Weather Forecasting

Weather systems are chaotic, which limits the degree of forecasting accuracy. The study of chaos theory and chaotic systems has allowed the development of innovative techniques that give insight into weather predictability.

Development of Weather Forecasting

From the earliest days of recorded civilization, people have actively attempted to forecast the weather, and for good reason: The weather directly affects everyone. Predicting the weather has been a priority for reasons of simple comfort to protection of property and lives and the success of agriculture. Among uneducated and superstitious people, weather forecasting has been widely associated with lunar phases, rainbows, halos, and the behavior and appearance of animals and insects. With the advent of weather instruments such as thermometers, anemometers, hygrometers, and barometers, the science of meteorology evolved. The first recorded use of the term “weather forecast” was by Englishman Robert Fitzroy in 1860, a man with enough perception to avoid any nonscientific term such as “prophecy” or even one so bold as “prediction.”

Record keeping and sharing of information across states, nations, and the world by telegraph and radio during and after World War I finally brought the science to a coherent body of knowledge. By the start of World War II, the invention of the high-altitude balloon for weather soundings (the radiosonde) enabled the tracking of upper air currents. The term “atmospheric science” began to encompass the grand, global scheme of weather forecasting. “Meteorology” technically accounts for the action in the lower atmosphere.

The earliest meaningful attempts at forecasting the weather scientifically came with the advent of the barometer, a device that measures atmospheric pressure, and the realization that changes in the weather seemed to follow changes in the barometric pressure. It was also noted that weather changed with the passage of air masses, which often dramatically altered it from what it had been. The passage of air masses was also often related to the barometric pressure. The accuracy of weather prediction increased with such observations.

Forecasting Methods

Predicting the weather from present conditions to two days forward is called a short-range forecast. A medium-range forecast can be stretched out to only seven days forward. Beyond seven days, it is called a long-range forecast, of which thirty days forward are all that can be predicted with any statistical accuracy. The National Weather Service gives a ninety-day outlook. Prediction method accuracy varies widely with the condition being predicted. Accurately forecasting damaging weather such as local high winds, heavy snow, dangerous thunderstorms, or tornadoes beyond twelve hours is generally impossible. Very short-range forecasting for periods of one hour or less is a relatively new technique primarily used for issuing warnings. Such immediate forecasts are called nowcasts.

Satellites offer a wide view of the planet as they orbit. The advantage of these instruments is a global view of weather patterns and their movement, which offers a wide range of immediate forecasting options when related to ground weather observations. Radar offers a view of precipitation in most forms and is useful in forecasting imminent weather patterns. Doppler radar provides a means of pinpointing tornadoes as they form inside thundercloud formations.

The global weather scheme is plotted by powerful computers, which collect data from all over the globe and integrate them into a planetary weather picture. From this picture, forecasters adopt mathematical models to calculate what will happen to the system over time. Supercomputers predict the weather using a global scheme that simulates weather worldwide based on a vast grid of data points covering the globe at various resolutions. These forecasts are calculated at billions of operations per second. Yet forecasters have found that no single model or combination of models, using even the most powerful computers, accurately predicts weather schemes beyond one week. The reason for that is an intrinsic condition called chaos.

Chaos accounts for the breakdown of forecasting accuracy over time. It was first related to weather patterns by Massachusetts Institute of Technology meteorologist Edward Lorenz in 1961. By 1963, Lorenz had constructed a computerized model of the atmosphere. Lorenz discovered that regardless of the point at which the computer begins calculating the equations for predicting the weather, every data point is potentially unstable, so very small errors are magnified over time. That is called sensitive dependence on initial conditions or the “butterfly effect.” The butterfly effect refers to flapping a butterfly’s wings in, for example, Beijing, China, which initiates a tiny instability in the atmosphere, ultimately magnified through chaotic repercussions into storm systems in New York. This rather fanciful analog of the effect of chaos in weather systems is the real problem faced in weather forecasting. The future of weather forecasting probably lies in understanding and modeling chaotic systems. Such work is underway at the National Weather Service.

Applications of Forecasting Methods

The three ranges of weather forecasting—nowcasting (one hour or less), short-range forecasting (less than two days), and long-range forecasting (up to ninety days ahead)—all use different techniques for the development of the forecast. Nowcasts are usually based on ongoing observations from radar for predictions for very specific locations. For example, a tornado warning, hurricane landfall storm surge warning, or severe thunderstorm warning may be issued for a specific city based on actual images of the storm bearing down. The only mathematical methods applied in these cases would be tracking equations based on speed and distance.

Short-range forecasting is much more complex. Mathematical modeling equations are developed from four different types of mathematical techniques. The National Weather Service uses combinations of all these equations in its computer analyses of the weather’s short-term outlook. These mathematical techniques are computations of physical parameters, computations of displacements of large weather systems, regression analysis, and statistical (time series) analyses. Computations of physical parameters are pure physical formulas: hydrodynamic, aerodynamic, thermodynamic, and classical physical laws. Examples are Sir Isaac Newton’s equations for motion, rules for the conservation of mass, heat transfer equations, and the laws for gas states, all applied to dynamic (moving) weather systems. In computations of displacements of large weather systems, the rates of movement of well-defined frontal systems are calculated based on a series of equations developed for just these systems and their movements. Using a statistical analogy between past events and present conditions, a forecaster can relate past weather patterns to future conditions. Statistical (time series) regression analyses called model output statistics (MOS) enable the forecaster to make predictions of future weather events. These equations account for the possible variations in weather patterns over time and the physical range of the weather system itself. In such forecasts, the weather is predicted to occur on a chance basis. For example, the forecast may give an area a 50 percent chance of measurable precipitation. All these parameters may be utilized to develop a single forecast.

Extended forecasting is developed using methods different from those employed in short-range forecasts. Meteorologists still use a battery of computerized tools. Still, the effect of chaos (instability at every point in the atmosphere) is so significant beyond two days that extending the computer tools used for short-range forecasting out many days is all but useless for specific areas. Yet seasonal weather patterns, worldwide circulations, oceanic currents, and historical norms account for making long-range forecasts possible. These nearly cyclic conditions are called aperiodic and enable some degree of forecasting. In the study of chaos itself, there are predictable, recognizable elements: all the nuances of aperiodicity that ultimately lead to the chaotic state. It is precisely the investigation of these elements that atmospheric scientists are now pursuing.

One of the most elementary, subjective elements of such an approach surfaced in the early 1950s, a decade before chaos was quantitatively linked to the weather. In such an approach, called weather analogs or weather typing, weather charts of previous seasons were categorized and cataloged. Current charts were then compared with old charts to indicate possible future trends. In the days before a complete understanding of the power of even minute atmospheric instabilities, these analyses proved useless. They were all but abandoned in favor of powerful computerized analyses.

In using computers to make long-range forecasts, meteorologists have noted that there are times when their automated models prove correct, often even surprisingly accurate. Chaos, again, proves to be the central determinant of accuracy. The stability of the weather pattern determines the accuracy of the computerized forecast. To weed the good forecasts from the bad, atmospheric scientists have developed a method called the ensemble approach. In this technique, they predict the forecast's reliability by testing the conditions' stability: They run the same forecast analysis at least ten times using a supercomputer, altering each run very slightly to mirror tiny instabilities in the atmosphere and then compare results. If the end result is widely variable forecasts, then the forecasting reliability for that day is low. Using such techniques, atmospheric scientists can not only predict the ultimate accuracy of their forecasts but also decide on the degree of their reliability.

Significance

Weather is a phenomenon of universal relevance, from the detrimental effect of storms to the periodic need for rain. Predicting weather has developed into a science involving individuals and governments and the pooling of an immense amount of information. Some forecasting abilities have resulted from ordinary and practical applications. At the same time, other methods are so complex that they are understood only by a very few highly skilled meteorologists in the scientific community.

Yet without regard for human progress and understanding, the deserts of equatorial Africa have been encroaching northward at tens of kilometers per year, rendering once fertile land into useless desert; hundreds of people have died daily from famine unfolding across the vast African savannah. The frequency of hurricanes is changing along the Atlantic seaboard, threatening millions of seashore inhabitants. The sunbelt population in the United States is rapidly growing along the paths of hundreds of potentially deadly tornadoes. The world’s atmospheric scientists are faced with the task of predicting the potential for the weather to directly affect the lives of millions of people. From giving imminent warnings to plotting the centimeter-by-centimeter alteration of long-term weather patterns, the significance of prediction has never been greater.

The discovery of the effect of chaos has dramatically facilitated the task at hand. Many scientists have since declared the task of accurate weather forecasting to be impossible. Others continue struggling to discern the science's meaning with instabilities at every point. The study of chaos itself has spread from meteorology to nearly every other discipline. In the twenty-first century, meteorologists continue to tackle the issues related to weather forecasting, especially regarding the integration of chaos theory. The development of Artificial Intelligence (AI) and AI models that forecast global weather patterns may answer the inherent unpredictability of weather. Technology giant Google has developed an AI model called Scalable Ensemble Envelope Diffusion Sampler (SEEDS), which can more accurately predict extreme weather events. Similarly, DeepMind's GraphCast system is an AI weather prediction model that can generate predictive forecasts far faster than traditional models. Although this science is still evolving, AI holds the potential to create more accurate weather forecasts more efficiently. 

Principal Terms

aperiodic: describing any phenomenon that occurs at random rather than at regularly occurring intervals, such as found in most weather cycles, rendering them virtually unpredictable

chaos: an emerging scientific discipline that attempts to quantify and describe the properties of seemingly random, aperiodic events

nowcasting: a weather forecast made and disseminated within an hour or less for a specific area for an approaching weather system

weather analog: an approach that uses the weather behavior of the past to predict what a current weather pattern will do in the future

Bibliography

Ackerman, Steven, and John Knox. Meteorology: Understanding the Atmosphere. 3d ed., Sudbury, Mass.: Jones and Bartlett, 2012.

Bader, M. J., et al., editors. Images in Weather Forecasting: A Practical Guide for Interpreting Satellite and Radar Imagery. New York: Cambridge University Press, 1997.

Barrett, E. C. Viewing Weather from Space. New York: Frederick A. Praeger, 1967.

“Chaos Theory Provides Hints for Controlling the Weather.” ScienceDaily, 28 Mar. 2022, www.sciencedaily.com/releases/2022/03/220328101303.htm. Accessed 21 July 2024.

Collier, Christopher G. Applications of Weather Radar Systems: A Guide to Uses of Radar Data in Meteorology and Hydrology. 2d ed., New York: Wiley, 1996.

Dolan, Ed. The Old Farmer’s Almanac Book of Weather Lore. Dublin, N.H.: Yankee Books, 1988.

Dunlop, Storm. The Weather Identification Handbook. Guilford, Conn.: The Lyons Press, 2003.

“Generative AI to Quantify Uncertainty in Weather Forecasting.” Google Research, 29 Mar. 2024, research.google/blog/generative-ai-to-quantify-uncertainty-in-weather-forecasting. Accessed 21 July 2024.

Gleick, James. Chaos: Making a New Science. 2d ed., New York: Penguin Books, 2008.

“Google Releases GenAI Model for Efficient Weather Forecasting.” CDO Magazine, 4 Apr. 2024, www.cdomagazine.tech/aiml/google-releases-genai-model-for-efficient-weather-forecasting. Accessed 21 July 2024.

Hamilton, Kevin, and Wataru Ohfuchi, editors. High Resolution Numerical Modeling of the Atmosphere and Ocean. New York: Springer Science+Business Media, 2008.

Hodgson, Michael. Basic Essentials: Weather Forecasting. 3d ed., Chester, Conn.: Globe Pequot, 2007.

“How Technology Pushes Us into a New Era of Weather Prediction.” Xweather, 2 May 2024, www.xweather.com/blog/article/how-technology-continues-to-change-the-weather. Accessed 21 July 2024.

Monmonier, Mark S. Air Apparent: How Meteorologists Learned to Map, Predict, and Dramatize Weather. Chicago: University of Chicago Press, 1999.

Vasquez, Tim. Weather Analysis and Forecasting Handbook. Garland, Tex.: Weather Graphics Technologies, 2011.

Wallace, John M., and Peter V. Hobbs. Atmospheric Science. 2d ed., Burlington, Mass.: Academic Press, 2006.