On one of the busiest travel weekends of the year, the northeastern U.S. was socked with a powerful low-pressure system that brought blizzard and whiteout conditions to a huge region extending from northern New Jersey to Maine. Although a winter storm had been forecast, few were prepared for the intensity of the post-Christmas storm, which included a phenomenon known as thundersnow in some areas, shut down railways and left thousands stranded at airports across the region.

The storm originated in the Pacific Ocean, sending a low-pressure system into the Southwest that swept across the southern states before combining with a system "diving southward from out of the Midwest to produce snow in the South and the whopper snowstorm along the East Coast," according to The Weather Channel's Web site.

People have been fascinated with weather forecasting at least dating back to the ancient Babylonians, who tried to use cloud formations and other atmospheric conditions to predict short-term weather changes. Computers have made the job a lot easier in recent decades by allowing meteorologists to plug massive amounts of data culled from land, sea and air-based sensors into sophisticated algorithms that help produce forecasting models. Still, there will always be an aspect of unpredictability in the atmosphere that foils even the most high-tech attempts to predict the weather.

To find out more about how meteorologists use forecasting models to get ahead of Mother Nature as well as the limitations of this technology, Scientific American spoke with Thomas Else, chief meteorologist and director of meteorological services and computer programming for Hackettstown, N.J.-based weather forecasting consulting firm Weather Works LLC.

[An edited transcript of the interview follows.]

What tools do meteorologists have at their disposal when trying to forecast storms?
We use different computer forecast models that feed initial conditions—including temperatures, humidity, wind speed and wind direction from around the United States and around the world, from the surface all the way up to the jet stream—into different equations. The models are run on National Weather Service supercomputers. Some of the short-term forecasting models are run hourly while other models are run every six hours or twice a day. But the models are only as good as the data we can get into them. One of the big problems with computer models comes when you can't collect accurate data from a particular area. Sometimes a storm will pass through an area where there are fewer sensors, with possibly 50 to 100 miles between each reading. So it may be that the model doesn't have any data for a particular spot on the map and has to use interpolated data.

Where do the data used in these models come from?
We get observations from ground stations, satellites and radiosondes, which are balloons that National Weather Service forecast offices launch twice a day to measure temperature, pressure, humidity, wind direction and wind speed. Satellites are becoming very important because they can do remote sensing, taking and reporting atmospheric measurements. Other data come from commercial airplanes, ships at sea and ocean buoys. So we're getting data from the surface all the way up to 30,000 or 40,000 feet.

Which technological advances will have the greatest impact on improving weather forecasts?
The basic technology is already in place. The main advances in forecasting come from more powerful computers and faster software that can ingest more data coming from a greater number of sources.

What made the Christmas weekend blizzard in the Northeast difficult to accurately forecast?
All coastal storms like this are very tough because you have to do very short-term forecasting called nowcasting as they're developing. You do this because you're trying to predict a couple of days ahead of time something that hasn't formed yet. The best way to explain the behavior of a storm moving from the sea across land is to use the theory of chaos, where any little change in an air mass (its direction, speed or temperature, for example) can throw the whole atmosphere out of balance. The weather that eventually hit the Northeast began developing in the Pacific Ocean, where there isn't a lot of sensor data. It was only as the disturbance moved inland the computer models started picking up on it. Depending upon when they ran their computer models, some meteorologists were predicting a big coastal storm while others were taking the weather disturbance back out to sea. This was fluctuating back and forth until about Friday afternoon.

After such a snowy start, what can people in the northeastern U.S. expect from this winter season?
Short-term, the current cold snap is becoming more moderate as we head into New Year's weekend and into early next week. But this will quickly be followed later next week by another cold blast of air. Longer range, we're going to see more wintry mix events (rain, freezing rain and sleet) this winter instead of these big snowstorms. The ongoing La Niña pattern, where there are colder than normal sea surface temperatures in the central and eastern equatorial Pacific Ocean, favors these types of conditions.