Skip to main content

Can Hurricanes Be Predicted Decades in Advance?

The rapidly growing field of paleotempestology lays the foundation for reliable hurricane predictions a decade or more into the future.

In 2007, a reporter for the Post & Courier of Charleston, S.C., was tired of doing straight stories on hurricane forecasts. So he hired a medium to predict the forthcoming storm season. "The sense we got from emergency-management people here," the reporter wrote, "is that the forecasts had been so wrong that they were hearing from the public, 'Why should we pay any attention to this stuff?'"

At the end of the hurricane season, it turned out the medium had been more accurate than the scientists who took it upon themselves to make storm predictions. But research seems likely to soon make weather mediums a thing of the past.

University of Massachusetts researcher Jon Woodruff putters his plastic boat near enough to the rest of his team so the others can hear when X marks the spot. Slowly, he relays the string of numbers: 30.01136N and 84.42989W. Sam Zipper, a research assistant from Woods Hole Oceanographic Institute, enters them into his GPS device and directs us west and then north.

Dana McDonald, another Woods Hole researcher, and I row the pond-worthy vessel — two aluminum canoes with a square of plywood lashed between them — making a few passes until we're over the right coordinates. McDonald sets three anchors to hold the ungainly catamaran against the cutting wind on this uncommonly cold 27-degree morning in St. Mark's National Wildlife Refuge, 25 miles south of Tallahassee along Florida's Gulf Coast.

Woods Hole research assistants Dana McDonald, Skye Moret, Sam Zipper and Phil Lane on a makeshift catamaran used in paleotempestology research. (Photo by Jim Morrison) Click to enlarge.

Woods Hole research assistants Dana McDonald, Skye Moret, Sam Zipper and Phil Lane on a makeshift catamaran used in paleotempestology research. (Photo by Jim Morrison) Click to enlarge.

The team has named this unassuming freshwater pond Lake Patricia after a favorite aunt of researcher Christine Brandon, who helped scout it yesterday. Thick brush, dry and brown with the season, surrounds the water. Lily pads ring the shallows of the shoreline. The water is surprisingly clear down to 3 or 4 feet. We're just a few miles from several small towns and million-dollar beach houses, but this is coastal wilderness, the gates to the refuge unlocked only for Woodruff and the team.

Earlier this morning, Brandon and Woodruff, who resembles a young Ron Howard, motored slowly back and forth watching the results from ground-penetrating radar on his laptop. The profile: a flat-bottomed bowl.

It's the perfect shape for a pond to be history's repository.

Woodruff and fellow principal investigator Jeff Donnelly of Woods Hole are seeking sediment samples, long tubes of sand, muck and other material that serve as a muddy analog to tree rings, giving evidence of weather and other physical phenomena going back thousands of years. The deeper into the muck they can drill, the longer the historical record. Each core is like a many-layered cake. They're looking for thin white slivers in the dark peat revealing major storms that picked up sand and deposited it far inland.

By reading the history in those sedimentary cores, these scientists hope to uncover patterns and, by knowing more about those patterns and the hurricanes they produced, to be better able to predict the future, saving lives and billions of dollars in damage.

Woodruff and Donnelly's quest for sediment cores is one example of the innovative techniques scientists are using to reconstruct hurricane history in a rapidly growing field called paleotempestology. Others are studying different proxies for prehistoric weather: the chemistry of corals, tree rings, oxygen isotopes in coastal cave stalagmites and the deposition of marine microfossils.

While researchers in the field are collecting specimens, others at select supercomputer sites across the globe are tuning massive models they say will be able to forecast the likelihood of hurricane strikes decades in advance. Most important, they may help settle the question of whether climate change will spawn a period of more frequent and destructive storms.

Greg Holland, the director of the Mesoscale and Microscale Meteorology program at the National Center for Atmospheric Research in Boulder, Colo., and one of the éminences grises of the hurricane forecasting community, says that 20 years ago, predicting the track of hurricanes headed toward land wasn't terribly reliable. Then a burst of research dramatically improved these real-time forecasts. He sees the same thing happening now with forecasting hurricane frequency and intensity, years and even decades in advance. "It's a rapidly evolving field," he says. "There's this intensive effort. You can see the curve bending down towards better and more precise predictions. It's an exciting time."

Last summer, Congress held hearings and called for a national hurricane research initiative similar to one that exists for earthquakes, which have caused only a fraction of the economic losses that hurricanes have in the U.S. in recent years. Nearly 30 million Americans live in coastal areas, from North Carolina to Texas, that are threatened by hurricanes. More than 3,000 people have died since 2003 in hurricanes. The National Science Board estimates that from 2002 to 2007, economic losses due to hurricanes totaled $180 billion. According to the Reinsurance Association of America, insured property exposure is $9 trillion. AIR Worldwide, a catastrophe modeling company, estimates that disaster losses will double every decade because of growing residential and commercial density along the coasts.

But predicting the number and intensity of hurricanes before they spawn in the Atlantic has been a hit-or-miss proposition, even just a few months before the season begins.

Every spring, researchers from Colorado State University, the National Oceanic and Atmospheric Administration, and the Center for Ocean-Atmospheric Prediction Studies at Florida State University issue forecasts for the coming hurricane season. They rely mostly on statistical models guided by recent history.

None projects hurricanes beyond the next season. Risk Management Solutions, a company that helps insurers manage their exposure from natural disasters, took a stab in 2006, issuing a five-year forecast predicting that insured hurricane losses in the United States would be 40 percent higher than the historical average. The projection has been a colossal bust, greatly overestimating losses.

Holland and others, including Roger Pielke Jr., director of the Center for Science and Technology Policy Research at the University of Colorado and a senior fellow at the Breakthrough Institute, a progressive think tank, say the current seasonal forecasts exhibit no more skill than Punxsutawney Phil predicting the end of winter. "These efforts to predict are scientifically worthwhile exploring and add to the understanding of hurricane dynamics," Pielke says. "But from the standpoint of being practical, skillful and useful, they are not there yet, and, who knows, they may never be."

In early 2005, hurricane predictions ranged from 11 to 14 tropical storms. Instead, there were 27 named tropical storms and more hurricanes — 15 — than any year in the past half century. Katrina, Rita, Dennis and the other storms did more than $100 billion damage. In the next two years, forecasters predicted more storms than materialized. In the last two years, they've been roughly on the mark.

It's easy to see why understanding hurricane risk is important driving with Woodruff along Route 98, the Coastal Highway, through the tiny town of Panacea, Fla. It seems as if every third building is for sale. As the road veers along the beach and past expensive homes down private roads, the for-sale signs only multiply.

A Navarre Beach, Fla., home damaged by Hurricane Dennis. Click to enlarge.

A Navarre Beach, Fla., home damaged by Hurricane Dennis. Click to enlarge.

We pass Angelo & Sons, a restaurant perched on stilts at the base of the bridge across the Ocklockonee Bay. The longtime eatery has been rebuilt since Dennis battered it into submission on July 10, 2005, the earliest major hurricane on record. The storm surge through here, expected to be only a few feet, reached 10 to 12 feet. As Dennis spun up the Gulf of Mexico, it pushed a rising wall of water. Winds skimmed the Apalachee Bay, which is only 5 to 10 feet deep even miles offshore, lifting the water and forcing it into the coast where rain had already swelled the St. Marks and Wakulla rivers. In Florida alone, Dennis killed 10 people and caused an estimated $1.5 billion in damages.

Part of the record-breaking 2005 season, Dennis was the second punch to Florida's northwest coast. Nine months earlier, Hurricane Ivan, nicknamed Ivan the Terrible, landed a left hook with 130 mph winds and extensive flooding, although the major damage was further west in Santa Rosa County and Pensacola, where the storm surge reached 15 feet. The storm destroyed a quarter mile of the bridge across Escambia Bay. At sea, Ivan sunk seven oil rigs and set five others adrift. Even more costly was the damage to underwater pipelines. Some were moved 3,000 feet while others were lost, buried under 30 feet of mud, according to a study by the United States Naval Research Laboratory. A monitoring buoy 100 miles south of Mobile, Ala., recorded a wave of more than 53 feet as Ivan approached. The hurricane killed 14 people in Florida and caused $13 billion in damages across the southern United States. (Three other hurricanes — Charley, Frances and Jeanne — struck other parts of Florida in 2004).

From 2000-2009, 53 tropical storms or hurricanes struck Florida, causing an estimated $64 billion in damages. Homeowners in high-risk areas like the Gulf Coast continue to pay the cost through soaring insurance rates. Some say they have been unable to sell their homes because the buyers can't obtain insurance, so getting a mortgage is all but impossible.

One afternoon, as Woodruff prepares to survey a nearby lake, a resident comes by and says the value of his beachfront home has dropped to $500,000 from $1 million. A friend, he adds, is paying $2,000 a month for insurance. Claiming it was losing $20 million a month, State Farm Florida threatened to withdraw from the state last year before agreeing to stay, but at a cost. The insurer said it would drop 125,000 policyholders in high-risk areas and raise rates on the remaining 685,000 by about 41 percent. Florida insurers say it took them 11 years to recover from the catastrophic losses of 1992's Hurricane Andrew, a rare Category 5 storm.

Franklin W. Nutter, president of the Reinsurance Association of America, says the industry has been encouraging and funding research on three main fronts: longer-term regional forecasts looking to understand the effects of climate change, a better understanding of storm surges in coastal areas and better forecasts of storm intensity. "There are so many unresolved questions," he says.

The most relevant for insurers is the outlook 10, 20 or 30 years in the future. "The big concern is whether the past is prologue, or whether, because of climate change, we're in for a period of not just natural variability, but a change in the dynamics of the oceans and ocean currents and winds — El Niño, La Niña — that might make the past less relevant going forward," Nutter adds. "The problem for insurance companies is you're pricing a product today, but you're paying losses based on future developments."

Better forecasting would give insurers, homeowners and policymakers an opportunity to adapt by changing land-use policies, strengthening building codes and preserving or restoring natural mitigation by wetlands, floodplains, and beach and dune systems. Several studies have shown investing in better forecasting and mitigation has significant returns. One study by the Multihazard Mitigation Council of the National Institute of Building Sciences concluded that each dollar spent on mitigation saves an average of $4 in damage costs.

"If you know what's happening, you can at least get ready for it or have the capacity to get ready for it," Holland says. "It's your choice."

Anecdotal information about hurricanes goes back a few hundred years but is considered unreliable until the last century or so. In 1943, the "Surprise Hurricane" (the current naming system wasn't in place) struck Texas. British pilots were being trained for World War II at Bryan Field in College Station. When they saw Americans evacuating their AT-6 Texan trainers, they chided them about the plane's durability. The lead instructor, Col. Joseph Duckworth, hopped into a trainer and flew into the storm. After Duckworth returned safely, the base's weather officer, Lt. William Jones-Burdick, demanded he be taken into the storm. Soon, there were regular flights through hurricanes to gather meteorological data.

Today, NOAA planes disperse dropsondes, foot-long probes that parachute through storms transmitting data every half-second. The first weather satellite was launched in 1960, but it wasn't until about a decade later that satellites began collecting hurricane data.

In 1984, William Gray, a meteorologist and professor at Colorado State University, issued the first seasonal hurricane forecast, predicting that 10 tropical storms and five hurricanes — existing for a total of 30 "hurricane days" — would strike the Atlantic and Caribbean basins.

He was close. That year produced 12 storms, five hurricanes and 18 hurricane days. Since then, Gray, an imposing 6 feet 5 inches tall with piercing eyes and white hair, has become the most quoted hurricane forecasting expert in the world. He's 80 now and has turned over the primary responsibility for the forecasts to his assistant, Philip Klotzbach, but is still revered for his work uncovering the effect conditions in the Pacific Ocean and in West Africa have on hurricanes spawning in the Atlantic.

He and Klotzbach use 60 years of data on sea surface temperature, atmospheric pressure and winds to make annual predictions. "We assume that the future is going to behave similarly to how the past has behaved," says Klotzbach, who has worked with Gray since 2000.

In 1998, the NOAA began offering an outlook that gives a range of named storms, hurricanes and major hurricanes. At NOAA, statistical analysis of climate factors and history, aided during the last two years by predictions from computer models, form the basis of the hurricane forecast. While Gray's team issues a forecast based on numbers, NOAA issues an "outlook" that reports the likelihood of a range of storms.

Three main factors, NOAA meteorologist Gerry Bell says, influence the number of hurricanes. They are the El Niño/La Niña cycle in the Pacific Ocean; what are known as multidecadal cycles, 25-to-40-year periods of higher or lower hurricane activity (we've been in a higher activity phase since 1995); and Atlantic Ocean temperatures. "Based on the climate patterns and their combined strength, we have a variety of statistical tools that say given this set of climate conditions, what historically have we seen?" Bell says. "You would say this combination has produced anywhere from 13 to 16 named storms or three to five hurricanes, things like that. Based on historical record, you can go back and link these climate conditions to level of activity."

Bell, who furthered Gray's work by creating an index to measure a hurricane season's strength, has heard the critics, and he acknowledges the outlook's limitations and the barriers to better forecasts. "There's a lot of skill in these forecasts," he says. "But we don't have the knowledge to make every outlook correct."

It's often not clear until April or May, for instance, whether it will be an El Niño or La Niña year. In an El Niño, waters in the Pacific Ocean are warmer than usual; the warm waters affect wind patterns in the atmosphere, increasing wind shear in the Atlantic, which depresses hurricane formation. La Niña, with cooler Pacific water temperatures, has the opposite effect, increasing the number of Atlantic hurricanes and their frequency. Because of uncertainties about the arrival and strength of El Niños and La Niñas, forecasters issue an early season prediction in April or May and a second forecast in August, just before the height of the season.

There is also no understanding when an active hurricane era, like the one we're in, will end. Cycles of low and high activity can last for decades, then suddenly shift. "We really don't have an ability to say here's our active era, and we expect it to end in year X," Bell adds.

Where Holland and Bell agree is that the future of forecasting lies in computer modeling. "The models get the entire complex chain of events that lead into the season in detail," Holland adds.

At the National Center for Atmospheric Research in Boulder, researchers have married a global climate model with a weather model, coupling the long view of changes over decades in oceans and wind shear that control hurricane activity with the finer regional resolution of shorter-term weather modeling. The project is looking at experimental predictions for three decades in detail: the current decade, 2020-2030 and 2045-2055.

Holland says NCAR researchers are still learning how to use the model properly, and it will take another five to eight years before it becomes fully operational. "I'm not going to forecast a squall line through New York in 2050," he adds. "But what we want to do is be able to say: 'What are the statistics of squall lines going through New York in 2050?'"

Running 20 or more forecasts with slightly different conditions gives forecasters a measure of confidence. "That's where the insurance industry and the emergency managers are starting to go," Holland says. "If they can get an indication next year will be 15 tropical storms and the uncertainty is plus or minus 1, or 15 and the uncertainty is plus or minus five, [rather] than just saying 15 [storms], tells them a lot more. They have an idea on what the bounds will be."

Holland and Bell think predicting landfalls — what matters most to residents, insurers and emergency planners — will one day be possible. NOAA is researching East Coast landfalls, working out the climate patterns controlling them and looking at the kind of statistical models needed to produce a meaningful forecast, Bell says.

Holland thinks it's just a matter of time. "I genuinely think as the models get better able to predict the general characteristics [of storms] in places like the Atlantic, you'll be able to get skillful at which ones are going to make landfall," he says.

The raging question is what effect climate change will have. Gray denies climate change is occurring. Bell says the effect is unclear. Holland says look out. "The consensus looks like there will be more tropical storms and hurricanes combined — a small amount," he says, "but the research starting to emerge is there will be a substantial increase in the really intense ones, half as many again."

The most severe hurricanes are so rare that it's almost impossible to detect a pattern in their occurrence. Only three Category 5 storms have pounded American shores in the last century: Hurricane Andrew, which killed 23 when it struck Florida and Louisiana in 1992, Hurricane Camille, which plowed through Mississippi in 1969, and an unnamed storm that smashed through the Florida Keys in 1935.

Finding patterns that link such storms is one of the aims of the Woodruff and Donnelly team. That's why they've come to Florida's Forgotten Coast, a sparsely populated area of longleaf pine flatlands, searching for the right spot to transport them back through time.

Together and separately, Woodruff and Donnelly have extracted cores in Belize, Yucatan, St. Kitts, St. Croix, Grenada, Brazil, Puerto Rico, New York, Massachusetts and Japan. "Given the limited scope of climate change experienced over the instrumental record, the paleorecord is a critical archive that allows us to explore a fuller range of climate conditions in order to tease apart what the important climatic controls are," says Donnelly, who has been extracting and analyzing cores from across the globe for 14 years.

In 2007, a Woodruff-Donnelly paper in Nature reported that a Playa Grande, Puerto Rico, sediment core revealed a 5,000-year storm record that identified weak El Niños and strong monsoons in West Africa as key factors that heightened Atlantic hurricane activity and that should be incorporated into computer models.

Last year, in another Nature paper, the researchers reported that the frequency of intense hurricanes in the Atlantic over the past 1,500 years has been closely linked to changes in El Niño and sea surface temperatures. The work suggested that a warming world may face more intense hurricanes, something that has become a consensus among scientists starting to analyze the data from computer models. The past decade, they noted, had the highest number of North Atlantic hurricanes since a similar period 1,000 years ago when El Niños were also weak and sea surface temperatures were high.

"Hurricane activity has responded noticeably to past climate shifts," Woodruff said in the careful words of a researcher. "When considering future climate change over the next century, our results indicate that measurable changes in hurricane activity could occur."

Holland, the National Center for Atmospheric Research official, sums up their efforts another way: "Some of the work those guys are doing is just brilliant stuff. The detective work would make Sherlock Holmes proud."

When he went to Japan for the first time to meet his future in-laws, Woodruff dragged them and his parents to a small island 12 miles southwest of the main island to dig in the muck. On a return trip in 2006, he drafted his wife and mother-in-law for long days of pulling sediment cores from jellyfish-infested waters onto a craft named The Unsinkable, fashioned from two Wal-Mart rafts. (It wasn't, by the way.) His Japan work formed the basis for his doctoral thesis, the first long-term coring record for North Pacific typhoons.

Woodruff found his research passion in the mud while earning a master's degree and doctorate in a joint Massachusetts Institute of Technology and Woods Hole program. He has a reputation as being the team's MacGyver, someone who can fashion tools from this and that in the field. He's made boats out of children's toy rafts and a beer cooler.

He and Donnelly returned this year to Florida's Apalachee Bay area to survey ponds farther from the coast to get a record of intense hurricanes that washed deep inland. The idea is to add pieces to the puzzles that portray both regional hurricane inundation and global conditions and activity.

The area is ideal for such research because hurricane strikes are common, storm surges are high and the sinkhole ponds offer the possibility of deep sedimentary records. Using Google Earth, Woods Hole researcher Phil Lane identified more than 100 potential ponds to scout. One night, after dinner in the cinderblock beachfront dormitory of Florida State University's Coastal and Marine Laboratory on St. James Island, the team sits at tables arranged in a T, poring over map printouts and the computerized results of ground-penetrating radar and sonar scouting expeditions.

They've spent the last two days pulling cores from some small ponds, but they're relatively short cores, not the kind promising a long sedimentary record. The next morning, the first site they look at is Lake Patricia. While Woodruff and Brandon slowly criss-cross the pond, using ground-penetrating radar to look for an area of deep sedimentation, the others lash the plywood to the canoes to create a platform for coring. Then they mount a homemade pipe assembly spanning the plywood. From it hangs a small winch used to raise the core.

By the time we're on the water, Woodruff has found a better coring location than the one scouted by Donnelly the day before. After he relays the coordinates, he swings back around as we're preparing the equipment.

"There's a boatload of sediment right under you," he shouts.

"How much?" McDonald asks.

"Twenty meters."

It's the best news of a research expedition, now in its second week, that has had more than its share of misses.

To get deep, the team is vibracoring, a technique using a vibrating drill head, designed to rid concrete forms of voids and air pockets, to sink the coring tube deep into the muck. A piston assembly in the tube creates a vacuum, holding the sediment when the tube is raised, like pressing your finger to the top of a straw in a drink. By the time we assemble the equipment, Woodruff has joined us. The gas motor to drive the head starts on the first pull, and, with some effort, we raise the 30-foot coring tube over the spot.

We steady it, then — with a "one-two" count — push down. It plows through the sediment like a spoon through yogurt, slowing only after the top end of the tube dips below the surface. Zipper, the Woods Hole research assistant, reaches into the frigid water to cap the tube. Woodruff attaches a winch to the homemade crossbar mounted on the plywood, and the tube is winched from the bottom until the four of us can muscle it onto the boat.

As it breaks the surface, Woodruff reaches into the bottom of the tin cylinder. "That's sand, my friend," he says holding out a smear with white grains highlighted against the dark organic matter. Sand is hard to move. Only a high-energy event like a hurricane, a big one, would blow it here.

"Paleo hurricane," McDonald says, offering the quickie caption.

Back in Massachusetts two weeks later, Woodruff notes that a previous core 6 meters long contained a historical record covering 4,000 years. How far back the layers in the Lake Patricia core will transport researchers depends upon the rate of sedimentation. More rapid deposition means a shorter record, but more detail. It will take months to begin radiocarbon dating the core and perhaps a couple of years to analyze it. "Each of these lakes is recording hurricanes in slightly different fashion. You sort of have to get a feel for how to read the book," Woodruff says. "It's a slow process, and there are few eureka moments."

For quicker answers, there's always the local medium.