Colombo's flood: a ten-year storm, at or past what the drains were built for

Two hours, minute by minute

Nearly all of the day's rain fell between 17:30 and 19:30. NBRI's automatic gauge at its Colombo office in Thimbirigasyaya reports every 15 minutes, and its record shows how the storm built.

A midday shower had left 19 mm by 13:15, then nothing fell for more than four hours. The storm arrived at 17:30 with 11 mm in its first quarter hour. The heaviest 15 minutes brought 27.6 mm, a rate of about 110 mm an hour. After 19:30 it faded to drizzle, and by 22:15 the gauge had stopped at 146.6 mm for the day, the figure the Disaster Management Centre reported for Thimbirigasyaya.

Heaviest over the city centre

The heaviest totals were in the city centre. Four gauges there caught 118 to 147 mm in the 24 hours. Gauges in the suburbs got about half that, and gauges 35 km inland got almost nothing. These are day totals, which include the midday shower, so they show where the rain was heaviest rather than the exact size of the storm.

The four gauges with more than 118 mm lie within about 9 km of one another, from Thimbirigasyaya to Mattakkuliya by the Kelani. Gauges 5 to 6 km east and south of Thimbirigasyaya, at Battaramulla and Dehiwala, had about 85 mm, and those further south less again.

That concentration matters for the flood. Heavy rain at one gauge is common enough; similar depths across much of the city's drainage at once put more water into the same canals at the same time.

How rare was it?

At its wettest point, Friday's storm was about a once-in-ten-years downpour over two hours. Its sharpest bursts were probably ordinary for Colombo. What was unusual was how long the heavy rain kept going.

Engineers judge storms with rainfall frequency curves: how much rain falls in a given time, in a storm that comes round once every so many years. A "ten-year storm" has about a one-in-ten chance of happening in any year; it is not a promise of ten quiet years after it. Colombo's curves were published in 2017 from the Department of Meteorology's recording gauge at the Colombo observatory, 1.2 km from Thimbirigasyaya.

• Short bursts: probably ordinary. The best estimates for the heaviest 15 and 30 minutes fall below even the lowest curve, rain Colombo sees in most years. Readings every 15 minutes cannot rule out a sharper burst inside them; at most it would have been about a six- to seven-year burst.
• One hour: about a six-year storm, about 92 mm (between 90 and 101 mm).
• Two hours: about a ten-year storm, 125 mm, roughly a 10% chance in any year. In the observatory's 1981–2010 record only 2010 and 1992 had wetter two hours, though that record comes from a different gauge 1.2 km away and does not cover the years since 2010.
• The whole day: about a two-year day. The 146.6 mm is close to Colombo's average wettest day of the year.

These figures are for one gauge, the wettest we found. At the observatory itself the whole day brought 123 mm. Similar depths at several city-centre gauges suggest the event over the city as a whole was rarer than a single point shows, but the gauges are too few to say by how much.

Colombo has seen worse

Friday's storm had about a third of the rain of Colombo's two great floods, but came close to them over its worst two hours. It was a short, sharp storm, not a night of soaking rain.

The two-hour figures for 2010 and 1992 are the highest in the Colombo observatory's 1981–2010 record, published with the city's frequency curves. Friday's two hours, measured at Thimbirigasyaya, would have ranked third in that record; storms since 2010 are not in it.

The "30 mm drains" and the "129 mm hour"

The day after the flood, officials explained it with two figures: drains built for about 30 mm of rain an hour, and 129 mm falling in one hour. The rain total is real, but it fell over about two hours. The drain figure has no design document behind it that we could find.

What was said

The 30 mm figure was not coined for this flood: the Council's engineering head gave it in January 2025. But it has not been tied to a design document, and its wording changes from telling to telling.

What the design documents say

No document we found sets a standard of 30 mm an hour. The figure may describe the oldest drains, or what blocked drains carry today, but nobody has shown which. What the documents do show is that much of the network was designed for storms of two to ten years: the 2003 master plan sets five to ten years for urban drains, the Phase II drains were built for a two-year storm, and the city's newer micro-drains for a ten-year one.

What the gauges say

The 129 mm is a real figure. It is the Department of Meteorology's total for its Colombo station, reported as falling "within a few hours", and it is close to the 125 mm the Thimbirigasyaya gauge caught in its two-hour burst. What the record does not show is that it all fell in one hour. The "one hour" may well be a shortening, made in a busy briefing or in the reporting of it.

At the wettest gauge with 15-minute readings, the worst hour was about 92 mm. A storm can be heavier a kilometre or two from the nearest gauge, so a somewhat wetter hour somewhere in the city cannot be ruled out; but nothing measured points to 129 mm in an hour. The length of time matters: 129 mm in one hour would be about a 60-year storm on Colombo's curves, while 129 mm in about two hours is close to a ten-year storm.

Design rates depend on how long the rain lasts: a street drain fills in minutes, a canal over hours. The ten-year rate of 80–100 mm an hour matches roughly one to one and a half hours on Colombo's curves, so the comparison below is fair for that row. The other rates are quoted without a duration, so their comparison is only a rough guide.

On the standards in the documents, a storm like Friday's is at or beyond what much of the network was built to carry, so water standing at low spots was to be expected even with clean drains. That supports the officials' central point, that the rain exceeded what the drains carry, even though the figures given for it were wrong. Whether the drains also fell short of their own standards cannot be told without the design and maintenance records and a map of where the water stood.

What else officials named

The same meeting named waste dumped in drains and unauthorised construction obstructing drains and canals (Newswire). Reports blamed the hospital flooding on the Norris Canal not being cleaned (Sri Lanka Mirror). JICA's 2003 study had already found many channels unable to carry even a two-year flood because of encroachment, silt and weeds. The Director General of Meteorology also said blocked drains kept the water standing longer.

The Kelani was not in flood: the Irrigation Department's gauge at Nagalagam Street stayed below its Alert level all night. We have not checked the state of the tide during the storm, which affects how fast the canals can empty into the sea. From Saturday the Land Development Corporation and the Council began clearing canals, starting with the Serpentine Canal at Dematagoda and the Torrington Canal (Newswire).

The record does not support a 30 mm design standard, and the 129 mm, true as a total, was attached to too short a time. Together they make the storm sound more extraordinary than it was. But the drains' design documents point the same way as the officials: a storm like this is at or beyond what much of the network was built for. The figures do not show anyone misled on purpose: a remembered figure repeated in good faith, or a total shortened in a busy briefing, would look the same. The Council could settle the 30 mm question by publishing the design report behind it.

Why the water had nowhere to go

Drains are only the last step. Where Friday's water went depended on how much of the city lets rain soak in, where water can be held, how fast channels can carry it, and whether it can get out to the sea or the river. No single tunnel or bigger pipe covers all four.

How much water there was

Central Colombo (the Colombo and Thimbirigasyaya divisional secretariats, about 44 km²) took roughly 100 mm (95–110 mm) in the two-hour burst on average, going by the city's gauges: around 4.5 million cubic metres of rain, enough to fill about 1,800 Olympic pools. On ground that is mostly roofs, roads and paving, most of that runs off, roughly 3.3 million cubic metres. In the wettest hour the water leaving the city's surfaces was in the order of several hundred cubic metres a second. These are rough figures, good to about a quarter either way, meant only to show the scale.

Capacities as published by the Sri Lanka Land Development Corporation, which built these works. Each serves a particular part of the city. Most of central Colombo drains by gravity through open canals and outfalls to the sea, such as the Wellawatte and Dehiwala canals and Beira Lake's outlets, whose capacities we have not found published. No drainage system carries a storm's peak away as fast as it falls: pipes, canals, lakes and, in the end, streets hold water for a while, so these capacities cannot simply be set against the peak.

Tunnels and bigger canals

A tunnel moves water quickly from one place to another; it does not make water disappear. It helps where a low spot sits close to the sea or a canal with spare room. Even a city-scale tunnel is no cure on its own. Kuala Lumpur's SMART tunnel, 9.7 km long, can take up to three million cubic metres, its builder says (Gamuda, from The Star, 2022), a volume of the same order as what fell on central Colombo on Friday. It was built as one part of a wider flood scheme, not as the whole of it.

Wider canals help only if the water can leave them. Much of Colombo is low: large parts of central Colombo lie within a few metres of sea level (a rough reading of the Copernicus elevation model, which measures the surface, buildings included). With so little fall, canals flow slowly, and their sea outlets need keeping open. The canals that drain east and north reach the Kelani and need pumping when the river is high. On Friday the Kelani was not in flood.

On this approximate map, the places reported flooded are not on the lowest ground in the city; that lies along the Kelani and in the marshes to the east. In a storm this short, water stands where it cannot get away fast enough, which depends on the drains and canals that serve each street as well as on its height. A map of where the water stood, and for how long, would show which.

Road drains and manholes

Road inlets have to be many and clear of silt and plastic to take water off the street in a heavy hour. But more inlets only help when the pipe below has room. When the drain is full, or the canal it empties into is high, water comes back up through the manholes instead. Inlets set how fast water leaves the road; the pipes, canals and outlets set whether it can.

Buildings, roads and lost marshes

Satellite maps of built-up land (the European Commission's Global Human Settlement Layer, made from Landsat and Sentinel images) show where the change has happened.

On the maps the difference between the two areas is plain. Central Colombo looks much the same in 1975 and 2020. In Kolonnawa and Kotte, land that was largely open in 1975 is mostly built on by 2020.

Central Colombo was already a dense city in 1975 and, on these maps, has changed little since; the early figures are the least certain, and the small change in the centre is within their margin. Roofs cover over a third of its ground, and roads and paving much of the rest: in ESA's 2021 land-cover map about 74% of central Colombo is built-up, and OpenStreetMap lists about 840 km of road and lane there, very roughly an eighth of the land. In a storm like Friday's every road becomes a channel, carrying water along to the lowest junctions.

The big change is to the east. The share of the marsh belt under roofs has more than doubled since 1975. Those marshes are the land the 2003 drainage master plan counts on to hold floodwater (JICA, 2003), and as much as 60% of Colombo's wetland area has been lost since the 1980s (IWMI). That is a long-term change to the wider drainage system. We have not shown that it played a part on Friday: the marsh belt had less rain than the centre that evening (88 mm at Battaramulla over the day), and we do not know which canals the flooded streets drain into. Building also narrows the channels themselves: officials named unauthorised construction on drain and canal reservations, and every raised plot or wall across a low path pushes the water on to the next property.

What would work

There is no single fix. Cities that cope with storms like this use several layers, each for a different part of the problem.

1. Keep what exists working. Clear inlets, drains and canals before each monsoon, not after the flood, keep the sea outlets open, and publish the cleaning records. It is the layer officials themselves named on Saturday.
2. Hold water before it reaches the drains. Protect what is left of the marshes; let parks, playgrounds and car parks hold a few tens of centimetres of water for a few hours; ask new buildings to store or soak in their own roof water. For scale, 3.3 million cubic metres is about 165 hectares of water 2 m deep: storage on that scale needs land, not pipes.
3. Make room ahead of a storm. Every 10 hectares of lake or retention pond lowered by a metre before forecast heavy rain makes room for about 100,000 cubic metres.
4. Fix bottlenecks, then build outlets. Find where narrow culverts and bridges, encroachments and blocked outfalls hold water back before choosing new tunnels, which make sense for specific low spots near the sea.
5. Plan for the storm that is bigger than the drains. Some storm will always exceed the design; on the documents' own standards, Friday's already did for much of the network. Decide which roads carry the water when that happens, and raise what must keep working: hospital emergency units, rail signals, electricity substations.

What the satellite estimates show

Satellite rain estimates, such as NASA's IMERG that our rainfall pages use, describe rain over an area, not at a single point. A storm whose heaviest rain falls on a few kilometres of city in two hours is spread across that area and that time, so it ranks much lower on a satellite estimate than it feels on the street.

That is a limit of what satellite estimates measure, and it applies to our own pages too: our satellite rain rankings can understate short, very local storms like this one. For a downpour over a city, the rain gauges say more. A low satellite ranking never means a place had little rain or is out of danger.

What it means

A storm like Friday's has about a two-in-three chance of happening at least once in any ten years. On the drains' own design standards, it will flood low spots again when it does; how badly depends on the state of the drains and canals, which the public record does not yet show.

1. Hold the explanation to the record. The official figures overstated the storm, while the design documents show it was at or beyond what much of the network was built for. Both the design figures and the canal-cleaning records should be public, with a map of where water stood, so the next flood can be judged against them.
2. Colombo's rainfall statistics need renewing. The city's frequency curves stop at 2010, and their authors found Colombo's rainfall intensities had increased compared with the older curves. Designs drawn from older curves may understate today's storms.
3. Gauges measure these storms. NBRI's automatic gauge in Thimbirigasyaya recorded this one in 15-minute steps and posted each reading on NBRI's public map as the rain fell. A denser, open network of such gauges in the city is what lets anyone check the next official explanation.

About the figures

• The frequency curves belong to the Colombo observatory; the Thimbirigasyaya gauge is 1.2 km away and was the wettest gauge we found. At the observatory the day's total was 123 mm.
• The curves end in 2010, and rain intensities have risen since, so today's return periods may be somewhat shorter than shown.
• Automatic gauges tend to read a little low in very heavy rain.
• The volumes, built-up shares, land cover, road length and heights in "Why the water had nowhere to go" are estimates from public maps and the gauges, meant to show scale; the elevation model measures the surface, including buildings and trees, so heights are approximate, and the 1975 built-up figures are the least certain.
• We have no map of where the water stood, how deep or for how long, so this page cannot say which part of the system (road inlets, pipes, canals, sea outlets or the tide) held the water back. One pattern is worth testing: the durations that load street drains (15 to 30 minutes) were probably ordinary, while those that load canals (one and a half to two hours) were about a ten-year storm. That points more towards the canals and outlets than the street inlets, but it is a question, not a finding.
• The design standards we found are from documents up to 2012. The Council may hold later ones.
• Rain gauge readings are as published by NBRI, the Department of Meteorology, the DMC and the Irrigation Department, credited above. This page is not a warning; for warnings, follow NBRI, the DMC and the Department of Meteorology.

Sources

• NBRI automatic rain gauge network; Department of Meteorology 24-hour rainfall report, 7 November 2026; Irrigation Department river levels
• Nandalal, K. D. W. and Ghnanapala, P. (2017), Development of IDF Curves for Colombo, ENGINEER 50(1), 33–39
• World Bank (2012), Project Appraisal Document, Metro Colombo Urban Development Project
• JICA (2009), Ex-post evaluation, Greater Colombo Flood Control and Environmental Improvement Project (II)(III); JICA and Nippon Koei (2003), Storm Water Drainage Plan for the Colombo Metropolitan Region
• NASA GPM IMERG precipitation estimates
• European Commission Joint Research Centre, Global Human Settlement Layer (built-up surface, 1975–2020); ESA WorldCover 2021; Copernicus DEM (elevation); OpenStreetMap contributors (roads, ODbL)
• Sri Lanka Land Development Corporation, ongoing sub-projects and flood and drainage management
• Kuala Lumpur SMART tunnel: Gamuda / The Star, Feb 2022; Colombo wetlands: IWMI
• Reporting: Ada Derana, Newsfirst, Daily Mirror, Newswire, LNW, The Island