What is a climate model?
A climate model is a computer program that copies how the climate system works. It follows the air, the oceans, the ice and the land, and how they swap heat and water.
The model cuts the whole Earth into a grid of boxes. In a typical global model each box is about 100 km wide. The air is stacked in many layers (say 40), and so is the sea. In each box the computer works out temperature, wind, humidity and more from the laws of physics.
Sums: Earth's surface is about 510 million km². One box of 100 km × 100 km covers 10 000 km². So there are about 51 000 boxes per layer. With 50 layers that is about 2.5 million boxes.
How the model runs
The computer moves forward in small time steps, for example 30 minutes. In each step, every box passes heat, air and water to its neighbours, and the sun adds energy. Then it takes the next step. One year is 365 × 48 = 17 520 steps; 80 years is about 1.4 million steps. That is why climate models need supercomputers.
Some things are smaller than a box: clouds, tiny waves, forests. Scientists describe these with simple rules based on measurements. Such rules are called parameterisations. They are a main source of uncertainty.
Scenarios: what if we release more or less gas?
No model can tell what people will choose to do. So scientists make scenarios: stories about how much greenhouse gas the world releases, based on population, energy use, technology and policy. The best-known set is the Shared Socioeconomic Pathways (SSPs).
- Low emissions (for example SSP1-2.6): fast cuts. Warming by 2100 is about 1.8 °C above pre-industrial.
- Middle emissions (SSP2-4.5): slow cuts. About 2.7 °C.
- High emissions (SSP5-8.5): emissions keep rising. About 4.4 °C.
These are rounded values. A scenario is not a prediction. It is a "what if".
Can we trust climate models?
Scientists test a model by hindcasting: they start it in the past (say 1850) and check whether it reproduces the temperatures, ice and rainfall we measured. Good models do this well. Also, models built by dozens of different teams give the same big message.
Three kinds of uncertainty remain: (1) which scenario we follow, (2) differences between models, and (3) natural ups and downs, like El Niño years. Scientists run many models together (an ensemble) and show the range. Models are better at average temperature than at local rainfall.
Weather vs climate: weather cannot be predicted beyond about two weeks, but the 30-year average can be, in the same way that we cannot say which day of summer is hottest but we know summer is hotter than winter.
Try it: be the model
Draw a 4 × 4 grid on paper. Write a temperature in each box (make a few hot, a few cold). Now the rule: new value = the average of the box and its four neighbours. Do one round for all boxes, then another. See how heat spreads and the grid smooths out. You just ran a tiny model with a time step. Now add a rule of your own, like "box 7 gets +2 each round" (a heater). That is a scenario.
Key formulas and definitions
- Number of boxes ≈ Earth surface ÷ box area × layers
- Time steps = years × 365 × (24 × 60 ÷ step in minutes)
- Warming = temperature after − temperature before
- Climate = average weather over about 30 years
Worked examples
1. Earth's surface is 510 million km². How many 100 km × 100 km boxes cover it in one layer?
Box area = 10 000 km². Number = 510 000 000 / 10 000 = 51 000 boxes.
2. A model uses 30-minute steps. How many steps are needed for one year?
2 steps per hour × 24 hours = 48 steps per day. 48 × 365 = 17 520 steps.
3. In 2100 the low scenario gives +1.8 °C and the high one +4.4 °C. How much more warming does the high scenario give?
4.4 − 1.8 = 2.6 °C more.
4. If the boxes are made 50 km wide instead of 100 km, how many times more boxes are needed in one layer?
Area of a box becomes (50 × 50) = 2 500 km², a quarter. So 4 times as many boxes.
Common mistakes
- Saying a scenario is a prediction. It is a "what if" story, and the real path depends on our choices.
- Mixing up weather and climate. Weather is today; climate is the 30-year average.
- Thinking a model is wrong because it cannot say if it will rain on one day. It is not built for that.
- Forgetting that smaller boxes need many more calculations.