Back in 1818, when the US was young and James Monroe was president, a New England farmer named David Young published a tome that revolutionised agribusiness forever. The Farmers' Almanac was a compendium of facts and figures intended to provide readers with moral guidance and the wherewithal to make "sound and healthy" life decisions. Young organised this information into three major categories: astronomy, gardening and, most important, weather. Farmers all across the fruited plains swore by the Almanac. Through a secret calculation involving sunspots, moon phases and other astronomical phenomena, the Almanac purported to predict the weather years in advance. These forecasts didn't call for specific weather events on specific days; rather the Almanac predicted broad changes in temperature and precipitation for a particular region at a particular time. Farmers who bought the book based business decisions on what they read. If the Almanac called for an atypically dry and chilly summer in New England, farmers would plant less corn. If it predicted wetter-than-usual conditions along the Mississippi, farmers would plan on a smaller grain harvest. Historians at the Hart Wright , the Lewiston, Maine, outfit that publishes the Almanac today, report that 19th century readers claimed the book's forecasts were accurate 85 per cent of the time, and not coincidentally the book sold more than 100,000 copies each year, a figure topped only by the Bible. When the Almanac prognosticated, its readers listened. And business listens today. But not, for the most part, to the Almanac. Long-range forecasts, made possible by advances in supercomputing and other emerging technologies are driving corporate decision making. Companies in almost every industry now inform their long-term strategies with weather data from the US government. What's more, an increasing number of businesses are paying private forecasting companies to add value to this data by customizing and particularizing it in order to predict how weather patterns both broad and narrow can and will affect the company's overall performance. Experts say that business leaders who have paid attention to the wind and the rain, El Nio and the Gulf Stream, have saved millions in overhead. Weather forecasting has a way to go before it becomes an exact science, but the companies that utilise long-range predictions certainly have a leg up on those that don't, says Ants Leetma, director of the government-funded Climate Prediction Centre. "These forecasts," he says, "let you manage risk in ways you never even dreamed about. For a business, that's a huge step in the right direction."
If anybody understands how long-range weather predictions can affect business performance, it's Leetma. His Climate Prediction Center, a joint venture run by the National Weather Service (NWS) and the National Oceanic and Atmospheric Administration (NOAA), is charged with the task of watching climate fluctuations, diagnosing them and predicting their effects. Together with the National Centers for Environmental Prediction (NCEP), the organisation is responsible for accumulating and disseminating the nation's official weather information every day. Since 1993 the same two organisations have issued free 30-day, 90-day and 13-month forecasts as well. These are the forecasts with which modern businesses are most concerned.
To grasp how the NWS makes these long-range predictions, Leetma insists one must first understand the technology behind short-range forecasts. Every second, government-owned satellites, buoys and weather balloons record data from the earth's atmosphere and send it electronically to NCEP's headquarters in Camp Springs, Md. There, technicians collect the information, format it and ship it over to an IBM RISC 6000 supercomputer. At speeds exceeding 30 gigaflops (30 billion floating point operations per second), this machine runs the data through a number of models that correlate certain weather elements with time, location and atmospheric conditions. The computer then plots these relationships on two-dimensional maps and forecasts weather systems for the immediate future.
According to Wayman Baker, NCEP's deputy director of central operations this method of prediction is known as synoptic forecasting. It enables scientists to forecast weather with a fair degree of accuracy for as many as 17 days into the future. Beyond that point, however, Baker says that "the margin of error becomes too great to forecast with any degree of skill." (See "The Butterfly Effect.") "You can forecast with accuracy only so far," Baker says. "Outside that two- to three-week period, even the most sophisticated forecast becomes nothing more than an educated guess."
Some guesses, however, are more educated than others. Government meteorologists base long-range predictions on a climatological rather than a synoptic approach. Here, Baker says, instead of focusing on the details of day-to-day weather patterns, meteorologists take that data, analyse it and focus on the broad weather picture over a period of time. The resulting forecasts attempt to predict the departures from normal conditions for a given month or season. These departures are called anomalies and are represented in terms of probability.
For instance, a long-range forecast might state that spring temperatures in Minneapolis have a 65 per cent probability of being above normal. Leetma explains that this kind of forecast would likely be based on a map that shows temperature anomaly patterns. These maps would not attempt to predict the weather for a particular day but, like the Almanac, the trend for an extended amount of time. Leetma says this information can be critical for businesses. His conservative estimate is that roughly 20 per cent of the country's $7 trillion economy is vulnerable to the weather shifts and events.
"Talk about government helping business," Leetma says. "We get in there and provide businesspeople with information that enables them to prepare for their futures. If you asked me, I'd say that's a pretty incredible service we provide."
Actually, You Do Need a Weatherman to Tell Which Way the Wind Blows Still, to the untrained eye, these government-sponsored forecasts can be a bit confusing. What do the predictions mean by "normal"? And how much warmer is "warmer"? One degree or 10? Instead of trying to answer these questions alone, an increasing number of companies have started looking for help. In response, an entire industry has sprouted up.
Some of the firms in this nascent industry interpret only long-range forecasts from the government, consulting on information already in the public domain. Others provide long-range forecasts of their own; predictions based on government data, artificial intelligence and proprietary technology. Todd Glickman, who was the assistant executive director of the American Meteorological Society for five years and is now an industrial liaison officer for MIT, says that while the government issues long-range forecasts on a macro level, many of these forecasting companies predict the weather on a number of different micro levels.
"Data from the government is just that: data," says Glickman. "That's why it's free. But forecasts from companies like these are tools companies can use to get ahead."
At Atmospheric and Environmental Research (AER), meteorologists specialize in providing utilities with long-range forecasts they can use to plan for extreme weather conditions. With advance information about temperatures and precipitation in the winter and summer months, clients of this 23-year-old company can reorganise resources to gear up for increased demands for power or heat. Companies in this industry can also mitigate risk via the burgeoning weather derivatives market (see "Pennies from Heaven,"), betting against how much energy they will use during the busiest times of year.
Most of AER's forecasts are based on a computerized formula developed by staff scientist Judah Cohen. The statistical model, written in Fortran, applies NWS data to a series of mathematical equations. While Cohen declines to reveal the nature of these equations, he admits they are based loosely on general circulation models (GCMs), some of the earliest statistical analyses meteorologists used to predict the weather. Right now, Cohen says the information gleaned from this formula enables him to issue predictions as many as six months in advance, forecasting ice and snow in June. Eventually, Cohen expects to be able to issue nine- and 12-month forecasts as well.
Down the road from AER,, Weather Services (WSC) specialises in providing forecasts to companies in other industries. The organisation sells long-range forecasts to businesses in the commodities and retail industries, as well as shorter-range forecasts to companies in the entertainment and sporting industries. Rob Carolan, director of custom product development, says that one of the company's biggest clients, the Boston Red Sox, purchases daily and weekly short-range forecasts to determine when the weather might force the team to postpone a game. Carolan adds that the America's Cup, another sporting organisation, purchases WSC medium- and short-range forecasts to plan its yacht race.
Farther west, in Wichita, Kansas, a company named WeatherData offers clients in the manufacturing industry short- and medium-range warnings about storms that could cause disruptions in production. By notifying clients such as Boeing and Burlington Northern/Santa Fe when and where a storm will hit, WeatherData saves these organisations big bucks in operation and recovery costs. Last year alone, officials at Daimler Chrysler estimated that early warnings from WeatherData saved the company more than $10 million.
"It costs us more than $100,000 a minute to shut down our facilities for a tornado," says Mike Corcoran, fire chief at the automobile giant's factory in Auburn Hills, Mich. "If the government issues a warning, [but it isn't necessary for us] to shut down, WeatherData tells us. If we do need to shut down, they'll tell us exactly when, so we don't lose any time we could have spent making cars."
WeatherData CEO Mike Smith won't talk about his organisation's proprietary formula for weather prediction, but he does say that the model incorporates data from the NWS, the National Lightning Detection Network and weather observation bureaus in more than 30 of the 50 states. That explanation appeared to be enough for the American Meteorological Society. In January 2000 the organisation presented WeatherData with its Outstanding Service to Meteorology by a Corporation award.
Getting the Weather Word from Mr. Frost
While all of these forecasting firms are blazing new trails in the area of business technology, experts say that none of them employs applications as sophisticated as those at Strategic Weather Services (SWS). Using a formula developed by meteorological legend Irving P. Krick (see "The Father of Long-Range Forecasting"), SWS has dabbled in long-range forecasting since the mid-1980s. Earlier this year, it launched Planalytics, a wholly owned subsidiary devoted to developing a technology that enables retail businesses to determine how their customers are influenced by weather. Dubbed "Weathernomics," the new technology uses artificial intelligence modeling of historical sales and weather data to chart consumer behavior over time. In other words, instead of telling clients it will be colder or warmer, Planalytics tells them whether it will be colder or warmer to a degree that drives sales.
Along with Chief Operating Officer Steve Beck, Senior Vice President David Frost masterminds the Planalytics operation from SWS headquarters in Wayne, Pa. On a chilly November afternoon (colder than normal, for sure) last year, the men explained how they help clients get a sense of what drives customer purchases. The secret, says Frost, is to correlate the effects of weather against retail sales by item and location. By presenting clients with this information, companies can better respond to the wants and needs of customers in different regions of the country.
"Typically businesses will blame the weather when sales are bad and credit themselves when things are good," Frost says in a practiced manner. "That's not always the case, though. Sales and customer behaviors are tied a lot more to weather than people realise. We want to help them see how."
So far, Planalytics customers say the company has done just that. At Duraflame, the California-based retailer of fabricated hearth logs, company executives estimate they've saved hundreds of thousands of dollars in distribution and buyback costs since they signed aboard in 1996. Before Duraflame purchased long-range forecasts from Planalytics, it prepared for the future by shipping thousands of extra logs to distribution centers across the country. If winter in a particular region ended up being colder than normal, these distribution centers had enough logs to meet demand. If the winter ended up being warmer, the centers had overstock. And that was expensive. Duraflame had already paid to ship the extra logs to the centers. Now it had to pay to ship the logs back to headquarters and pay to warehouse the logs on its shelves.
When Planalytics approached Duraflame in 1996, company officials were skeptical but figured they didn't have much to lose except their warehouse and distribution costs. (And, of course, Planalytics' fee.) Chris Caron, Duraflame vice president of marketing, says he purchased forecasts for the 1996-1997 season and sat back to see if they were correct. Of the 12 one-month forecasts the company bought, nine were close or right on target. Duraflame officials were amazed. The following season, the officials put the forecasts to work, basing the company's strategic plan on Planalytics predictions for 1997-1998. This time, the company sent extra logs only to areas where Planalytics had predicted colder than normal temperatures. Sure enough, Caron says, the move paid off. Caron estimates that by the end of the 1998 season, Duraflame had saved more than $100,000.
"The forecasts didn't help us increase sales, but they certainly helped us control and reduce our costs," Caron says. "We've been able to manage our inventory so much more efficiently, I have managers calling me and telling me they can't believe how well we've planned."
Planalytics forecasted a colder and wetter than normal winter in the Southwest this year. So Duraflame plans to keep logs on the shelves well into April in K-Marts and Wal-Marts in New Mexico, Arizona and northwestern Texas. Conversely, in the Northeast, where Planalytics predicted the winter won't be so bad, Duraflame won't sell as many logs as it did last year to supermarket clients such as Stop 'n' Shop and Victory.
Duraflame is one of SWS's star clients. Either Beck or Frost speaks to Caron every week. Frost admits that few clients use his company's data as well as Duraflame does. Beck, however, isn't so modest.
"Duraflame is a testament to how well we do what we do," he boasts. "Someday, with Weathernomics, or something else, every business will incorporate weather into its strategic plans."
The Long-Range Forecast for Long-Range Forecasting Someday Beck may be right. But many scientists are still skeptical about the reliability of long-range weather forecasting, and experts say the industry will go nowhere until it shakes some of the controversy that surrounds it. Some of these critics call long-range weather prediction quackery and say it amounts to nothing more than highly efficient data processing. Other critics allege that businesses could save thousands of dollars by hiring full-time meteorologists to analyse NWS data instead of outsourcing the function to vendors such as SWS.
Bob Livesy is one of those naysayers. Livesy, a senior scientist and research meteorologist at NOAA's Climate Prediction Center, has followed the art of long-range weather prediction since the mid-1960s. Livesy has even done some long-range forecasting of his own over the years and says that he sees no way to predict further than 30 or 90 days into the future.
"These [forecasting firms] are using science, but it's incredibly simplistic science," he says. "They're basically making money off sure bets. Most of the areas of the country are undergoing warming trends. So you hem and you haw and you play little games with the forecasts here and there. Before you know it, sheer luck makes you right 50 per cent of the time."
Livesy says many of the firms that specialise in long-range forecasts walk a fine line between actually forecasting the future and making educated guesses. Roger Pielke Jr., a social scientist at the National Center for Atmospheric Research in Boulder, Colorado, agrees. Pielke says that database technology has become so advanced that anyone could search 30 or 40 years of weather information for patterns and make predictions based on them. He calls this reference forecasting, and reference forecasting, he says, is not science.
"The key to the success of this industry will be a customer's ability to recognize a good prediction when he or she has one," Pielke says, conceding that some companies do show skill in making long-range predictions.
Other scientists are less negative. Alan Eustis, director of Digital Earth and Space Applications for the government's National Environmental Satellite Data and Information Services, says that as long as forecasting firms use accepted scientific methods to arrive at their predictions for the future, the forecasts are welcome additions to the "vague and cumbersome" data produced by the NWS. Phillip Arkin, deputy director of the International Research Institute for Climate Prediction, offers yet another opinion. While Arkin doesn't disparage the science behind these long-range forecasts, he says he simply can't understand why companies would pay good money for information they could get for free. Arkin suggests companies should eschew for-profit entities and redirect their money toward full-time meteorologists who can analyse weather data in-house.
"If there's a client who's foolish enough to pay someone for something they can get for nothing, I guess that's what capitalism is all about," he says. "For somebody like me, though, somebody who knows all the details, I look at companies selling predictions and chuckle."
Because many of these forecasting firms are selling something people can get for free, some experts say the companies will have to diversify in order to survive. For consumers, this might mean being able to purchase predictions that will be tailored to their needs, such as humidity levels and pollen counts.
SWS recently launched Weatherplanner.com, a subsidiary that offers medium-range forecasts for a particular day in the future. For $14.95, an individual can purchase predictions for days up to six months away in order to plan, say, that company picnic or family reunion. A newer company, the Edmond, Okla.-based WeatherBank, offers the same service at www.weatherbank.com for only $4.95 a month. Other companies, such as AccuWeather and WeatherData, are contemplating similar services as well.
As for the Farmers' Almanac, it too is looking to the future. The book launched a Web site early last year (www.farmersalmanac.com) and now boasts free long-range forecasts right there, online. Think the Almanac has gone the way of the NWS? Think again. A visit to the site reveals only two months' worth of long-range predictions. For a complete year of forecasts, the site directs users to its store, where the 2000 edition retails for $3.19. Still cheap by any reckoning.
The Butterfly Effect
Chaos theory explains why long-range weather prediction is so hard Why is it difficult to make long-range forecasts? Because the global weather system is dauntingly complex. It is, in fact, chaotic. And what that means was defined by Massachusetts Institute of Technology meteorologist Edward Lorenz almost 30 years ago.
Over a series of weeks in 1961, Lorenz gathered and analysed years of information on weather patterns and the predictions that proceeded them. By running a simple computer-generated model of the earth's weather, he discovered that tiny predictive errors that seemed inconsequential from one day to the next became more and more significant as time went on. That is, a small change anywhere in the system produced an enormous effect down the line. This ran counter to Newtonian physics that assumed that every action should have an equal and opposite reaction. There was nothing equal about the reactions Lorenz observed. Nor did Lorenz's data fit classical mathematical theory that predicted that a small change on one side of an equation ought to produce a similarly small change on the other. The anomalies Lorenz identified in the way the global weather system operated formed the basis of what we now call chaos theory.
Lorenz presented his ideas in 1972 during a speech in which he alleged that the "flap of a butterfly's wings in Brazil [a small change] can set off a tornado in Texas [a big effect]." And if that's the case, the idea that anyone or anything could predict the weather in a small area of the globe two or three weeks in advance becomes a practical impossibility. No amount of data, no supercomputer no matter how powerful can measure the energy generated by that butterfly's wings, let alone quantify its influence around the globe.
Lorenz continued his studies throughout the '70s and '80s. In 1993 he published The Essence of Chaos, which today is viewed as the bible of chaos theory.
The Father of Long-Range Forecasting: Irving P KrickCompanies that say they can predict the weather one, two and three years in advance owe their technology to Irving P. Krick. Krick, who died at the age of 89 in 1996, launched the science of long-range forecasting more than 60 years ago.
"If it weren't for Krick, nobody could have made any strides in the area of weather prediction," says Alan Eustis, director of Digital Earth and Space Applications for the National Environmental Satellite Data and Information Services. "He showed all of us the way, and proved that you didn't necessarily have to be part of the government to get things done."
When Krick burst onto the national scene in 1933, he was chairman of the meteorology department at the California Institute of Technology. There, through his work with synoptic weather patterns, Krick developed a method for classifying sequences of weather systems based on his views on the "dynamics of centers of action" in circulation patterns over the northern hemisphere. Along with his colleague Robert D. Elliott, Krick extended this method into a series of six-day weather types and an analog system for long-range forecasting.
Krick was a strong believer that the atmosphere was orderly and repetitive by nature, and much of his effort was devoted to finding the predictive clues required to prove this theory. To this end, he was an early advocate of the budding science of computer technology and incorporated one of the first computers in his work for the American military during World War II. Krick served as chief of the Weather Information Section at General Dwight D. Eisenhower's Supreme Allied Headquarters; his forecasts were used to design the D-day invasion of Europe at Normandy and to plan the Allied atom bombing runs on Nagasaki and Hiroshima.
After the war, in 1948, Krick founded Irving P. Krick & Associates, a private company dedicated to promoting, advancing and marketing his work in the forecasting field. Krick sold forecasts to companies in almost every industry, including retail, agriculture and utilities. As a consultant to the Motion Picture Producers Association, his forecasting services became an integral part of the daily decision-making process of when to film and where. In 1954 Krick published a detailed account of his approach to forecasting in Sun, Sea and Sky: Weather in Our World and in Our Lives. The book, now out-of-print, sold nearly 600,000 copies in six years.
Krick continued to predict the weather into this decade. Finally, toward the end of 1990, Krick sold his company and its proprietary long-range weather forecasting technology to Strategic Weather Services , a Wayne, Pa., company that today boasts more than 2,000 customers. Krick remained active in the company until his death.
Join the CIO Australia group on LinkedIn. The group is open to CIOs, IT Directors, COOs, CTOs and senior IT managers.