Water surge

Flooding is one of the most frequent and costly natural perils in Canada, but to date, the catastrophe modelling companies have not developed flood models for the country.

The flood peril is very challenging to model using the standard catastrophe modelling approach. Relative to perils such as hurricanes and earthquakes, flood models require higher resolution data and have several unique complexities.

The catastrophe models have four primary components:

1) an event catalogue defining where, how severe and how frequent future events are likely to occur;

2) scientific formulas to estimate the event intensity at each impacted location;

3) vulnerability curves to estimate the damages for different types of buildings and their contents based on the intensities; and

4) a financial module to translate the damages to insured losses, accounting for policy conditions such as deductibles and limits.

While the financial module does not change significantly, the first three components are unique to each peril region.

Ideally, these components are built using historical data and other scientific information, and the more data that is available, the more credible the model will be. Where data are scarce, expert judgment is used to develop the model components and assumptions.

DEFINING THE PERIL

The first complication with modelling floods is defining the peril. Separate and very different models are required to cover the various types of floods, such as the following:

• storm surge from hurricanes;

• inland precipitation caused by tropical cyclones;

• tsunamis generated by earthquakes;

• river flooding resulting from heavy rains and excessive snowmelt; and

• flash floods from heavy precipitation over a short period of time.

Storm surge flooding is the most straightforward to model because the driving events are hurricanes, and the hurricane models are well-developed and based on a wealth of historical data and scientific information. The event catalogue can be the same as the hurricane catalogue.

The intensity formulas will also utilize many of the same parameters used to estimate the hurricane wind intensities.

The minimum central pressure in a hurricane determines, to a large extent, the peak wind speeds and the peak surge heights. The peak surge will also be influenced by the coastal bathymetry, tides and the presence of inlets and bays.

High-resolution coastline data is required to capture the storm surge footprint along the coast and high-resolution elevation data is used to calculate the water depths inland – data that are generally available from scientific and government organizations. Studies on the flooding induced from storm surge, along with data on past events, inform the vulnerability curves for damage estimation.

Hurricane Katrina and Superstorm Sandy are two recent events with significant losses from storm surge.

Along the U.S. coastline, Tampa is the most vulnerable to storm surge flooding. For example, the surge footprint of a 1,000-year hurricane – a strong Category 4 storm – could lead to a loss of more than US$100 billion.

TSUNAMI RISK IN CANADA

Storm surge flooding does not pose a significant threat to Canada, but coastal areas such as Vancouver Island, could be inundated by tsunamis.

A tsunami is a series of waves in the ocean caused by the displacement of a large volume of water. The wavelengths are much longer than ocean waves associated with storm surge, and whereas a strong hurricane is not likely to cause storm surge heights greater than about 10 metres, a large-magnitude earthquake can generate a tsunami wave tens of metres high at the coast.

Many types of underwater disturbances can cause a tsunami, but the most destructive tsunamis in Canada will most likely result from large-magnitude earthquakes, particularly those generated on thrust faults associated with major plate boundaries. These types of events create more vertical displacement and, therefore, generate larger waves.

The Atlantic, Arctic and Pacific coasts of Canada are all susceptible to inundation from tsunamis, but the risk is highest on the west coast. Large-magnitude earthquakes generated by the Cascadia Subduction Zone where the Juan de Fuca plate subducts beneath the North American Plate pose the biggest threat.

A recent report from the Canadian Geological Survey, A Preliminary Tsunami Hazard Assessment of the Canadian Coastline, suggests that there is a 10% chance over the next 50 years of a megathrust event in this region. To estimate the inundation area and damages from such an event, many of the same parameters used in a storm surge model apply.

For example, high-resolution elevation and bathymetry information are used to estimate the peak wave height at the coast and the water heights inland. Tsunamis travel at much greater speed and, therefore, will impact coastal properties with much greater force and will cause flooding inundation further inland. The scientific formulas underlying the intensity calculations and the vulnerability curves can be refined to account for these impacts.

ESTIMATING EVENT FREQUENCY, SIZE

The more difficult and most challenging aspect of modelling tsunamis under the catastrophe model paradigm is estimating the frequencies and sizes of future events. Defining the event catalogue for tsunamis is more challenging than for storm surge, because while all hurricanes generate some storm surge, not all earthquakes generate tsunamis. And because there is more historical data for scientists to work with, the estimated frequencies and severities of hurricanes are more robust than for earthquakes – particularly large-magnitude events.

The best that can be done currently is a scenario-type model that provides possible deterministic tsunami scenarios without specific probabilities. Tsunami scenarios have been created by the catastrophe modellers, and newer, open loss modelling platforms enable users to create their own events. These events can be superimposed on a book of business to estimate the resulting damages.

While not fully probabilistic, this method can still inform underwriting and risk management decisions. For example, an insurer can estimate how much property value they insure within a possible footprint and decide if that is too much. Insurers can evaluate specific locations within the footprint to make individual risk decisions.

This approach is also recommended for river and flash floods, which are the most frequent flood events in Canada, but also the most difficult for catastrophe modellers. Because each event is so different in terms of severity, spatial extent and duration, Canadian insurers cannot expect to have a credible fully probabilistic model for inland flooding. That said, high-resolution data can enable insurers to get robust scenario loss estimates.

For example, Image 1 (upper right) shows the flood footprint for the Alberta floods of June 2013. This footprint can be superimposed on high-resolution elevation data available in the open loss modelling platforms. It can be “floated” along the river to see how losses are impacted by different scenarios.

Canadian government agencies have invested in creating high-resolution maps for different return period floods in significant floodplains. For example, Image 2 (upper right) is a 100-year flood footprint for Alberta, presented at the 2014 Association of State Floodplain Managers Annual Conference.

While not quite the methodology employed by the catastrophe models, insurers can leverage the detailed work that has been done.

Historically, flood perils have not been covered by the catastrophe models because of several unique complexities. Even though catastrophe modelling technology is expanding to cover new perils that require more complex and higher-resolution models, it will take a lot more time and scientific knowledge for the cat
astrophe modellers to create fully probabilistic models covering all types of flooding in Canada.

In the meantime, Canadian insurers can evaluate flood risk using new, open modelling platforms that provide insight into potential future losses from a wide range of scenarios.

Image credit: Massive Flooding in Calgary, Canada – 1D/2D Models, Inundation Mapping and Reality by Peter Onyshko, P.Eng., CFM, Government of Alberta; Bryce Haimila, CFM, Government of Alberta Association of State Floodplain Managers Annual Conference, 2014