Hail
Pick the size of the hail you remember. See how fast it fell, how hard it hit and how strong an updraft held it up. Then grow a stone in a storm and cut it in half.
How big was the hail you remember?
A 5 cm stone (a hen's egg) falls at 89 km/h on average and hits with about 7.4 J, like a 1 kg object dropped from 75 cm. To hold it up at 6 or 7 kilometres, where the air is thinner, the updraft has to rise at least 125 km/h.
Grow a stone in the stormHail grows in an updraft inside a thunderstorm cloud. It needs three things: an ice embryo smaller than a centimetre, supercooled water (liquid water below 0 °C) and time. The embryo collects the water droplets it meets, and the water freezes on its surface. The stone grows for as long as the updraft holds it up. When it gets too heavy for the updraft, or leaves it, it falls, and if it does not melt on the way down we find it on the ground.
Most of the growth happens between −10 °C and −25 °C, at 5.7–8.0 km up in the atmosphere used below. Studies cited in the review by Allen and colleagues (2020) suggest a large stone spends up to 10–15 minutes or more in the growth zone. Its path depends on its fall speed, which rises as it gets heavier, on the speed of the updraft and on the horizontal winds in the storm.
Choose an updraft, how much water the cloud holds and the height where the embryo enters. The stone starts by itself.
How fast air rises in the middle of the cloud. Air cannot hold up a stone that falls faster than it rises.
The embryo enters the updraft at its edge, where the air rises only just fast enough to hold it. Lower down it is warmer, higher up the air is thinner.
Supercooled water is the food of hail. With little water the stone stays small and milky.
What is model here. The storm in the picture is simplified. The atmosphere is a textbook one for middle latitudes in summer: 0 °C at 4.0 km, −20 °C at 7.2 km. The updraft is a tilted column, and the stone is carried downwind. The model follows the stone only vertically and downwind, and the stone's heat balance is a simplified version of the one used by Kumjian and Lombardo (2020). In the model a stone spends between 20 and 40 minutes in the storm. Real trajectories are three-dimensional and are not known for any particular stone.
What is measured: fall speed, energy and weight, from thousands of stones collected on the US Great Plains (Heymsfield and colleagues, 2018, with the 2020 correction), valid up to about 7 cm. Below 5 millimetres there are no measurements, which is why the embryo starts at 5 millimetres.
The layers show how the water froze on the stone. When droplets freeze at once, they trap air bubbles and the ice comes out milky (dry growth). When the stone catches water faster than it can freeze it, the water spreads over the surface, the bubbles escape and the ice comes out clear (wet growth). The stone switches from one to the other as the water in the cloud, the temperature or its own size changes.
The layers cannot be counted like tree rings. The idea that each layer means one trip up and down the storm is widespread, but the US National Severe Storms Laboratory (NSSL) and the review by Allen and colleagues (2020) reject it: the layers depend on the conditions the stone meets, and these change when the stone crosses the updraft sideways too. The layers can tell the order of dry and wet growth. They cannot tell how many times the stone went up.
Cut the stone you grew. Open it like a book and put a finger on a layer: the stone's path lights up in the storm.
Drag the stone to turn it.
Size is measured along the longest dimension of the stone. The table puts everyday objects next to fall speed and impact energy, computed from measurements on thousands of stones. Above 7 cm there are too few measurements, and the figures there are extrapolations.
| Object | Size | Speed at the ground | Impact energy | Updraft needed at 6.5 km | ESSL class |
|---|---|---|---|---|---|
| a pea | 0.6 cm | 17 km/h | 0.00 J | 25 km/h | under 2 cm |
| a cherry or a hazelnut | 2 cm | 48 km/h | 0.26 J | 65 km/h | large |
| a walnut | 3 cm | 63 km/h | 1.2 J | 90 km/h | large |
| a table-tennis ball | 4 cm | 77 km/h | 3.3 J | 105 km/h | large |
| a hen's egg | 5 cm | 89 km/h | 7.4 J | 125 km/h | very large |
| a tennis ball | 7 cm | 111 km/h | 24 J | 155 km/h | very large |
| a peach | 8 cm | ≈ 122 km/h | — | ≈ 170 km/h | very large |
| a grapefruit | 11 cm | ≈ 151 km/h | — | ≈ 210 km/h | giant |
| Romania's record: Sânandrei, 2016 | 15 cm | ≈ 186 km/h | — | ≈ 260 km/h | giant |
| Europe's record: Italy, 2023 | 19 cm | ≈ 217 km/h | — | ≈ 300 km/h | giant |
| the US record: Vivian, 2010 | 20.3 cm | ≈ 227 km/h | — | ≈ 315 km/h | giant |
≈ = extrapolation above 7 cm, where measurements are too thin. Speeds are medians for natural stones (Heymsfield and colleagues, 2018, with the 2020 correction). Between stones of the same size, speed differs by about ±50% with shape and weight, and energy varies even more: for 5 cm, between about 3 and 23 J (10th–90th percentiles). “Updraft needed” is the least speed at which air must rise to hold the stone at 6.5 km, where thinner air lets it fall about a third faster than at the ground. ESSL classes: “large” hail from 2 cm, “very large” from 5 cm, “giant” from 10 cm. The object comparisons come from ESSL's European chart (Kühne and Castellano, 2019), with the pea and the grapefruit from the US National Weather Service chart. Charts disagree with each other: a walnut is between 2.7 and 4.1 cm.
At ground level a 2 cm stone falls at about 48 km/h on average, a 5 cm stone at about 89 km/h and a 7 cm stone at about 111 km/h. At 6 kilometres up, where the air is thinner, the same stone falls a third faster (a factor of about 1.35), so an updraft needs more strength to hold a stone of a given size than its ground speed suggests. A natural stone falls more slowly on average than a solid ball of ice of the same size, because it is lighter and less round.
Hail forms high up, where the temperature is below 0 °C even on a summer day: in the textbook atmosphere used here, 0 °C is at 4.0 km and −20 °C at 7.2 km. On the way down the stone passes through air warmer than 0 °C and melts in part. Small stones melt more often and faster; the review by Allen and colleagues (2020) notes that a warmer lower atmosphere has little effect on the melting of large stones. In Romania, 94.2% of hail days fall between April and September (weather stations, 1961–2014).
Yes, but less often. In Romania most hail falls in the afternoon: weather stations record it mostly between 13:00 and 21:00 local summer time (Burcea and colleagues, 2016). It is not a rule. Europe's largest hailstone, about 19 cm, fell at Azzano Decimo in Italy at around 23:00 on 24 July 2023 (ESSL).
The largest hailstone ever recovered in the United States fell at Vivian, South Dakota, on 23 July 2010: 20.3 cm and 879 g. The WMO (World Meteorological Organization) keeps no diameter record, only a weight record: 1.02 kg at Gopalganj, Bangladesh, in 1986, not formally evaluated. In Europe the largest photographed stone is about 19 cm (Italy, 2023). Romania held Europe's record until 2023: 15 cm at Sânandrei, in Timiș, in 2016.
These stones, at the same scale as each other
| Stone | Size | When | Who measured it and its status |
|---|---|---|---|
| Vivian, South Dakota (USA) | 20.3 cm, 879 g | 23 July 2010 | NOAA's National Climate Extremes Committee (vote 3–0, 28 July 2010): the largest in diameter and the heaviest hailstone recovered in the United States. Diameter is measured tip to tip, the day after the fall; the stone had already partly melted, and the weight was taken several days later. |
| Gopalganj (Bangladesh) | 1.02 kg | 14 April 1986 | The heaviest hailstone in the World Meteorological Organization (WMO) archive. WMO has not formally evaluated it; the sources are popular books from 1996–2000. |
| Azzano Decimo (Italy) | about 19 cm | 24 July 2023, about 23:00 | ESSL: the largest photographed hailstone in Europe. The size is estimated from photos with reference objects beside the stone. |
| Carmignano di Brenta (Italy) | about 16 cm | 19 July 2023 | ESSL: estimate from photos. It held Europe's record for five days. |
| Sânandrei, Timiș (Romania) | 15 cm | 20 June 2016 | ESSL: the largest hailstone reported in Europe until July 2023. A 2024 paper gives the date as 26 May 2016; the measuring method is not documented. |
| Răducăneni, Iași (Romania) | 10 cm | 4 June 2024 | Romania's National Meteorological Administration (ANM). The value comes from a Facebook group of observers and was validated by ANM; the European database ESWD has 9 cm at Bohotin the same day. |
Size records are estimates from photographs or measurements on partly melted stones; the wording follows the body that checked them. Report databases such as ESWD show where stones were reported, not where hail falls most often. “Diameter” here means the largest distance between two points.
Nobody knows. No independent or randomised evaluation of the Romanian system has been published, and the official 75% figure does not say what it measures. The World Meteorological Organization (WMO) writes that experiments on seeding hailstorms are not yet sufficient to accept the ideas behind the technology. The idea of the rockets is simple: the silver iodide in the rocket would make the cloud freeze too early and stop stones from growing. Whether it works in practice remains open.
Officials and operators have said the rockets are “over 75%” or “75–80%” effective. No document read says what the percentage measures: hail frequency, diameter, energy or affected area.
Law 173/2008 requires comparison zones next to each protected zone and monitoring by non-governmental organisations. The only independent monitoring the Court of Accounts found was a 2021 contract, in which “efficiency” was a ratio of land areas, not a reduction in hail.
In a 2025 study (Istrate and colleagues, Atmospheric Research), 20 seeded storms had somewhat lower radar values than 20 unseeded ones: median maximum reflectivity 58 against 60 dBZ, and vertically integrated liquid water 19 against 27 kg/m². The comparison is not randomised, the authors say natural fluctuations complicate interpretation, and six of the seven authors are from the group that says it operates the launch points.
An earlier study by the same group, called exploratory by its authors, found radar values rising during seeding and falling afterwards.
The review by Allen and colleagues (2020) writes that the experiments of the 1970s and 1980s gave no conclusive evidence for hail suppression, and research has continued regionally. WMO says a sound evaluation needs randomisation.
In France (the ANELFA programme) and Spain, researchers of a ground-generator programme report about 50% less hail energy on the most severe hail days; the result does not come from a randomised experiment and concerns generators, not rockets. An analysis from the Republic of Moldova (1967–2015) found more hail days, not fewer, as the protected area grew; it is a correlation, not proof.
Growers of vines and orchards say hail losses have grown since the launches stopped. Farmers of field crops say it rains more since they stopped. Neither claim has been tested by an independent study. WMO says no significant health or environmental effects of silver iodide were found in past operations, and that downwind effects have been suggested but need research.
Situation on 29 September 2026. This page does not advise growers to rely on the system or to ignore it.
Fall speed, energy and weight of the stones come from published measurements on natural stones, valid up to about 7 cm; above that size they are extrapolations. The storm on this page, its atmosphere, updraft, cloud water and the stone's path are a simplified model; what is model is said where it appears. Records follow the wording of the body that checked them.
Published on 29 September 2026.