|Above: Damage caused the day before by Hurricane Maria in Roseau, Dominica, on September 20, 2017. Arriving with Category 5 winds, Maria was the strongest hurricane on record to hit Dominica, inflicting more than $1 billion US in damage and taking 31 lives. Image credit: Photo credit should read STR/AFP/Getty Images.|
As we discussed on Wednesday in the first part of this two-part post, it turns out that land-based observations seldom live up to a hurricane’s Saffir-Simpson ranking at landfall. There are some very good reasons for the difference, as well as potential implications for stakeholders, from insurance companies to the public at large. Note: This is Part II of a two-part post. Part I was published on Wednesday, April 11.
The sudden drop-off
When a hurricane passes over open water toward the U.S. coast, wind observations come from five main sources:
—Reliable flight-level wind data measured directly from Hurricane Hunter aircraft, adjusted to the surface from measurement levels that are typically around 700 mb (3000 meters or 10,000 feet)
—Dropsondes that descend from Hurricane Hunter aircraft
—The Stepped Frequency Microwave Radiometer (SFMR) aboard Hurricane Hunter planes. The SFMR infers near-surface wind speed by passively sensing microwave radiation from the ocean surface.
—Satellite imagery, which allows forecasters to use the Dvorak technique to obtain estimates of winds and central pressure
—Ground-based NEXRAD Doppler radars (assuming that one or more are located near the landfall region)
|Figure 1. Flight track for a Hurricane Hunter mission that explored Harvey on the afternoon of August 24, 2017, around the time Harvey intensified to hurricane strength in the northwest Gulf of Mexico. Each wind flag points toward the direction the wind is blowing. Barbs on each flag correspond to the strength of the flight-level wind, which is also denoted by the colors shown at right. Top flight-level winds were 80 knots (92 mph), and the strongest estimated surface winds based on SFMR data were around 70 knots (80 mph). Image credit: NOAA Atmospheric Oceanographic and Meteorological Laboratory.|
Satellite- and aircraft-based techniques aren’t perfect, but they can provide a solid overview of wind distribution patterns up to landfall. From that point on, land-based reports—collected from volunteer observers, NWS and military weather stations, state-based instrument networks, and portable radars—become increasingly important. Personal weather stations can also play a role: one PWS measured a 199-mph gust from Hurricane Irma on St. Barts, as we reported in January.
The catch is that a storm’s character can change dramatically in the several hours just before and especially just after landfall. Peak wind speeds tend to drop off quickly, especially if a hurricane’s structure becomes compromised in the hours before landfall.
“Within less than a mile of the coast, the sustained winds drop by about 20% due to greater surface friction,” noted Chris Landsea, science and operations officer at the NOAA/NWS National Hurricane Center, in an email. “Substantial weakening of the peak winds is also occurring over the first several hours after landfall as the center of the hurricane continues moving inland.”
Several other factors also reduce the odds a hurricane’s official strength at landfall will be verified over land.
Peak winds are used to categorize an entire hurricane. The strongest winds in a given hurricane would only be expected to occur in a small part of that hurricane's total wind field. Take the analogy of a tornado: a block of well-built homes swept cleanly off their foundations can dictate an entire twister’s rating on the Enhanced Fujita Scale (e.g., an EF5). Similarly, the strongest wind speeds in a hurricane will determine its Saffir-Simpson rating, even if those peak winds occur only in a very small area.
My colleague Dr. Jeff Masters, who encountered Hurricane Hugo first hand in 1989 as a NOAA flight meteorologist, pointed out how densely packed a hurricane’s peak winds can be. “Major hurricanes sometimes concentrate their Cat 2 and stronger winds exclusively in the eyewall,” he told me. “Flying into Hurricane Hugo at 1,500 feet, we found Cat 1 winds in a spiral band as we approached the eyewall, but no other hurricane-force winds until we hit the eyewall, when we encountered Cat 5 winds. Hugo’s flight-level winds were only 60 mph in the moat region just outside the eyewall.”
|Figure 2. Hurricane Hugo as it approached South Carolina at 1844Z (2:44 pm EDT) on September 21, 1989. Image credit: NOAA/NESDIS.|
Observing points are limited. “Over land, there often are very limited (or no) observations of the winds in the right front quadrant of the eyewall where the peak conditions reside,” said Landsea. It’s unlikely that a high-quality weather instrument will just happen to be located right at the spot where a hurricane’s strongest post-landfall winds occur.
“Given the small number of surface stations, and the even smaller number that survive an extreme event, it is no surprise that the measurements don't get near the maximum NHC states,” said Sim Aberson, a meteorologist at the NOAA/AOML Hurricane Research Division.
On top of these spatial issues, there are plenty of observational challenges in trying to gauge surface winds near ground level amid the maelstrom of a hurricane in progress.
Station height. Surface friction will lead to weaker winds at lower heights and stronger winds at higher heights. The standard instrument height specified by the World Meteorological Organization for wind measurement is 10 meters (33 feet). As noted in Part I for Hurricane Irma, many tower-mounted anemometers are positioned much higher than this, and some instruments are lower, which makes it challenging to reconcile reports across a landfall region.
Instrument siting. Buildings and trees can block winds from reaching a weather station. The World Meteorological Organization recommends that anemometers be positioned at least 10 times their height from any obstruction—i.e., a horizontal distance of at least 100 meters (330 feet) from a 10-meter-high tower. “In practice, it is often difficult to find a good or even acceptable location for a wind station,” the WMO notes.
Instrument quality. A poor-quality anemometer won’t be able to accurately capture the peak winds at its location, no matter how well situated it is.
Vulnerable instruments. Even high-quality weather sensors can be knocked out of service in the midst of a hurricane, whether it’s because of damage to the instrument, power failure, or connectivity issues. During Irma, as reported by Kim Miller (Palm Beach Post), “Either complete failure, or loss of multiple functions, occurred at eight sites monitored by the National Weather Service office in Miami. The two official gauges of record in the Keys also died during the storm. And the Melbourne station, which was waiting for a part when Irma struck, failed to collect wind speeds on Sept. 10 and 11.”
|Figure 3. VIIRS infrared satellite image of Hurricane Maria moving just west of St. Croix while at Cat 5 strength at 2:13 am EDT on Wednesday, September 20, 2017. Image credit: NOAA/CIMSS/UM-Madison.|
Maria in Puerto Rico
The catastrophic course of Hurricane Maria across Dominica and Puerto Rico illustrates several of the points above. Before it reached Puerto Rico, Maria slammed into Dominica and produced catastrophic damage just hours after having rocketed to Category 5 strength. In this case, a resilient surface anemometer was ideally positioned to capture the storm at its worst. The Douglas-Charles Airport on Dominica’s northeast coast was hit squarely by the stronger northern side of Maria’s eyewall. The top 10-minute sustained wind reported at the airport was 130 knots (150 mph), which corresponds to a standard 1-minute sustained wind of at least 143 knots (165 mph). This is very close to the NHC estimate of winds at landfall of 145 knots, which was based on SFMR-derived surface winds.
“The once-lush tropical island was effectively reduced to an immense field of debris,” observed NHC in its official report on Maria, issued on April 6. A weather.com roundup by Jonathan Belles features ten of the most jaw-dropping aspects of this horrific hurricane.
|Figure 4. Damaged houses are seen on a barren hillside in Yabucoa, eastern Puerto Rico, on September 29, 2017, nine days after the storm made landfall in this vicinity. Image credit: Hector Retamal/AFP/Getty Images.|
Maria made landfall near Yabucoa in southeast Puerto Rico at 4:15 am AST on September 20, 2017, with an NHC-estimated peak wind of 135 knots (155 mph), just below Category 5 strength. In its Maria report, NHC stated: “The landfall intensity of the cyclone in Puerto Rico, 135 kt, is based on an extrapolation of the weakening trend noted in the aircraft data after the eyewall replacement began several hours earlier. There were no believable Doppler-derived winds from the San Juan WSR-88D radar that supported a higher intensity. It should be noted, however, that in Puerto Rico, winds of category 5 intensity were almost certainly felt at some elevated locations on the island.”
The near-complete loss of Puerto Rico’s power grid, which was already limping before the arrival of Maria, meant that many weather stations were knocked out of service when the storm hit, and there were no mobile radar deployments to help fill the gaps. The strongest sustained surface wind reported in Puerto Rico was 101 knots (116 mph) at a WeatherFlow site at Isla Culebrita Light, located about 20 miles east of the main island of Puerto Rico on the island of Culebra. This wind was measured from a tower-mounted anemometer at a height of 83 meters (272 feet), so winds near the surface at this point were most likely weaker.
Among the few wind measurements from the main island cited in the NHC report, the highest was the 94 knots (108 mph) measured at a standard height of 10 meters (33 feet) at a WeatherFlow site at Las Mareas, on the southeast coast of Puerto Rico. Las Mareas was about 20 miles southwest of the landfall site, on the weaker side of the storm’s inner core.
We may never know exactly how strong Maria’s top winds were in Puerto Rico, but the immense amount of wind damage to structures and trees, especially at higher elevations, speaks for itself. The radome of the main Doppler radar serving the island—reportedly designed to survive winds of more than 130 mph—was destroyed during the storm, and the interior radar dish was blown off its pedestal.
|Figure 5. The National Weather Service WSR-88D Doppler radar serving Puerto Rico took a direct hit from Hurricane Maria. Image credit: NWS/San Juan, via weather.com.|
How do people think about hurricane wind speeds?
As discussed in Part I, there were only two measurements in Florida of winds topping Category 1 strength during Hurricane Irma, even though it came ashore as a Category 3 storm. Some Floridians who got no more than Cat 1 winds might well think of the experience as having endured a major hurricane. If so, this would be consistent with work carried out by Robert Meyer (Wharton School, University of Pennsylvania) and Kenneth Broad (University of Miami). Along with colleagues from Columbia and Florida State, Meyer and Broad examined public perception of hurricane risk in a 2014 BAMS article.
The team carried out field studies during the approaches of Isaac and Sandy in 2012, comparing the NHC-generated odds of receiving hurricane-force winds to what people in those locations saw as their actual risk of getting such winds. They concluded: “The data suggest that residents threatened by such storms had a poor understanding of the threat posed by the storms; they overestimated the likelihood that their homes would be subject to hurricane-force wind conditions but underestimated the potential damage that such winds could cause, and they misconstrued the greatest threat as coming from wind rather than water.”
“We see the over-estimation bias as problematic from a preparedness standpoint,” Meyer said in an email. “People don’t seem to get what ‘sustained hurricane-force winds’ actually means, and believe that if they are in the hurricane-warning area they will get the maximum winds, or at least hurricane winds (why else would they give the warning?). Of course, the worry is that many will feel they have been through a Cat 1 when in fact they went through a ‘Cat 0’.”
|Figure 6. A tree downed by winds from Hurricane Irma blocks a road in Coconut Grove, Florida, on September 11, 2017. Image credit: Saul Loeb/AFP/Getty Images.|
Implications for insurers
Even if there’s good reason for the differences between landfall strength estimates and surface wind reports, it’s something that those who insure against hurricane losses have to reckon with.
James Done, a scientist at the National Center for Atmospheric Research and a Willis Research Fellow, works closely with the insurance broker Willis Re on modeling tropical cyclones and their damage potential. “I've noticed that observed winds just onshore are far lower (by at least one category) than the NHC rating at landfall,” said Done in an email. “We know that winds respond immediately to changes in surface friction, with the effect deepening upwards through the atmosphere with time, but I've always been surprised how abrupt this is.” High-resolution modeling with the WRF model depicts the dramatic drop-off, he added.
“This is a rather complex topic for the insurance industry,” said Steve Bowen, director of meteorology at insurance broker Aon Benfield. Bowen noted that many insurance contracts are based on “triggers” linked to such events as the issuance or cancellation of hurricane watches and warnings, the official NHC advisory intensity at landfall, and the timing of pre- and post-landfall damage.
“In the aftermath of storms, you can run into challenges of not finding observed onshore surface wind speeds that match the NHC landfall intensity,” said Bowen. “This is particularly important when there are triggers based on whether a storm comes ashore as a major hurricane. The difference between a Cat 2 or Cat 3 can be the deciding factor of which triggers are hit."
|Figure 7. The iconic Princess Cottage, built in 1855, remained standing for months after being ravaged by floodwaters associated with Hurricane/Superstorm Sandy in Union Beach, New Jersey. The bulk of Sandy’s damage was associated with storm surge rather than high wind. Image credit: Photo by Mario Tama/Getty Images.|
Perhaps the most famous recent example of a trigger with big implications is with 2012’s Hurricane/Superstorm Sandy, which caused more than $71 billion in damage (2017 dollars). There were no hurricane watches or warnings for Sandy anywhere north of North Carolina, and the storm was reclassified from a Category 1 hurricane to a post-tropical cyclone less than 3 hours before its New Jersey landfall, based mainly on its loss of thunderstorms near its center and its absorption into a frontal zone. To reduce the potential for public confusion, NHC now allows for hurricane watches and warnings to continue to landfall even if a hurricane has a chance of becoming post-tropical before that point. Because Sandy was not classified as a hurricane at landfall, many homeowners did not have to pay deductibles that would have otherwise kicked in—a boon for residents and a hit to insurers.
Would it make sense for the hurricane community to explore ways to address the winds-at-landfall mismatch? “The question you raise is certainly a good one, and important to ask,” Bowen told me. “If this hypothetical were to occur, it would likely require the NHC working closely with the insurance industry to clearly define how such a process would be implemented.”
The bottom line: Landfall winds may be weaker yet more damaging than we expect
One of the most striking aspects of Irma is the amount of mayhem its winds produced despite the lack of sustained wind reports above Category 1 strength. Especially in a densely settled state like Florida, the financial impact of downed trees, power lines, and ravaged roofs alone can add up very quickly.
NOAA ranks Irma as the nation’s fifth costliest hurricane (adjusted for inflation) in its dataset of billion-dollar disasters going back to 1980, with U.S. damages at $50 billion (2017 dollars). Of course, there may have been pockets of especially high wind that went undetected. And the amount of wind-related damage produced by a landfalling hurricane depends on many factors other than peak wind—including the size and duration of the wind field, the extent of development along a hurricane’s path, and the amount of rainfall before and during the storm, since saturated soil makes it easier for wind to uproot trees and bring down power lines. Wind can also exacerbate damage inflicted by storm surge and inland flooding. (Water-related hurricane damage is a hugely important topic outside the scope of this post.)
|Figure 8. On May 20, 2013, a two-mile wide EF5 tornado struck Moore, Oklahoma, killing 24 people and leaving behind massive damage to homes and businesses. Image credit: Tom Pennington/Getty Images.|
There’s another type of tempest where researchers found that it didn’t take as much wind as once thought to produce a given amount of damage: tornadoes. Back in the mid-20th century, some experts thought that tornadic winds might howl at speeds up to 600 mph. In the original Fujita scale, the topmost rating of F5 corresponded to damage believed to be caused by top wind gusts of 261-318 mph. Eventually, comprehensive engineering and damage analyses made it increasingly clear that F5 damage could be produced by weaker winds. After a thorough revision process involving meteorologists and engineers, the Enhanced Fujita Scale was adopted by the National Weather Service in 2007. The EF5 damage rating for damage on par with the previous F5 rating now corresponds to peak wind gusts greater than 200 mph, with no upper bound.
A tornado does not get its official rating until days after it strikes. Hurricane forecasters have no such luxury: they must rate the intensity of a hurricane before it makes landfall, based on the best data and best science at hand. If there’s any silver lining to the horrific 2017 hurricane season, perhaps it’s the light it could shed on how estimated winds at landfall relate to surface wind measurements and damage, and how these relationships are best characterized for those whose lives and livelihoods are at risk each hurricane season.
Part I of this two-part post was published on Wednesday, April 11.