The AIR Tropical Cyclone Model for the Carribean In October 212, Hurricane Sandy wreaked havoc across Jamaica, Cuba, and the Bahamas. The AIR Tropical Cyclone Model for the Caribbean incorporates the latest science and an understanding of local building practices to provide a highly granular view of risk across all 28 territories. With average annual aggregate insured losses approaching USD 3 billion, companies writing business in the Caribbean need sophisticated tools to support better risk selection and risk transfer decisions.
Tropical cyclones take a wide variety of tracks through the Caribbean, although the risk generally increases from south to north. The AIR Tropical Cyclone Model for the Caribbean features a large catalog of simulated events that appropriately characterizes the frequency, tracks, and meteorology of potential future storms, as well as the observed correlations of risk between countries an important criterion for companies that insure properties across the region. Incorporating high resolution hazard data and local expertise in countryspecific building practices, the AIR Tropical Cyclone Model for the Caribbean provides an incomparable level of detail, enabling more precise risk differentiation based on such factors as geography, construction, occupancy, year built, and individual building characteristics. Category 5 Category 4 Category 3 Category 2 Category 1 Tropical Storm In 27, Dean crossed the Lesser Antilles as a tropical storm, impacted Jamaica with Category 1 hurricane winds, made landfall in the Yucatan at Category 5 strength, and finally came ashore for a second landfall in central Mexico as a Category 2 hurricane. The AIR Model allows companies to seamlessly assess the risk to policies and portfolios that span multiple countries. (Source: AIR, HURDAT). The Advantage of the Basinwide Approach The AIR Tropical Cyclone Model for the Caribbean features a unified catalog of nearly 5, simulated events affecting the United States, Mexico, Central America, and offshore assets in the Gulf of Mexico. AIR s basinwide catalog allows companies to more accurately and seamlessly model losses to policies and portfolios that span multiple countries a key feature in light of the fact that a significant percentage of storms in the North Atlantic impact multiple regions. I am very impressed with the detailed engineering treatment and the sound science incorporated into the latest Caribbean model update. The [AIR Tropical Cyclone Model for the Caribbean] is an impressive piece of work. It is clear that substantial resources have been utilized in developing the model. Tony Gibbs, Engineer and Caribbean risk expert, peer reviewer of the AIR model. 2
A MEASURED APPROACH TO MODELING TROPICAL CYCLONE RISK IN A WARM OCEAN CLIMATE Tropical cyclone activity has fluctuated during the past century, alternating between periods of high and low frequency that correlate with long-term warming and cooling trends in the Atlantic. Among the many strengths of the AIR model is its measured approach for capturing the impact of variable climate conditions on hurricane risk and insured losses. To capture the range of observed recent activity, AIR provides clients with two catalogs an ensemble of simulated events. While both AIR views are scientifically valid, the warm SST view is about 12% higher for the Caribbean. It is not meant to be a forecast of hurricane activity in the medium term nor is it intended to replace AIR s standard view of hurricane risk, which is based on all years of hurricane activity since 19. Built using AIR s transparent catalog design, both catalogs capture the range of observed recent activity over the current period of warm SSTs and allow companies to test the sensitivity of a range of risk assumptions. High-Resolution Land Use and Land Cover Data Captures Directional Effects on Surface Wind Speeds To capture surface winds at any location with realism, the model must account for the surrounding environment. Winds arriving from the open ocean, for example, will be faster than winds that have traveled over a mountainous island interior or forest. Using the latest satellite-derived, high-resolution land use/land cover and elevation data from the United States Geological Survey (USGS), the AIR model captures the effects of surface friction based on the direction of the wind at each location. Knowing the precise location of exposure becomes more critical when the effects of wind direction are explicitly modeled. For example, insured exposure in the smaller Caribbean islands tends to be fairly evenly distributed along the coast, while exposure in the larger islands is more concentrated (in urban centers). A narrow band of exposures around the entire coastline will moderate the wind speeds very differently than a built-up urban environment. Flood Component Incorporates Detailed Information on Soil Type, Land Use/Land Cover, and Topography In the Caribbean, tropical cyclone induced flooding can be a key driver of loss. In 1979, Hurricane David a seminal event in the region s history dropped nearly 21 inches of rainfall in Puerto Rico, triggering massive flooding. David went on to make landfall in the Dominican Republic as a Category 5 hurricane, bringing not only damaging winds but additional precipitation that caused severe flooding and killed thousands of people. More than 5% of the insured losses were flood-related. But even storms with relatively low wind speeds can be accompanied by significant flooding, exacerbated by the inland mountains that characterize many islands throughout the region. The AIR Tropical Cyclone Model for the Caribbean incorporates a flood module that estimates the probability and severity of a flooding event and its associated damage. Based on storm size, intensity, and rainfall characteristics, as well as terrain and land use, the model captures an hourly precipitation footprint at each location over the entire duration of the storm. Slower moving storms will subject any given location to higher rainfall totals. The total accumulation is then redistributed based on the porosity of the soil, land use, and slope. W NW SW N San Juan AIR explicitly models the directional effects of surface friction on wind speeds. In northern Puerto Rico, a northeast wind will be relatively unobstructed as it comes in off of the Atlantic Ocean, while winds from the south will degrade as they travel over the mountainous interior of the island. (Source: AIR, USGS) S NE SE E 3
Rainfall Parameters Hourly Precipitation Storm Duration First Building Code Higher Standards Post-Hurricane Floyd 1987 Code Change and Increase in Bunker Type Construction Post-Hurricane Hugo Higher Standards Post-Hurricane Georges Topography Storm Intensity and Size High Relative Vulnerability Pre-1971 1971-1999 Post-1999 Pre-1972 1972-1997 Post-1997 Pre-1989 1989-1999 Post-1999 Low Storm Total Precipitation Bahamas Bermuda Puerto Rico A separate flood module estimates the probability and severity of a flood event, and distributes total rainfall accumulations based on topography and the ability of soils to absorb precipitation. (Source: NASA) The AIR Model accounts for variability in vulnerability by year built and territory for better risk differentiation. (Source: AIR) Accounting for Regional Building Codes and Construction Practices Some islands in the Caribbean are subjected to higher levels of wind speeds more frequently than others. Over time, this has resulted in variations in building vulnerability across the region, as each country develops its own building codes and construction practices that reflect its historical storm experience and regional building inventory. For example, in Puerto Rico, after Hurricane Hugo in 1989, concrete bunker style buildings usually built of reinforced concrete walls and concrete slabs roofs became more prevalent. The AIR Tropical Cyclone Model for the Caribbean incorporates exhaustive research by AIR engineers into the local and temporal variations in the construction practices and building codes, as well as the record of their enforcement success. The model takes into account the specific year of construction, by territory, to assess building vulnerability at a highly granular level. A building s response to tropical cyclone winds and performance during extreme flooding can vary significantly depending on its construction type, occupancy class, and height. To reflect the vulnerability of the building stock in the Caribbean, AIR engineers have developed separate wind and flood damage functions for 37 different construction classes and 11 occupancy classes, including hotels, single-family homes, and industrial facilities. Business interruption damage functions vary by occupancy and account for physical damage level, building size and complexity, and business characteristics, such as resiliency and the ability to relocate. Leveraging AIR s Detailed Industry Exposure Database for the Caribbean To produce the most reliable estimates of industry losses, the AIR Tropical Cyclone Model for the Caribbean incorporates a comprehensive, high-resolution industry exposure database (IED). AIR builds the IED by compiling detailed information about risk counts, building attributes and replacement values, as well as information on standard policy terms and conditions. Because the raw data can vary considerably in resolution, AIR uses a sophisticated algorithm that takes into account topography, satellite-derived land use/land cover information, and the Defense Meteorological Satellite Program s ISA (Impervious Surface Area) data from NOAA to disaggregate all risk counts to a 1 km grid. AIR is the only modeler to feature a separate industry exposure database for every Caribbean island. Companies can validate their own loss distributions for each island by comparing detailed loss results from Touchstone against AIR s CATRADER industry losses. 4
A COMPONENT-BASED APPROACH TO MODELING COMPLEX INDUSTRIAL FACILITIES The AIR model employs a component-based approach to estimate potential losses to industrial facilities. AIR assesses the overall vulnerability of various kinds of facilities (chemical plants, oil refineries) based on the vulnerability of the individual assets the components and sub-components that the facility comprises. The model implicitly accounts for more than 55 distinct industrial components based on detailed, site-specific engineering-based risk assessments conducted through AIR s Catastrophe Risk Engineering (CRE) service. Industrial plants often have similar interconnected components including tanks, flares, cooling towers, transmission systems and transportation assets but are present in different proportions depending on the individual plant type. The distribution of component classes and their response to hurricane winds is utilized by AIR to develop the vulnerability of an entire plant. Millions USD Preciptation Levels (mm) Antigua 3 25 2 15 1 5 Precipitation Level (mm) Dominica 3 25 2 15 1 5 Precipitation Levels (mm) Guadeloupe 6 5 4 3 2 1 Precipitation Levels (mm) Martinique 6 AIR modeled precipitation totals consistently reproduce observed patterns. Here, simulated precipitation for Hurricane Marilyn is validated using historical records. (Source: AIR) 8, 7, 6, 5, 4, 3, 2, 1, Puerto Rico Observed Modeled Georges Hugo David Hortense Total 5 4 3 2 1 Millions USD AIR Validates its Models from the Bottom Up and Top Down To ensure the most robust and scientifically rigorous model possible, the model is carefully validated against actual loss experience. However, validation is not limited to final model results. Each component is independently validated against multiple sources and data from historical events. For example, the AIR modeled wind speeds and precipitation totals are validated against observation data from actual storms. Millions USD 8 7 6 5 4 3 2 Bermuda Observed Modeled Modeled damage ratios and footprints are validated against actual observations from both published reports and from AIR post-disaster surveys conducted in the aftermath of hurricanes. Modeled losses have been validated not only at the company level, but by event, by line of business, and by coverage using significant amounts of claims data. 1 Fabian Emily Total AIR modeled losses compare well to industry losses for historical hurricanes. (Source: AIR) 5
Achieving More Accurate Model Results by Separating Wind and Flood Losses The ability to model flood losses with unique flood policy conditions can greatly impact a company s high exceedance probability (low return period) losses. This includes losses from tropical cyclones with relatively low sustained wind speeds storms that can still deliver significant precipitation. Determining the insured flood loss from these storms is vital to overall risk assessment. The AIR model enables companies to estimate losses to policies that contain multiple terms and conditions. Using Touchstone, AIR s detailed modeling software, companies can generate loss estimates for flood only, wind only, or wind and flood combined. When information about policy-level flood coverage is not available, Touchstone also allows users to apply average flood take-up rates to the portfolio. The ability to model the two perils separately and apply appropriate, peril-specific policy conditions leads to more accurate loss results and a better understanding of tropical cyclone risk. AIR has been conducting damage surveys in the aftermath of Caribbean hurricanes since Hurricane Floyd in 1998. Survey findings are used to validate the model s damage functions and to inform future model enhancements. In line with expectations, this masonry residence sustained only non-structural damage from Hurricane Irene s 211 path through the Bahamas. (Source: AIR) 6
Model at a Glance Modeled Perils Supported Geographic Resolution Stochastic Catalog Supported Construction Classes and Occupancies Supported Policy Conditions Tropical cyclone winds and precipitation-induced flood CATRADER: CRESTA Touchstone: CRESTA, municipality, parish, and postcode; can also accept user-provided latitude and longitude The catalog includes 1, simulated years containing more than 5, simulated events; the catalog is shared with other AIR regional hurricane models, allowing companies to accurately estimate losses to portfolios that span multiple countries. Supported Construction Classes: Separate wind and flood damage function for 37 construction types that account for the impact of height and year built on building vulnerability Supported Occupancy Classes: 11 Unknown Damage Function: When detailed exposure data (e.g., construction type or height) is unavailable, the model applies an unknown damage function that takes into account countryspecific construction characteristics Touchstone allows a wide variety of location, policy, and reinsurance conditions, including limits and deductibles by site or by coverage, blanket and excess layers, minimum and maximum deductibles, and sublimits. Reinsurance terms include facultative certificates and various types of risk-specific and aggregate treaties with occurrence and aggregate limits. Touchstone also provides information used to support risk retention and transfer alternatives, growth strategies, and real-time event planning. Model Highlights Peer-reviewed model provides comprehensive coverage of all 28 countries and territories in the Caribbean Standard and warm sea surface temperature (WSST) catalogs provide two robust and scientifically defensible views of risk Incorporates the latest United States Geological Survey (USGS) land use/land cover data to capture the directional effects of surface friction Explicitly models precipitation-induced flood loss by assessing the accumulated runoff from a storm s precipitation totals Accounts for differences in regional vulnerability arising from each territory s historical storm experience, construction practices, and building code evolution and enforcement Damage to industrial facilities modeled using an objective, engineering-based, component-level approach, instead of traditional approaches based on expert opinion Wind and precipitation fields validated against observations for all available data sets in the region, including more than 3 historical Caribbean hurricanes; modeled losses validated against significant amounts of claims data 7
ABOUT AIR WORLDWIDE AIR Worldwide (AIR) provides risk modeling solutions that make individuals, businesses, and society more resilient to extreme events. In 1987, AIR Worldwide founded the catastrophe modeling industry and today models the risk from natural catastrophes, terrorism, pandemics, casualty catastrophes, and cyber attacks, globally. Insurance, reinsurance, financial, corporate, and government clients rely on AIR s advanced science, software, and consulting services for catastrophe risk management, insurance-linked securities, site-specific engineering analyses, and agricultural risk management. AIR Worldwide, a Verisk (Nasdaq:VRSK) business, is headquartered in Boston with additional offices in North America, Europe, and Asia. For more information, please visit www.air-worldwide.com. 218 AIR Worldwide A Verisk Business