How to Aggregate Climate Risk Across a Property Portfolio Without Losing Local Detail

Why the portfolio average is the wrong answer

The most common mistake in climate risk reporting is to compute a single portfolio score and call it a day. Physical climate risk is not normally distributed across a book of property. A diversified portfolio can look calm on average while a handful of assets quietly concentrate flood, wildfire or heat exposure. Once those individual buildings are the source of lost tenants, uninsurable premiums or impaired collateral, the average did not warn anyone in time.

This is why experienced investors treat climate risk through several lenses at once—regional, market and asset level. A broad regional index can act as an early 'yellow flag' that directs underwriting effort to markets that deserve a deeper look, but it can never substitute for the asset-level judgement that determines whether a specific building survives a hazard intact. Aggregation should therefore be a deliberate act of rolling bottom-up detail upward, not a top-down smoothing exercise.

The practical consequence is structural: you need a portfolio risk framework that can show both a defensible aggregate number for disclosure and the granular information that managers actually act on. Building that two-speed picture begins with how you record and score each asset.

• Report the mean only together with the 90th/95th percentile and the list of highest-risk assets.
• Treat each hazard separately; a combined 'physical risk' score can hide the hazard that matters most.
• Never let a market or regional average replace a site-level assessment.

A bottom-up data model that survives roll-up

Aggregation is only as trustworthy as the per-asset records underneath it. Every property needs a stable identifier linked to precise coordinates, not just a postcode or administrative region. Evidence from mortgage stress testing shows why location resolution matters: two loans in the same NUTS3 region can carry very different flood risk depending on proximity to water, so models anchored only on regional house-price shocks miss the material contributors to credit risk.

For each asset, record the hazards that are genuinely material to that location—river and pluvial flood, coastal and sea-level rise, heat stress, wildfire, storm, drought, subsidence—rather than applying one generic list. Capture building characteristics that change vulnerability: elevation, ground-floor construction, roof and cladding, mechanical equipment placement, drainage, basement use and occupancy type. Finally, log protective infrastructure such as sea walls, levees, pumping stations and municipal flood management, and note over what horizon that protection is assumed to hold.

Industry guidance increasingly reflects this structure. The IIGCC Physical Climate Risk Appraisal Methodology brings climate science, engineering and finance together precisely so that physical risk is assessed at the place and scale where it materialises, then aggregated for an investor, fund or portfolio view. Building your asset registry with this discipline means the local detail is never lost when you scale up.

Scoring hazards at asset level with scenarios and time horizons

For each material hazard, use hazard-specific models rather than one blunt score. Flood, heat and wildfire have entirely different spatial logics and damage mechanisms. Location-driven scores, often expressed on a 0–100 scale for a given peril, allow a distribution of outcomes across the portfolio rather than a categorical yes/no. The key is consistency: apply the same model family, scenario and reference period to every asset so the roll-up is comparable.

Scenario selection should follow the purpose of the assessment. Under the EU Taxonomy climate adaptation objective, companies must consider the lifespan of eligible activities and bring in future IPCC scenarios for activities with a lifespan of at least ten years—which covers most real estate. Related reporting rules ask firms to identify and assess material physical and transition risks and their financial effects. In other jurisdictions the same logic appears through IFRS S2 and its TCFD lineage. Choose a short-horizon 'today plus near term' view for operations and insurance, and a long-horizon high-emissions scenario for capital allocation and lending decisions that lock in risk for decades.

Be explicit about uncertainty. Climate projections differ across providers, and many raw datasets ignore adaptation measures, so apply an overlay of judgement. Ask what mitigation infrastructure exists now, how long it protects the asset, and whether future plans are funded and likely to be delivered.

Aggregating without erasing tail risk

Once every asset has hazard scores and a loss model, aggregate in a way that respects correlation. Summing exposure-weighted scores gives a rough portfolio temperature, but true financial loss in a single storm or flood depends on how many vulnerable assets are hit at once. Correlated events mean portfolio loss is not the simple sum of independent per-asset expected losses—it can be larger in the tail, or smaller if hazards genuinely diversify across geographies.

In practice, build the roll-up in layers. First produce an exposure-weighted view by value at risk and insurance value. Second, report the distribution: median, high percentiles and concentration metrics such as the share of portfolio value in high-hazard locations. Third, stress the portfolio against a small number of severe, correlated event scenarios—for example a 100-year flood across one basin or a heat wave across a region—to see which clusters drive the worst outcomes.

Keep the tail visible in reporting. Disclosing a single 'portfolio in a high-risk zone' percentage hides the concentration that matters. The most useful reporting combines a headline aggregate for regulators and investors with an annex of asset-level scores, the assumptions behind them, and the mitigation measures already in place.

• Layer exposure-weighted totals, loss distributions and correlated-event stress scenarios.
• Track the share of portfolio value in the highest-hazard band as a concentration metric.
• Pair every portfolio figure with the underlying assumptions and per-asset detail.

Regulatory and market drivers to design for

The pressure to aggregate responsibly is no longer voluntary. Central banks have moved physical risk into prudential supervision: stress-testing guidance and finalized requirements increasingly expect lenders to bring flood and other hazards into credit assessment, and insurers and investors face parallel expectations. Regulators and supervisors acknowledge that no single universal methodology exists, so the burden is on each organization to run a defensible, documented process.

For property owners reporting in the EU, the practical baseline is a climate risk and vulnerability assessment per asset that feeds both EU Taxonomy 'do no significant harm' checks and CSRD-style materiality disclosure. Outside the EU, equivalent expectations are emerging from IFRS S2-aligned frameworks and local supervisors, sometimes via mortgage-lending rules. Where a company operates across jurisdictions, design one asset-level assessment that can be tagged and re-aggregated to satisfy several regimes at once.

Rules continue to evolve, so treat the current requirements as a moving target. What will not change is the underlying direction: supervisors want to see that your headline climate number is built from defensible, location-specific analysis rather than a coarse regional average.