Climate risk modelling is becoming actuarial work, not specialist climate-science work. The frameworks borrow from cat modelling, the disciplines borrow from reserving, and the people best placed to make the numbers defensible are increasingly insurance actuaries. This is a shift worth naming.
The frame change
Until about two years ago, climate risk inside an insurer typically lived in the sustainability or risk-strategy function. It produced narrative reports, scenario decks for the board, and references to TCFD and ISSB frameworks. The numbers were directional. The models were spreadsheet-ish or vendor-ish. The actuarial team was rarely the owner.
That is changing. Three forces are pulling climate-risk numbers into the actuarial estate.
Reserving exposure. Climate-attributable losses are now a material component of central reserve estimates. They cannot be ringfenced as a stress overlay. The actuary signing the financial position cannot defer the question; they have to own a position on it.
Capital framework integration. SAM, Solvency II and increasingly the IAIS framework on climate risk all expect climate to be visible in capital. ORSA narratives are being held to a higher numerical standard. Stress tests are being graded on assumption defensibility, not whether they are run.
Pricing pull-through. The pricing actuary on a property book in a cat-exposed market cannot price 2026 business off 2020-2022 history without owning a view on the climate signal. That view is actuarial work.
What actuarial climate work looks like
The well-run actuarial climate practices we see have four characteristics in common.
Peril-by-peril decomposition. A single climate-loading factor on the whole book is not actuarial work; it is bookkeeping. Real climate work decomposes the book by peril (windstorm, flood, hail, fire, drought, heat-stress mortality), runs the climate signal through each peril independently, and recombines with explicit dependence assumptions. The dependence assumptions are themselves actuarial judgement.
Tied to underwriting and claims. The numbers only matter if they connect to what the underwriting team is writing and what the claims team is paying. Actuaries who do their climate modelling in isolation tend to produce defensible numbers that nobody uses. Actuaries who sit in the underwriting committee end up with numbers that change the book.
Versioned and reproducible. Climate models are revised frequently — vendor updates, IPCC framework changes, new regional studies. An ORSA or an annual report that quotes a number from a one-shot Excel run cannot be defended six months later. The actuarial discipline of versioned, reproducible runs applies here as much as anywhere; we have argued elsewhere that this is now table stakes.
Honest about uncertainty. Climate signal at the next-five-year horizon is much more confident than at the 2050 horizon. Most actuarial climate work conflates the two. The well-run practice publishes a near-term, well-bounded number and a long-term, deliberately wide range — and is explicit about which is which.
Where this leaves the climate-science specialists
This is not about replacing climate scientists with actuaries. The science is genuinely deep and outside actuarial training. But the integration into a capital, pricing or reserving framework is actuarial work, and most insurers are realising that they need both — a climate-science partner (often academic or consultancy) and an actuarial owner who decides how the science gets used in the numbers that matter.
The pattern that is working: a climate-science partner provides the underlying peril views and uncertainty bounds; the actuarial team integrates those into the existing reserving, pricing and capital pipelines, applies dependence assumptions, calibrates against actual experience, and signs off the result.
What we see go wrong
Three failure modes recur.
Outsourcing the position. The insurer commissions a climate-risk report from a consultancy, files it, and writes “see report” in the ORSA. The number is not owned internally. When the regulator probes, nobody can defend it. The fix is to insist on internal ownership — the consultancy can do the analysis, the actuarial team must own the conclusion.
Conflating physical and transition risk. Physical risk (cat losses, mortality) and transition risk (asset impairment from policy or technology shifts) are radically different in horizon, in mechanism and in the data needed to model them. Treating them as one number obscures both. The two need separate models and separate ownership; they meet only when the numbers are combined into capital.
Optimism on the long horizon. Long-horizon climate models are often run on benign middle scenarios (RCP4.5, “well below 2°C”) because the more severe scenarios produce numbers nobody wants to defend. This is not actuarial conservatism; it is the opposite. The honest version runs both, shows the spread, and lets the board carry the actual uncertainty.
Where this is heading
Within five years, climate-risk modelling will be a routine part of the actuarial qualification syllabus and a standard part of any signed actuarial report. The transition is already underway. The actuaries leading it are not climate scientists; they are insurance actuaries with a working understanding of the science and a strong grounding in reserving and capital. That combination is rarer than it should be, and it is what the next generation of insurance and reinsurance work increasingly requires.
If you are scoping the actuarial side of a climate-risk programme, our Finance Modernisation practice does this work — pricing, reserving and capital integration, with the lineage and reproducibility the regulator now expects.