EN
Home Child Prodigies Raul Victor Magalhães Souza
SCIENCE PRODIGY

🇧🇷 Raul Victor Magalhães Souza

At 16 he turned the rain prophets of Ceará into a model five times more accurate than a commercial forecast

Winner, 31st Prêmio Jovem Cientista • Brazilian national winner, Stockholm Junior Water Prize 2026

In the semi-arid interior of Ceará, farmers called profetas da chuva — rain prophets — forecast the coming season by reading the flowering of a cactus, the behaviour of ants and the brightness of the Pleiades. Most data science treats that as folklore. A sixteen-year-old from Iracema treated it as a dataset. He structured six prophets' indicators, trained a model on forty-four years of meteorological records, and measured the result: a 5.7% mean monthly error against a commercial platform's 30.8%.

RV🇧🇷Raul Victor Magalhães Souza

Raul Victor Magalhães Souza — Geniuses.club editorial portrait

💬 Chat with Raul
Share on X Facebook LinkedIn WhatsApp Reddit

Raul Victor Magalhães Souza comes from Iracema, a town in the Vale do Jaguaribe region of Ceará, in Brazil's semi-arid Northeast — a place where rainfall uncertainty determines whether smallholder families plant. On 26 November 2025 he was announced as the winner of the secondary-school category of the 31st Prêmio Jovem Cientista, Brazil's national young-scientist award, run by the National Council for Scientific and Technological Development (CNPq) with the Fundação Roberto Marinho. He was sixteen, and studying at a state secondary school in the Benfica district of Fortaleza.

The winning project carries a long title: "Development of a hybrid machine-learning tool for rainfall prediction based on the knowledge of the rain prophets of the Vale do Jaguaribe, Ceará." It was supervised by his teacher, Helyson Lucas Bezerra Braz. The profetas da chuva are farmers of the Northeast who forecast the rainy season by reading signs in nature, a practice passed down over generations. Raul set out to record, structure and test that knowledge rather than dismiss it.

The system he and his supervisor built is called the IA dos Profetas da Chuva. It was constructed with machine-learning libraries behind a Streamlit interface running on Python, and fed with observations from six rain prophets in the municipalities of Iracema, Limoeiro do Norte, Morada Nova, Quixeré and Russas, all within the Vale do Jaguaribe. The point of the design is that the traditional observations enter the model as structured variables, not as anecdote.

The indicators were organised into three families. Atmospheric phenomena included the lunar halo, cloud bars at the start of the year, and the brightness of the Sete-Estrelo, the Pleiades. Botanical indicators included the flowering of the mandacaru cactus, observation of the embiratanha, and the flowering of the juazeiro. Animal-behaviour indicators included the black butterfly, the tarantula, frogs and anthills. Each was converted into data the model could learn from.

Against those signals the team trained the system on meteorological records from Ceará's state weather foundation Funceme and the national institute Inmet. The tool integrates forty-four years of data, from 1981 to 2024, comprising more than forty thousand records. All of it went through cleaning and standardisation before being cross-referenced with the traditional indicators — the unglamorous half of the work, and the half that makes the result testable.

The validation numbers are specific, which is what raises this above a science-fair narrative. During the 2025 rainy season, from January to March, the model estimated 422.1 mm of accumulated rainfall against an actual observed volume of 431.5 mm — a mean monthly error of 5.7%. A commercial reference platform, over the same window, estimated 318 mm, an error of 30.8%. On classification performance the system reached 94.5% accuracy.

On 1 June 2026 Raul won the Brazilian national stage of the Stockholm Junior Water Prize, at a ceremony held at the Escola de Guerra Naval in Rio de Janeiro — the tenth Brazilian edition, organised by the Brazilian Association of Sanitary and Environmental Engineering with the Brazil-Sweden Chamber of Commerce. The win sends him to the international final in August 2026, during World Water Week in Stockholm.

He traces the project to his grandfather, the farmer Luiz Maia. He has been interested in the subject since primary school, won a CNPq scholarship in his first year of secondary school, and developed the research in partnership with the Laboratory of Pharmacology of Venoms, Toxins and Lectins at the Federal University of Ceará. His stated next step is to expand the database and offer the tool to the municipal governments of the Vale do Jaguaribe — which would move it from a prize-winning model to an operating forecast.

Two national competitions sit behind this project, and both are worth naming because they are what turn a school science fair into a citable result. The Prêmio Jovem Cientista is run by Brazil's National Council for Scientific and Technological Development, the federal research funding agency, with the Fundação Roberto Marinho; it has run for thirty-one editions and its secondary-school category is judged against submitted methodology, not a presentation. The Stockholm Junior Water Prize national stage is organised by the Brazilian Association of Sanitary and Environmental Engineering and feeds a single national entrant into an international final held during World Water Week.

The more interesting methodological point is what the project did not do. It did not assert that traditional knowledge works, and it did not assert that it does not. It converted six practitioners' indicators into structured variables, trained against forty-four years of instrument data from two national meteorological bodies, and then reported a measured error against observed rainfall in a specific season — alongside the error of a commercial benchmark over the same window. Publishing the comparison is the whole contribution. A model without an error figure is a claim; a model with one is a result.

“Our next step is to expand the database to get even more precise forecasts. With the system improved, it would be possible to offer it to the municipal governments of the Vale do Jaguaribe, because I believe the use of this tool is very important.”
— Raul Victor Magalhães Souza, Diário do Nordeste, 26 November 2025 (translated from Portuguese)
“I have to thank my grandfather, my main reference and the foundation for building the project.”
— Raul Victor Magalhães Souza, ABES, 11 June 2026 (translated from Portuguese)
2024
Data window closesThe training set runs 1981-2024: forty-four years and more than forty thousand records from Funceme and Inmet.
2025
Field validationDuring the January-March rainy season the model predicts 422.1 mm against 431.5 mm observed — a 5.7% error.
2025
National science awardWins the secondary-school category of the 31st Prêmio Jovem Cientista on 26 November, aged sixteen.
2026
Stockholm prize, Brazil stageNamed grand winner of the Brazilian national stage of the Stockholm Junior Water Prize on 1 June.
2026
International voteConfirmed for the prize's international public vote alongside young scientists from other countries.
2026
Stockholm finalDue to represent Brazil at the international final during World Water Week in Sweden in August.
PersonCountryMilestoneAge / Stat
IA dos Profetas da ChuvaCeará, BrazilPredicted 422.1 mm vs 431.5 mm observed5.7% mean error
Commercial reference platformPredicted 318 mm over the same window30.8% mean error
Training dataFunceme and Inmet1981-2024, 40,000+ records44 years
Classification accuracyModel performance on the validation set94.5%

Most climate-modelling stories treat folk forecasting and data science as opposites; this one measured them together and published the error bars.

On a like-for-like test window the model was roughly five times more accurate than a commercial reference platform for the same region.

It is a low-cost tool built for a place where rainfall uncertainty decides whether a family plants — and the stated goal is to hand it to local government, not to publish and stop.

Discover More Child Prodigies

50+ stories of contemporary young geniuses changing chess, music, science, and art.

Explore All Prodigies →

Compare com os grandes

John Von Neumann vs Karl MarxLeonhard Euler vs PlatoHippocrates vs Wolfgang PauliAndrew Wiles vs David Hilbert
Ver Ranking →Todas as comparações →

Crianças prodígio

Ethan BortnickEthan BortnickGuinness World Record — Youngest Solo Musician to Headline a…Tathagat Avatar TulsiTathagat Avatar TulsiEarned a BSc at 11, an MSc at 12, and became an IIT professor…Lydia SebastianLydia SebastianMensa Score 162 at Age 12 — Indian-British Doctor's Daughter…Colin CarlsonColin CarlsonBegan university courses at nine and now forecasts the next…
Crianças prodígio →

Jogue e volte amanhã

Desafio diário do gênio · Guess the genius
English naturalist whose 1859 book On the Origin of Species introduced evolution by natural selection. Who is it?
Qual gênio é você? Teste de QI grátis
Which genius shares your birthday? Get one unforgettable mind in your inbox every week. Free, unsubscribe anytime.
or create a free account →