After spending money installing numerous sensors in canals and large drains during the last few years, we’ve now discovered that they were “overwhelmed” by last Sunday’s (5 June 2011) storm. The sensors were designed to monitor levels every two minutes in wet weather. When the water level in a drainage system hits 75 percent of capacity, the sensors are supposed to send an alert to subscriber buildings. They do this again when the level hits 90 percent of capacity.
Last Sunday however, the water rose so fast, the system failed. Alas, it wasn’t clear from the Straits Times why exactly it failed, though we may be able to make an educated guess.
Try to parse this report for yourself:
In a matter of minutes during Sunday’s heavy storm, Stamford Canal at Orchard Road was overwhelmed.
So was the PUB’s water level sensor software, which sends out SMS flood alerts to building owners in flood-prone areas.
The software was programmed to send out alerts when water levels in canals hit 75 per cent, indicating a ‘moderate’ risk of flooding; then at 90 to 99 per cent, indicating a ‘high’ risk. It was not programmed to send out text alerts if the water level went beyond 100 per cent.
However, at the Stamford Canal in front of Forum The Shopping Mall that day, water levels surged from 50 per cent to overflowing – or above 100 per cent – in two minutes or less.
The result? The building managements of Tanglin Mall and St Regis Residences and Hotel did not receive any alerts. Both the mall and St Regis Residences suffered serious flooding, although the hotel was spared that.
However, on Sunday, water levels at that segment of Stamford Canal rose ‘three or four times faster than anything we have seen before’, said Mr Peng Kah Poh, director of PUB’s infocommunnications department.
‘It was so high that it bypassed the trigger points of 75 per cent and 90 per cent, so the SMS alert was not sent,’ he said.
It will also look at shortening the current two-minute time interval for monitoring water levels during wet weather.
— Straits Times, 8 June 2011, Alert system overwhelmed by storm, by Grace Chua
A graphic accompanying the newspaper’s article revealed that at the 50-percent mark, the system notifies the Public Utilities Board (PUB). Then when it crosses 75 and 90 percent, alerts are sent to buildings that subscribe to the alert system. The report above indicates that after crossing 50 percent, the water rose to overflowing within two minutes; in other words, by the subsequent sampling, it was over 100 percent already. Perhaps the system then failed to send out an alert because alerts are only triggered when water levels are in prescribed ranges (75 – 89% for the first public alert and 90 – 99% for the next one), not triggered by measurements that return a True finding when tested for >75% or >90% levels. If you understand a little about software programming you will immediately see the difference and slap your head over the folly of using ranges instead of threshold tests!
I wished the reporter had the presence of mind to pin the official down on this.
* * * * *
Anyone who has ever struggled with software glitches in our government systems will tell you this is bloody typical of the Singapore bureaucracy that keeps on boasting how wired they are but actually is bedevilled by amateurish incompetence.
For example, I’m told by a friend that filing annual returns for a registered society is a pain. The system requires attachments (e.g. financial statements) to be scanned and linked to the filing, except that it imposes a maximum file size that effectively limits it to only two pages of attachments. Anybody who has handled audited financial statements will tell you that if they come in only two pages, they are very dodgy. So who in the Accounting and Corporate Regulatory Authority came up with software that only wants dodgy statements?
In 2007 I wrote about trying to lodge an application for a Professional Visit Pass for a foreign academic, Douglas Sanders. The online form that I had to fill allowed me a maximum of 25 characters (characters, not words) to declare ALL his previous employers and their addresses. In another box, where I had to declare ALL his previous residential addresses, I was also allowed the grand luxury of 25 characters.
Am I surprised that storm surge software is not able to handle a storm surge? No.
* * * * *
Was it such an impossible event last Saturday that the water rose ‘three or four times faster than anything we have seen before’, as Peng Kah Poh said (quoted above)?
A total of 124mm of rainfall was recorded for the central area, about 77 per cent of the average monthly rainfall for June.
About 65mm was recorded within 30 minutes yesterday compared with 100mm within two hours on June 16 last year, which had led to serious flooding of Orchard Road.
— Straits Times, 6 June 2011, Govt to review drainage after year’s worst flood, by Grace Chua
Already, those numbers suggest something troubling. Why is it that the data for the 16 June 2010 flood is based on a two-hour reading (100 mm), and that for 5 June 2011 based on a half-hour reading? Why not compare the 65 mm of 2011 with the worst half-hour reading from June 2010? Apples with apples? And if the PUB didn’t volunteer the information, why didn’t the reporter ask?
Is it possible that in fact we do not have data pertaining to half-hourly precipitation until recently? Because if we did, it would be the easiest thing in the world to compare like with like. And if we do not have historical data pertaining to half-hour spans, then how would the PUB know that the flow that was experienced was three or four times faster than anything we’ve seen so far?
It’s like this: Unless we have historical data about surges, we really can’t plan for them.
Consistent with my guess that we have not been collecting such data, I cannot easily find references to precipitation peaks in previous years on a half-hour or even one-hour basis. I see reports like this one, mentioning rainfall on a 24-hour basis:
To date, the highest rainfall recorded in Singapore over 24 hours was 512 mm in 1978, the second highest was 467 mm in 1969, while the third highest amount of rainfall to drench the island was on 20 December 2006, when 366mm of rain fell over a 24-hour period. Flash floods also occurred in various parts of the island republic in 2007 and 2009, raising the disturbing possibility that such floods are becoming a regular feature.
— Asia Sentinel, 17 June 2010, Hip Deep on Singapore’s Orchard Road. Link.
Every search I tried about the 2 December 1978 flood returned the same figure of 512 mm in a 24-hour period, such as this sunny story in The New Paper (click the thumbnail) titled Why Orchard Road will never flood. No other figure regarding peak rainfall for the worst one-hour or two-hour period could be found. I find it hard to believe that the data is not available (at least internally within the Meteorological Department) but hey, stranger things have happened.
So, might it be possible that planners simply used the figure of 512 mm over 24 hours for planning purposes? Dividing it by 24 hours, they would come up with 21.3 mm per hour, then make some adjustment on the fly and say to themselves: Let’s plan for 50 mm per hour. Only to be caught out when rain came down at a rate of 65 mm in 30 minutes.
That’s the danger with relying on averages when trying to make projections for surges. There is no short cut. One has to go out and collect data about surges themselves.
If this was what happened, it would not the first time. Remember the huge controversy a few years ago about crunches on trains, with the management of SMRT Corp issuing press release after press release giving average crowd densities, in effect denying that things were as bad as commuters reported them to be? I wondered then, how did they measure what they measured? When did they measure what they measured?
If only we had a Freedom of Information Act, I would be asking for historical data of previous rainfall peaks. I would also be asking what assumptions the PUB used in planning for drainage, not just in terms of precipitation, but also in terms of run-off from built-up areas. Then hydrological academics in our universities can crunch those numbers and tell us whether they are realistic.
If only. . . .
How do we monitor the quality of our government unless citizens have access to information?