On 27 November 2014, a hailstorm swept across Brisbane and southeast Queensland. The storm lasted approximately 45 minutes. Hailstones ranging from golf ball to softball size fell across a densely insured suburban corridor containing hundreds of thousands of residential and commercial properties. The insured loss reached AUD 1.4 billion — at the time, one of the costliest hailstorm events in Australian history. No one died. No structures collapsed. The damage was almost entirely cosmetic and envelope-related: dented vehicle roofs, shattered skylights, punctured metal cladding, destroyed solar panels, and stripped roof tiles.
In the United States, severe convective storm — the meteorological family that includes hail, tornadoes, and straight-line wind — has in recent years generated annual insured losses exceeding those of the Atlantic hurricane season in most years. Individual hailstorm events regularly produce USD 1–3 billion in insured losses in a single afternoon across the Great Plains and Midwest. Yet for many years, hail was the peril that major cat modelling vendors developed last and modelled least accurately — because the physical scale of the damage mechanism (individual hailstones, centimetres in diameter) sits far below the resolution at which any atmospheric model or satellite observation can directly observe the hazard.
Hail forms within deep convective storm cells — severe thunderstorms — where strong updrafts carry water droplets high into the atmosphere, above the freezing level, where they accumulate ice in repeated cycles of uplift and partial descent. Each cycle adds a layer of ice around the growing hailstone. When the stone becomes too heavy for the updraft to support, it falls. The final size of the hailstone is governed by updraft intensity, the liquid water content of the storm cloud, and the residence time of the stone in the growth region. Updrafts in severe supercell thunderstorms can exceed 150 mph — sufficient to support hailstones larger than a cricket ball before release.
Hail is always a component of a broader severe convective storm (SCS) system. The same storm cell that produces hail may simultaneously produce tornadoes and damaging straight-line winds. In practice, the insured loss from an SCS event is an aggregate of all three sub-perils — hail damage to roofing and cladding, tornado damage to structures in the tornado path, and straight-line wind damage across the broader storm footprint. Separating these components in post-event loss data is difficult, which complicates model calibration.
Ice pellets causing surface impact damage to roofing, cladding, glazing, vehicles, and solar panels. Damage is proportional to kinetic energy at impact — a function of stone size and terminal velocity. Footprint can be tens of kilometres wide and hundreds of kilometres long along the storm track.
Violently rotating column of air extending from storm cloud to ground. Extreme wind speeds (EF0–EF5 scale) cause structural damage along a narrow track typically 50–500m wide and up to 50km long. High potential for total loss of structures in the direct path.
Damaging surface winds from downdrafts and outflow boundaries — not rotating. Can produce wind gusts exceeding 100 mph over a wide area around the storm. Derecho events (long-lived bow echoes) produce straight-line wind damage across hundreds of kilometres.
The primary intensity metric for hail damage is the kinetic energy of the hailstone at impact — a function of stone diameter (mass) and terminal velocity. Kinetic energy scales with the cube of diameter: a 5cm hailstone carries roughly eight times the kinetic energy of a 2.5cm stone. This non-linear relationship means that large hailstones cause disproportionately more damage than small ones, and that the tail of the hail size distribution — supercell storms producing stones above 5cm — accounts for a large fraction of the total insured loss despite being relatively rare events.
Maximum hail diameter at a specific location is the metric most commonly used in cat models as a proxy for kinetic energy, because diameter is the observable quantity in post-event field surveys and weather radar estimation. The practical intensity scale for insurance purposes runs from approximately 2cm (damage threshold for most roofing types) through 5cm (significant damage to metal cladding and solar panels) to 7cm and above (structural risk and total loss of vulnerable roofing systems).
May cause minor damage to vulnerable glazing or exposed crops. Generally below the threshold for building insurance claims.
Threshold for damage to vulnerable roofing materials. Motor vehicle paint and soft-metal cladding susceptible. Solar panels at risk.
Significant damage to asphalt shingle and tile roofing. Metal cladding dented. Skylights and polycarbonate panels breached. Vehicle damage widespread.
Severe roofing damage across most residential types. Metal decking and purlins damaged. Solar panel glass shattered. Industrial roof cladding punctured.
Total loss of vulnerable roofing systems. Structural damage to light-gauge steel framing. Catastrophic damage to vehicle fleets. Rare but possible in Great Plains supercell events.
Hail hazard modelling faces a fundamental observational challenge that no other major peril shares: the damage-causing mechanism is sub-grid in all practical atmospheric models. Individual hailstones are centimetres in diameter; weather model grid cells are kilometres across. Radar can detect the presence of hail-producing updrafts and estimate maximum hail size within a storm cell, but it cannot directly observe the surface hail field — the spatial pattern of stone sizes at ground level — with the resolution that damage functions require.
Hail stochastic catalogues are constructed from three primary data sources, each with specific strengths and limitations:
The stochastic catalogue for hail is generated by fitting statistical models to the historical hail occurrence, track, and size distribution data from these sources, then sampling synthetic events from the fitted distributions. The key parameters are: storm genesis location, track direction and speed, the swath width of damaging hail, the spatial distribution of hail sizes within the swath, and the event frequency at each location. Each of these carries meaningful uncertainty for regions where observational records are sparse or short.
A single hailstorm produces a narrow but elongated damage footprint — the hail swath — that can be 10–50km wide and 200–500km long, oriented along the storm track. Within the swath, hail size and density vary substantially — the core of the swath may receive golf-ball size hail while the margins receive pea-size stones that cause little or no damage. Capturing this internal structure is essential for accurate loss estimation: a model that treats the entire swath as receiving uniform maximum hail size will overstate losses; one that assumes uniform sub-threshold hail will understate them.
Hail generates more annual insured loss in the US than most other perils in most years — yet the hail cat model market did not develop the depth of US hurricane or earthquake modelling until the 2010s, and in many international markets mature hail cat models are still not available. Three factors explain the lag: the sub-grid observational challenge makes ground-truth data thin; the narrow swath footprint means that small shifts in a synthetic track produce large changes in which properties are hit; and hail losses accumulate from many moderate events rather than a small number of catastrophic ones — making portfolio pricing sensitive to the accuracy of the frequency distribution, not just the severity tail. These features make hail harder to model well and easier to get wrong silently.
Hail damages buildings through surface impact — kinetic energy transferred to roofing, cladding, glazing, and external equipment at the point of contact. Unlike wind or earthquake, which load the building as a structural system, hail attacks the building envelope one stone at a time. The resulting damage pattern is granular and surface-oriented: dents, cracks, perforations, and fractures in the exterior skin of the building rather than structural deformation or inundation of the interior.
| Component | Common Materials | Sensitivity at Golf-Ball (3.5–4cm) | Key Failure Mode |
|---|---|---|---|
| Roof covering | Asphalt shingle, clay/concrete tile, metal sheet, single-ply membrane | HIGH | Asphalt granule loss, tile fracture, membrane puncture. Metal denting without perforation — functional but cosmetically damaged, triggers replacement claims. |
| Roof skylights / rooflights | Polycarbonate, acrylic, wired glass, toughened glass | HIGH | Polycarbonate crazing and fracture at moderate hail sizes. Glass fracture at larger sizes. Breach of roof envelope leads to water ingress — secondary water damage often exceeds direct hail damage. |
| Wall cladding | Aluminium composite, metal sheet, fibre cement, EIFS/render | MEDIUM | Metal panels dent without perforation. Render and EIFS crack, allowing moisture ingress. Aluminium composite panels require full replacement once dented — high replacement cost relative to structural impact. |
| Glazing (walls) | Float glass, toughened glass, laminated glass, double glazing | MEDIUM | Toughened glass highly resistant to hail below 5cm. Float glass more vulnerable. Laminated glass retains integrity on fracture. Mullion-supported curtain wall systems distribute impact differently from patch-fixed glazing. |
| Solar panels (PV) | Tempered glass over silicon cells | HIGH | Tempered glass fractures at 2.5–3cm from oblique impact despite IEC 61215 certification testing. Cell cracking reduces output without visible surface damage. Emerging and rapidly growing exposure class with limited loss history for vulnerability calibration. |
| HVAC and rooftop plant | Aluminium fins, copper coils, plastic housings | HIGH | Aluminium condenser fins collapse under hail impact, reducing heat transfer efficiency significantly. Replacement of rooftop HVAC units is a major cost driver in commercial hail losses. |
| Motor vehicles | Sheet steel body panels, glass | HIGH | Body panel denting across the entire exposed surface at golf-ball size. Windscreen and rear glass fracture. Vehicle fleet losses are often the largest single component of urban hailstorm events. |
Solar photovoltaic panels represent one of the fastest-growing exposure classes in hail-exposed regions, and one of the least well-calibrated in existing cat model vulnerability functions. Installed PV capacity has grown exponentially across Australia, the US, Germany, and other hail-exposed markets. Large-scale solar farms — utility-scale installations covering hundreds of hectares — create concentrated, single-asset exposures with TIVs in the hundreds of millions of dollars that sit entirely in the open, with no structural shelter, in regions with established hail climatology.
The vulnerability of solar panels to hail is complex and not fully captured by standard property vulnerability functions. PV glass is typically tempered to IEC 61215 standard, which requires resistance to 25mm hailstones at 23 m/s — a test condition that does not capture oblique impact, large stone size, or high-density multiple impacts. Field evidence from hail events affecting solar farms shows that panels can fail at hail sizes and velocities below the IEC threshold, particularly under oblique impact at lower angles. Cell-level microcracking — invisible to surface inspection — reduces panel output without triggering conventional claims processes, meaning that actual economic loss may exceed insured claims for impacted solar farms.
Hail is typically covered under standard commercial property and homeowner policies without specific exclusion in most markets. Unlike flood, which faces adverse selection challenges that require government intervention, hail's geographic distribution — broad enough that it cannot easily be excluded without creating gaps across entire regions — means it is generally included in all-perils property coverage. The key insurance market dynamics for hail are driven by accumulation, demand surge, and the emerging exposure in solar and high-value roofing systems.
In most markets, hail is included under the windstorm or all-perils section of commercial property policies without a separate hail deductible or sub-limit. This is distinct from the US named storm deductible structure for hurricane, which creates a separate deductible threshold for tropical cyclone wind events. For hail, the standard policy deductible applies — which means that the large volume of moderate hail events (small-stone events causing cosmetic damage below the deductible threshold) do not generate claims, while events above the deductible threshold generate full claims on the policy.
Motor vehicle coverage for hail is distinct from property coverage. Vehicle fleet hail damage — often the largest single component of a major urban hailstorm loss — is covered under comprehensive motor vehicle policies rather than property policies. Large vehicle fleet exposures (car dealerships, rental fleet operators, logistics companies) create concentrated hail accumulations that may sit in a different insurer's book from the commercial property at the same location.
Hail events generate some of the most acute demand surge of any natural peril. A major hailstorm affecting a large suburban area simultaneously damages tens of thousands of roofs, creating immediate demand for roofing contractors, tiles, shingles, and solar panel replacement capacity that far exceeds local availability. Following major Australian hailstorm events, roof repair waiting times of 6–12 months have been documented, with contractor rates 30–50% above pre-event levels. This demand surge effect inflates the cost of claims relative to the pre-event repair cost baseline that vulnerability functions are calibrated to. Most hail models apply some demand surge loading, but the calibration is imprecise and the actual surge magnitude varies significantly between events depending on geographic concentration and pre-existing contractor capacity.
| Region | Hazard Character | Insurance Significance | Modelling Maturity |
|---|---|---|---|
| US Great Plains & Midwest |
The world's most active hail region — "Hail Alley" spanning Kansas, Nebraska, Colorado, Oklahoma, and Texas. Supercell thunderstorms produce the largest hailstones and most frequent significant hail days globally. Season April–September, peak May–June. | The single largest hail insurance market globally. SCS events regularly produce USD 1–3B insured losses per event. Annual US SCS insured losses frequently exceed USD 20B — often surpassing the Atlantic hurricane season. Commercial, residential, agricultural, and automotive exposure all significant. | High and improving rapidly. NEXRAD radar network provides excellent historical hail data. Multiple vendor models available. Significant investment following run of large SCS loss years in 2010s. Best-developed hail model domain globally. |
| Australia (SE Queensland, NSW, Victoria) |
Afternoon and evening convective storms produce significant hail across southeast Australia. Brisbane and Sydney corridors particularly exposed — dense suburban areas with high insured value. Season October–March. Less frequent than US but events can be catastrophic. | Very high relative to exposure area. Sydney and Brisbane hailstorms (1999 Sydney storm: AUD 5.4B; 2014 Brisbane: AUD 1.4B) are among the most significant single-event losses in Australian history. Motor vehicle and residential roofing are primary exposure classes. | Good and improving. BOM radar network covers key exposure areas. Multiple vendor and specialist models available. Post-2014 and post-2020 (Canberra, SE Queensland) events have improved calibration. Solar panel vulnerability calibration still developing. |
| Central Europe (Germany, Austria, Switzerland, France) |
Convective hail events occur across Central Europe in summer months (May–August). Alps and pre-Alpine zones particularly active. German and Austrian agricultural and property exposure significant. Events tend to be smaller footprint than US but can be intense. | Significant — Germany and Switzerland in particular have experienced multiple major hail loss years. Agricultural crop hail is separately insured and a major market in its own right. Commercial property hail losses are growing as exposure values rise and event frequency appears to be increasing. | Moderate to good. National radar networks provide reasonable coverage. JBA, Perils, and vendor models cover European hail. Sparser historical loss data than the US constrains calibration. Crop hail models more developed than property hail models in some markets. |
| India (North and Central) |
Pre-monsoon convective storms produce hail across northern India (March–May). Events can be intense — large hailstones with significant agricultural and property damage. Urban exposure growing rapidly with expanding insured asset base in tier-1 and tier-2 cities. | Growing but still low insurance penetration for hail. Agricultural hail losses are significant and partially insured under government crop insurance schemes. Commercial property hail coverage is improving but urban hail cat models for India are not mature. | Low. Limited radar data coverage, sparse loss history for calibration. Hail is a known exposure risk but quantified primarily through historical loss reports rather than probabilistic modelling for most of the country. |
The damage-causing mechanism — individual hailstones — is fundamentally below the resolution of any operational weather observation or atmospheric model. Radar-derived hail size estimates carry uncertainty of 1–2cm in diameter — large relative to the damage threshold range of 2–5cm. Ground station records are too sparse to reconstruct the full spatial hail field. This observational limitation propagates into the stochastic catalogue: the catalogue represents the best statistical fit to available data, but the underlying data is sparse and imprecise relative to the sensitivity of the damage functions to hail size.
Because hail damage is confined to a narrow swath, small shifts in synthetic track placement produce large changes in which locations are hit and what size hail they receive. This creates very high location-level loss uncertainty — the modelled AAL for an individual property can vary by an order of magnitude depending on how nearby synthetic tracks are positioned. At the portfolio level this uncertainty averages out, but for individual risk pricing and for concentrated portfolios in narrow geographic corridors, the model's single-track-per-event approach understates the true range of outcomes the location could experience.
Utility-scale solar farms, rooftop PV systems, and large vehicle fleet exposures are growing rapidly in hail-exposed regions. Standard commercial property and residential vulnerability functions were not calibrated to these exposure classes, and the loss history needed to derive robust solar-specific vulnerability functions is still accumulating. Models that apply standard property functions to solar panel exposure are likely understating hail vulnerability for this class. The solar energy cat model market is developing, but specialist vulnerability functions for PV are not yet universally available across vendors.
Hail losses accumulate from many moderate events rather than a small number of catastrophic ones — the frequency-severity distribution looks different from hurricane or earthquake, where the tail events dominate. This means that pricing accuracy requires the model to correctly reproduce not just the tail severity but the full frequency distribution of moderate events above the deductible threshold. Small errors in the annual event frequency, compounded across many events, produce cumulative AAL errors that can be as large as large-event calibration errors. Many hail models have been assessed as more accurate at the tail (1-in-100 year) than at the mean (AAL) for this reason.
Outside the United States, mature hail cat models exist for Australia and parts of Central Europe, but many hail-exposed regions — India, South America, South Africa — have limited or no commercial cat model coverage. For insurers writing hail-exposed risk in these markets, model-based pricing is either unavailable or relies on adapted US or European models with uncertain transferability. The consequence is that hail pricing in emerging markets relies more heavily on historical loss ratios and expert judgement than on probabilistic loss modelling.
The meteorological family encompassing hail, tornadoes, and damaging straight-line winds generated by organised deep convective storm systems. The peril category under which hail is modelled. Individual SCS events produce losses from all three sub-perils simultaneously, complicating post-event attribution.
A deep, rotating convective storm — the most severe type of thunderstorm — characterised by a persistent mesocyclone (rotating updraft). Supercells are the primary producer of large hail, significant tornadoes, and extreme straight-line winds. Most common in the US Great Plains during spring and early summer.
The narrow, elongated damage footprint produced by a hailstorm along its track — typically 10–50km wide and 200–500km long. Hail size and density vary significantly across the swath, with the damage core concentrated in a narrower zone within the total swath. The narrow width makes individual location loss estimates highly sensitive to track placement.
The primary physical driver of hail damage — the energy of the hailstone at impact, proportional to mass (cube of diameter) times velocity squared. Scales non-linearly with hail size: a 5cm stone carries ~8× the kinetic energy of a 2.5cm stone. This non-linearity means that the tail of the hail size distribution is disproportionately damaging.
The proxy intensity metric for hail in most cat models — the largest stone diameter recorded or estimated at a specific location during an event. Used as a direct input to hail damage functions. Estimated from radar reflectivity signatures or ground observations. Radar-derived estimates carry ±1–2cm uncertainty.
The US national weather radar network (Next Generation Radar) — a network of WSR-88D Doppler weather radars providing continuous, high-resolution coverage of the contiguous United States. The primary data source for US hail hazard model calibration. Provides dual-polarisation capability that improves hail detection and size estimation relative to older single-polarisation systems.
A long-lived, rapidly moving band of severe thunderstorms producing widespread, straight-line wind damage over distances exceeding 400km. Distinct from rotating supercell storms that produce tornadoes. Produces an elongated swath of straight-line wind damage that can cross multiple states or countries in a single event.
The international standard governing design qualification testing for terrestrial photovoltaic modules. Includes a hail impact test using 25mm ice balls at 23 m/s perpendicular impact. The test standard does not replicate oblique impact, larger stone sizes, or high-density multiple impacts — conditions observed in damaging hail events affecting solar farms.