A Risk Professional’s Guide to Physical Risk Assessments

Climate change is driving an increase in the frequency and severity of physical risk events, such as floods, storms, and heatwaves. Yet quantifying the impacts on individual firms and assets remains complex and uncertain, requiring granular spatial analysis. Regulators are setting expectations for firms to embed climate risk considerations, prompting many to turn to third-party vendors for expertise and data. However, even leading vendors can deliver strikingly different results.

GARP conducted a benchmarking study of 13 physical risk data vendors to understand the differences in their approaches to physical risk assessment at the asset level and to quantify the variation in their outputs. The study is the first comprehensive, comparative assessment of vendor physical risk models, and focuses on how vendor estimates differ, the reasons for such differences, and practical lessons for financial institutions.

This study was undertaken on behalf of the Climate Financial Risk Forum and was published as part of their 2025 outputs.

Key Takeaways:

• There is a large spread of hazard and physical risk estimates. For the 13 vendors in the survey, this dispersion was observed for both estimates of the physical hazards and the damages associated with those hazards.
• Complex modelling underlies the dispersion. Although dispersion is not surprising given how complex the underlying modelling and methodological choices are, the scale of the dispersion is significant. For example, a property can be assessed as either highly exposed to flooding by one or more vendors, but not at all exposed by other vendors.
• Estimates of vulnerability are highly variable. When we looked at the relationship between the mean estimates of hazards and the mean damage ratios, there was a sensible and intuitive link. However, this relationship was extremely variable at the level of the individual vendor. Estimates of damages vary considerably, with each vendor having their own perspective on vulnerability.
• There is little standardisation of metrics. For some hazards, there is a degree of comparability: metrics for flooding were the most comparable, followed by cyclones and windstorms. However, there are several hazards where there is no standardisation, which makes comparisons across vendors challenging. However, different metrics may be more suitable for different use cases.
• Measuring how dispersion changes over time and with changes in severity is challenging. When measured by standard deviation, the dispersion of physical risk assessments increases for time horizons further in the future and by the severity of the return period. But when dispersion is measured by the coefficient of variation (standard deviation divided by the mean), we get the opposite result, since the mean of the distribution is rising faster than the standard deviation. Overall, this may simply indicate that all vendors are anticipating rising physical risks on average, driven by climate change.
• Asset location accuracy can make a big difference to the results. The quality of asset location information impacts both the estimate of a hazard’s severity and the associated physical risk assessment. Even small differences in location can make a difference to estimates for some perils — for example, flooding, but is far less important for perils like heat.
• There are broader impacts of climate change that this study does not address (although some vendors do examine these). This study offers useful insights into vendor estimates of physical risk, but it’s important to recognise broader uncertainties. Some arise from material assumptions — such as asset vulnerabilities, static flood defence standards, and simplified exposure pathways — which can drive significant differences in estimates. Beyond this, systemic climate impacts on food systems, health, and migration are difficult to model but increasingly relevant. Financial institutions will need to build internal capacity and collaborate with vendors and the broader research community to assess risks that extend beyond asset-level hazards and losses.

About the authors

Jo Paisley, President, GARP Risk Institute, has worked on a variety of risk areas at GARP Risk Institute, including stress testing, operational resilience, model risk management, and climate and environmental risk. Her career prior to joining GARP spanned public and private sectors, including working as the Director of the Supervisory Risk Specialist Division within the Prudential Regulation Authority and as Global Head of Stress Testing at HSBC.

Dr. Maxine Nelson, Senior Vice President, GARP Risk Institute, currently focuses on climate and environmental risk management. She has extensive experience in risk, capital, and regulations gained from a wide variety of roles across firms including Head of Capital Planning at HSBC. She also previously worked at the UK Financial Services Authority, where she was responsible for counterparty credit risk during the last financial crisis.