Reinsurance-Grade Flood Loss Models in South Africa: What That Actually Requires

“Reinsurance-grade” is a phrase that gets attached to flood models more often than it is earned. When a reinsurer or a treaty actually relies on a loss model, it is holding capital against the numbers — so the bar for what the model has to demonstrate is genuinely higher than for a screening layer. Here is what the phrase should mean, and what it demands of a flood loss model built in the South African data context.

Holding capital against a South African flood model? We build locally calibrated hazard layers and are explicit about where the data supports the answer and where it does not.

Hydraulic modelling

The three pillars of any loss model

Every catastrophe loss model rests on three components, and reinsurance-grade means each is defensible in its own right:

• Hazard — the physical flood: extent, depth, and velocity across the relevant return periods, ideally as a full set of events rather than a handful of return-period maps. This is the flood hazard mapping layer, and it has to be calibrated, not just plausible.
• Exposure — what is actually at risk: the locations, values, and characteristics of the insured assets. A precise hazard layer applied to poorly geocoded exposure produces a precise-looking but wrong answer.
• Vulnerability — the damage functions relating flood depth (and velocity) at a property to the financial loss. These curves are where a lot of models quietly fail, because a curve borrowed from a different country’s building stock may not represent South African construction at all.

Reinsurance-grade means all three are sound and combined correctly — a strong hazard layer cannot rescue weak exposure or the wrong vulnerability curves.

It is a loss distribution, not a single number

The defining feature of a reinsurance-grade model is that it produces a distribution of losses, not a point estimate. Reinsurance is priced on the tail — the rare, severe events — so the model has to represent the full range of possible outcomes and their probabilities, typically as an exceedance probability (EP) curve and metrics like the average annual loss and losses at specific return periods. A model that outputs “the 1:100 loss is X” without characterising the uncertainty and the tail around it is not doing the job reinsurance needs. This is the same portfolio-level thinking behind quantifying flood exposure across a portfolio.

The South African data reality

Building this in South Africa means confronting the local data honestly. Hydrological records are uneven, national high-resolution terrain is incomplete, and detailed local vulnerability data is scarce. Reinsurance-grade does not mean pretending these gaps do not exist — it means characterising them and being transparent about the resulting uncertainty. A model that acknowledges where it is well-supported and where it is extrapolating is more trustworthy, not less, than one that projects false precision everywhere.

Correlation and accumulation

The reason flood is hard for reinsurers is spatial correlation: a single storm can flood many properties at once, so losses accumulate rather than diversifying away. A reinsurance-grade model has to represent how a single event hits a whole portfolio together — which is precisely the accumulation risk that individual property assessments miss. Getting the correlation structure right is often what separates a model that reprices a treaty sensibly from one that mis-states the tail.

What it comes down to

Reinsurance-grade is less a badge than a standard of evidence: calibrated hazard, well-geocoded exposure, locally appropriate vulnerability, a full loss distribution, honest uncertainty, and correct accumulation. Each is a place a model can fail quietly, which is why the methodology has to be open to inspection. It is the engineering-grade end of flood risk intelligence, and it is built the same careful way as any defensible model.

See our portfolio exposure quantification offering for how we build and document loss models, or get in touch to discuss the evidence standard your treaty or capital decision needs.