Training Climate & Weather Weather Prediction — Differential Equations in the Sky
1 / 1

Weather Prediction — Differential Equations in the Sky

20 min Climate & Weather

Weather Prediction

When you check tomorrow's forecast, you're seeing the output of billions of calculations — differential equations solved on supercomputers, predicting how air, moisture, and heat will move across the planet.

The Primitive Equations

Modern weather models solve a set of equations called the primitive equations of atmospheric motion:

Core Weather Equations

Momentum: $$\frac{D\mathbf{v}}{Dt} = -\frac{1}{\rho}\nabla p - 2\mathbf{\Omega} \times \mathbf{v} + \mathbf{g} + \mathbf{F}$$

Continuity: $$\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \mathbf{v}) = 0$$

Thermodynamic energy: $$c_p \frac{DT}{Dt} - \frac{1}{\rho}\frac{Dp}{Dt} = Q$$

Each equation is built from the same math concepts you're learning:

  • Derivatives ($\frac{\partial}{\partial t}$) — rates of change
  • Vectors ($\mathbf{v}$, $\nabla$) — direction and magnitude
  • Multiplication and division — combining physical quantities
  • Cross products ($\mathbf{\Omega} \times \mathbf{v}$) — the Coriolis effect that creates cyclones

How Weather Models Work

Meteorologists divide the atmosphere into a 3D grid of millions of cells. For each cell, they:

  1. Know the current temperature, pressure, humidity, and wind (from observations)
  2. Use the equations above to calculate how much each value changes in a tiny time step (say, 10 minutes)
  3. Update all cells simultaneously
  4. Repeat thousands of times to reach the forecast time

This is numerical integration — approximating derivatives with arithmetic.

Example: Simple Temperature Forecast

A grid cell has temperature $T = 20°C$. The rate of temperature change is $\frac{dT}{dt} = -0.5°C/\text{hr}$ (cooling after sunset). What will the temperature be in 6 hours?

$$T(6) = T(0) + \frac{dT}{dt} \times \Delta t = 20 + (-0.5)(6) = 20 - 3 = 17°C$$

Real models do this calculation for millions of cells with interacting quantities — but the core operation is this simple: multiply the rate by the time and add.

Chaos and the Butterfly Effect

In 1961, Edward Lorenz discovered that weather equations are chaotic — tiny differences in starting conditions grow exponentially:

$$|\delta(t)| \approx |\delta_0| \cdot e^{\lambda t}$$

where $\lambda$ is the Lyapunov exponent. This exponential growth is why weather forecasts become unreliable after about 10 days — and it's the same exponential function from algebra!

The Scale of the Math

Weather Model FactNumber
Grid cells~1 billion
Variables per cell~7 (T, p, humidity, u, v, w, ...)
Time steps per day~150
Calculations per forecast~1015 (1 quadrillion)
Supercomputer speed needed~10 petaflops
Key Insight

Every weather forecast is a massive exercise in arithmetic, algebra, and calculus — performed a quadrillion times. The "multiply and add" you learn in arithmetic is literally the operation running on supercomputers to predict tomorrow's weather.

Interactive Explorer: Simple Temperature Forecast
Predicted Temperature = 17.0 °C
Change = -3.0 °C