Importance of exposure data quality versus uncertainty in vulnerability and hazard for catastrophe modelling

Catastrophe (cat) models are widely used to combine information on the probability distribution of hazard intensity, exposure location, and exposure vulnerability to quantify risk, usually expressed in terms of financial loss. While substantial attention has been paid to improving hazard and vulnerability components (including incorporating climate change), exposure data often lags in terms of quality and detail and may vary widely in granularity and reliability. For instance, reinsurers frequently receive aggregated portfolios from insurers, which may lead to loss of critical information about location-specific risks. This lack of detail undermines the precision of loss estimates, even if hazard and vulnerability components are highly refined. This raises an important question: how influential is the level of detail exposure information on risk estimates with respect to uncertainties in vulnerability and climate change model?

In this presentation we will answer this question via a global sensitivity analysis (GSA) of the JBA flood cat model. GSA is a methodology to systematically investigate the propagation of input uncertainties through mathematical models and quantify the relative importance of those uncertainties on the variability of model outputs. Differently from local sensitivity analyses, in GSA all input uncertainties are varied simultaneously within their plausible variability ranges, instead of being varied one at the time from a baseline. This enables us to capture interaction effects between uncertain inputs and ensure that sensitivity results are not conditional on the chosen baseline. In our application, the three input uncertainties are hazard (including climate change), vulnerability, and exposure data and we quantify their relative influence on financial loss estimates.

Overall, the analysis and the results will highlight how hazard, vulnerability and exposure data quality impact loss estimates guiding cat model developers to prioritize their efforts on model improvement and reinsurers to leverage better quality exposure data.