With the recent devastating flooding in Pakistan, estimates that between 10% and a third of the country has been submerged by flood waters have been circulating in the media.
Whatever the true figure for the scale of the catastrophe, this is a huge humanitarian disaster for the country. The disparity in estimates highlights the complexity of flood as a peril to model, requiring significant computing power and detailed, high-resolution inputs.
But as Nalan Senol Cabi, AVP of Catastrophe Research for Arch Insurance, told attendees at a recent InsuranceERM and ICEYE webinar on flood risk management, “It’s not just the type of data that is used in flood modelling that is key, it’s how the data is utilised.”
For many years, she said, flood has been treated by the insurance industry as an unmodelled peril. Flood hazard maps and deterministic scenarios were used for underwriting until stochastic flood models began to emerge just over a decade ago. These models were first produced for more developed countries, due to the greater availability of data and historical records of major flood events.
With increasing demand over the years to quantify flood risk in emerging markets, said Senol Cabi, more flood models have become available, expanding from individual countries to continental and even global products.
“It’s not just the type of data that is used in flood modelling that is key, it’s how the data is utilised.”
Stochastic modelling of flood hazard begins with calculating the frequency and intensity of each event at a location level. But with climate change expected to increase both the frequency and intensity of flood events, the nature of the risk is changing rapidly, making modelling more challenging.
“Arch Insurance invests a lot of time and resources in understanding these models and making sure that we use them wisely. We choose fit-for-purpose models, and then build our view of risk around the available tools and latest technology to capture the current risk,” explains Senol Cabi.
But it’s not only the hazard that is crucial for assessing potential insured losses from flooding, she added. Vulnerability plays a significant role in relating flood intensity to damage ratios for buildings and contents and the impact of business interruption.
“Knowing exactly where your exposure is, what type of building and content you’re looking at, and what kind of information is available on the exposure all play a part in calculating potential losses,” she said.
Current flood models are still in their infancy when it comes to capturing the impact of climate change on flood risk, she added, and over the longer term this needs to be built into models.
“At Arch, we choose fit-for-purpose models, and then build our view of risk around the available tools and latest technology to capture the current risk.”
In addition to modelling natural processes, human interventions — such as the maintenance of flood defences, operation of dams and reservoirs, and decisions on building or upgrading structures — need to be factored in. These are complex scenarios that are lacking in current flood models, adding uncertainty to the modelled output.
If flood risk modelling and management are to improve, argued Senol Cabi, the industry needs to address the lack of a complete and detailed global exposure dataset.
“While high-resolution terrain data is available at a country level, the global picture is far from complete. River gauge stations, which provide an important input to flood models, are missing in many countries,” she explained.
The industry has come a long way in modelling and quantifying flood risk, said Senol Cabi, but with the development of state-of-the-art stochastic models, where location-level outputs are available with multi-peril correlations, the industry should be able to evaluate different sources of flooding – both precipitation-induced inland flooding and coastal storm surge. Artificial intelligence and machine learning techniques will also become increasingly important for assessing flood risk in industry exposure datasets and developing near-real-time flood forecasting and rapid disaster response.
“Capturing multi-peril correlations, understanding the impacts of current risks and assessing how they are changing, and using the latest technologies will help us move forward as an industry,” she concluded.