Skip to main content
    JOJonas Osman
    All articles
    Climate
    Modelling
    Insurance
    Banking

    From Hazard to Loss: A Financial Framework for Physical Climate Risk

    A calibrated hazard–exposure–vulnerability–loss chain that turns climate heat maps into balance-sheet numbers regulators and boards will accept.

    By Jonas Osman Abdelghafour · · 11 min read
    From Hazard to Loss: A Financial Framework for Physical Climate Risk — technical illustration by Jonas Osman Abdelghafour, climate and modelling and insurance and banking risk modelling
    From Hazard to Loss: A Financial Framework for Physical Climate Risk — Climate · Modelling · Insurance · Banking

    In March 2024, the European Environment Agency published the first European Climate Risk Assessment — EUCRA — a continent-wide stocktake identifying 36 major climate risks to Europe's energy and food security, ecosystems, infrastructure, water resources, financial stability and public health, and warning that many have already reached critical levels (EEA, 2024). The assessment's underlying facts are stark: Europe is the fastest-warming continent, having warmed at roughly twice the global rate since the 1980s, and European Commission analysis cited alongside EUCRA suggests annual coastal-flood damages alone could exceed EUR 1 trillion by 2100 under adverse pathways (EEA, 2024; European Commission, 2024).

    Numbers of that scale get attention in boardrooms. But here is the uncomfortable truth from the modelling trenches: most financial institutions still cannot translate a report like EUCRA into a number on their own balance sheet. They can point to a heat map — this region is red, that portfolio is amber — but a heat map is not a loss estimate. It cannot price a policy, provision a loan, set a capital buffer or justify a divestment. Somewhere between the climate science and the financial statement, the analysis breaks.

    The business problem: an unbroken chain or no number at all

    The reason is structural. Physical climate risk only becomes a financial quantity when four distinct questions are answered in sequence, and most institutional analyses answer only one or two of them:

    • Hazard — what physical events, at what intensity and frequency, can occur at each location, now and under future climates?
    • Exposure — what assets (buildings, collateral, infrastructure, insured objects) sit at those locations, and what are they worth?
    • Vulnerability — when a hazard of given intensity meets an asset of given type, how much damage results?
    • Loss — how does physical damage translate into a financial outcome for this institution, given insurance terms, loan structures, or ownership shares?

    This is not a proprietary framework; it is the risk concept of the IPCC's Sixth Assessment Report, in which risk emerges from the interaction of hazard, exposure and vulnerability — and it is the concept EUCRA itself explicitly applies (IPCC, 2021; Climate-ADAPT, 2024). What financial modelling adds is the fourth link, and the discipline of calibrating and validating each link separately. When an analysis produces no defensible number, the diagnosis is almost always a broken or missing link, not a failure of climate science.

    Overlaid physical hazard intensity layers for flood, wind and heat exposure, illustrating "The business problem: an unbroken chain or no number at all" in From Hazard to Loss: A Financial Framework for Physical Climate Risk
    Figure 1. Overlaid physical hazard intensity layers for flood, wind and heat exposure, in the context of the business problem: an unbroken chain or no number at all.

    The mathematics of the chain

    The canonical formulation is simple to state. For a portfolio of assets indexed by a, the loss from a hazard event with local intensity h(x_a) at asset location x_a is:

    L = Σₐ Vₐ · Dₐ(h(xₐ))

    where Vₐ is the value of asset a (the exposure link), and Dₐ(·) is the damage function for that asset type (the vulnerability link) — a curve mapping hazard intensity, such as flood depth in metres or peak gust speed, to the fraction of the asset's value destroyed. Summing over assets gives the event loss; running the calculation over a large catalogue of simulated events, each with an annual occurrence rate, gives the full loss distribution, from expected annual loss to tail return periods. Readers of the first article in this series will recognise the structure: it is exactly the four-module catastrophe-model chain, generalised beyond insurance.

    Each link carries its own data, its own calibration problem, and its own uncertainty:

    Hazard is where climate science enters. Historical hazard maps must be adjusted for non-stationarity: a "1-in-100-year" flood defined on 20th-century data is not a 1-in-100-year event in the climate of 2040. Forward-looking hazard requires climate projections — and for long horizons, scenarios. The NGFS Phase V long-term scenarios, published in November 2024, are the reference set for financial institutions; notably, Phase V introduced a new damage function calibrated on a broader set of climate variables, resulting in substantially larger estimated physical-risk impacts than earlier vintages (NGFS, 2024). That revision is itself a lesson: physical-risk numbers are conditional on modelling choices that remain in flux, and an updated methodology is already announced for the next long-term release (NGFS, 2025).

    Exposure is where most practical projects fail quietly. Asset registers with missing coordinates, collateral geocoded to postcode centroids, values recorded at historical cost rather than reinstatement cost — every such defect propagates directly into the loss estimate. Exposure work is unglamorous and decisive.

    Vulnerability is, in my experience, the largest and least acknowledged source of uncertainty. Damage functions are typically fitted to sparse claims or engineering data, transferred across regions and building codes where they were never validated. Two analyses using identical hazard maps and identical portfolios can diverge by multiples purely on the choice of damage curves. Any physical-risk result presented without sensitivity analysis on the vulnerability link should be treated as incomplete.

    Loss translation is institution-specific. An insurer applies deductibles, limits and reinsurance. A mortgage lender asks a different question: does the damage impair collateral value and the borrower's ability to pay — and is the property insured, shifting the loss elsewhere? The same physical event produces different financial risk for different balance sheets, which is why borrowed "climate scores" rarely survive contact with a validator.

    Prior and posterior densities illustrating Bayesian parameter updating, illustrating "The mathematics of the chain" in From Hazard to Loss: A Financial Framework for Physical Climate Risk
    Figure 2. Prior and posterior densities illustrating Bayesian parameter updating, in the context of the mathematics of the chain.

    A worked example: flood risk in a mortgage book

    Consider a bank assessing flood risk in a residential mortgage portfolio — a workflow of the kind Quantica Risk is developing in its climate-analytics frameworks:

    1. Geocode every collateral property to coordinates, flagging records that only resolve to postcode level for uncertainty treatment.
    2. Overlay hazard: intersect locations with current fluvial and pluvial flood maps at several return periods, then apply climate-conditioned adjustments for the scenario horizons under review.
    3. Apply vulnerability: depth–damage curves by property type produce a damage ratio per property per return period — with a documented sensitivity range on the curves themselves.
    4. Translate to loss: damaged value feeds updated loan-to-value ratios; combined with insurance-coverage assumptions, this yields the shift in loss-given-default, and, through affordability effects, a view on default probability.
    5. Aggregate and stress: portfolio results per scenario, reported with the vulnerability and hazard sensitivities alongside the central estimate.

    Nothing in this workflow is exotic. Its value lies in the fact that every step is explicit, every assumption is inspectable, and the final number can be decomposed back into its drivers when a supervisor — or a sceptical CFO — asks where it came from.

    Implications for risk leaders

    Three practical conclusions. First, audit your analysis against the chain: if you cannot state what hazard data, exposure register, damage functions and financial translation produced your climate number, you do not yet have a climate model — you have a graphic. Second, direct investment where uncertainty is largest, which for most institutions means exposure data quality and vulnerability calibration rather than ever-more-detailed climate projections. Third, treat scenario outputs as conditional statements, not forecasts: the material revision of physical-risk estimates between NGFS vintages shows how sensitive results are to methodological choices, and governance frameworks should present ranges and sensitivities rather than single points.

    Capital ratio trajectory under baseline and adverse stress paths, illustrating "Implications for risk leaders" in From Hazard to Loss: A Financial Framework for Physical Climate Risk
    Figure 3. Capital ratio trajectory under baseline and adverse stress paths, in the context of implications for risk leaders.

    Conclusion

    EUCRA has given Europe's institutions no excuse to claim the risks are unknown; 36 of them are now catalogued, several rated critical. The remaining gap is translational, and it is the province of quantitative risk teams: building the calibrated, validated chain from hazard through exposure and vulnerability to institution-specific loss. That chain is the foundation on which every subsequent article in the climate pillar of this series — flood modelling, compound perils, scenario governance — will build. Institutions that construct it well will find that climate risk stops being a reporting burden and becomes what it should have been all along: an input to pricing, lending and capital decisions.

    About the author. Jonas Osman is a risk-modelling and financial-risk professional and the founder of Quantica Risk, an AI-driven modelling company focused on insurance, banking, climate risk, actuarial analytics, and model validation.

    Quantica Risk develops transparent, data-driven modelling frameworks for financial institutions and risk-sensitive businesses. To discuss climate analytics, catastrophe risk, or model validation, visit the Quantica Risk website. [website — to be inserted when verified]

    • Next in this pillar: Flood Risk at Asset Level: Return Periods in a Non-Stationary Climate (Article B2) — the hazard and vulnerability links examined in depth for Europe's most consequential peril.
    • Related: The Future of Catastrophe-Risk Modelling: From Black Boxes to Transparent Frameworks (Article A1) — the insurance-native origin of the four-module chain.
    Schematic of a credit risk parameter chain linking exposure, default probability and loss given default, illustrating "Series links" in From Hazard to Loss: A Financial Framework for Physical Climate Risk
    Figure 4. Schematic of a credit risk parameter chain linking exposure, default probability and loss given default, in the context of series links.

    References

    Related articles

    Work with Jonas

    For partnerships, modelling reviews, advisory roles, institutional risk projects or technical collaboration.

    Contact Jonas