The Planet in Motion

PREDICTION

In 1904, Vilhelm Bjerknes proposed that tomorrow's weather could be treated as an initial-value problem: describe the atmosphere now, apply the laws governing it, and calculate the next state. The proposal was mathematically respectable and operationally useless. There was not enough data and there was nowhere near enough computation.

The important idea survived. Prediction became the problem of evolving state through time. Weather was first because the equations were known. Water, forests, crops, animal movement, electricity systems and freight followed with different mathematics and different kinds of uncertainty. The planet turned out to contain many futures, none of them running on the same clock.

Every useful model simplifies. Weather models filter processes they cannot afford to resolve. Hydrological models approximate what happens below the ground. Forest models compress millions of organisms into cohorts or functional types. Food models eventually meet prices, policy and conflict, where physical law becomes much less helpful. A complete model is not the goal. A model that leaves out the right things is.

This volume follows that engineering problem from Richardson's hand calculations to ensemble forecasts, reanalysis, digital twins and coupled Earth-system models. The question is not whether prediction works. It plainly does. The interesting question is where it stops working, why it stops, and what kind of uncertainty should survive all the way to the person using the forecast.