2017 Hurricane Season: Review and Analysis

KCC WHITE PAPER 2017 Hurricane Season: Review and Analysis

JUNE 2018 KCC White Paper: 2017 Hurricane Season: Review and Analysis

RiskInsight® is a registered trademark of Karen Clark & Company.

Document Date

June 2018

Contact Information

If you have any questions regarding this document please contact:

Karen Clark & Company 116 Huntington Avenue Boston, MA 02116 T: 617.423.2800 F: 617.423.2808 [email protected]

Contents Executive Summary ...... 3 Meteorological Highlights and Forecasts ...... 3 Modeled Loss Estimates...... 5 Hurricane Harvey ...... 6 Meteorological Development ...... 6 Real Time Loss Estimation ...... 7 Post-Event Damage Survey...... 9 Residential Structures ...... 9 Commercial structures ...... 12 Hurricane Irma ...... 16 Meteorological Development ...... 16 Real Time Loss Estimation ...... 19 Post-Event Damage Survey...... 20 Residential Structures ...... 21 How much does it cost? ...... 24 Commercial Structures ...... 26 Storm Surge Damage...... 27 Damage Observations from the Caribbean ...... 29 Hurricane Maria ...... 31 Meteorological Development ...... 31 Real Time Loss Estimation ...... 32 Post-Event Damage Survey...... 33 Residential Structures ...... 33 Commercial Structures ...... 35 Hurricane Nate ...... 42 Meteorological Development ...... 42 Real Time Loss Estimation ...... 43 Conclusions ...... 44

Executive Summary

The 2017 hurricane season was a notable one in many respects. From a meteorological perspective, several storms exhibited extreme values for intensity, duration, and/or rainfall amounts. Several islands in the Caribbean experienced their worst losses in decades, and there is still uncertainty around the final insured losses. Perhaps most notable was the wide discrepancy in the catastrophe model loss estimates for Harvey, Irma, and Maria (HIM).

Meteorological Highlights and Forecasts

. 2017 was the most active hurricane season since 2005 based on the Accumulated Cyclonic Energy (ACE) index, which is a measure of seasonal intensity and duration. Ten hurricanes, six of which became major hurricanes, formed in the Atlantic basin.

. Two major hurricanes made landfall in the US, ending a 13 year drought of major landfalling storms.

. Every state along the Gulf Coast was impacted by at least one hurricane. . Two Category 5 hurricanes impacted Puerto Rico and other Caribbean Islands. . Multiple storms, including Jose, Irma, and Maria, exhibited extreme wind speeds and/or duration.

Named storms of the 2017 North Atlantic hurricane season.

In 2017, sea surface temperatures (SSTs) throughout the Atlantic were above average by 0.5-1.0°C. In addition, in the region most conducive to Atlantic hurricane genesis, SSTs were 0.23°C higher than the rest of the tropics.

In May, NOAA forecast a 45 percent probability of an above-average season and a 35 percent probability of a near-average season. El Niño conditions had been anticipated to develop as the season progressed. This would have introduced characteristics, such as increased wind shear and lower SSTs, that would have inhibited storm formation. However, El Niño conditions did not materialize.

In August, the outlook was updated to indicate a 60 percent chance of an above- average season. The minimal wind shear over the North Atlantic due to weak trade winds combined with the African Easterly Jet allowed tropical disturbances off the west African coast to gain cyclonic vorticity and increased the likelihood of tropical storm formation. By this time, the 2017 hurricanes were expected to be larger, longer lasting, and more numerous than average seasons.

The 2017 North Atlantic hurricane season was characterized by fewer tropical cyclones than projected but there were more hurricanes and major hurricanes than forecast.

2017 Hurricane Season Statistics

20 18 16 14 12 10 8 6 4 2 0 Tropical Storms Hurricanes Major Hurricanes

NOAA May Forecast NOAA August Forecast 2017 Season 30 Year Average (1980‐2011)

Comparison of NOAA forecasts, the 2017 hurricane season, and the 30 year average.

Modeled Loss Estimates

Perhaps the biggest surprise of the 2017 hurricane season was the wide disparity between the catastrophe model loss estimates. The chart below shows the three primary modelers’ industry loss estimates for HIM. These numbers are for the US only for Harvey and Irma, and they exclude NFIP losses.

Harvey (US) Irma (US) Maria (PR) 80 72 70

? 50

40 35 $ Billions 33 32 30 34 26 25 28 20 20 18 15 10 12 13 >10

0 KCC (9/1) AIR (9/6) RMS (9/9) KCC (9/13) AIR (9/15) RMS (9/20) KCC (9/28) AIR (9/25) RMS (9/29) Comparison of modeled loss estimates for Hurricanes Harvey, Irma, and Maria.

Several things are notable about the industry loss estimates.

. KCC provided a best estimate for each storm; other modelers provided wide ranges.

. The actual losses aligned well with the KCC projected loss estimates for the industry as a whole and for individual insurers. KCC estimated that the privately insured loss, including the US and Caribbean, would total to $70 billion for the 2017 hurricane season.

Hurricane Estimated Industry Loss ($ Billions) Harvey 15 Irma 25 Maria 30 Nate 0.5 Total 70.5

Hurricane Harvey August 17-September 3

. First major landfalling US hurricane since Hurricane Charley (2004).

. Made landfall near Rockport, Texas, as a Category 4 hurricane with 132 mph winds.

. The storm’s slow motion, eventual stall, and location near the Gulf of Mexico led to historic flooding in and around Houston with rainfall in excess of 50 inches.

Meteorological Development

After forming as a tropical depression in the Atlantic, Harvey traveled westward south of the Caribbean and across the Yucatan Peninsula. The storm rapidly intensified over the Gulf of Mexico due to warm SSTs. Prior to making landfall near Rockport, Texas, Harvey became a Category 4 storm with 132 mph sustained wind speeds.

KCC inland flood footprint for Hurricane Harvey generated on September 1.

Hurricane Harvey’s track through the Caribbean Sea and Gulf of Mexico.

High pressure regions over the US and the Atlantic caused Harvey to stall near Houston. Much of the storm’s circulation remained over the Gulf of Mexico, which prolonged the storm’s duration and provided a constant source of moisture for precipitation. Rainfall in excess of 50 inches from August 25-29 produced historic flooding in Houston and the surrounding areas.

After four days over Texas, Harvey slowly moved eastward and back out to the Gulf as a tropical storm before turning northeastward and making a second landfall near Lake Charles, Louisiana. Harvey finally dissipated on September 3.

Real Time Loss Estimation

KCC’s high resolution models are used by insurers and reinsurers to estimate losses for storms as they unfold. KCC clients were able to update their loss estimates multiple times a day starting several days before Harvey’s initial landfall in Texas. On September 1, before Harvey had totally dissipated, the KCC industry loss estimates were released publicly.

The first step in real time loss estimation is producing accurate intensity footprints. For most hurricanes, wind and storm surge footprints are sufficient, but for Harvey it soon became evident that an inland flood footprint would also be required. The images below show the wind, storm surge, and inland flood footprints created by KCC high resolution models for Hurricane Harvey in real time based on the final National Hurricane Center (NHC) advisories.

KCC’s real time wind (left), inland flood (center), and storm surge (right) footprints for Hurricane Harvey.

Once the high resolution footprints are created, insurers use the KCC models to estimate their numbers of claims, average claim severity, and total losses by intensity band. The KCC industry losses are shown in the table below.

KCC Breakdown of Hurricane Harvey Insured Loss Provided to Industry on September 1 Commercial and Hazard Residential Auto Total Industrial Wind 1.4 0.9 0.2 2.5 Storm Surge 0.2 0.3 0.0 0.5 Inland Flooding* 1.1** 8.1 3.2 12.4 Total 2.7 9.3 3.4 15.4

*These estimates do not include the NFIP. **Assumed “leakage” on wind policies.

These estimates include losses to buildings and other insured structures, contents, business interruption, and automobiles. They exclude crop damage and NFIP losses. The majority of losses from Hurricane Harvey were due to its record breaking flooding in and around the Houston.

Post-Event Damage Survey

KCC engineers conducted post-event damage surveys along the central Texas coast shortly after Harvey dissipated. The major findings for Harvey were:

. Significant wind damage was isolated to areas of Category 3 and 4 winds near Aransas Pass and Rockport.

. Structural damage was more frequently observed in masonry and metal constructions than wood frame buildings, and light metal industrial buildings sustained the most damage.

. Harvey observations confirmed that newer construction generally performs much better than older buildings.

Residential Structures

Harvey’s wind speeds decreased rapidly after landfall, so only the properties right along the coast experienced Category 3 and 4 winds. In this wind speed region, there were instances of severe structural damage to single family homes. Most of the residential damage was to roofs and siding, particularly for areas away from the coast.

The dominant roof covering in this area is asphalt shingle. Damage to asphalt shingle roofs typically begins at the corners of buildings and propagates upwards. As soon as one shingle peels away, the next becomes more vulnerable to detachment. An estimated forty percent of shingle roofs sustained damage in areas of Category 3 and 4 wind speeds.

Newer homes performed significantly better than older homes. An example of this is shown below. These two rows of similar wood frame buildings with shingle roofs were exposed to the same wind speeds. The top row of homes was constructed in 2006 and exhibit the classic damage pattern, which will require complete replacement of the roof covering. The homes in the bottom row were constructed in 2015 and did not suffer any observable roof covering damage.

Two rows of wood frame houses each experiencing the same wind speeds demonstrate relative vulnerabilities between older (top) and newer (bottom) roofs.

Siding damage on single family homes occurs less frequently than roof covering damage due to the differences in the wind pressures exerted on the different elements. Limited siding damage was observed from the Category 3 and 4 wind speeds right along the coast.

The damage that was observed, such as that shown below, was usually as a result of poor installation practices. The outer layer vinyl siding on this home was installed over an older existing layer of wood shingle siding, creating weak attachments between the walls and siding.

Poor installation increases the wind vulnerability of residential siding.

Volatility in wind speed and loading and uncertainty in the resistance of building components results in similar buildings experiencing varying levels of damage. The three residential buildings shown below are the same age and are constructed similarly, including roof height and covering. The building to the left experienced gable-end wall damage, the middle building experienced minimal shingle damage, and the building to the right experienced severe damage to its roof structure. This variability is interesting to observe and is captured in the KCC Hurricane Reference Model’s financial module.

Three very similar buildings sustained different levels of damage when exposed to similar wind speeds, demonstrating the importance of accounting for secondary uncertainty.

Mobile homes are highly vulnerable to hurricane force winds. Due to their light weight and lack of foundation, they can be overturned and their envelope elements compromised by high winds. The use of tie-downs increases stability and prevents mobile homes from overturning. Most mobile homes observed in the region were not tied down and sustained severe damage. Many were total losses.

Mobile homes sustained severe damage that could have been mitigated by hurricane tie-downs.

Commercial structures

Right along the coast in the highest wind speed region, commercial structures sustained severe damage. Total building collapse and structural damage were particularly common in older buildings. The apartment building complex shown below, located in Rockport, was built in 1985 and constructed using wood frame with brick veneer siding. The building did not have hurricane straps on the roof-to-wall connections or hurricane shutters. As a result, when the windows were breached causing internal pressurization, the building failed.

Lack of hurricane prevention measures, such as hurricane straps or hurricane shutters, and age of structure contributed to structural collapse.

Total structural failure occurred to the upwind units of each apartment block, while the other units suffered roof deck and gable end wall failures. The fallen “Wrong Way” sign indicates the direction of the wind.

Upwind units (identified by red circles) sustained total structural failure.

The importance of roof-to-wall connection in maintaining the structural integrity of a building cannot be overemphasized. The photo below shows before and after images of a building in downtown Rockport. The entire roof, including the covering and decking, was displaced from the main structure which led to the collapse of the building. Weak roof-to-wall connections can lead to catastrophic failure of the structure, and the use of hurricane straps, specifically double wraps, would have significantly decreased the probability of collapse.

Source: Google

Before and after photos of an art gallery in Rockport that collapsed due to weak roof-to-wall connections.

Damage to masonry walls of commercial buildings was frequently observed in multi- story buildings. In the example below, a two-story unreinforced masonry building with a low slope gable roof, the roof structure is intact, but a full section of wall has collapsed. The reason for this failure is weak connections between the wall elements that were unable to withstand the wind pressure. Older masonry structures are particularly vulnerable due to cement losing attachment properties with age. In general, such wind damage to masonry structures is not expected but was consistently observed in the high wind areas in Texas.

Unexpected masonry damage was consistently observed in Category 3 and 4 wind speed areas.

Structures with a steel frame and a light metal envelope typically experience damage when the metal siding peels away from the underlying frame.

Lightweight steel and metal structures can serve a variety of purposes from a market (top), school gymnasium (bottom left), warehouse (bottom center), and storage facility (bottom right), and all were damaged from wind speeds of Category 1 and above.

Damage to industrial properties typically results from specialized equipment on the site. Industrial facilities that are made of lightweight material, such as metal, are also easily damaged by hurricane force winds.

Structural damage to lightweight scaffolds and grain elevators (left) and damage to aluminum roofs on silos (right) can occur even at Category 1 wind speeds.

Aerodynamically, gable end walls experience high wind pressure. As a result, damage to gable end walls and their siding is commonly observed. A safe construction practice is to provide additional bracing to gable end trusses and additional reinforcement to the sidings.

Damage to gable end siding due to high wind pressures frequently located at gable ends

In general, Hurricane Harvey exposed weaknesses in the Texas building code and quality of constuction, particularly in older commercial structures. Fortunately, Harvey’s Category 3 and 4 wind speeds were limited to areas right along the coast. Storm surge damage was severe near the the landfall point and in localizd areas. The majority of the insured loss resulted from inland flooding to commercial and industrial properties.

Hurricane Irma August 30-September 11

. Maintained peak wind speed of 185 mph for a record breaking 37 hours throughout much of the storm’s track

. Passed directly over Barbuda, St. Martin, and the British Virgin Islands as a Category 5 hurricane

. Strongest hurricane to impact the Bahamas since Hurricane Andrew (1992)

. Two Florida landfalls: Cudjoe Key (Category 4) and Marco Island (Category 3)

. First Category 4 hurricane to make landfall in Florida since Charley (2004)

Meteorological Development

Irma started out on August 30 as a typical Cape Verde storm. Irma steadily intensified as the storm progressed west, ultimately strengthening to a Category 5 hurricane near the northern Leeward Islands. Warm SSTs allowed Irma to maintain this intensity for nearly two days.

Hurricane Irma’s track through the Caribbean.

As the storm tracked through the Caribbean, forecast models disagreed regarding the size and strength of the subtropical ridge located over the Atlantic that provided a steering flow to Irma. The model disagreement introduced uncertainty regarding the storm’s northward turn and subsequent US landfall.

On September 7, ensemble forecasts indicated that Irma could make landfall on either coast of Florida or miss the state entirely.

The mean forecast track was through Miami as a borderline Category 4/5 hurricane. KCC estimated that this track would have caused over $180 billion in insured losses.

Global Ensemble Forecast System (GEFS) tracks (left) and KCC wind footprint (right) for NHC Advisory 35 three days before Irma’s Florida landfall.

As Irma got closer to Florida, the projected tracks shifted to the west. At this time, Irma became a storm surge threat due to unique characteristics of the Florida west coast.

KCC wind (left) and storm surge footprints (center) and NHC inundation maps (right) generated on September 9 one day before Irma’s Florida landfall.

In addition to uncertainty regarding the storm’s landfall in Florida, there was also variation in the forecast peak wind speed at landfall, which was highly dependent on SSTs and the storm’s potential decay in the Caribbean. The warm ocean temperatures were unable to compensate for decreased wind speeds due to land interaction with Cuba, and the storm made its first Florida landfall near Cudjoe Key as a Category 4 hurricane followed by a second landfall at Marco Island with peak wind speeds of 115 mph.

Irma also tracked farther inland than expected as the storm traveled up the Florida peninsula. By September 11, Irma was categorized as a tropical depression, and by September 13, the storm had dissipated completely.

KCC final wind footprint for Hurricane Irma.

Real Time Loss Estimation

KCC’s High Resolution Caribbean and US Hurricane Reference Models were used to estimate losses from Hurricane Irma for each NHC Advisory. To generate the intensity footprints, two sources for the projected storm track were used pre-landfall—the NHC and the European Center for Medium-Range Weather Forecasts (ECMWF).

When the tracks disagreed significantly, KCC clients could obtain multiple estimates of claims and losses. The chart below shows the industry losses for each advisory and where appropriate the range from the two different forecast tracks.

KCC Pre‐landfall Florida Loss Estimates for Hurricane Irma

180

160

140 ECMWF 120

100 NHC 80

Loss ($Billions) NHC NHC ECMWF 60

ECMWF 40 ECMWF

20 NHC

0 30 31 33A 34 35 37A 39 41A 43 45A 47 49A 54 NHC Advisory KCC industry loss estimates by advisory showing the range from different projected tracks (NHC and ECMWF) when applicable.

Estimates include losses to buildings, other insured structures, contents, business interruption, and autos. Estimates do not include crop or NFIP losses. As soon as Irma dissipated on September 13, the following loss estimates were released publicly by KCC.

Breakdown of KCC Real Time Insured Loss Estimate ($ Billions) US 18 Caribbean 7 Total 25

Post-Event Damage Survey

KCC dispatched several teams to the most impacted areas within a week of Irma’s landfall. Key observations from the KCC damage survey in Florida included:

. While major structural damage was limited, moderate to low-level winds resulted in damage throughout the state.

. Irma confirmed that properly installed metal roofs on homes perform significantly better than asphalt shingles or tile.

. Detailed investigations with KCC client claims adjusters revealed significant differences in claims handling practices between insurers.

. Commercial structures typically sustain more cladding damage than residential structures from low category wind speeds because of building height and popular commercial cladding materials.

. Architectural features, such as overhangs and signage, increase the damage on commercial structures.

KCC wind footprint with hurricane winds indicated by category (left) and wind speed observations (right) for Hurricane Irma in Florida.

Residential Structures

Category 3 wind speeds from Hurricane Irma were experienced along the southwest Florida coast in localized areas around Marco Island and Goodland. Ten percent of single family homes exposed to these winds sustained structural damage, typically in the form of roof failure.

Structural damage in residential buildings was consistently observed in older buildings. In the following example, the two older buildings sustained damage while the newer home, even though it experienced higher wind loading due to its height, sustained minimal damage.

Buildings built in the 1970’s (left) and 1980’s (center) sustained significantly more damage than newer homes (right).

More frequently observed damage included loss of roof covering and siding. Roofing materials in this area are predominantly clay tiles and asphalt shingles.

Asphalt shingles are fastened to roof decks and are subject to uplift when exposed to high winds. Because of their overlapping nature, damage to one section of the roof results in progressive failure and often requires the entirety of the roof to be replaced. Over 30 percent of asphalt roofs experienced some damage from the Category 2 and 3 wind speeds.

Typical damage to asphalt shingle roofs from Category 2 and 3 wind speeds.

Tile roofs are slightly less vulnerable but are still subject to progressive failure due to their interlocking nature. Ridge, hip, and perimeter tiles are typically the first point of failure during a hurricane. When a tile becomes detached, it exposes the roof deck, increases the vulnerability of nearby tiles, and the dislodged tiles become windborne debris potentially causing additional damage.

Typical damage to tile roofs in Marco Island.

Roof covering damage can also lead to water infiltration, which will significantly increase the insured claim amount. Depending on the level of roof covering damage, the building interior can experience water leakage and ceiling collapse. In the majority of homes with roof covering damage, there was some water leakage, and approximately ten percent of those experienced partial ceiling collapse. Water infiltration can also result from damage to soffits.

Wind damage to roofs and soffits (left) can lead to water infiltration and damage to ceiling and internal elements (right).

Properly installed metal roofs are the most resistant to high winds. Overall, a much smaller proportion of metal roofs experienced damage from Irma’s highest winds. The images below show the resistance of metal roofs; these homes right along the coast lost a significant portion of the siding, but the metal roofs are completely intact.

These residential buildings along the coast all sustained substantial siding damage from Irma’s high winds but minimal roof damage demonstrating the strength of properly installed metal roofs.

However, if the metal roof is not properly installed or if the fastenings become loose and a panel peels away, the entire roof typically needs to be replaced.

Older construction (left) and metal roofs with poor installation or weak fastenings (right) sustained damage from Category 2 and 3 wind speeds.

How Much Does it Cost?

KCC detailed damage investigations with claims adjusters provided additional insight into how damage translates to claims estimates. Many factors influence the final paid loss.

Determining the cause and extent of damage is the first step in the claims adjustment process. For example, claims for roof damage depend on decisions on whether the roof can be repair or should be replaced, and the accuracy of measurements and cost estimates. This means the same level of damage can translate into widely varying claims estimates depending on the adjuster.

Adjusters are typically given wide latitude in determining claim estimates. For large events, adjusters with varying levels of experience, from full-time professionals with years of experience to those working on their first event, are deployed.

Insurers have different philosophies on how to manage the balance between keeping losses low versus keeping policyholders happy. Additional factors, such as practices regarding the exclusion of preexisting damage and policy terms, can have large impacts on claim estimates and are specific to each insurer.

Appurtenant structures, screen enclosures, and fences are highly vulnerable to high winds. These structures are typically lighter weight and built using less resilient construction practices.

Damage to screen enclosures (left), fences (center), and sheds (right) was observed in high and low wind speed areas throughout Florida.

Nearly all mobile homes sustained significant damage from the Category 3 wind speeds. Mobile homes are made of lightweight materials, which makes them susceptible to high winds. Additionally, when siding and other material becomes detached from the underlying structure, it can become windborne debris and cause additional damage to downwind structures.

Extensive damage to mobile homes near Hurricane Irma’s landfall point in Florida.

Moving farther inland and to lower wind speed areas, the damage to roofs and siding becomes spottier, and structural damage to residential homes was typically limited to homes impacted by fallen trees. However, damage to mobile homes is common even in low wind speed regions.

Commercial Structures

Commercial structures also experienced significant roof and siding damage. Additional sources of damage to commercial buildings are typically the result of eaves, overhangs, and signage.

Low levels of commercial roof damage were frequently observed in multiple wind speed regions, and similar damage was observed in commercial and residential roof coverings.

Metal (left), shingle (center) and tile (right) commercial roofs sustained damage when exposed to Category 1 and 2 wind speeds.

Envelope elements, including cladding and windows, sustained observable damage when exposed to high wind speeds. A popular form of siding for commercial buildings is External Insulating Finishing System (EIFS). EIFS is a series of coverings mechanically attached to the exterior of a building and is popular for commercial buildings because it is easy to install, durable, water resistant, and can be cut to any shape. EIFs provides insulating and finishing in one installation but is relatively less wind resistant than other types of cladding, and it begins to exhibit damage at relatively low wind speeds.

EIFS damage from Category 1 wind speeds at the front (left) and corner (right) of a mid-rise office building.

Architectural features including eaves, overhangs, and canopies are also highly susceptible to damage even from Category 1 winds.

In addition to damage to roofs, commercial buildings also experienced damage to eaves, overhangs, and ceilings of canopies from Category 1 winds.

Damage to signage was pervasive throughout regions experiencing hurricane force winds. Gas stations are particularly vulnerable because they experience wind loading above and below their canopies, and damage can begin at low wind speeds.

Ten percent of gas stations sustained significant damage to overhangs and canopies in areas of high wind speeds.

Storm Surge Damage

The majority of storm surge damage was located along the north shore of the Keys, where Irma was a Category 4 hurricane.

Storm surge can destabilize the foundation of a building and the underlying soil. This causes severe structural damage and was observed in isolated cases.

Severe storm surge damage along the north coast of the Keys.

More common and less severe storm surge damage occurred to siding on lower levels of buildings and their supporting elements. Storm surge beneath a building can also cause structural damage to supporting beams.

More common forms of storm surge damage included damage to siding (left) and damage to the underside of raised buildings (right).

In addition to structural and envelope damage, storm surge can infiltrate a structure and cause extensive contents damage. Along the coast in Marco Island and Goodland, storm surge was around three feet and caused contents and interior damage but minimal structural damage.

Contents damage in Goodland and Mar co Island was frequently observed but few instances of structural damage from storm surge.

Damage Observations from the Caribbean

Key damage trends in the Caribbean included:

. The most impacted islands were Antigua and Barbuda, Anguilla, the British Virgin Islands, and Saint Martin.

. Housing, tourism, and agriculture were the most impacted sectors with insured losses coming largely from islands with strong tourism economies

. Damage to one area caused problems in others, such as damaged airports limiting supplies in the region and environmental damage impacting tourism and agriculture

KCC wind footprint for Hurricane Irma in the Caribbean.

Hurricane Irma caused disruption throughout the Caribbean. Upwards of ninety percent of the buildings on several islands sustained damaged, including hotels and resorts. The majority of islands sustained significant damage to their electrical infrastructure. Most islands restored at least half of their electrical grid by the end of 2017, but some areas are still without power.

Tourism was the most impacted industry because, in addition to decreased lodging capacity, Caribbean attractions closed for renovation or restoration immediately following the hurricane. Furthermore, travel to and from the islands was limited due to damage to airports in the region.

While the tourism industry began to rebound by late 2017, a full recovery is not expected until late 2018 or 2019. Resorts in hard-hit areas need extensive repairs, and the infrastructure in many areas is still being restored.

In general, residential take-up rates are low on the Caribbean islands. Many homes that are covered by insurance policies are not insured for their full replacement values.

At the time of the event, KCC estimated $7 billion of insured losses throughout the region. The table below shows the estimated economic losses tabulated from various sources.

Island Estimated Economic Loss ($ millions) Antigua and Barbuda 300A Anguilla 290B St. Martin 1,500A St. Barthelemy 480A Bahamas 130C British Virgin Island 3,600B US Virgin Islands 7,500D Cuba 650E Turks and Caicos 500A St. Kitts and Nevis 20F A NHC Tropical Cyclone Report: Hurricane Irma B United Nations Development Program (UNDP) Regional Overview Impact of Hurricanes Irma and Maria C Inter-American Development Bank D USVI government, includes impacts from Irma and Maria, appeal to Congress December 18, 2017 E UNDP Cuba Hurricane Irma Three Month Report F St. Kitts and Nevis government, press conference September 14, 2017

Hurricane Maria September 16-30

. Second Category 5 storm in one month to impact the Caribbean

. Intensified from a Category 1 to a Category 5 hurricane in 15 hours, which is second only to Hurricane Wilma’s 12 hour record

. Passed over Dominica at peak Category 5 intensity

. Made landfall in Puerto Rico with wind speeds of 155 mph, the strongest hurricane to make landfall there since the San Felipe II Hurricane (1928)

Meteorological Development

On September 16, Maria was identified as a tropical storm, and by September 17, Maria was upgraded to a Category 1 hurricane. Maria intensified rapidly to Category 5 status on the approach to Dominica and passed over the island on September 18 with wind speeds of 160 mph.

Hurricane Maria’s track through the Atlantic and northeastern Caribbean.

The storm continued a northwestward track over the warm waters of the northeastern Caribbean Sea, which allowed the storm to remain a strong Category 4 hurricane. Maria passed to the south of the US Virgin Islands, and made landfall near Yabucoa, Puerto Rico, with wind speeds of 155 mph on September 20.

Much of the island was subject to hurricane force winds. Maria returned to sea along Puerto Rico’s north coast and passed east of Turks and Caicos and the Bahamas before veering out into the Atlantic.

KCC wind footprint for Hurricane Maria.

Real Time Loss Estimation

The KCC Caribbean Hurricane Reference Model provided the loss estimates for Hurricane Maria’s impacts.

Breakdown of KCC Real Time Insured Loss Estimates ($ millions)

Dominica 445

Guadeloupe 119

Puerto Rico 28,350

US Virgin Islands 789

Other Territories* 94

Total 29,797

*British Virgin Islands, Dominican Republic, Martinique, Saint Kitts and Nevis, and Turks and Caicos

Modeled estimates include wind losses to residential, commercial, and industrial properties. Modeled property losses were increased by 20 percent to account for auto losses, insured flood damage, and additional demand surge.

Post-Event Damage Survey

KCC engineering teams conducted post-event damage surveys in Puerto Rico which included independent surveys as well as detailed investigations with KCC client claims adjusters. Key observations included:

. Damage to insured single family homes was minimal. . The most extensive damage was to commercial properties, particularly chain stores, shopping malls, and hotels.

. Damage to roof, cladding, and other external elements on mid and high rise office buildings often led to water infiltration further damaging contents, ceilings, and systems such as HVAC.

. Observed damage to industrial facilities was relatively minimal, except for a few specific cases.

Residential Structures

Single family homes in Puerto Rico are of three dominant construction types: concrete (over 70 percent), wood, and mixed. As expected, very few concrete homes sustained significant damage. More than half of wooden structures experiencing Category 2 or higher wind speeds sustained damage. In cases of mixed construction, where the ground floor is concrete and the upper floor wood, the wood portion of the structure can suffer significant structural damage.

AAAA

Concrete bunker-style homes (left) typically withstood Maria’s high winds while wood frame homes (right) sustained severe structural damage.

Concrete single family homes typically have concrete roofs which further enhance their strength against high winds. In contrast, many condominiums and apartment buildings have tile roofs. Tile roofs are vulnerable to both wind pressures and debris

impacts. When tiles break, they can become additional debris in the windfield that can cause further damage to downwind structures.

Tile roofs sustained damage from Category 2 and above wind speeds.

The height of a building impacts the wind speed that will be experienced and sustained damage. In the following image, the roofs on the structures at the top of the hill experienced damage while those at the bottom of the hill remained intact due to the increased winds at higher altitude.

Differing damage states to roofs due to hurricane wind speeds increasing with altitude.

Other instances of damage to concrete homes and apartment buildings resulted from stucco siding. A technique common in Puerto Rico is to apply stucco plaster to the exterior of concrete buildings. As stucco ages, it cracks and becomes loose. Even lower wind speeds can detach cracked stucco from a wall, and detached stucco becomes additional debris in the windfield.

Damage to stucco was predominantly observed in mid rise buildings, although isolated instances of damage to low-rise buildings (less than one percent) were observed.

Stucco can be peeled from the underlying surface causing damage and adding debris to the windfield.

Commercial Structures

Apartment complexes typically have architectural attributes, such as balconies, that are vulnerable to high winds. High rise apartment buildings and hotels sustained significant damage to window and door openings. Most of this damage occurred because of weak door to wall frame systems, which were unable to withstand the pressure exerted by the high winds.

Damage to the balconies of apartment complexes occurred under different wind conditions, including in Palmas Del Mar (left) and San Juan (right). Weak frame to wall connections led to door and window damage (right).

The Manilla Healthcare Building in San Juan demonstrates the typical vulnerabilities of high rise office buildings. The glazing exhibits evidence of both pressure and impact damage. Pressure impacts are most common at the corners of buildings due to negative suction pressures, and debris damage can be seen in the patterns of broken glass.

Glazing damage can occur because of extreme pressure (left) and by debris impacts (right).

Extensive exterior damage allows a significant amount of wind and water into the building. This results in internal damages to ceilings, electrical systems, and HVAC. Water infiltration causing a domino effect on other interior elements was commonly observed throughout the damage survey.

Water and wind infiltration can cause interior damages to ceilings (left), electrical systems (center) and HVAC (right).

High-rise buildings often have external equipment which can sustain significant damage when exposed to high winds and heavy rain. For example, the elevator room for the building is located on the roof and was significantly damaged by Hurricane Maria. This allowed water infiltration into the elevator shafts requiring the replacement of six elevators costing over $2.5 million and taking months to complete.

Water infiltration into elevator shafts results in costly and lengthy repairs. On this high rise building, damage to the elevators alone is costing $2.5 million to repair.

Hotels are particularly vulnerable to high winds due to their large windows, balconies, and other architectural features. Over half of the hotels in San Juan along the coast

sustained various states of damage. Water infiltration greatly increased claims amounts when contents and interior elements were damaged.

EFIS is again quite a prevalent material for hotel cladding and highly vulnerable to strong winds.

EIFS siding on San Juan hotels exhibited significant damage.

Well engineered long-span buildings are not more vulnerable than other structures. However, poorly designed long-span structures are highly vulnerable due to their light mass, high flexibility, and low damping characteristics. Long-span roofs are exposed to high uplift loading under strong winds, and if not specifically accounted for in the design, they will experience damage from hurricane force winds.

Long span light steel frame shopping plaza sustained severe roof damage when exposed to high wind speeds, which caused further contents and interior damage.

Free standing retail stores with large openings and windows are particularly vulnerable to high winds.

Before and after Hurricane Maria photos of a furniture store in Manati, which suffered significant damage due to large window openings.

During a hurricane, when an opening occurs on the windward side of a structure, high positive internal pressure occurs, which when combined with negative external pressures, produce high net pressures. In rare cases, the frequency of the internal pressure fluctuations equals the natural period of the building and leads to resonant or “explosion” failure (damage to openings and roof simultaneously) as occurred to this car dealership during Hurricane Maria.

Explosion-type building failure at this car dealership destroyed the roof and windows simultaneously.

Shopping plazas throughout Puerto Rico exhibited consistent modes of damage. Eave caps, roofs, signage, and the majority of verandas were damaged. Approximately seventy percent of shopping plazas in the areas surveyed sustained some damage to exterior elements.

Damage to eave caps (top), the underside of verandas (bottom left), and siding (right) all can lead to water and wind infiltration.

The San Juan international Airport also sustained damage to envelope elements which led to interior and contents damage. Notably, most damage to the airport occurred in the older section, demonstrating how age of structure impacts wind vulnerability.

Water infiltration led to damage to acoustic and plaster ceilings throughout the airport, and most noticeably in the older section near Terminal D.

Additional structures vulnerable to hurricane force wind speeds include warehouses and sports arenas. These structures share similar characteristics, such as lightweight exterior elements, which make them vulnerable to wind damage.

Zinc roofs (left) to warehouses and light metal envelope elements (right) like those in sports arenas often did not withstand Maria’s high wind speeds.

Damages were also observed in the sustainable energy sector. All turbines at the Punta Lima wind farm near Humacao sustained damage to their blades due to Maria’s Category 4 wind speeds exceeding their design speed. Solar panels, due to their near horizontal position and broad surface area, are vulnerable to wind loading at even relatively low wind speeds. More than 75 percent of solar panels installed near Humacao sustained damage.

Wind turbines (left) whose design speed was exceeded by Hurricane Maria, and solar panels (right) that are vulnerable to hurricane force winds.

No other instances of significant damage to industrial facilities were observed.

Hurricane Nate October 4-8

. Made landfall as a Category 1 hurricane near Biloxi, Mississippi, with wind speeds of 85 mph

. Dissipated to a tropical depression within 10 hours of landfall

. Forecast intensification to a Category 2 storm did not materialize

Meteorological Development

Tropical storm Nate developed in the southern Caribbean Sea on October 4 and followed a northwest track across Nicaragua and Honduras.

Hurricane Nate’s track through the Caribbean Sea and Gulf of Mexico.

Nate continued northwest and skirted the Yucatan Peninsula before entering the Gulf of Mexico and becoming a Category 1 hurricane. Due to storm organization and warm SSTs, there was some uncertainty if Nate would strengthen to a Category 2 hurricane

or remain a Category 1 storm. Warm SSTs were not enough for strengthening, and Nate remained a Category 1 hurricane.

Real Time Loss Estimation

KCC’s US Hurricane Reference Model estimated losses from Hurricane Nate to be close to $500 million. Modeled losses include insured wind and storm surge losses to residential, commercial, and industrial properties as well as automobiles.

Conclusions

The 2017 hurricane season was well above average in terms of the number of hurricanes and the intensity of the storms. There were as twice many major hurricanes than in an average season. Based on the ACE index, 2017 was the most active season since 2005.

Utilizing KCC high resolution models and advanced tools, insurers were able to track these storms as they were unfolding to obtain timely and accurate loss estimates. KCC’s real time loss estimates total to $70 billion in privately insured losses for the US and Caribbean. These industry estimates along with the modeled loss estimates for individual insurers are proving to be quite accurate.

Post-event, KCC scientists and engineers conducted extensive field surveys in the most impacted areas including detailed investigations with KCC client claims adjusters which revealed significant differences in claims handling practices between insurers.

This highlights the value of the KCC open models and advanced new tools that insurers are using to track hurricanes in real time and to leverage their detailed claims data. With the ability to customize the model damage functions to their unique policy conditions and claims handling practices, insurers can further enhance the accuracy of their modeled loss estimates.