Climate Risk Assessment for Manufacturing: A Worked Example

Portfolio headline (ssp585, 2050)

Where the portfolio sits

Dot color is the 2050 composite rating under the high scenario. Dot size scales with flood EAD. The financial exposure clusters on the US riverine sites; the coastal and emerging-market sites carry their risk in hazards, not in this flood number.

Which hazards threaten the most sites (materiality-weighted, ssp585 2050)

Start with the portfolio question: which hazards threaten the most sites? Breadth is the count of
sites where a hazard is both highly rated and material to operations (a High+ rating and a real operational dependency),
high scenario, 2050. It pairs with the worst-case grid further down. Breadth says what to tackle portfolio-wide,
peak says where to look site-by-site. Materiality weighting is defined below; hover a bar for its critical-severity and
pre-materiality (raw) counts.

Site screen (ranked by composite, ssp585 2050)

A high composite does not always mean high flood loss. Pune and Chennai screen near the top on composite (water stress, storm) yet carry zero modelled flood EAD. The financial flood exposure and the hazard exposure are two different lenses.

What the composite score is, and why we re-weight it

The composite is a single 1-to-5 number that blends all twelve hazard ratings for a site into one
headline. It is useful for ranking, but it has a built-in weakness: averaging twelve hazards flattens the signal.
A site with one Extreme hazard and eleven Lows averages out to a calm-looking Moderate. And a flat average treats a
foundry’s exposure to heat exactly like a warehouse’s, even though heat threatens the foundry’s furnace cooling and its
workers while barely touching the warehouse.

So we keep the platform’s number as the anchor and apply materiality as a tilt on top of it:

• Platform composite: the score the assessment returns (its own roll-up of the twelve hazards, with geographic modifiers). The starting point.
• Materiality-adjusted: the platform score shifted by the materiality tilt, so heat counts more at a foundry, storm and sea level rise more at a coastal port, and so on.

The tilt is measured cleanly: take an equal-weight average of the twelve ratings, then a materiality-weighted average
of the same ratings. The difference between those two is purely the effect of the weights, with the engine’s method cancelling out.
We add that difference to the platform composite. The two lenses don’t double-count: the platform’s modifiers are geographic,
materiality is operational.

Materiality map: which hazard matters to which site

Each cell is how much a site’s operations depend on being spared a hazard, on a 0-to-2 scale
(0 not material, 2 a critical dependency). These weights describe the business, not the climate. They are the same
whatever the scenario. Read a row to see a site’s operational profile; read a column to see which sites a hazard would hurt most.

Weights encode operational dependency only, never dollar consequence. The money lives in the
flood EAD layer below. Keeping them separate stops the two lenses from double-counting. On a real engagement these come from
the asset’s own profile (water withdrawal, process heat, single-source status); here they are illustrative, anchored in the
kind of dependency materiality ISSB and SASB describe for the sector.

The composite, re-weighted for what each site depends on

Flood Value-at-Risk: where the money is, and what kind of number it is

Loss-exceedance curve and loss composition

Area under the curve: how the annual flood loss is computed

The headline EAD (expected annual damage) is the area under the loss-exceedance curve:
the average yearly loss you would book if you ran this site for a very long time. We have the loss at four return periods;
EAD is the area they enclose, found by joining the points with straight lines and adding up the strips (trapezoidal
integration). The x-axis is annual probability (not return period). That is what makes the shaded area
equal the expected annual loss. Two sites, ssp585 2050, with very different curve shapes:

The shaded region is the EAD. Each dashed line splits it into a trapezoid.
The table below works it strip by strip for Valencia (the sloping curve, where the method is easiest to see).

Hazard matrix for (12 hazards, 4 horizons)

The three pathways bracket the range. SSP1-2.6 is the Paris-aligned, well-below-2C low case and shows the exposure that is locked in even if the world decarbonises. SSP5-8.5 is the high stress case. Near-term horizons can flip one band between pathways because a single model carries internal variability; the ordering settles by 2050.

Portfolio worst case: highest rating across all 10 sites

Read this as the portfolio envelope. For any hazard, horizon and scenario it shows the worst single site, which is the screening signal for where to look first.

Site detail:

Projection vs record: what is modelled vs what has happened

How to read this, and what it does not cover

Related reading

• What a physical climate risk assessment is
• How to interpret a climate risk assessment report
• Expected annual damage (AAL/EAD)
• Loss-exceedance probability
• Return periods explained
• Probable maximum loss (PML)
• HAZUS depth-damage method
• Climate Value-at-Risk
• Climate risk methodology
• Flood depth methodology
• Manufacturing disclosure example (SB 261)