Better model to predict floods likely

It will include 3-D land-height map of Marina catchment area for a start
Feng Zengkun Straits Times 21 Jan 12;

NATIONAL water agency PUB could be using a better flood-prediction computer model in the near future.

It will include a 3-D land-height map of just the Marina catchment area for a start, to predict the direction in which rainwater will flow at ground level during storms, and where flooding might occur.

Such a map, which the PUB has commissioned, will depict land height in that area to within 10cm accuracy.

Computer models now in use only predict how rainwater flows within drains and canals, and the intensity of rainfall they can handle.

PUB called for proposals for the land-height map last month and is reviewing the submissions.

A panel of drainage specialists appointed by the Government to tackle the flood problem recommended a national map last week.

But PUB told The Straits Times it will focus on the Marina area first.

The Marina catchment area makes up a sixth of Singapore and has been hit by floods in recent years. It includes Orchard Road, which has been hit by floods recently.

According to PUB's tender document for the map, obtained by The Straits Times, the work will cover some 100km of roads in low-lying and flood-prone areas of the zone.

The agency said it came up with the requirement that the map show the lay of the land to within 10cm accuracy by examining floods, which run between 5cm and 30cm deep.

The terrain will be captured in 3-D for the map through a technique known as ground-based light detection and ranging (Lidar).

Using this involves outfitting a vehicle with a Global Positioning System (GPS) and 360-degree laser-scanning technology.

The scanners emit laser pulses, which bounce off surfaces back to the scanners. By recording the travelling time of the pulse, a computer system calculates the distance between the car and the ground.

This data is combined with the GPS position of the car to produce a 3-D map.

Experts say this technique produces the map faster and cheaper than manual land surveys can, but it suffers in urban areas, where tall buildings may interfere with the GPS signals. The pulses may also bounce off cars and pedestrians, distorting the land-height information.

In its tender, PUB specified that the vehicle should have cameras that take high-resolution photographs, which can be used to evaluate the laser data.

The tender also calls for the Lidar- equipped vehicles to be supplemented by ground-survey crews in areas where poor GPS reception may affect accuracy.

If the data collection is distorted by parked cars or pedestrians, the surveys should be repeated at night, when such distortions are less likely, it said.

Dr Armin Gruen, a mapping specialist with Future Cities Laboratory, said one shortfall of Lidar is it does not collect details on land use which is important for flood-prediction computer models.

Whether the ground surface is soil or concrete, for example, will determine if water seeps into the ground or stays above it, he said.

To this, PUB said the Lidar car's camera system can capture land-use data.

The Future Cities Laboratory, a collaboration of the National University of Singapore, Nanyang Technological University and the Swiss Federal Institute of Technology Zurich (ETH), does research on urban planning, including the use of aerial photographs to produce 3-D maps, a field called stereophotogrammetry.

Dr Gruen, who is the chair of photogrammetry in ETH's Institute of Geodesy and Photogrammetry, suggested combining the ground approach with an aerial survey to get a comprehensive data set for the computer model.

His team demonstrated to The Straits Times how a digital land-height map could be used to predict floods.

They showed a map of Singapore divided into squares, with each square assigned an arrow and a number. The arrow shows the direction water will flow over that area based on the lay of the land; the number represents the number of other squares that feed water into this square.

An '86' square, for example, gets rainwater from 86 other squares, which identifies the area in it as more flood-prone than one with a lower number.