NOAA Magazine || NOAA Home Page

January 7, 2002 — During the past 50 years, the science and technology of meteorology has undergone revolutionary changes. In the 1940's, weather analysis and forecasting relied almost exclusively on surface observations, collected at six-hour intervals, to make a forecast out to 24-to-36 hours. These forecasts were very subjective and based almost entirely on the long-term experiences of individual forecasters. The forecasts beyond 12-to-18 hours were often inaccurate. For extreme weather events, forecasts were noted more for misses than successes, even one day in advance.

In the late 1940's and early 1950s, Jule Charney from Massachusetts Institute of Technology and Jon Von Neumann from the Institute of Advanced Studies at Princeton led a team of scientists from around the world to describe the global atmospheric circulation patterns with a mathematical system of equations that could also be used for weather forecasting. They transformed a simplified version of these equations into mathematical expressions (called a "numerical model"), which could be integrated by the first computer to make the first numerical weather prediction (NWP). These simple NWP models calculated changes in the Earth's atmospheric circulation at specific points positioned around the globe. The models were started or "initialized" with an array of vertical wind, temperature and moisture profiles as measured by balloon borne weather sensors, which could attain altitudes of 35,000 ft. These early models were successful (at times) producing representations of the Earth's atmosphere in large scale circulation patterns. However, the models were unsuccessful in predicting storms and related weather features any better than the subjective forecasters of the 1940s.

Nevertheless, over the next 50 years more sophisticated models run on powerful computers were developed and applied to weather forecasts. By the mid-1990s, numerical models were capable of simulating a wide range of physical processes, including the effect of solar heating on different types of land surfaces, ocean-atmosphere interactions, the effects of latent heating as precipitation forms in clouds, and radiative processes including reflection and absorption in the clouds themselves. The newer models have been designed to work off a global data set that include: commercial aircraft observations collected every five minutes, weather balloon data, ship data, standard hourly surface observations, along with a multitude of satellite, radar, and wind profiling data. These models are run on one of the most powerful computers in the world, the IBM SP, with global model forecasts run out to 16 days in advance everyday.

Over the past 10 to 15 years, newer models have produced spectacular results. The "Perfect Storm" in 1991, as depicted by the movie of the same name, was predicted four days in advance. The "Superstorm" of March 1993, which dumped more than 30 inches of snow along the Appalachian mountain chain, was predicted five days in advance. The January blizzard of 1996 was predicted two to three days in advance, allowing forecasters to issue warnings for several feet of snow 24 hours before the first flake of snow fell in the urban corridors from Virginia to Boston.

But, as meteorologists were reminded on March 6-7, 2001, forecasts are not perfect. Numerical models accurately predicted that a Nor'easter would develop along the East Coast, seven, six, five, four, and three days in advance with remarkable consistency. But critical forecast issues confronted the forecasters just one day before the storm. It turned out that this slow-moving storm system took 12-to-24 hours longer to develop over the Atlantic and produce the heaviest snow 120 miles further northeast than the models had earlier predicted. The most noticeable result was two-to-three feet of snow fell in Albany, N.Y., the Catskills of eastern New York, western Massachusetts, and the mountains of southern New Hampshire and Vermont instead of the densely metropolitan areas along the East Coast.

Given the complexity of the Earth's atmosphere and the finite ability to measure all the key elements that influence the initial conditions that drive weather prediction, there will always be storms that either are not predicted, predicted to take the wrong track, or predicted—but fail to develop. The goal with today's model development is to minimize these errors and produce a confidence measure for the predictability of any specific impending event.

NOAA weather forecasters have had their successes and challenging moments. Today's one-day forecast exceeds 90 percent accuracy, and five-day forecasts are as good as three-day forecasts were 15 years ago. These results highlight the revolutionary advances in the science of weather forecasting since the pioneering efforts of the 1950s. With the increasing reliance on weather forecasts and warnings for life-saving and economic decision-making, increases in forecast accuracy and lead time are demanded from individuals, corporations and government agencies worldwide.

So where do we go from here? NOAA forecasters are constantly working on a variety of methods to increase the accuracy of predicting the development and track of East Coast snow storms. This year, NOAA's National Weather Service will:

• Institute a short-term (one-to-two day) ensemble forecast system with the ability to quantify the uncertainty associated with a given forecast for a developing storm system. This information can then be used by the forecaster in stating the confidence of expected snowfall, rainfall, winds etc. Decision makers and the public in general can then make more informed decisions on how to react to a forecast.
• Evaluate the use of ensemble models from day seven to day 16 forecasts for determining forecasts; and having national forecasters collaborate with field forecasters who have the knowledge and experience to tailor forecasts to their local area, thus providing the best products and services to the public.
• Continue the implementation of higher resolution global and regional numerical weather prediction models with improved physics, which can better predict the development, movement and precipitation associated with both winter and summer storms.