"To the often-heard question, 'Why can't we make better weather forecasts?' I have been tempted to reply, 'Well, why should we be able to make any forecasts at all?' " - Edward N. Lorenz in The Essence of Chaos.

Lorenz is the MIT atmospheric science researcher whose work led to the development of the idea of chaos in physical systems. In meteorology, one of the important implications is that small differences in the initial conditions of the atmosphere can lead to big differences in the weather that results only a few days later. These differences can be too small to detect. This is why scientists say that day-to-day forecasts of the weather for more than about two weeks ahead will never be possible. Today, five days is about the limit of useful day-to-day forecasts. And, forecasts for five days ahead are, at best, little better than using "normal" climatic values.

Forecasters rely on computer models

With the explosion of computer technology scientists have developed many computer models to help forecasters analyze and predict the weather.

Computer models depend on the fact that mathematical equations can describe the physical changes that govern the weather, just as equations describe movements of the solar system well enough for solar and lunar eclipses to be predicted years, even centuries, in advance.

Equations describing the solar system are complicated. Still, mathematicians and astronomers have been able to predict eclipses for centuries. The atmosphere's equations are much more complex. Solving them had to wait not only for more knowledge, but also for computers.

And, even with the fastest computers, meteorologists can't forecast day-to-day weather for more than about a week ahead of time. The atmosphere is just too complex, among other reasons.

When researchers began developing the first computer models for the earth's atmosphere in the 1950s, they worked with computers that were extremely limited compared with today's. As a result, the first models were overly simplified, but still provided valuable insight into the atmosphere's future state. As computer technology advanced, the complexity of the forecast models increased and more of the dynamical and physical factors influencing the atmosphere were taken into account. This improved forecasts.

These models were never intended to replace human forecasters. Instead, the models were developed as aids. Human forecasters study the output from models over a long time period and compare the forecast output from the models to the actual verification for the forecasted time period. This is how model biases, model strengths, and model weaknesses are determined. Often, human forecasters will modify model output based on past experience in forecasting the weather and physical and dynamical reasons.

Many new models are currently being developed. Some of the models being developed will help forecast tropical features, such as hurricanes. Other models of the future will help forecast smaller scale features, such as thunderstorms and severe weather outbreaks. Once these models are developed, forecasters will be able to issue better and more timely warnings and advisories.

Long range forecasts

Meteorologists agree that forecasting day-to-day weather changes for more than five or six days ahead isn't possible now. In fact, they have good theoretical reasons for thinking that no matter how much more they learn about the atmosphere and no matter what improvements are made in weather observations, computer modes and computers, they'll never be able to make day-to-day forecasts for more than about three weeks ahead of time.

Still, people want to know what's going to happen farther in the future than the next few days. Atmospheric scientists have developed various ways to give generalized outlooks of what to expect farther ahead. But, these outlooks are different in important ways from ordinary daily forecasts.

The key differences are that the long-range outlooks don't attempt to forecast day-to-day changes. Instead, they attempt to say whether the average temperature and precipitation will be above normal, below normal or near normal. And, these averages are for the entire period, not parts of it.

Copyright © Malta Weather Services 2003