Economic losses from severe convective storms (SCS) have increased by 425% since the early 2000s, now averaging over $40 billion annually in the U.S (Gallagher Re). This systemic increase in peril specific loss in recent years has placed a heavy focus on how model vendors have traditionally approached the peril and client feedback consistently points to models that do not satisfy the market and show real cracks when placed under scrutiny.
Convergence as the Benchmark of Model Integrity
One area in which traditional model vendor solutions have shown considerable fragility is model convergence. Convergence occurs when further increases in sample size or grid resolution no longer change a model’s outputs. In other words, model convergence means that a solution is sufficiently robust and captures the entire statistical parameter space and the modeler is now accurately representing the risk and all its potential permutations. It demonstrates that the model faithfully reproduces the physical behavior of perils rather than masking deficiencies through smoothing or approximation. The best example amongst the SCS perils are tornadoes which have typical path widths near 0.5 km and require high-resolution grids to capture sharp intensity gradients and localized damage patterns.
“Convergence isn’t a theoretical benchmark; it’s what gives companies the confidence to act. If your model doesn’t converge or properly resolve the underlying hazard, it’s not built for real-world risk.”
— Brandon Katz, EVP of Strategy, KatRisk
KatRisk’s Converged Model Principles
A fully converged SCS model must:
- Maintain stable outputs as sample size and resolution increase
- Reflect detailed physics and the spatial structure of tornado, hail, and straight-line wind events
- Deliver consistent results across portfolio sizes, deductible layers, and exposure types
KatRisk’s tornado simulations run at 100 m resolution, applying atmospheric and geophysical fluid-dynamical equations. We avoid coarse averaging which, by definition, overestimates low losses and underestimate high losses, by leveraging a large catalog of unique, high-resolution events that ensure loss variation aligns with regional climatology. A tornado doesn’t strike in 5-km blocks. Your model shouldn’t either.
A simulated tornado from KatRisk’s stochastic model passing through Dallas Texas, demonstrating the need to simulate SCS hazards at a high-resolution to capture sharp gradients in intensity.
The Pitfalls of Coarse Resolution
Grids with cell sizes of 5 km or greater, as often utilized by traditional SCS models, may reduce file sizes and speed processing, but they compromise accuracy by:
- Misrepresenting exposure in urban and rural contexts
- Distorting severity distributions and damage footprints
- Producing loss estimates misaligned with historical events
In May 2024 alone, tornadoes caused over $10 billion in U.S. losses. Broad footprints and unstable outputs translate into financial blind spots for underwriting, pricing, and reinsurance planning.
Prioritizing Convergence Before Performance
KatRisk’s development process enforces convergence as a non-negotiable requirement. Once we achieve stability and scientific peril fidelity, only then do we optimize performance. Compared with industry norms, KatRisk’s models offer:
- 1 km hail and wind grids versus typical 5 km resolutions
- 100 m realistic tornado paths versus generalized swaths
- Climate-state correlation (e.g., ENSO) for authentic inter-peril dependence
- Ultra-fast runtimes even with 2.5 million years of simulated events
Unified, Operationally Ready Risk Insights
By simulating tornado, hail, and straight-line wind concurrently (something lacking in traditional SCS models), KatRisk delivers a single coherent view of SCS risk that supports:
- Reliable pricing and capital planning through stable outputs
- Enablement of underwriting decisions powered by high-resolution insights
- Board-level reporting with granular hazard details
Pressure-Test Your SCS Model
Challenge your provider with these questions to verify true convergence:
- What grid resolution underlies your model?
- Is convergence tested separately for tornado, hail, and wind?
- How many years of simulation are used?
- How do your modeled tornado footprint widths compare to observed events?
- What is the full-event set runtime on a medium-sized portfolio?
Adopting a converged SCS model means leaving behind the uncertainty of fragmented data and uncalibrated assumptions. KatRisk’s converged SCS model unites high-resolution data, advanced physics, and consistent peril interactions to reveal a complete, realistic view of risk. For insurers, reinsurers, and modelers alike, that means fewer blind spots, stronger correlations across perils, and greater confidence in every decision.
