Earth models can predict the planet's future but not their own

"I n the 1960s, meteorologist Edward Lorenz was running weather simulations on an early computer system when he realized that a small rounding difference led to extremely divergent weather predictions. He later called this idea the butterfly effect to communicate that small changes in initial conditions, like a butterfly flapping its wings in Nepal, could produce wildly different outcomes, like rain in New York."

"Today, computers are much more powerful than when Lorenz was working, and scientists use a special kind of simulation that accounts for physics, chemistry, biology, and water cycles to try to grasp the past and predict the future. These simulations, called Earth system models, or ESMs, attempt to consider the planet as a system made up of components that nudge and shove each other."

""It's coupling together usually an atmosphere model, an ocean model, a sea ice model, land model, together to get a full picture of a physical system," said David Lawrence, a senior scientist at the National Center for Atmospheric Research's Climate and Global Dynamics Laboratory, which he noted was recently changed to the CGD Laboratory to remove the word climate. The models also move beyond the planet's physical components, including chemistry and biology."

One of the world's foremost climate models faces funding threats that risk undermining integrated climate prediction capabilities. Small differences in initial conditions can produce large divergences in weather outcomes, an effect identified in the 1960s and known as the butterfly effect. Modern Earth system models (ESMs) integrate atmospheric, oceanic, sea-ice, land, chemical, biological, and water-cycle processes to simulate past and future climate. ESMs evolved from early physical climate models as computing power and environmental knowledge increased. Coupling these components enables more realistic system behavior and can yield surprising conclusions about climate impacts, variability, and regional projections.

Read at Ars Technica