In the narrow channel leading into the Port of Houston, ships traveling in opposite directions sometimes employ a tactic called the Texas Chicken. The vessels steer straight toward each other, and at the last possible moment, veer to the side. But what sounds like a maritime version of reckless hot-rodding is actually a precise technique for minimizing the chance of collisions and groundings along the shallow banks.
Taking a less dramatic tact, Nicholas Patrikalakis, a professor in the Department of Ocean Engineering, and Hauke Kite-Powell, a research specialist at the Marine Policy Center at the Woods Hole Oceanographic Institution (WHOI), are also hoping to reduce the risk of groundings and collisions. In a recently completed three-year study, the researchers and their associates examined historical data from four US ports. The goal of the study, funded by MIT Sea Grant, the US Coast Guard, and the US Army Corps of Engineers, is to reduce the number of accidents by understanding why they occur.
Annual costs of groundings and collisions in US waters run between $2 billion and $3 billion -- a figure that includes direct costs such as damage to ships and loss of cargo, and indirect costs such as delays for other ships and environmental damage. However, agencies responsible for maritime safety have only limited resources to battle this costly problem. According to Dr. Kite-Powell, "a better understanding of what leads to accidents can help us spend that money more wisely."
Since most accidents occur in ports -- where traffic is heaviest and water is shallower -- Drs. Patrikalakis and Kite-Powell and their colleagues and students examined data on groundings and collisions of commercial vessels in the ports of New York, Tampa, Houston and San Francisco. Specifically, they looked at data on accidents (from the Coast Guard), data on the number of ships going in and out of ports (from the Army Corps of Engineers) and data on environmental conditions at the time of accidents (from NOAA).
They were assisted by Di Jin, an assistant scientist at WHOI's Marine Policy Center; Stephen Abrams, a research engineer in the Department of Ocean Engineering; and three graduate students in that department: Johan Jebsen, Vassilis Papakonstantinou and Shu-chiang Lin.
"Risk is the probability that an accident will occur," said Dr. Kite-Powell. To ascertain that probability, the researchers used their data to create what's called a Bayesian model. That model, notes the researcher, "is based on a statistical equation that allows you to calculate how the probability of an event changes when surrounding factors change or how an estimate of the probability of an event is influenced by new information that you get about the event."
The study's findings are significant because they indicate both what does and doesn't contribute to an accident. One suspected factor was inaccurate nautical charts based on ocean surveys taken 50 to 100 years ago. Back then, lead-line soundings were taken at considerable intervals and chart makers simply estimated the depth of the sea floor between those soundings.
However, the researchers found no correlation between where groundings occurred and where water depth is less certain. This suggests that even modern surveys with multibeam technology would not necessarily have prevented the groundings.
Even more surprisingly, the study showed that ships did not run aground primarily during low tide. According to Dr. Kite-Powell, this may be because "many ships, particularly deeper draft ships, tend to move in and out of ports during high tide." In addition, errors in water level forecast also did not correlate with groundings -- perhaps, he said, "because people who pilot ships are aware of errors in forecasts and compensate for them."
Factors that do seem to trigger accidents are those associated with visibility, including night-time transits. But Dr. Kite-Powell is cautious in making a definitive statement about the dangers of traveling in the dark because the available data do not indicate the total number of ships transiting during daylight and dark. Yet the probable link and the dearth of information leads to one strong recommendation from the investigators: more comprehensive and accessible data.
"It's possible that data are kept somewhere," he said. "It's just that we couldn't find them and they're not readily available in digital form."
In addition to better data, Dr. Kite-Powell reported that one way to reduce risk would be to restrict port operations when visibility is poor. But he added that such restrictions would not win favor with shippers, and he offers another option: better navigational tools and aids.
"One example would be an electronic chart system coupled with a GPS [global positioning system] that gives them real-time information about exactly where they are in relation to the navigation hazards and to other ships," he said. Such devices, which he said are currently under commercial development, would ideally have systems to warn ships of potential problems.
This finding about the role of visibility may also help direct the Coast Guard as it makes decisions about navigational aids. With an increase in modern navigational tools on board ships, explained Dr. Kite-Powell, the Coast Guard has considered reducing the number of costly physical navigational aids such as buoys, lighthouses, light towers and on-shore range markers. But that might be a mistake. "People use visual cues extensively to navigate -- even today," he said.
The researchers also found maneuverability figures into groundings and collisions. Larger, less maneuverable vessels have higher risks than smaller ones. Also at higher risk are barge trains -- several barges rafted together and pulled or pushed by a single tug.
Yet here, too, the researchers found data scarce. "Important information about maneuverability -- draft of ship and presence of tugs -- was not recorded," Dr. Kite-Powell said. And Professor Patrikalakis noted gross errors in data that are recorded: some accident coordinates in the Coast Guard database actually indicate that accidents occurred on dry land. "These are probably recording errors during the database creation or data collection," he noted.
For its part, the Coast Guard hopes to correct such glitches and is currently revising its data collection procedures. And Dr. Kite-Powell said both the Coast Guard and the Army Corps of Engineers are keen on building a more complete model of risk. That model would include port configuration, said Professor Patrikalakis, who noted that the study also underscored the need to "invest more in rational methods for design of ports."
But aside from port design, limited or iffy data, visibility conditions and maneuverability, there's also the human element. And while this study did not address human factors (many Coast Guard studies have examined those), Dr. Kite-Powell pointed out one unavoidable human component: acceptable levels of risk.
"If you ask anybody in agencies or industry what an acceptable level is, they might be reluctant to give you an answer -- it's like asking people who regulate aviation safety what the acceptable number of crashes is in commercial aviation. I don't think there is a non-zero number that people shoot for or will admit to shooting for. In fact, though, there's one in aviation and it's not zero. There's one in shipping and it's clearly not zero. The acceptable risk is more or less what society accepts."
A version of this article appeared in the April 7, 1999 issue of MIT Tech Talk (Volume 43, Number 25).
