The distribution of aggregate storm risk in a changing climate

TL;DR

This paper develops a theoretical framework to understand how aggregate storm risk distributions change over time in a changing climate, incorporating new equations and applying them to climate model data.

Contribution

It introduces a new equation relating storm frequency and aggregate risk correlation, accounting for climate change effects, and applies the theory to simulated climate data.

Findings

Covariance between event occurrence and risk is the product of expected risk and dispersion.

The new 'J-equation' links correlation to storm intensity distribution shape.

Theory validated with simulated climate model data.

Abstract

The financial losses from extreme weather events can have a disastrous effect, often costing billions of pounds. While changes in the disposition of individual events is of importance to both the insurance and re-insurance industries, these companies are often concerned with the aggregate risk posed in a season. This project explores how the statistical properties of aggregate risk measures may change when, to reflect the earth's changing climate, models are made time dependent. Historical random sum equations by Wald (1945) and Blackwell and Girshick (1947) are used to develop a relationship between the frequency of events and the aggregate risk. The covariance between the occurrence of events and aggregate risk is found to be the product of the expected value of the aggregate risk and the dispersion statistic. Furthermore, a new equation (the "J-equation") relates the correlation between the frequency of events and aggregate risk to the shape of the distribution of storm intensities and the dispersion statistic. This equation highlights that the correlation between the two variables is invariant to a change in the scale of the severity distribution. The theory is applied to a simulated future dataset from the 2020 Norwegian climate model NorESM2-LM. As the data observes the theory presented, this opens the door for these results to be applied to wider geographical regions in future studies.