Catastrophe model uncertainty is the range of possible error in the losses produced by catastrophe models, which simulate hurricanes, earthquakes, floods and other perils against insured exposures. Modellers usually distinguish primary uncertainty, about whether and where events occur and how intense they are, from secondary uncertainty, about the damage a given event causes to a given property and the resulting insured loss after policy terms.
Further layers add to this. Vendor models built on different hazard catalogues and vulnerability curves can produce very different PMLs for the same portfolio; exposure data may be incomplete; and some perils or loss sources, such as loss amplification or contingent business interruption, are not modelled at all. Insurers respond by blending results from several models, applying loadings for non-modelled risk, running scenarios and validating against historical losses, all within their model risk management. The uncertainty is particularly relevant to pricing catastrophe bonds and to capital measures based on exceedance probability curves.