A flood warning is not produced by one rain gauge.
Rainfall begins in the atmosphere, lands unevenly across neighbourhoods, runs across roofs and roads, enters drains, raises water levels in canals and finally tests whether each downstream pathway can carry the arriving flow quickly enough.
Singapore therefore watches several stages of the same event at once.
PUB combines weather radar, rain gauges, more than 1,000 water-level sensors in drains and canals, more than 500 CCTV cameras, field observations and the locations of flood-response vehicles. PUB has also announced another 650 water-level sensors by end-2028, bringing monitoring coverage to roughly one sensor per kilometre of major waterways and canals.
The operating loop is: observe approaching rain → measure rain actually falling → watch drainage levels respond → compare the pattern with flood-prone locations and system capacity → deploy Quick Response Teams → issue alerts → verify conditions on the ground → update the operating picture as the storm moves.
Quick answer: why are several sensor types needed?
- Weather radar: shows where rainfall is developing and moving across Singapore.
- Rain gauges: measure rainfall that actually reached the ground at specific locations.
- Water-level sensors: show how drains and canals are responding hydraulically.
- CCTV: provides visual confirmation where water levels, roads and local conditions need interpretation.
- Drainage maps and asset data: explain where the water is expected to flow next.
- Response-vehicle locations: help operations teams send people and equipment toward developing hotspots.
- Public reports: provide additional local observations that can be checked against instrument data.
1. Flood forecasting begins before rain reaches the drain
Weather radar detects precipitation over a broad area and helps show how storm cells are developing and moving.
This gives drainage operators an earlier signal than waiting for a canal to become visibly high.
The forecast is still probabilistic. A convective storm can intensify, weaken or shift over short distances. Radar therefore extends preparation time without turning tropical weather into a perfectly known future.
2. Rain gauges turn a forecast into measured rainfall
Radar estimates rainfall over space. A rain gauge measures what reached one point on the ground.
The two measurements complement one another.
A gauge can confirm an intense local downpour that may be difficult to interpret precisely from radar alone, while radar shows whether the heavy cell extends beyond the gauge or is moving toward another catchment.
3. Singapore rain can vary sharply across neighbourhoods
Tropical convection can produce intense rain over one part of the island while another area receives much less.
An islandwide average therefore says little about whether one particular drainage catchment is under stress.
Flood monitoring has to preserve geography: which cloud, which rain gauge, which drain and which downstream waterway belong to the same local event?
4. Water-level sensors measure the drainage system’s response
Rainfall intensity is only the input.
The flood question is whether water is accumulating faster than the drainage pathway can move it away.
Water-level sensors in drains and canals show that response directly. A rapidly rising level can indicate that the local system is approaching a more critical operating condition even before overtopping occurs.
5. More than 1,000 sensors create a distributed picture
PUB stated in its 2025 flood-resilience campaign that more than 1,000 water-level sensors were installed across Singapore.
The agency is adding 650 more by end-2028.
The expansion matters because a drainage network can change state locally. A sensor far downstream may be too late to explain what is happening in a smaller upstream canal or side catchment.
Denser sensing shortens the distance between the physical event and the nearest instrument capable of seeing it.
6. CCTV adds visual meaning to a number
A water-level reading says how high the water is at the sensor.
A camera can show whether a nearby road is passable, whether flow is turbulent, whether debris or a local obstruction is visible and how fast conditions appear to be changing.
PUB’s network includes more than 500 CCTV cameras for flood monitoring.
Numbers provide scalable monitoring. Images provide context when operators need to interpret what the number means on the ground.
7. The Catchment and Waterways Operations System joins the feeds
PUB has described its Catchment and Waterways Operations System, or CWOS, as an integrated flood-operations dashboard.
The system brings together rain gauges, water-level sensors, CCTV, weather-radar information and response-vehicle locations so operators do not have to compare separate screens manually during the most time-sensitive part of a storm.
The value of integration is not merely convenience. It makes relationships visible: heavy rain upstream, rising canal level downstream and the nearest response team can appear inside one operating picture.
8. Forecasting flood risk is harder than forecasting rain
Two storms with similar rainfall can produce different outcomes.
The catchment may already be wet. A downstream canal may be high. Tide or coastal conditions may constrain discharge. A local drain may be partially obstructed. Construction may have temporarily changed runoff paths.
Flood forecasting therefore has to combine meteorology with drainage state and local geography instead of assuming millimetres of rain translate directly into centimetres of floodwater everywhere.
9. Drainage capacity is a threshold, not a promise of zero flooding
Every drain has finite hydraulic capacity.
Singapore upgrades drains and uses Source–Pathway–Receptor measures, but exceptional rainfall can still exceed local conveyance or storage capacity.
A flood forecast therefore helps manage residual risk after infrastructure has done as much as its design allows.
10. Quick Response Teams turn data into field action
Sensors do not place barriers, direct road users or inspect a drain physically.
PUB deploys Quick Response Teams to locations where monitoring indicates elevated flood risk or actual flooding.
Knowing team locations allows operations controllers to consider travel time as part of response planning rather than merely asking which site has the highest water reading.
11. Public alerts are another operational output
Monitoring becomes more useful when the people exposed to risk receive timely information.
PUB issues flood and heavy-rain information through public channels including myENV and other official alert routes.
The objective is not only warning after a road is already impassable. Earlier information can change travel, site protection and operational decisions before exposure increases.
12. A sensor alarm still needs interpretation
An unusual reading can indicate real rising water, a temporary local disturbance or an instrument problem.
Operators therefore compare neighbouring sensors, radar, cameras and field reports rather than treating one anomalous value as a complete flood diagnosis.
The strongest warning comes from converging evidence.
13. Flood operations manage time as much as water
Rain can intensify faster than permanent infrastructure can be changed.
During the storm, operators therefore manage the short horizon: where the next heavy cell may go, which waterway is rising, which road may become vulnerable and where field teams should move first.
Long-term engineering reduces the probability and severity of flooding. Real-time operations manage the remaining minutes.
14. Radar can be early and still be uncertain
A useful forecast is not the same as an exact prediction.
Short-horizon radar nowcasting can show where rainfall is likely to develop or move, but tropical storm cells can change rapidly.
Singapore therefore keeps measuring after the forecast is issued. The next rain-gauge reading and water-level sensor update can confirm or contradict what the earlier radar suggested.
15. Local flooding can occur between sensor points
No sensor network observes every metre of every street.
A blocked inlet, very local depression or sudden micro-scale runoff concentration can create a problem between instrument locations.
This is why cameras, field teams and public observations remain relevant even as sensor density increases.
16. More sensors improve detection, not the size of the drain
Instrumentation does not increase hydraulic capacity by itself.
It buys information and decision time.
Drain widening, detention, pumping, coastal protection and local flood barriers remain engineering measures. Monitoring helps Singapore decide when and where those systems are under stress and how to respond while longer-term upgrades are planned.
17. A worked example: afternoon convective storm
Imagine radar shows an intense storm cell growing west of a dense urban catchment.
Rain gauges begin recording high local intensities. Upstream water-level sensors rise first, followed by downstream canal sensors. CWOS shows the storm track, drainage response and nearby QRT locations. CCTV confirms ponding near one low road. PUB deploys field teams and issues public information while continuing to watch whether downstream levels stabilise or keep rising.
The flood response began before anybody could describe the event with one final number.
18. Common misconceptions
Misconception: Radar can tell exactly which street will flood.
No. radar mainly describes rainfall; local flooding also depends on drainage state, topography and site conditions.
Misconception: A high water-level sensor proves flooding has already occurred.
No. it shows hydraulic state at that location; the critical threshold and local context determine the consequence.
Misconception: More sensors prevent floods.
No. sensors improve detection and response; physical drainage and flood-protection infrastructure provide capacity.
Misconception: One dry neighbourhood means an islandwide warning was unnecessary.
No. tropical rainfall can be highly local and warnings may target catchments experiencing very different conditions.
19. The deeper idea: flood forecasting is a chain of shrinking uncertainty
Radar sees a possible future first.
Rain gauges confirm what reached the ground. Water-level sensors reveal how the drainage network absorbed it. Cameras show visible consequences. Field teams return local evidence. The next sensor update tells operators whether the intervention and changing weather are moving the system toward recovery.
No single layer knows the whole flood.
Singapore’s flood-monitoring system works because each layer reduces uncertainty left by the one before it—and because the information returns quickly enough to change what people do before the water finishes moving.