Claire Souch (pictured) is vicepresident of Model Solutions at Risk ManagementSolutions. Jeff Waters is a senior risk analyst with thecatastrophe risk modeler. 

For recent hurricanes, including Sandy in 2012, storm surge wasthe principal driver of U.S. hurricane insured losses. Formany of these hurricanes the damage caused by the storm surge faroutweighed the wind-driven impacts. And yet, many storm surgemodels have been shown to underestimate the potential insuredlosses of precisely these types of disproportionately large stormsurge events.

To understand why, we must first understand the physics of stormsurge. The magnitude and extent of storm surge flooding isinfluenced by a series of factors. These include hurricanecharacteristics such as wind speed, forward speed, centralpressure, and storm size in the days preceding landfall, as well astidal levels, offshore bathymetry, and coastal topography.

Traditional storm surge models lack the necessary compute powerto capture this information in enough detail, meaning that thesemodels rely on simplified assumptions on the relationship betweenthe storm's intensity and corresponding surge levels. The resultingoutput tends to under predict the potential for disproportionatelylarge and catastrophic storm surges, leading the industry tounderestimate the potential magnitude of storm surge loss, for manyyears.

Rapidly developing high-performance computing has enabled RMS tocreate a high-resolution dynamic storm surge model that solves thefundamental challenges of modeling the flood impacts fromcatastrophic storm surge. Previously, there had not been a crediblemodel to do so. Our North Atlantic Hurricane model is the onlycommercially available model that quantifies the risk fromcatastrophic hurricane-driven storm surge.  The version 13model, which incorporates still more scientifically advanced andhigh-resolution data, will be released on July 31, 2013.

Using advanced technology to improve storm surgemodeling

In 2011, following several major surge events such as Katrinaand Rita in 2005, and Ike in 2008, RMS incorporated one of theworld's most advanced storm surge models, MIKE 21 from the DHIGroup, into its North Atlantic Hurricane model suite. RMS remainsthe only catastrophe modeling company to use such a modern, fullytime-stepping hydrodynamic storm surge model across a largegeographic range. This fully probabilistic catastrophe modelcaptures wave damage potential for offshore risks, as well asdetailed onshore flood levels.

What makes the model unique is that it is dynamically linkedwith the hurricane wind field, so that the progress of the surgecan be tracked throughout the entire lifecycle of the storm, notjust at landfall.

It also models onshore surge dynamically, using a flexible meshgrid system that enables high-resolution modeling around complexcoastlines like New Orleans or New York City. In addition, theonshore impacts of surge can be modeled using some of the highestresolution data possible, for features such as topography, landuse, land cover, and coastal bathymetry.

The overall value of this model is its ability to capture thelocalized nature of flooding and damage, enabling users to assessdamage and quantify loss to individual locations at resolutions ashigh as street level. This modeling technology has been thoroughlyverified by our ability to capture Hurricane Sandy's extensiveflooding: demonstrated through comparisons to more than 200 actualstorm surge gauge recordings of flood heights.

Factors affecting loss-payouts

One key factor that influences insured storm surge loss payoutsis the elevation of properties in high-risk coastal floodzones.  For our version 13 of the North Atlantic Hurricanemodel, we have analysed and incorporated new and higher resolutionversions of the Flood Emergency Management Agency (FEMA) Base FloodElevation (BFE) data, and National Flood Insurance Program (NFIP)Flood Zone data to refine our modelling assumptions.

The line of business also has a significant impact on the losspayout.  The loss payout dynamics for commercialproperties are very different to residential, and cannot be treatedwith the same assumptions. For commercial and industrialproperties, flood coverage is often included, but sub-limited insome way. These properties often have high-value contents locatedin the basements, and are vulnerable to business interruption (BI)and contingent BI impacts from wide-scale flooding andinfrastructure outages.

Sandy demonstrated the real potential for industry storm surgelosses for commercial and industrial lines of business: more than 65 per cent of the total insured loss for Sandycame from these lines of business, and mostly for storm surgedriven loss. Learning from Sandy's industry losses, we have refinedthe loss calculations for commercial and industrial policies andvulnerability.

Residential and small commercial flood insurance is often notoffered by the private market, and is provided by the NFIP schemefor properties only in high-risk zones. The private market does,however, still have a small amount of exposure to residential surgeloss: through high-value homeowners policies, XS-NFIP policies, andclaims for uninsured surge losses that are not covered by NFIPpolicies, but are being made under wind policies. Sandy hasrevealed that these issues are less of a concern now, because ofimproved claims handling practices and recent court rulingsthat reduce the potential for"coverage leakage". We have conducted furtherresearch into this potential, and reflected the findings into theversion 13 model, which will be released this month.

Uncertainty still present

When estimating storm surge losses, insurers and reinsurers mustacknowledge that there are several uncertainties that continue topersist. For instance, the degree to which buildings comply withFEMA stipulated base flood elevations – BFEs in force since 1974 –as well as the presence of flood defenses, is still unclear.

Hurricane Sandy revealed that the industry does not know theoverall proportion of commercial and industrial exposure that isinsured for flood loss. There is also considerable uncertaintyaround the proportion of uninsured storm surge losses that may bepaid out under residential wind-only policies following a majorstorm surge event. In light of recent court rulings and theindustry's strengthened claims handling processes, the futurepotential could be quite different to payouts from past events.

The resolution of input exposure data also has a significantimpact on the range of storm surge loss output. Generally, usingcoarser resolution input data, such as zip code level, to model ahigh-resolution hazard like storm surge, results in a higher degreeof uncertainty because the localized surge impacts are not capturedas accurately. Establishing more confidence around this type ofinformation is critical to establishing credible insured lossestimates.

RMS continues to advance the development of loss estimates, byinvestigating actual building elevations in modeled locations,incorporating additional claims data from recent events – such asHurricane Sandy – to assess flood coverage assumptions andvulnerability, and encouraging the industry to improve theresolution of input exposure data.

The RMS North Atlantic Hurricane model continues to be known asthe industry's leading and most scientifically advanced hurricanemodel: our models enable the industry to understand its losspotential, with more accurate data and more research, and; reduceuncertainty in potential uninsured loss payouts and claimsinflation.