pins are verified incidents, coloured by service. A cluster is pins too close together to draw apart. A soft ring means the exact spot is uncertain. Stations turn amber, then red, as the water rises, and a faded pin is something already resolved.

💧 FloodSensor stations on drains and rivers, AI nowcasts, street-level warnings 30–120 minutes ahead — the deepest service, covered in detail below.
🔥 FireStructure and bush fires, gas leaks, smoke sightings — reviewed by the fire service, mapped live so neighbours evacuate early.
🛡️ CrimeTheft, assault and suspicious activity, visible to responders by default and published only when the police approve — trust without vigilantism.
♻️ WasteIllegal dumping and choked gutters — the slow emergencies that cause the fast ones when the rains come.
⛅ WeatherLive conditions, 24-hour rain outlook, air quality and rain radar, fused into every flood-risk score.
🧡 SafeNetYour family's safety net: consent-first live location for people and assets and danger-zone alerts, on the same live map.

Tap anything and you get the category, who verified it, how long ago, and a Confirm button if you are standing in front of it. The time bar above the map is the same map at another hour — the last few hours, yesterday, the week. Nothing is hidden; it is filtered.

Open the live map →

How it reaches you

Alerts are for places, not accounts. Save home, work and the school, and you are told when a verified incident or a forecast touches them: a push notification first, then email or Telegram if you have linked them and the phone stays silent.

10languages. The alert is translated as it is sent, and so is the button on it — that is the one you have to press.

The card carries your household plan and the emergency number with it, so the page you would have had no time to look for is already open.

What an AlertCommunities afternoon looks like
2:15 pm — Rain begins

Heavy rainfall starts over Achimota. Upstream sensors log rising intensity.

2:38 pm — Threshold crossed

Water at the Avenor bridge node passes the warning threshold, climbing 4 cm per minute.

2:39 pm — AI predicts flooding at Circle

The model projects the flood wave will reach Kwame Nkrumah Circle in ~55 minutes.

2:40 pm — Alerts go out

Warnings reach residents, traders and drivers across the corridor in the app and by push notification. NADMO's dashboard flashes the risk heat-map.

3:05 pm — NADMO dispatches

Response teams pre-position at Circle and Odawna. Market traders move goods to higher ground.

3:20 pm — Roads closed

Police divert traffic away from the underpass before it becomes a trap.

3:35 pm — The flood arrives

Water covers the road at Circle — but this time, the people and goods are already gone.

56 minof warning delivered before the water arrived — enough to move people, goods and vehicles
4The app, push notifications, email and the public dashboard — no single point of failure
Check this device →

Your family

SafeNet is the same alerts, for the people you watch over: a child's phone, a parent, the family car. Each one is added with consent and shares only what both sides agreed to. Zones have hours. A share you gave for one journey expires on the day you set.

Those alerts ride through silent mode, on purpose.
SafeNet →

What you do next

The warning is the easy part. Your community's check-in asks for one tap — fine, or need help — and shows the people leading it who has not answered yet. Volunteers whose skills, radius and hours match are paged with one question: can you go?

Local resources show the nearest shelter, clinic and fire station with directions. And the learning centre is ten things, each with a short lesson, that make all of this land when it matters: the plan, the go bag, the numbers.

School visits & open dataStudents explore live water-level data from their own neighbourhood and learn how rainfall becomes a flood — with free classroom datasets.
Build-a-node workshopsHands-on electronics workshops where students assemble and test a real sensor node with our engineers and Ghana Planetarium educators.
Flood-warden trainingCommunity volunteers learn to validate sensor readings, report ground truth, and lead their street's response when a warning lands.
How prepared are you? →

The week after

Every flood has a drain behind it. A community's issues list turns one complaint into a cluster: reported, supported by neighbours, queued for the assembly, and you are told when it is resolved.

The stories feed keeps the record of what happened and what was done about it. The next rainy season starts with the gutter that was cleared in March.

Read the stories →

Doesn't Accra already have flood warnings?

Yes — regional forecasts and radio announcements. They matter, but they were never designed to tell a trader at Kaneshie that her street floods in 50 minutes. That's the layer we add.

Today's systemsAlertCommunities
City-wide weather forecastsStreet-level sensors on the actual drains and rivers
Regional warnings, hours-old dataHyperlocal alerts refreshed every 5 minutes
Manual observation and phone callsAutomated real-time monitoring, 24/7
Reactive — respond after flooding startsPredictive — 30–120 min ahead, targeting 3–6 hrs
Limited public visibilityOpen live map anyone can check on any browser
Commercial flood systems built for rich citiesCommunity-owned, ~$3,200 all-in per station

Common questions

Do I need a smartphone to get alerts?

Alerts come through the app or the website, by push notification, and by email or Telegram if you link them. Any phone with a web browser works; a basic phone without internet cannot receive them yet.

How far ahead can AlertCommunities warn me?

Today the pilot delivers 30–120 minutes of lead time depending on where rain falls in the catchment. As sensor density and the AI model grow, we're targeting 3–6 hours.

What stops the sensors being stolen?

Every node has built-in tamper detection and GPS location, uses low-resale-value enclosures, and is adopted by community flood wardens who own its upkeep. Rapid low-cost replacement is built into the budget.

What happens when the mobile network goes down in a storm?

The primary link is LoRaWAN radio — independent of the mobile network — with LTE-M/Cat-M1 cellular as automatic failover. Nodes also cache readings and retransmit when reconnected.

Who pays for this long-term?

Grants fund the build-out; operations are sustained by government platform licences, insurance data feeds and corporate risk subscriptions — so the network outlives the grant cycle. More on the model →

Ada
An unhandled error has occurred. Reload 🗙