MTG satellite
Use of Meteosat Third Generation data to monitor the development and rapid evolution of convective systems.
thunderstorms · hail · heavy precipitation
AI that anticipates severe weather.
AI4VIGIL combines MTG satellite, radar and lightning data in a deep learning model to forecast the trajectory and intensity of hazardous weather every 10 minutes.
10 min
update frequency
3 h
operational forecast horizon
MTG · Radar · Lightning
multi-source fusion
France / Europe
European data foundation
Active monitoring
10 minupdate
0–3 hForecast
MTG + Radar+FoudreAI fusion
the concept
From weather observation to predictive vigilance.
The neural network does not simply enhance an existing forecast: it directly learns the rapid-evolution signatures of convective cells from complementary observation streams.
AI4VIGIL
A short, clear and operational chain: real-time weather data, deep learning fusion, then production of spatialized vigilance for the immediate forecast horizon.
Detect intensification signals earlier
Identify where risk is escalating
Update the forecast at the pace of observations
Not a static extrapolation.
Not a forecast recalculated only once an hour.
A living forecast.
Updated every 10 minutes to follow the real evolution of weather phenomena.
technology
Three data streams, fused in real time.
Satellite, radar and lightning observations are assimilated by a deep learning model trained on episodes of heavy precipitation, thunderstorms and hail.
Weather radar
Analysis of reflectivity, cell organization and signatures associated with heavy precipitation.
Lightning
Monitoring electrical activity as an indicator of intensification, maturity and thunderstorm severity.
Multi-source fusion
Precipitation
Forecast des zones de pluie et des intensités, y compris hors contexte strictement orageux.
Severe thunderstorms
Anticipation of high-potential cells, their trajectory and evolution.
Hail
Detection of signatures favorable to hail events and monitoring of their movement.
forecast cycle
A short cycle designed to support decisions.
Each cycle transforms recent observation data into an actionable very-short-range forecast.
T=0
Input data
Acquisition of satellite, radar and lightning streams over the monitored area.
T + 5
AI fusion
Assimilation and processing of signatures by the deep learning model.
T + 10
Forecast publiée
Production of a quantitative and spatialized output for the monitored phenomena.
0–3 h
Forecast
Continuous monitoring of trajectory, intensity and associated risk over 3 hours.
team
Where weather expertise meets deep learning.
Yann Amice
Meteorologist and oceanographer
33 years of experience in operational meteorology, expertise in hazardous weather, short-range forecasting and decision support.
Adrien Bufort
Deep Learning & AI
Specialist in deep learning modeling, design, training and optimization of models applied to multi-source meteorological data.
a Weather'n'Co brand
35 years of expertise serving performance.
From operational meteorology to artificial intelligence, one constant requirement: turning weather information into decisions.
Meteorological expertise
A benchmark for offshore racing weather
Weather'n'Co, founded by Yann Amice, is recognized in the offshore racing world as an essential performance partner. For more than twenty years alongside top skippers, the brand has stood for strategic analysis, reliability in critical situations and operational precision.
AI4VIGIL is the technological extension of this expertise: an AI-powered nowcasting platform designed to anticipate severe weather before impact, across Europe.
from weather analysis to routing decisions
Anticipate severe weather before impact.
AI4VIGIL aims for faster, more spatially precise weather vigilance that follows the real pace of observations.
Download the app
