Simple catastrophe model: Risk assessment from a cyclone case

Introduction

Components of Catastrophe Model

Hazard

Vulnerability

Introduction

Imagine a powerful cyclone makes landfall near a densely populated coastal region. Infrastructures like homes, businesses, and other critical buildings lie in its path. How can we assess the potential damage and financial losses from such a devastating event? This is where catastrophe models come in.

These models are essential tools for the insurance and reinsurance industries, helping them understand and quantify the risk posed by cyclones. By estimating potential losses, catastrophe models enable insurers to set appropriate premiums, manage their exposure to risk, and ensure they have the financial capacity to pay out claims in the event of a disaster. Re-insurers, who provide insurance to insurance companies, also rely heavily on these models to assess and manage their own risk portfolios.

In this blog post, we'll explore a simplified catastrophe model and apply it to a real-world cyclone case study. We'll see how these models help us quantify and manage risk, playing a vital role in building resilience against the devastating impacts of cyclones.

Components of Catastrophe Model

Catastrophe models are built on a foundation of four key components: hazard, vulnerability, exposure and financial. These work together to give us a comprehensive picture of risk. The first three components form the classic "Risk Triangle", a fundamental concept in understanding and managing risk.

Hazard

This module contains a vast catalogue of simulated events, each representing a possible peril scenario such as earthquake, cyclone/hurricane, outbreak of a war or disease etc. Consider it as a library of "what-if" scenarios, covering a wide range of intensities, locations, and paths. The hazard module tells us how often different types of perils are likely to occur and where they might strike. Each event comes with its own unique "footprint" – a map showing the extent and intensity of the hazard over its path.

Vulnerability

This module is all about how susceptible assets are to damage. It uses clever equations called "damage functions" to estimate the expected damage based on the asset's characteristics (for example in case of buildings; construction material, age, and height) and the local intensity of the cyclone (e.g., wind speed).

Exposure

Here's where we bring in the real-world assets: in case of cyclones/hurricanes, the buildings and infrastructure that might be affected. The exposure module contains detailed information about their location, value, and any existing insurance policies, including deductibles, limits, and reinsurance coverage.

Financial

This module crunches the numbers to calculate the financial impact of each simulated peril event. It takes into account the damage estimates, insurance policy details, and any reinsurance arrangements to generate a range of possible financial losses. By aggregating the losses from all the scenarios, the model produces a loss probability distribution, a powerful tool for understanding the likelihood of different loss levels.

Figure 1: Catastrophe modelling framework discussed in Mignan, A. (2022).

Catastrophe Model

I have created a simple catastrophe model that assumes Asset Value (AV) or Total Insured Value (TIV) (as it is not available freely) for the buildings exposed to strong winds during a landfall event of a cyclone leading to damage to infrastructures. Landfall event is the event when cyclone crosses the coastal boundary in land regions.

Meteorological data to assess landfall conditions is taken from ERA5 reanalysis (Hersbach et al. (2020)). Building data as a geospatial information in the exposed region is assessed from OpenStreetMap (OSM; Boeing (2024)). There can be place to place variation in quality and completeness of the data this data. However, a wealth of information about the characteristics of those features can also be found with the data such as building type, building levels, building material etc. This "metadata" is what makes OSM so powerful for analysis and exposure modelling.

Maximum wind location (in this model but not limited to in reality) is assumed to be the location of landfall event and damage to infrastructure is considered by strong winds and heavy precipitation alone.

Maximum wind location approximation can limit model's ability to calculate damage before the landfall and also can differ in real landfall event as the strongest winds aren't always located precisely at the centre. Hence, in reality strongest finds can be located on the land even before eye reaches the land.

Damage is assumed to be linearly varying with both wind and precipitation, which is another assumption in this model. However model calculates the decay in wind-speed as we move away from the eye using a linear decay function. As mentioned already this is a Simple model just for the understanding and hands-on. This model calculates deterministic losses only. It will be interesting to add probabilistic features to this model in meteorological conditions to get probabilistic loss metrics.

Model can be found at https://github.com/JiveshDixit/Simple-Catastrophe-Model/tree/main/CAT_model. I have also shared a Jupyter-Notebook at https://github.com/JiveshDixit/Simple-Catastrophe-Model/tree/main/Test_Haiyan to test the model for Haiyan typhoon in Philippines (Nov. 2013). Model produces a list of buildings that had to face damage during to the event and saves them into a CSV file. It also produces an interactive map of the geotagged buildings that faced damage or no-damage to the assumed criteria. More complex criteria can give more realistic results.

Using the model

Jump in to get a hands-on to the model using the jupyter notebook link given in last section. You can also have a look at the test conducted on Typhoon Haiyan. If you want to try out other cyclone case, you'll need to change some parameters such as cyclone_name, range of latitude and longitude to change the analysis region for the cyclone, time period of interest, Place_name for vulnerability and exposure analysis.

Test_Haiyan-1.pdf

I hope this exercise gives you a start on catastrophic modelling. Feel free to contact me for any questions and clarifications.

References

Mignan (2022). Categorizing and Harmonizing Natural, Technological, and Socio-Economic Perils Following the Catastrophe Modeling Paradigm. International Journal of Environmental Research and Public Health, 19(19), 12780. https://doi.org/10.3390/ijerph191912780

Hersbach et al. (2020). The ERA5 global reanalysis. Q.J.R. Meteorol. Soc., 146, 1999–2049, https://doi.org/10.1002/qj.3803.

Boeing (2024). Modelling and Analysing Urban Networks and Amenities with OSMnx. Working paper. https://geoffboeing.com/publications/osmnx-paper/