Advances in numerical weather prediction, data science, and open-source software herald a paradigm shift in catastrophe risk modeling and insurance underwriting

Recent advances in numerical weather prediction, combined with the new generation, high-resolution climate simulations, and open-source loss modeling frameworks, herald a move beyond the limited statistical representation of catastrophe risk based on past observations. In this new forward-looking view of risk, an appreciation that our observed record of past natural catastrophes represents a limited sample of possible events, and that the statistics of weather and climate are changing as the planet warms, highlights a key limitation in traditional catastrophe modeling approaches that are built on defining statistical relationships using the observed record. Instead, ensembles of new spatially and dynamically consistent simulations of weather and climate provide physically plausible, but as-yet-unseen events at scales appropriate for making effective risk management and risk transfer decisions. This approach is especially useful in locations around the world where observational records are unobtainable or of short historical duration, such as in low-income countries. We take a forward-looking approach at the way that future catastrophe modeling and insurance underwriting could occur in response to these technological and scientific advances, using open-source loss model frameworks.