Last year, the early forecasts overestimated the number of named storms and hurricanes. (Bigstock)

For some forecasters gathered last week for the National Hurricane Center’s annual conference in Orlando, looking back at last season proved a painful reminder of just how tricky their jobs can be.

“It’s hard to go all these years and then make your worst forecast after 30 years,” said climatologist William Gray, 84, the man who pioneered the science of preseason predictions. “It’s not very good progress.”

Gray, founder of Colorado State University’s Tropical Meteorology Project, and colleague Philip Klotzbach recently released their closely watched forecast for 2014, which calls for just three hurricanes, with only one becoming a major storm reaching Category 3 or higher.

The annual forecast from Gray and Klotzbach — followed by a handful of other expert predictions, including from the federal government — has become a kind of opening bell for the hurricane season, which begins June 1.

But for the last two years, predictions have been off — way off in the case of 2013, which Gray called a “bust.” Colorado State University had predicted an active season, with 18 named storms and eight to nine major storms. Instead, it was slow, with 13 named storms and only two hurricanes.

Technology and ever-improving computer models have dramatically improved the ability to forecast the path of storms once they have formed. But, as last year illustrated, there are still flaws in long-range, preseason predictions.

In fairness, just about everybody else in the preseason prediction business also overshot the mark last year — including Florida State University, Columbia University and the National Oceanic and Atmospheric Administration.

Gray attributed the botched forecast to the dramatic weakening of a deep current called the Thermohaline Circulation, which carries denser, saltier water from the Earth’s polar oceans. When it is strong, more hurricanes tend to form.

The collapse of the current last year, Gray said, was the biggest observed since 1950.

“This is another illustration of how brilliant meteorologists are, after the fact,” he joked. “I’ve been working very hard the last three or four months to explain our bust. So we’ve eaten our crow, especially me.”

Making matters trickier is the inability to predict the current, which scientists describe as the ocean’s conveyor belt, said Klotzbach, also a Colorado State University climate researcher.

“You can’t really see it until it’s happening. That’s the problem,” he said.

Forecasters basically use two models to predict storms. Gray and Klotzbach rely on historical data and statistics. NOAA and Florida State create climate models, using existing sea temperatures and crunching more than 200,000 lines of computer code in the week leading up to the season to factor in atmospheric conditions such as wind speed, rainfall, sunlight and cloud cover across the globe, said Tim LaRow, an associate research scientist at FSU’s Center for Ocean-Atmospheric Prediction Studies.

“It incorporates all the physical understanding that we have of the atmosphere,” he said. “And it gives us the equations of motion to calculate the flow of the air.”

In the six years LaRow has been running the complex climate model — which NOAA’s Miami chief of hurricane specialists, James Franklin, described as one of the most reliable around — it has proved about 70 percent accurate. But 2013 also foiled the model, even when he later ran it using actual conditions.

This year’s less-intense forecast stems largely from an expected El Niño — a recurrent global climate pattern — that has already started warming Pacific waters, Klotzbach said. An El Niño pattern generally produces stronger wind shear and fewer storms in the Atlantic.