Comment Fonctionne la Météo

What is Weather, Exactly?

Understanding how weather is forecast begins with understanding what weather actually is. Weather is the state of the atmosphere at a specific place and time. It describes everything from the temperature and humidity you feel when you step outside to the wind pushing against you and the rain soaking your shoes. Weather is not random — it is governed by physical laws. And because it follows physical laws, it can be predicted.

Weather is driven by four core ingredients working together: temperature (the amount of heat energy in the air), air pressure (the weight of the atmospheric column above any given point), humidity (the concentration of water vapour in the air), and wind (the horizontal movement of air from high-pressure zones to low-pressure zones). When these four variables interact across the surface of the Earth, they create every weather pattern we experience — from a gentle summer breeze to a devastating hurricane.

The Data Collection Phase: Eyes on the Sky

Before any forecast can be made, meteorologists need a precise snapshot of the current state of the atmosphere. This snapshot is assembled from a global network of observation systems operating continuously around the clock.

Weather satellites are the most familiar of these systems. Geostationary satellites (which hover at a fixed point 36,000km above the Earth) such as GOES-16 and GOES-18 (operated by the US) and the European Meteosat series provide continuous, real-time imagery of cloud cover, storm systems, sea surface temperatures, and atmospheric water vapour. Polar-orbiting satellites pass over the entire globe twice daily, capturing vertical profiles of temperature and humidity at different altitudes.

Weather balloons are launched twice daily from hundreds of stations worldwide — at midnight and noon UTC. Each balloon carries a radiosonde (a small instrument package) that transmits temperature, humidity, pressure, and wind data as it ascends from the surface to about 30km altitude. These vertical profiles are irreplaceable for understanding the three-dimensional structure of the atmosphere.

Ground-based stations — airports, automated weather stations, ocean buoys, and ships — add a dense surface-level grid of observations covering temperature, wind, pressure, precipitation, and visibility. Commercial aircraft also contribute temperature and wind measurements via systems like AMDAR (Aircraft Meteorological Data Relay).

Numerical Weather Prediction: The Mathematics of the Atmosphere

Once all this observational data is assembled (a process called "data assimilation"), the core of modern forecasting begins: Numerical Weather Prediction (NWP).

NWP works by dividing the atmosphere into a massive three-dimensional grid of cells — imagine stacking thousands of layers of horizontal grid squares from the surface up to the stratosphere. Each cell might be 10–25km wide horizontally and a few hundred metres deep vertically for a global model, or as small as 1–2km for high-resolution local models.

The current observed state of the atmosphere (temperature, wind, humidity, pressure) is loaded into each grid cell. Then a set of equations — the primitive equations, derived from Newton's laws of motion and thermodynamics — are solved for each cell to calculate how the state of the atmosphere will evolve over the next short timestep (typically 5–30 minutes). This process is repeated thousands of times, stepping forward through time to produce a forecast for 12, 24, 48, 72 hours or beyond.

The computing power required for this is immense. The European Centre for Medium-Range Weather Forecasts (ECMWF) — widely regarded as the world's best forecasting centre — operates one of the world's most powerful supercomputers, capable of more than 100 petaflops (100 quadrillion calculations per second) specifically for weather modelling.

Ensemble Forecasting: Understanding Uncertainty

Any forecast carries uncertainty. To quantify that uncertainty, modern forecast centres run ensemble models. Rather than running the model once, they run it 50 or more times simultaneously, each time introducing tiny, carefully designed variations in the starting conditions. If all 50 runs agree — say, they all predict heavy rain on Thursday — the forecast is highly confident. If the runs diverge wildly, the meteorologist knows there is genuine uncertainty in the forecast.

Ensemble forecasting is why you see probability-based forecasts ("70% chance of rain") rather than certainties. The probability reflects the proportion of ensemble members that agree on a given outcome.

The 7-Day Forecast Limit: Why Chaos Matters

The atmosphere is what mathematicians call a "chaotic system." This means that tiny differences in the initial state grow exponentially over time — the famous "butterfly effect." In practice, this limits useful deterministic weather forecasting to roughly 7–10 days. Short-term forecasts (1–3 days) now achieve accuracy over 90% of the time. Five-day forecasts are reliable roughly 80% of the time. Beyond 10 days, forecast skill drops sharply.

This is a hard physical limit, not a technological one. No matter how powerful computers become, the chaotic nature of the atmosphere means that perfect 14-day forecasts are theoretically impossible.

How SunorSnow Uses Weather Data

SunorSnow's weather dashboard displays live data from WeatherAPI.com, which aggregates output from multiple global and regional NWP models. When you add a city to your dashboard, you see temperature, humidity, wind speed, UV index, AQI, and hourly and 3-day forecasts — all derived from the modelling processes described above.

By tracking multiple cities on a single dashboard, you can observe weather patterns playing out in real time across different parts of the world — seeing the jet stream effects, pressure systems, and seasonal shifts that meteorologists analyse every day.