Anti-hail rockets · rain · 1995–2025

Did the anti-hail rockets change the rain around them?

Some say the rockets save crops. Others say they dry out the village next door. Rain can be measured, and Romania’s weather stations have logged it daily since 1995. This page asks the stations what happened around the country’s 119 rocket launch areas, and shows where the answer stops.

Lift the stone

Drag the cloud to turn it. The stone rises and falls in the updraft and gains a new layer, opaque or clear, depending on the temperature and wetness of the air it passes through.

Illustrative diagram: the heights and layers are typical, not a reconstruction of one storm, and the stone is enlarged tens of thousands of times. A real stone, cut open and read, is on the Grindina (hail) page. Grow a hailstone and cut it open

9 of 8010 km circles with a known start year hold a rain station with a public daily record

±10%the change in summer rain that the stations inside the circles can neither show nor rule out

134 of 216monthly values in the authority’s own table match the weather service’s records to 0.05 mm

#Two promises and a question

Romania’s national anti-hail system fires rockets carrying silver iodide into thunderclouds so that hail does not grow into large stones. Officials have spoken of an effectiveness above 75%; the Grindina page shows why that figure does not say what it measures. Farmers in Prahova county say something else (Observatorul Prahovean, June 2025): that the rockets take the rain and leave the village downwind dry.

The first claim can only be settled in the field, with hail measured on plots, and that data is not public. The second is easier to check: if rain moves, the stations around it log the change. This page follows the second claim.

The system’s own table gives the year each launch point counts as operating. The page compares May–September rain before and after that year, near the point and far from it. The calculation plan was written down and fingerprinted before the first result.

#Where the areas and stations are

Each orange dot is the centre of an airspace area for anti-hail rockets, taken from Romania’s aeronautical publication (119 areas). Its colour is the year in the system’s own table. For 33 areas the table has no row with a matching name, and six carry only the range 2022–2024.

The 10 km circle on the map is a working assumption. The printed airspace area has a radius of six nautical miles (about 11 km), and the crop-protection zones are not published. A dot on the map does not prove that rockets are fired there.

Areas Rain stations Hail (1961–1990 average)

Areas: AIP Romania, 17 April 2025. Years: the system’s dossier, 2025. Stations: ANM. Hail: WMO normals 1961–1990.

#The rain map cannot see inside the circles

ANM, the national weather service, also publishes a daily rain map with a 0.01° step (about one kilometre). It closely follows the 160 stations with public daily records, tens of kilometres apart: at a station’s own cell, its value matches the station’s to within 2% in 79% of the 4,856 station-summers.

Between stations the map smooths. Summer rain in two cells 10 km apart moves almost identically: the correlation is 0.99. At 50 km it falls to 0.85. A test run on the map inside a circle with no station really measures the stations around it.

How alike summer rain is in two map cells, by the distance between them (correlation across 1995–2022)

Of the 80 circles with a known year, only 9 hold a station with a public daily record. The median distance to the nearest station is 17.7 km, and 61% of circles have a station within 20 km. That is why the test below uses stations, not the map.

#Measured rain: before and after

The nine circles with a station hold seven stations in all (some circles overlap). One also lies near an area with no known year and was set aside, which leaves six within 10 km of the centre of an area with a known year. Their May–September rain, compared before and after that year with stations more than 25 km from any area (125 stations), changed by −0.3%. The 95% confidence interval runs from −10% to +10%.

Change in May–September rain after the area’s year, against distant stations, with the 95% interval

With six stations, an effect of ten percent either way is invisible. The data show no change and do not rule one out. With a 15 km radius (eleven stations) and 20 km (twelve), the picture is the same: −0.7% and −0.7%.

#The next village’s drought

If the rockets take the rain, it should be missing downwind. Downwind here means east (bearing 55° to 145°), the way storms carried by westerly winds travel. Upwind is the opposite side. For each start year, May–September rain was taken from 10 to 40 km out from the point, downwind and upwind, before and after.

The result: −5.0%. After the start year, the downwind side received almost five percent less than the upwind side, compared with before. If the same six groups of areas had been moved at random elsewhere in the country (2,500 simulations), a difference this large would have appeared in about 3 cases out of a hundred.

Downwind minus upwind in 2,500 simulations with the areas moved at random (bars) and in the real data (line)

How solid is the result

The checks below were added after the first result, so they are weaker than the test fixed in advance. With the two main tests run (local and downwind), the chance that at least one looks like this by accident is about six in a hundred.

Downwind minus upwind, each group on its own, and with shorter periods

Only two of the six groups show less rain downwind, Oltenia (−11%) and Prahova (−9%). In the other four, downwind is wetter. If the years after 2022 are left out (the ANM map changes in 2023), the difference falls to −2.5%. Without the points dated 2018, it falls to −1.2%.

A real effect of the rockets should show in every group. Here it shows in two. The east side of a group in Moldavia or Prahova slopes toward the plain and the west side climbs toward the mountains. If the years after the start were drier on the plain, the difference appears without any rockets. The test cannot tell the two explanations apart.

Conclusion: the measured rain does not show a steady drought downwind of the launch areas. A weak signal remains, in two groups, that only the firing logs can settle.

#The table the authority cites

The system’s dossier (annex 7) compares monthly rain inside the protected area with rain outside it, at three stations in Iași county, 2019–2024. The 216 monthly values were set against ANM’s daily records. If the ANM day carries the next day’s date, 134 match to 0.05 mm; under the other date convention, 74. Sixteen differ by more than five millimetres.

The largest gap: Roman station, July 2021. The table gives 84.2 mm, ANM’s records give 136.1 mm (167.4 mm under the other convention). The cause could be a data revision or a different measurement window; the table does not say. Even corrected, such a table cannot settle cause, because it lacks the firing days and pre-chosen comparison stations.

#Where hail falls

Few stations log hail, and a commune-by-commune hail map cannot be made from open data. The 11 stations with official averages for 1961–1990 still show what relief does: Vârful Omu, a peak at 2,504 m, has 5.9 hail days a year, Cluj-Napoca 2.4, Iași 0.6 and Sulina, on the Danube delta coast, 0.1.

Hail days per year, 1961–1990 average (Cluj: 1971–1990)

The study by Burcea and colleagues (105 stations, 1961–2014) finds more hail days in the mountains and fewer in the south-east. It does not demonstrate fixed corridors downwind of the Carpathians across the whole country. That would need years of radar and field reports, which are not public.

#Your place

Type the name of a commune, town or village. You get the nearest launch area, the rain stations around it and what they logged.

#What would settle the question

Three things the authorities hold and the public cannot see: the firing log of each point (the “Registrul de tragere”, model 3.F.5, from the 2009 regulation), the polygons of the protected and comparison areas that Law 173/2008 requires, and the daily rain series from each area.

Anyone can request them under Law 544/2001 on access to public-interest information. Below is the text of a request, filled in with the launch area nearest to the chosen place. You send it yourself, from your own address. It is in Romanian, the language the authority works in.

#Method in brief

Areas: the 119 centres in the aeronautical publication; the year comes from the table in the system’s dossier where names match exactly (86 areas, six of them carrying only the range 2022–2024). The table’s year is not a verified start date: documents show some points were announced before they were authorised. The first treated season is the one after the year.

Rain: the May–September total. The ANM map has 483×972 cells up to 2022 and a different, larger grid from 2023, resampled onto the first. The measurement window of the day changes in 2023 from 18–18 UTC to 06–06 UTC; seasonal totals barely change.

Downwind and upwind: the 10 km circles excluded, bands from 10 to 40 km. The comparison is before (ten years) and after (up to eight years) for each start year, weighted by the number of points. The simulations move each group of areas (Prahova, Vrancea, Iași, Timiș, Mureș, Oltenia) rigidly elsewhere, keeping the same years and the same distances between points.

Limits: hail and crop losses are not tested; the airspace centres are not verified launch pads; the table’s year is not the date of firings; the stations and the ANM map are not independent. The plan fixed before the first calculation, its fingerprint, the results and the extra checks made after the result are published separately, below.