People — perhaps most people — love to complain about weather forecasts almost as much as they like to complain about the weather itself. "The forecast," goes a common refrain, "is always wrong." In fact, forecasts are usually right. In many areas, the hourly forecasts can be uncannily correct. Forecast bashing simply does not reflect the reality of modern meteorology — a highly technical endeavor based on more than 100 years of ever-improving science and mathematics. Here Bill Streever of Anchorage, author of "Cold" and the just-released "And Soon I Heard a Roaring Wind," discusses Alaska forecasting with two of the state's most experienced meteorologists.
For years I drove and bicycled and walked past the National Weather Service building in Anchorage, not far from Kincaid Park. I passed it under clear blue skies, in rain, snow and sleet. I passed it in still air and in light breezes and in gusts that threw road grit into my eyes. I passed it when the forecasts were spot on and (far less frequently) when the forecasts let me down. Often, I passed it thinking I should drop in to say hello, to thank those inside for what they are doing, to see what I could learn about the sorcery that allows them to see tomorrow's weather today.
And so, eventually, I came to meet Sam Albanese, the facility's meteorologist in charge. After exchanging pleasantries, Albanese shows me technology that has been used since 1896: a weather balloon. It is a limp greenish-brown latex sack, hanging in a closet. But when filled with helium, hydrogen or some other lighter-than-air gas, the balloon will come alive.
At 3 p.m. and 3 a.m., day after day, every day of the year, the National Weather Service breathes life into 13 weather balloons in Alaska. Balloons rise above Anchorage and Fairbanks, they float up from Kodiak, Yakutat and Nome, they depart Kotzebue, Bethel, Annette Island, King Salmon, Cold Bay, St. Paul, McGrath, and, of course, Barrow.
Like thousands of other weather balloons launched at the same time each day all over the world, they leave the ground 5 feet wide. They pass 1,000 feet, 10,000 feet, 30,000 feet. Winds buffet them. Rain and snow pelt them. Sun shines upon their stretched skin. They move through layers of weather in ways no human will ever experience. And as they rise, the air around them thins.
In response to the thinning air, the balloons expand, growing as they ascend above the highest mountains and beyond the reach of commercial aircraft. At 50,000 feet, they float above the layer of atmosphere that holds most of what we think of as weather.
The weather factory is behind them.
And then, 18 or 20 miles above the earth's surface, with latex strained beyond endurance, in air far too thin to breathe, the balloons pop.
Just 3 percent returned
From those popped balloons, slowed by tiny parachutes, instrument packages fall to earth. Each is wrapped in white foam, small enough to fit in the palm of a hand. They land at sea, in tundra, on mountainsides and in spruce bogs. The tiny instrument packages carry a shipping bag. At no cost to the sender, they can be sent back to the government, to be refurbished and reused.
"Less than 3 percent of them are returned to us," Albanese says. Each kit costs about $400. With two kits going aloft from 13 Alaska locations each day, the daily cost exceeds $10,000 — or $3,650,000 a year.
"It seems expensive," he says, "but when you look at return on investment, it is nothing. The data help aircraft avoid strong headwinds and fuel savings alone pay for the program."
The information goes well beyond the identification of headwinds. All the data, along with information from ground stations, satellites, airplanes, ships and buoys, is combined with similar information from Canada, from the other 49 states, from Japan, Russia, Antarctica — from everywhere on earth.
Computers handle the information. These are not ordinary computers. These are supercomputers, room-filling monsters capable of trillions of operations every second. These are computers capable of handling the mathematics of the atmosphere. They start with information from the ground stations, satellites, airplanes, ships, buoys, and, of course, balloons, and they move forward from there, advancing fast enough to see what the future might hold.
And that future, the output from the churning of numbers that represents the real churning that occurs in our skies, comes back to Alaska. At forecast offices in Anchorage, Fairbanks, and Juneau, it gives government meteorologists a big picture of what the atmosphere has in store. It gives them a first glimpse of tomorrow's weather, a glimpse that allows them to provide, as the National Weather Service likes to say, "watches, warnings, and advisories for protection of life and property and enhancement of the economy."
First mathematical forecasts in the 1970s
Pause for a moment. Do not be fooled by old-fashioned weather balloons. Realize that weather forecasting now, mathematically based forecasting, is relatively new. The mathematics used in the computers was not even imagined until 1904. Primitive versions of the computers themselves, capable of clumsily managing the numbers in small doses for research purposes, appeared in the 1950s. The first mathematical forecasts did not reach the public until the 1970s. Anyone old enough to remember forecasts from before then remembers forecasts that were based on map interpretations.
Both the mathematical models that run in the computers and the computers themselves continue to evolve, to improve.
But realize, too, that the brute-force mathematics raging through the innards of super computers only gives the big picture.
And in some places, Alaska included, that big picture might be a little hazy.
"Terrain is the big thing in Alaska," Albanese says. "Look at how water flows down a streambed. The atmosphere is a fluid just like water. Watch how the clouds move around. Think about white water running over boulders — you can see the same thing in clouds coming across the tops of the mountains."
On one side of a mountain, a gentle breeze may blow, but on the other side, that gentle breeze may be funneled into a gale. Albanese compares winds blowing through Turnagain Arm — the "Turnagain Arm jet," in his words — to water coming from the nozzle of a hose. He talks of "model efficiencies," "grid density," and "barrier jets."
But from the specialized jargon of a professional forecaster, an important message emerges: The large-scale mathematical models offer a regional picture. But a local picture, the kind that matters to Alaskans, requires local interpretation.
Terrain is only one of many challenges. Another is the paucity of information coming from some parts of Alaska. And in some areas, the forecasts themselves may contribute to a certain kind of bias.
Say, for example, that a forecast calls for a gale, as it often does, at a particular location in the Bering Sea. Mariners, when they can, avoid gales. Weather reports coming from vessels in the Bering Sea miss the worst of the winds.
'Poor man's bias correction'
Rick Thoman, with the National Weather Service's Fairbanks office, echoes much of what Albanese says. The models coming from super computers offer a grainy image, a picture of what happens in squares something like 6 miles on each side. The weather — wind included — occurs at smaller scales.
Thoman talks about "operational history," code for good records captured over long periods. If the operational history for an airport or a harbor shows that the wind usually blows 15 mph faster than the super computer models suggest, the forecast for that airport or harbor can be dramatically improved.
"We use statistical techniques that link models to historical data," he says, describing a method that he calls "a poor man's bias correction."
Alaska forecasters deal with other wind-related issues that set Alaska apart. There is, for example, the frequency of high winds in the state. In Alaska, hurricane-strength winds, with speeds in excess of 74 mph, typically come five to seven times a year. These are winds capable of toppling cranes and sending ships onto rocky shorelines. Even lesser winds can carry freezing spray that accumulates on boats and ships, making them dangerously top heavy — a particular hazard in a state dependent on marine transport and with one of the nation's most productive fishing fleets.
Blowing dust, volcanic ash
On land, there is dust from glaciated valleys. Dust blowing off the Knik and Matanuska glaciers can pose a problem for small planes and for people with respiratory problems.
There is volcanic ash, too — most famously the ash from the 1989 Mount Redoubt eruption that forced KLM Flight 867 into a 14,000-foot drop, an incident that led to the creation of the Volcanic Ash Advisory Center (vaac.arh.noaa.gov).
And it is not just new ash that matters.
"Wind still suspends ash from the Katmai eruption, impacting aviation over Kodiak," says Albanese, referring to ash picked up from the ground around Katmai National Park. The Katmai eruption occurred in 1912.
A lack of wind creates another set of problems. When the wind rests, air over Anchorage and Fairbanks is trapped, allowing exhaust from cars and smoke from wood-burning stoves to accumulate, sometimes to the point that it's dangerously unhealthy.
From a forecaster's perspective, some weather is easier and some is harder. Sometimes the atmosphere is well mixed and well behaved. Sometimes it is layered and uncooperative.
And for forecasters, as for all of us, there is the issue of resource availability. If weather is particularly threatening in one area, forecasters may divert resources from areas with more benign conditions.
If, for example, the Anchorage area faces mild winds and maybe a little rain or snow, but Dutch Harbor faces the possibility of hurricane-strength winds, forecasters might put most of their effort into Dutch Harbor. If Fairbanks is likely to be cold and clear, but wind threatens to push ice hard against the shore in Barrow, forecasters might concentrate on Barrow.
Earlier in his career, Albanese sometimes worked as an incident meteorologist — a forecaster tasked with a very specific area during operations where weather played a critical role.
He has worked forest fires and he worked the oil spill from the Selendang Ayu, the cargo ship with 1,100 tons of fuel that ran aground on Unalaska Island in 2004. On fires, he could tell crews how the flames might be fanned and which way the smoke might blow. On the Selendang Ayu spill, his knowledge of future winds interested responders working from small boats and planners concerned about where the ship's leaking fuel might be blown to shore so responders knew where to best deploy booms that trap oil and protect vulnerable shorelines.
"I tried to explain to people," he says of his time as an incident meteorologist, "it's a forecast. Forecasts aren't perfect and never have been. You have to give some wiggle room, a way out." Forecasts can provide insights, but not absolutes.
"Are we going to miss a forecast?" Albanese asks, in reference to his current job as the Anchorage region's meteorologist in charge. "Sure. We've done it. But it's not like we forecast flurries and get 10 inches of snow."
Just as he did when he worked as an incident meteorologist, Albanese emphasizes the proper use of a forecast. Someone planning to hike a neighborhood trail could change plans on short notice if the forecast proves faulty, but someone planning a remote wilderness trip might be stuck with whatever nature has to offer.
"We're trying to do the best we can to provide you with good information so that you can make a good decision," he says. His emphasis lies clearly in the second half of that sentence — the half that puts decision-making in the hands of users.
Dramatic forecast improvements
Despite the caveats, Albanese has seen dramatic improvements in forecasting during a 30-year National Weather Service career that started in Yakutat in 1986. Forecasters once relied on mathematical models depicting the atmosphere in nine layers, but today they rely on models depicting the atmosphere in 20 layers. The Earth's surface, too, is divided into smaller pieces by today's models. And there is better real-world information streaming into the models, telling those models what nature is doing.
When he started, Albanese and his colleagues could offer a two-day forecast and a more-general five-day outlook. In the 1990s, they started offering five-day forecasts. Today, they offer a seven-day forecast for some locations.
"Today's five-day forecast," he says, "is as good as our two-day forecast in 1988."
Albanese's colleague Thoman points out that Alaska has something like 150 or 200 observation points showing temperature and wind speed from locations scattered around the state. "That," he says, "is quite a change in the last 25 years." Satellites have also improved during the last decade.
"On the other hand," Thoman says, "our weather radar set is from the mid-1990s."
He also points to gaps in information coming from high terrain. For example, there are no real-time observations coming in from the eastern Brooks Range north of Arctic Village. And while there is a tremendous amount of information coming in from airplanes, most of it comes from the Anchorage area.
What does the future hold? I ask both Albanese and Thoman for, well, a forecast.
One answer: Weather models have been built around a fairly stable climate regime, and climate change may throw some surprises our way. We know the climate is warming, and we have some ideas about what that will mean for rain and snow in different areas, but what does it mean for wind?
This, it turns out, is "an area of active research" — meaning, in plain language, that nobody has a very good idea of how warming will affect wind.
But there is another forecast about forecasting. The National Weather Service, despite the challenges of climate change, will push for longer-term forecasts. To succeed, they have to overcome the problems of chaos theory, a theory that applies well to the mathematics of the atmosphere.
According to that theory, small inaccuracies may not be important in the short term, but these same small inaccuracies will lead to ever-growing and unpredictable inaccuracies in the long term. A little mistake here, a missed data point there, and the future clouds over. But despite chaos theory, the kinds of improvements that both Albanese and Thoman have seen in their careers will continue to extend the range of useful forecasts, stretching them out weeks in advance of the actual weather.
And as to those old-fashioned National Weather Service balloons? The forecast is clear. Each afternoon and each morning, day after day, every day of the year, 13 of them will rise up above Alaska, expanding as they ascend, experiencing Alaska's weather in ways that no human will ever know firsthand. But despite the ongoing use of 19th-century technology, there is nothing old fashioned about weather forecasting.
Correction: An earlier version of this story miscalculated the yearly cost of sending aloft weather data instrument kits. The cost is $3,650,000 a year.
Bill Streever, a biologist and affiliate faculty member at the University of Alaska Fairbanks, is the author of the best-selling book "Cold" as well as "Heat" and the just-released "And Soon I Heard a Roaring Wind." As a scientist, he has worked on issues ranging from the environmental effects of underwater sound to the evolution of cave crayfish to the restoration of tundra wetlands to climate change.