FACTBOX: Hurricane tracks are hard to predict

(Reuters) - The science of hurricane forecasting has come a long way since the days when entire fleets of Spanish galleons sank in unexpected storms as they tried to haul South American gold and other treasures back to Europe.

Despite the launch of "hurricane hunter" flights in 1944, the advent of satellite imagery in the 1960s and expanding research into the powerful tropical cyclones, it remains impossible to say exactly where a hurricane will end up more than a day or so away from landfall.

The following are some facts about storm forecasting:

* Tropical storms -- called hurricanes in the Atlantic when their top sustained winds reach 74 miles per hour (119 km per hour), and typhoons or cyclones in the Pacific -- are influenced and steered by atmospheric conditions around them, such as high pressure or low pressure areas. Neil Frank, a former director of the U.S. National Hurricane Center, described the movement of hurricanes as like a leaf being steered by the currents in a stream, except that for a hurricane the stream has no set boundaries.

* The accuracy of forecasting the tracks of hurricanes has improved and 3-day forecasts are as accurate as 2-day forecasts were in the late 1980s.

* Despite the improvements, the average error in the National Hurricane Centre's track forecasts is near 260 miles (420 km) by day four and 345 miles (559 km) by day five. That means a storm predicted at the beginning of a week to strike New Orleans could easily end up hitting Houston on a Thursday.

* Intensity forecasts are even more difficult. The hurricane centre calculates that the error in its forecasts for a storm's top sustained winds averages 23 miles per hour (37 km per hour) per day. That margin of error means it is quite possible to go to bed expecting a manageable Category 2 hurricane on the five-step Saffir-Simpson scale of intensity, only to wake up to face a dangerously powerful Category 4 storm.

* Track forecasts are derived from an array of computer models. Some are statistical and make predictions using historical information about the behaviour of past storms. The best tend to be so-called "global" models which involve complicated mathematical calculations by supercomputers based on a huge amount of climatic and environmental data collated from around the planet.

Sources: U.S. National Hurricane centre at www.nhc.noaa.gov/ and Jeff Masters at The Weather Underground on www.wunderground.com/

(Reporting by Michael Christie in Miami, Editing by Jackie Frank)