Modelling flood on a global scale has long been needed within the re/insurance industry but, without global hazard and event set data, has previously been impossible.
Stochastic and historic events are a key component of catastrophe models, enabling probabilistic assessment of damages and loss as a result of flood events. At national level, event sets contain many thousands of plausible hazard events which are both statistically similar to events in the recent past and events that extend beyond the limits of the historical record, expressed by their severity and geographical extent.
But event sets constrained to a single country or geographical region have limited utility for international exposure assessment. JBA has developed and implemented a global methodology that not only benefits from a large pool of data but gives users a consistent and correlated event set for assessing international portfolios.
This global event set methodology is what makes our new Global Flood Model possible, the first to enable probabilistic flood modelling at any location worldwide – we explore the event set component of the Global Flood Model in this blog.
The need for something global
Capturing the spatial extent of flooding within an event set is challenging, yet important. Floods don’t stop at political borders and can span large geographical areas, especially along extensive river systems that flow through many countries. Flooding on the Danube in June 2013 impacted large swathes of Austria, Germany, Hungary, Slovakia and the Czech Republic, while the South Asia floods of August 2017 caused extensive damage and loss of life in India, Bangladesh, and Nepal (BBC, 2013; Guardian, 2017). Weather systems which give rise to heavy rainfall can be very large and travel for long distances leading to surface water flooding in multiple countries.
As a result, organisations with international exposure require models that can assess risk across whole catchments regardless of their size, using globally consistent methodology and data, for both river and surface water flooding in order to fully understand the impact of flood events on their business.
JBA has addressed this challenge and developed a Global Flood Event Set (GFES). Our GFES includes over 15 million river and surface water flood events across the entire globe and is used within our Global Flood Model.
Overcoming the challenge for a globally consistent methodology
Developing global datasets, especially ones based on complex statistical methods, is far from straightforward. Two immediate challenges arise:
- The need for consistent data with global coverage
- Computational demands given the vast scale of the problem.
With a unique combination of sophisticated statistical and physically-based models applied to carefully selected global data, our GFES methodology produces events which preserve the spatial and temporal patterns of flooding. Because the approach is driven by rainfall, the correlation structure of combined river and surface water flooding is maintained, ensuring that the estimated frequency and intensity of simulated events is robust. There are three main stages in our GFES methodology:
- Rainfall modelling
- Streamflow modelling
- Event selection.
The default event definition is for natural hazard events which can last days or even weeks, although events may also be re-defined to enable exposure analysis in accordance with reinsurance terms of contract such as a typical hours clause (e.g. 504 hours). Each event in the catalogue represents a cluster of river and surface water extremes which may be entirely contained within a single country or across international boundaries as shown in Figure 1. 
Figure 1: Three inland flood hazard events with GFES catalogue numbers 00803505, 00904753, and 00679464. Event 00803505 extends over 618 simulation points and impacts two countries, China and Mongolia. Event 00904753 extends over 766 simulation points and impacts four countries, China, Myanmar, Laos, and Thailand. Event 00679464 extends over 598 simulation points and impacts three countries, China, South Korea, and Japan.
What can the Global Flood Event Set be used for?
The GFES is vital in ensuring that portfolio and scenario analysis for multiple regions is consistent and robust.
When used in combination with JBA’s Global Flood Maps, it can be used for portfolio stress testing, optimisation and the identification of exposure hotspots across new markets. The stochastic nature of the GFES also makes it suitable for probabilistic modelling and inter-regional event correlation analysis, through JBA’s own Global Flood Model or through incorporation into an in-house model as the event set is platform agnostic. In addition to modelling inland flood, our event generation methodology is flexible meaning that it can be extended to include other perils such as coastal inundation, which is available in JBA’s UK Flood Model. Consultancy services using the GFES are also available.
For further information about the GFES and the Global Flood Model, please get in touch. We'll also be exploring how the GFES can be used to help close the protection gap, using China as a case study, in a future blog.
