Inland Damage from Hurricanes

In most cases, hurricanes deteriorate as they move inland away from oceanic sources of moisture or into the cooler waters north of the Gulf Stream. However, some hurricanes re-intensify as a result of topographic and/or meteorological factors in inland areas. The purpose of this paper is to describe the various conditions that can cause hurricane damage inland. The examples presented in this paper will enable people to better predict the inland behavior of some hurricanes.

This paper presents five different possible causes of hurricane re-intensification. The specific hurricanes considered here are Tropical Storm Agnes (1969), Hurricane Iniki (1992), Hurricane Camille (1969), Hurricane Hugo (1989), and the New England Hurricane of 1938.

Numerous studies have been carried out and published on the winds, storm surges, and damage of these storms at landfall in their coastal areas. This information is not repeated in this study. This paper deals only with the inland conditions in each of these storms. The references are also specific to the inland areas.

Five distinct conditions lead to re-intensification of hurricane damage inland:

1. A deteriorating hurricane may merge with a moist extratropical low-pressure system, causing massive rainfall and river flooding (Tropical Storm Agnes in Pennsylvania, 1969).

2. Winds can be channeled through passes on mountainous islands, like Kauai, to cause massive destruction on the lee side (Hurricane Iniki, 1992).

3. Mountains can induce orographic precipitation that can result in massive debris flows (Hurricane Camille in Nelson County, Virginia, 1969).

4. A decaying hurricane can have high convective centers inland that result in localized damage more typical of the hurricane at landfall (Hurricane Hugo, 1989).

5. Northern hurricanes can encounter polar air masses on their left sides as they move inland in the early fall (New England Hurricane of 1938).

Late on the afternoon of 19 June 1972, Hurricane Agnes crossed the northwest Florida coast. It rapidly weakened to a tropical depression and then turned into a tropical storm as it neared coastal waters in the Carolinas and Virginia. Rainfall intensity increased rapidly as tropical moisture streamed inland from the Atlantic Ocean. By the morning of 21 June, reports from the Virginia mountain areas recorded 10 in. (25 cm) of rain (National Weather Service, 2019). The storm moved northward into Pennsylvania. It had been blocked from turning eastward into the Atlantic by a high-pressure system (Figure 1). It then turned northward and then northeastward and exited the United States (Figure 1).

From the Carolinas northward, Hurricane Agnes was in contact with a midlatitude cyclone (National Weather Service, 2019) on its west side. The collision of these two low-pressure systems was to result in exceptional rainfall in Maryland, southwestern New York, and especially Pennsylvania (Figure 2). On 22 June, the James River at Richmond, Virginia, exceeded the flood levels set by Hurricane Camille in 1969. With a flood stage of 9 ft (2.7 m), the James River reached a flood crest of 36.5 ft (12.1 m). The previous high had been 28.6 ft (8.5 m) (U.S. Department of Commerce, 1973). Flood levels continued to rise, and the state capital at Harrisburg, Pennsylvania, was especially inundated on 24 June (Figure 3). Flood stage of the Susquehanna River at Harrisburg is 17 ft (5.2 m). On 24 June, the flood crest was 32.6 ft (9.9 m) (U.S. Department of Commerce, 1973).

Hurricane Agnes was noteworthy in that it caused contemporaneous flooding in a number of river basins, including the James, Potomac, Schuylkill, Susquehanna, Genesee, and Upper Ohio–Alleghany/Monongahela Rivers. Tropical Storm Agnes had a profound effect on America's largest estuary, Chesapeake Bay. A detailed discussion is available in U.S. Army Corps of Engineers (1973). The Susquehanna River discharge reached 1130 ft³/s (1919 m³/s), which is the greatest level in 185 years. A volume equivalent to 10 to 50 years of sediment was deposited in a few days. A considerable volume of raw and partially treated sewage was washed into Chesapeake Bay along with agricultural nutrients. The influx of nutrients started phytoplankton blooms, causing “mahogany tides.” Decreases in salinity also affected oyster and clam populations.

Hurricanes are common in the eastern Pacific, but they are far less common in the Hawaiian Islands. Only four hurricanes have impacted the islands since 1950. Hurricane Iniki was by far the strongest and most destructive. On 11 September 1992, Iniki was located just south of the island of Kauai moving northward at 30 mi/h (48 km/h) with winds of 145 mi/h (233 km/h) and gusts up to 175 mi/h (281 km/h) (U.S. Department of Commerce, 1993). It made a landfall at Port Allen (Figure 4).

In most coastal landfalling hurricanes, storm surge is a strong component of the water column. However, Hurricane Iniki had storm surge values of only 2–4 ft (0.6–1.2 m). Waves did most of the damage along the coast. The steep offshore slope around Kauai inhibits strong storm surge development, but it favors development of high waves. Mapped debris lines marked the furthest intrusion of waves at 10–18.5 ft (3–5.6 m) (U.S. Army Corps of Engineers, 1993).

The mountainous island of Kauai is a dissected volcano with numerous east-facing slopes and long valleys (Figure 4). Winds reached 130–160 mi/h (209–257 km/h), primarily in areas where winds were channeled through valleys (U.S. Department of Commerce, 1993).

The following analysis of wind dynamics is based largely on the research of Ted Fujita (Fujita, 1992), who produced detailed maps of the first (front of the eye) and second (back of the eye) wind fields on the island during the passage of Hurricane Iniki (Figure 5). Fujita (1992) stated that the wind field was very chaotic, and he identified 26 microbursts (Figure 6). In addition, his highly detailed maps show extensive interaction between the winds and the rugged topography of northern Kauai (Figure 5).

Normally, hurricane wind effects are more severe on the front side of an island, where a hurricane first makes landfall. In the case of Hurricane Iniki, the irregular topography inland (Figure 4) created more damage on the lee side. Buildings at the top of the island in Princeville were heavily damaged (Figure 7). The northwest (left) section of Fujita's wind map of Hawaii (Figure 5) details the interaction of Hurricane Iniki's winds with Kauai's rugged topography. Note that both the first and second wind fields were channeled down the valleys in the left part of the map (Figure 5).

A working hypothesis for the damage on the lee side of Kauai is given in Figure 8. As the first and second wind fields streamed into the valleys, wind speed was increased by the Bernoulli effect (Figure 8A). The damage at the top of the cliffs was a combination of compressed air flow at the top and the upslope return flow on the north side (Figure 8B). The wind damage seen in Kauai was similar to that observed in Hurricane Marylin (National Hurricane Center, 1996, updated version). Marylin caused severe wind damage on St. Thomas in the U.S. Virgin Islands. St. Thomas is a mountainous island, with a steep offshore slope similar to Kauai. Hillsides were littered with sheets of metal roofing, wood planks, and household debris (National Hurricane Center, 1996). The mechanism for the wind destruction on St. Thomas was probably the same as that proposed for Kauai (Figure 8).

Hurricane Camille made a landfall in Mississippi on 17 August 1969. Emanuel (2005a) described the intensification of the storm and the conditions at landfall. At the coast, damage from surge and wind was monumental, but the rainfall was only 5 in. (12.7 cm) (Figure 9). The hurricane then moved northward through the lower Midwest and then turned eastward towards the coast. As it crossed Virginia and West Virginia, on the night of 19 August 1969, rainfall reached a maximum of five times what it was at landfall (Figure 9).

This rainfall caused extensive mass movements and flooding in local stream basins (Figure 10). Thompson (1969) stated that, for Virginia, the amount of rainfall associated with this storm occurs, on average, only once in more than 1000 years. The damage was described in reports of the Virginia Department of Mines, Minerals, and Energy (1969a,b). The State of Virginia is now conducting surveys of the devastated areas to better understand the potential for slope failure in future storms (Virginia Department of Mines, Minerals, and Energy, 2019).

Debris avalanche chutes and fans (Williams and Guy, 1971) formed on the mountain slopes (Figure 10a). The size of the boulders in these flows indicates a high flow competence (Figure 10b). In Nelson County, Virginia, alone (Figure 10c), 150 homes and other buildings, 120 mi (193 km) of roads, 150 bridges and culverts, hundreds of cars and trucks, and 25,000 acres (101,175,000 m²) of cropland were destroyed (Williams and Guy, 1971). Emanuel (2005a) noted that the death toll was so high because most residents were asleep during the unpredicted storm. Even if forecasters had realized what was happening, there was no way to warn anyone because phone lines had been destroyed by the numerous flash floods.

A very detailed discussion of erosion, depositional features, and the characteristics of debris deposits was given by Williams and Guy (1971). They stated that 6 million cubic feet (169,800 m³) of sediment were eroded and deposited elsewhere. Avalanches flowed down existing chutes. Slopes facing north, northeast, and east had the greatest number of these debris avalanches.

The rainfall caused massive stream flooding as well. According to Camp and Miller (1970), the James River discharge at Richmond, Virginia, reached 222,000 ft³/s (377,178 m³/s). Camp and Miller (1970) considered that to be the second highest discharge on the James River since Jamestown was settled in 1607. According to Williams and Guy (1971), discharges as large as the 1969 flood at Richmond (Figure 10d) occur there about once every 180 years.

What caused such a major increase in rainfall over a thousand miles (1600 km) inland from hurricane landfall? Schwarz (1970) noted that Camille's maximum rainfall was within about 80%–85% of the maximum possible rainfall, for areas up to 1000 square miles (2,590 km²) over a 12 h period. Williams and Guy (1971) described the reasons for the excessive rainfall, based on research by Schwarz (1970). The quote below is from Williams and Guy (1971):

What contribution could orographic rainfall have contributed as Camille passed over the Appalachian and Blue Ridge Mountains? Williams and Guy (1971) believed that in some places, orography may have contributed to the intense rainfall by funneling the NW-flowing air into the mountain ridges. However, they concluded that the effect of orography on the rainfall is difficult to determine. The interpretation preferred in this paper is that the orographic effect must have been considerable as Hurricane Camille moved from the low elevations of the Mississippi Valley to flow over the Appalachian and Blue Ridge Mountains.

Hurricane Hugo, a category 4 hurricane, made a landfall on the South Carolina coast at Sullivan's Island around 0400 h on 22 September 1969. Conditions in Charleston at 0340 h were 1 min averaged winds of 76 mi/h (122 km/h), peak gusts of 94 mi/h (151 km/h), storm surge of 8.0 ft (2.4 m), and a total rainfall of 5.90 in. (14.7 cm) (National Hurricane Center, 1989).

One hundred miles (161 km) inland, Victor Jones, director of public safety for Sumter County (Figure 11), was making preparations for his county as well as provisions for the inevitable evacuees from coastal counties. Two hours later, Victor Jones was faced with massive wind damage in his county, far inland from Hugo's landfall. Almost 2 days would pass before people realized the major damage in inland Sumter County. All media attention was focused on coastal Carolina and the city of Charleston, because people believed “that hurricanes were just coastal events” (Victor Jones, personal communication, 1993).

National Weather Service radar at Charleston tracked Hugo's evolution as it moved inland (Figure 12). Strong areas of convection (reflectivity of 34–50 dB) were still occurring in the rain bands as the storm moved far inland. The radar picture as the storm was centered over Sumter County is shown in Figure 13. This radar image reveals that the eastern two-thirds of the county were showing very high reflectivity values of 26–50 dB. The weather station at nearby Shaw Air Force Base recorded sustained winds of 105 mi/h (169 km/h) (Sumter Item, 10 October 2019). The wind speed in Sumter County exceeded that in the landfall of Hugo in Charleston!

While Hugo's mean winds had subsided to near hurricane force in Sumter County, the periodic 10 min gusts caused extreme damage. The 10 min gust factor (Krayer and Marshall, 1991) approached a value of 2.0 because of the presence of the extremely high convective rain-band features (Figure 13) during eyewall passage (Powell, Dodge, and Black, 1991). Hugo's mean winds had subsided to below hurricane force by 93 mi (150 km) inland in Sumter County (Figure 11); however, maximum gust speeds were still above hurricane force as far as Hickory, North Carolina, 217 mi (350 km) inland along the hurricane track (Powell, Dodge, and Black, 1991). It is unfortunate that strong convection bands (Figure 13) were present at the same time that Hugo passed over eastern Sumter County. The Sumter County experience shows that, even in a decaying hurricane, local high convection centers can cause significant damage. Tree damage was extensive (Sheffield and Thompson, 1992). There was greater damage on the right side of Hugo's eyewall and great damage inland in Sumter County (Sheffield and Thompson, 1992). Treefall and uprooting were major causes of structural destruction and loss of gas, water, and sewage services.

The people in Sumter County were largely on their own for almost 2 days after Hurricane Hugo hit. The damage was extensive (Figure 14). Armed South Carolina National Guardsmen were sent to Sumter to maintain order. They soon put away their weapons and took up chainsaws, hammers, and nails to help get Sumter County back on its feet (South Carolina Palmetto Guard, 1990). South Carolina Governor Caroll Campbell flew over the area, and the American media soon realized that the damage in Sumter County was severe. Help soon poured in (The State, 1990). A major problem was in donated goods: Where could they be stored? What were priority items? How could they be distributed with tree-cluttered roads? There was an absence of predonation planning.

Sumter County was clearly a major disaster area. Eighty percent of the roads were impassable as a result of fallen trees. Surprisingly, the phones worked, although the lines were down. This ended when people started to drive over the lines and cut them. The massive tree uprooting (Figure 14C) tore up water, sewage, and gas lines. There was no gas, sewage, or water service for weeks as a result of tree uprooting. Home damage was extensive, with 2112 homes destroyed, 3946 homes with major damage, and 7110 homes with minor damage. Structural damage included a combination of roof stripping (Figure 14A,B), uprooted trees falling on structures (Figure 14C), and gable-end failure on homes (Figure 14D). Business damage in Sumter County alone (Figure 14A,B) was 25 million dollars. Restoration workers, media, and charities made the recovery effort more difficult. The area was inundated with out of town people coming to help. This created serious logistical problems. These people created shortages in food, lodging, fuel, rental cars, etc., without prior planning (U.S. Army Corps of Engineers, 1990).

Hurricane Hugo changed perceptions of hurricane damage and how to plan effective recovery after the event. It established the fact that some hurricanes can cause great damage inland. As soon as possible, affected communities must plan with relief agencies so needed work and supplies can be prioritized. Finally, relief workers must minimize their effect on devastated communities by providing their own housing, food, equipment, and vehicles.

On 4 September 1938, a French observer noted a gentle wind shift and an easterly wave passing over an oasis in the Sahara Desert. Seventeen days later, a high category 3 hurricane made a landfall on Long Island, New York. The storm then proceeded to devastate parts of six states as well as southeastern Canada. This event is still considered the greatest natural disaster to occur across New England.

Emanuel (2005b) presented a good summary of the meteorological changes with increasing latitude and the problems the National Weather Bureau had in accurately determining the landfall and intensity of the storm. The storm made a landfall with such ferocity that it was recorded on seismographs at Fordham University in New York City (Figure 15) and at other seismological stations. Brian Jarvinen, SLOSH expert (retired) at the National Hurricane Center, stated that the 1938 hurricane could have been a borderline category 4 storm just before landfall (Brian Jarvinen, personal communication, 1983).

Damage was reported at many locations in New York and New England (Figure 16). Clover (1939) reported extensive tree damage in eastern Long Island. Each major damage location is indicated on Figure 16A. Damage extended from New York City east to Montauk Point (Figure 16B). Homes on Fire Island, New York, were largely destroyed (Figure 16C). Barrier islands were swept clean of homes, and new tidal inlets formed in the Hamptons, New York (Figure 16D). Record flooding occurred on New England streams (Figure 16E). The waters of Narragansett Bay were pushed into Providence, Rhode Island, by the afternoon rush hour (Figure 16F). Ludlum (1988) reported that the tide reached 14 ft (4.2 m) above mean water. This was 2 ft (0.6 m) higher than the water level reached in the 1815 hurricane. Coastal homes in Connecticut, Massachusetts, and Rhode Island were totally destroyed (Figure 16G). Brown (1939) described the coastal damage in Rhode Island. Massive waves riding on a surge of 10–13 ft (3–4 m) caused cliff retreat of 36 ft (11 m) at Watch Hill, Rhode Island. Cliff recession of 27 ft (9 m) occurred in the cliffs of Block Island, Rhode Island.

The most exceptional regional damage was to trees all across New England and into southeast Canada (Figure 16H). For this paper, photos of tree damage across New England were examined in the archives of the White Mountain National Forest in Laconia, New Hampshire. It was surprising to see that there was little decrease in tree damage from Long Island across New England. Figure 16H shows the damage in Keene, New Hampshire, not far from the Canadian border.

This tree destruction was a function both of the wind strength and persistence, as well as the antecedent rainfall. In addition, the trees were still in total bloom in late September and offered maximum resistance to the winds. A detailed picture of tree damage in the 1938 Hurricane was given by Wessels (1997). He stated that gusts on ridge tops, slopes with SW orientations, and gaps that funneled the wind suffered the most severe blowdowns. Minsinger (1988) reported that the Blue Hill Observatory, just south of Boston, recorded sustained winds of 120 mi/h (193 km/h) for 5 min around 1800 h. Emanuel (2005b) reported that subsequent examination of the record showed wind gusts as high as 186 mi/h (299 km/h) were estimated to have occurred at that site.

From 18 to 21 September, New England had experienced unprecedented heavy rain. This rain caused massive flooding in several towns of Connecticut and Massachusetts (Colton, 1939). The hurricane itself added another 4–6 in. (10.5–15.5 cm) (Emanuel, 2005b). Brooks (1939) mapped the rainfall from the storm. This excessive rainfall softened the soil and made it easy for the strong winds to break and topple trees. As many as 250 million trees were downed in a few hours. Sea salt, thrown up by the surf, blew far inland as far as 20 mi (32 km) (Emanuel, 2005a,b).

The rugged topography of New England maximized rainfall in the 1938 hurricane. Coch (2012) showed how topography increased rainfall in southern New York and New England during Hurricane Irene (2011). As warm moist air was forced upward over the topographic barriers, it cooled at upper altitudes. This moisture in the air condensed and fell as rain. The rainfall map of Brooks (1939) shows higher rainfalls over the Berkshire Mountains of Massachusetts and the White Mountains of New Hampshire in the 1938 hurricane.

The quantitative relation between rainfall and topography is hard to define. However, the recent research by Cole and Yuter (2007) casts some light on the effect of topography on storm rainfall. They analyzed radar patterns to study precipitation as a tropical storm passed over the glacial hills of western Long Island, New York. The hills (a glacial moraine) average about 320 ft (100 m) high. They concluded that the precipitation increased by 30%–40% as the storm passed over the hills.

What were the meteorological conditions that led to the extraordinary regional damage in the 1938 hurricane? The air mass positions on 21 September 1938 are shown in Figure 17. Pierce (1939) showed that there was a continental high to the west and a semipermanent Bermuda High on the east. The normal position of the Bermuda High in September is between 30 and 35° N. However in September 1938, the Bermuda High was located southwest of Newfoundland at a latitude of 44° N. This abnormally high location blocked any normal recurving of the hurricane to the east and into the Atlantic Ocean and set it on a path across New England and into southeastern Canada (Wagner, 1988).

The 1938 hurricane moved north into the low-pressure trough between the two high-pressure ridges. The forward speed of the storm increased as the hurricane moved northward. This is because the hurricane was confined by the high-pressure masses on either side and was steered northward by the upper-level winds. This was the weather setup that made the 1938 hurricane the fastest moving hurricane in history (Pierce, 1939). Howard (1939) stated that the strong wind gusts along the South Shore of Long Island at hurricane landfall allowed storm surge and superposed waves to breach higher portions of the dunes on the barrier islands (Figure 10D).

New England experiences most of its hurricanes in September and October. At this time, there are occasional intrusions of polar air into the region. This is what happened on 21 September 1938, and it greatly energized the wind field. On that day, a polar air mass had moved south and was colliding with the tropical air in the hurricane to the east (Figure 18). The change in temperature was recorded in Amherst, Massachusetts, as the hurricane passed over (Figure 19).

Pierce (1939) stated that the temperature difference of 11 °C (20 °F) between the two air masses (Figure 18) energized the wind field and increased the translational velocity of the hurricane. The 1938 hurricane was moving at 112 km/h (70 mi/h) when it was 100 mi (161 km) east-southeast of New York City (Pierce, 1939). Tannehill (1956) believed that the speed was slower and that the storm had a forward velocity of 56 mi/h (90 km/h) as it crossed the Long Island shoreline. Either of these extremely high translational velocities can account for the extreme wind damage of the storm, especially on its right side. The forward speed of the 1938 hurricane is shown on Figure 20. The plot shows that the hurricane was moving forward at a speed of 55 mi/h (88 km/h) over the rugged Green and White Mountains of New England.

Pierce (1939) explained the role of the temperature difference across the wind field. Even though the moisture content was reduced as the storm moved inland, the potential energy of the air mass distribution overcame some of the frictional effects (topography and vegetation) and maintained the energy of the storm. These temperature contrasts intensified the wind field and enabled the storm to penetrate further inland at a higher speed than expected. As a result, the hurricane maintained a distinct eye into northern Vermont.

After the 1938 hurricane, the U.S. government sponsored a study of the storm by the Federal Writers Project, Works Progress Administration in the New England States (1938). The following quote from their report (pp. 219–220) gives their conclusions:

The Federal Writer's suggestions were good ones. However, how many have been implemented on our coasts today? When will we learn the lessons of the past?

The writer's suggestions in the last two sentences of the quote were to prove a serious problem in the future. After the storm, coastal engineering structures were seen as a way to increase beach width and stabilize storm-cut inlets (U.S. Army Beach Erosion Board, 1946). Hall (1939) recommended a combination of groins tied to a seawall to “build up and preserve the protective beach in front.” At that time, these suggestions seemed reasonable.

However, with time, these hardened structures would trap sand and increase downdrift erosion along the South Shore of Long Island (Coch, 2009). Terchunian and Merkert (1995) described the opening and closing of barrier breaches that resulted from groin construction at Westhampton Beach, New York. McCormick (1973) concluded that the coast between Moriches and Shinnecock inlets had receded at an accelerated rate in response to the tidal action at the stabilized inlets. Coch (2015) summarized all the factors that demonstrate that the Long Island shoreline today is in worse shape than it was before the 1938 hurricane.

While inland intensification is not common, it can occur under the specific conditions described in this paper. Although the frequency of inland intensification is low, the consequences can be very high. It is therefore important to consider the possibility of inland intensification in every future hurricane.

Many individuals helped to obtain information on the historic hurricanes described in this paper. These organizations included: U.S. Geological Survey Information Service, Smithsonian Institution Archive Center, and the Virginia Department of Mines, Minerals, and Energy. Fred Wolff provided valuable help in a reading of a preliminary manuscript. John Theret was of considerable help with the computer graphics.

The three Hurricane Hugo field studies were aided by the information and logistical support provided by Paul Gayes of the Center for Marine and Wetland Studies of Coastal Carolina University. Mark Powell, Hurricane Research Division, National Oceanic and Atmospheric Administration (NOAA), supplied the Hugo radar images used in this paper. Special thanks go to Victor Jones, director of public safety, Sumter County (retired), who provided a great number of records and photographs about Hurricane Hugo's impact in Sumter County and the problems in recovery.

Mel Nishahara, hurricane program manager for the State of Hawaii, set up meetings with emergency managers on Oahu and Kauai and briefed the writer on Hurricane Iniki damage. Glen Trapp, National Weather Service, Oahu, provided Iniki satellite images and preliminary reports of Hurricane Iniki damage. Tom Batey, administrator of Kauai County, suggested damage study sites during two research tours in Kauai. He graciously discussed the results and interpretations of the observations after each visit.

The author's decades-long study of the 1938 hurricane has been aided by a great number of people and organizations. The staff of the White Mountains National Forest in Laconia, New Hampshire, provided archive photographs of tree damage across New England. Mike Wyllie, Toni Gigi, and Harvey Thurm, U.S. Weather Service, provided information on the 1938 hurricane. Dr. William Gray, Colorado State University, provided information and interpretations on the 1938 hurricane on numerous occasions. The author is collectively grateful to the eight survivors who provided a first-hand view of the storm experience. Thanks also go to the several people who provided archival photos of storm damage.

Special thanks go to the weather experts in Miami and Colorado. These include Robert Sheets, Max Mayfield, and Brian Jarvinen, National Hurricane Center, and John Willoughby and Chris Landsea, Hurricane Research Division, NOAA, and William Gray, Colorado State University, for their assistance over two decades. They graciously provided information and explanations during many visits to their facilities.