Hailstorm hazard (size-resolved, global)
How often hail large enough to damage property falls, resolved by hailstone size, on a 0.05 degree global grid.
Source: Prein & Holland (2018) global backbone · GridRad-Severe v4.2 (contiguous-US radar) · MeteoSwiss Swiss Hail Climatology (footprint area law) Year: Historical climatology (no forward horizon) Category: Physical risks · Direct Physical Loss engine inputs Coverage: Global Format: Raster grid (0.05°, roughly 5 km) Used in risk analysis: Yes — carries the hailstorm proximity flag on Mitigating services (Climate Acute Risks) at 1.2 events per year of hail ≥ 25 mm
What it shows
How often damaging hail reaches the ground, resolved by hailstone size. The displayed band is the annual exceedance rate of hail ≥ 25 mm in diameter, counted at the scale of a grid cell maximum (roughly 1 to 2 km²): the headline hail-frequency picture.
Size is what drives the loss. The damage a stone causes rises steeply with its diameter, so a layer that separates a 25 mm stone from a 60 mm one says more about exposure (roofing, vehicles, solar modules, façades, standing crops) than an undifferentiated frequency index. The grid is 0.05°, 40 times finer than the 2° passive-microwave hail climatology it takes over from.
How it is built
Three open, commercially redistributable sources are fused onto a single 0.05° global grid:
- Global backbone, Prein & Holland (2018). An estimate of hail hazard derived from the large-scale storm environment, giving the annual rate of ≥ 25 mm hail on land worldwide.
- Contiguous United States, GridRad-Severe v4.2. Where high-quality radar exists, observed radar hail sizes are used directly.
- Footprint scaling, MeteoSwiss. Any hail grid reports the largest stone anywhere in a cell. A building is far smaller than a cell, so a size-versus-area law fitted on the Swiss hail climatology converts the cell maximum to the size expected on a 100 m² building footprint, the scale a damage function should work at.
The product ships as a 7-band raster. Darwin registers the ≥ 25 mm cell-maximum band as this map layer, and splits the three footprint bands (≥ 25, ≥ 40, ≥ 60 mm on 100 m²) into separate single-band Float64 rasters that stay hidden and feed the loss engine.
How to read it
The symbolised value is an annual rate in events per year: the number of times per year, on average, that hail of at least 25 mm is expected somewhere in the cell. Higher means more frequent damaging hail, and the hot spots are the mid-latitude continental hail belts. Away from the US radar coverage the backbone is a smoothed estimate built from the storm environment, so read a single cell as a regional signal rather than a site measurement.
The cell-maximum reading is deliberately generous: it answers "does a cell this size produce 25 mm hail", not "does a given roof get hit". That is the right question for screening and the wrong one for pricing, which is why the loss engine reads the footprint bands instead.
What the loss engine reads
The Direct Physical Loss engine does not price hail off this displayed band. It reads the three hidden 100 m² footprint bands and fits, per cell, both a severe-hail rate and the exponential tail of the size distribution, then integrates the loss over return periods from 2 to 1000 years. The damage curve is the Schmid (2024) cubic size-to-loss relation, whose intensity axis is the hail diameter in millimetres; see Damage functions.
Feeding the cell-maximum band to that cubic would overstate the loss by roughly a factor of five, since the cell-maximum to footprint size ratio enters the curve cubed. Where the footprint bands are absent or cannot be fitted (disagreeing no-data masks, a near-flat slope, or a rate too low for the return-period grid), the engine falls back to a calibrated bridge on the 2° normalised hail density from the NASA Passive Microwave Hail Climatology (Bang & Cecil 2019, TRMM and GPM combined). That same density raster is published for reading as the companion display layer Hailstorm density (continuous), filed alongside this one among the engine inputs.
Class thresholds
The layer publishes no class bands. The value is a continuous physical rate, and the map ramp runs linearly from 0 to 6 events per year.
The only threshold on it is the proximity flag: a site is flagged on the Mitigating services dimension (Climate Acute Risks) when the rate reaches 1.2 events per year of hail ≥ 25 mm, with the flag reading "High hailstorm frequency zone (≥25 mm hail ≥ 1.2×/yr)".
How this threshold was set. It is a Darwin calibration, measured rather than published. The flag used to sit on the 2° hail-density raster at a normalised density of 0.8. Moving it onto a 40 times finer field in a different unit meant no number could be carried over, so 1.2 events per year was chosen as the value that reproduces the land coverage of the old rule, about 5 % of land, on the new grid.
References: Prein & Holland (2018), global damaging-hail hazard; Murphy & Homeyer, GridRad-Severe v4.2; MeteoSwiss Swiss Hail Climatology; Schmid et al. (2024) hail damage function.
Risk analysis
A site is flagged on a dimension by combining a proximity trigger (this layer) with an activity trigger. Proximity only → Potentially material; proximity and a material activity → Very material; neither → Not material.
No ENCORE pressure indicator exists for hail, so the activity side is rated by site type rather than by an ENCORE score, and the flag needs at least the medium tier. Solar panels are the size-critical case, standing crops next, ordinary buildings (offices, shops, warehouses, factories, data centres, airports) medium; robust and linear assets (managed forest, mining, construction, roads, rail, onshore wind, land-based oil and gas) low; pipelines, hydropower and offshore assets carry no hail exposure.
| Dimension | ENCORE service / pressure | Proximity trigger (this layer) | Activity trigger (entity) |
|---|---|---|---|
| Mitigating services | Hailstorm (no ENCORE indicator) | Layer value above 1.2 events per year (hail ≥ 25 mm) | Site type at medium tier or above |
Source
- Prein, A. F., & Holland, G. J. (2018). Global estimates of damaging hail hazard. PANGAEA, doi:10.1594/PANGAEA.893160 (CC BY 3.0).
- GridRad-Severe v4.2 (Murphy & Homeyer). Zenodo 13887225 (CC BY 4.0), contiguous-US radar hail sizes.
- MeteoSwiss Swiss Hail Climatology (CHHC) (CC BY 4.0), the cell-maximum to building-footprint size law.
Comparison with the WWF Risk Filter Suite
There is no equivalent indicator in the WWF Risk Filter Suite: hail is an acute weather hazard rather than a biodiversity or water risk, so it sits outside the WWF methodology. This layer is a Darwin extension of physical-hazard screening.
Legend
Symbolised field: Hail ≥ 25 mm cell-max (events/yr), continuous ramp.
Generated from darwin/tools/darwin-geo/darwin-layers/hail_hazard_global_v2/layer.toml and risk_indicator_pairs.toml (develop).