FINDING the tanker responsible for an oil spill can be surprisingly difficult. Malaysian officials are wrestling with the problem after a huge spill in a busy shipping route on 25 June left a 5-kilometre slick off the country鈥檚 coast. A new software system developed in Australia could soon make the task easier.
To identify the culprit, environmental protection officers usually test the chemical composition of a slick and compare it with samples taken from the tanks of suspect ships. But deciding whether two samples match is not easy, says Brynn Hibbert of the University of New South Wales, who led the software design team.
False positive matches, where two different oil samples are thought to be the same, can happen because petroleum products often have a similar basic chemical composition, even if they come from different places and were refined differently. False negative matches, when two samples that should match do not, occur because the chemical profile of an oil slick can change significantly in a matter of days. 鈥淎s soon as oil or another pollutant hits the environment, it starts to change. The more volatile stuff comes off, bugs start to eat it, seabirds might shit in it. Establishing whether an oil sample really did come from a suspect source is often very difficult to do.鈥�
Advertisement
To make identification more reliable, Hibbert鈥檚 team has devised software that decides how likely it is that a pollutant sample and a suspected source really do match. They developed the system using decades of records of the results of investigations conducted by the Environmental Protection Agency of New South Wales. The software looks for similar case studies, their test results, and other pertinent information about the oil samples, before generating a probability that the suspect source was responsible.
鈥淪pilt oil can quickly change in the environment. Bugs eat it, while birds shit in it鈥�
The New South Wales EPA does not investigate oil spills at sea, but the software could easily be adapted to examine past oil spill records, says Hibbert, who presents his research at a Royal Australian Chemical Institute conference in Sydney this week. He thinks the software鈥檚 predictions are reliable enough to be used as evidence in court.
