Climate model temperature predictions sped up with new AI emulator
EC-EarthFlow: Probabilistic emulation of daily transient global climate model simulations with flow matching
Machine Learning
Summary
Climate models simulate how Earth's temperature changes over time, but they take a long time to run. The authors created EC-EarthFlow, an AI model that learns from a detailed climate simulator and predicts daily temperature changes much faster. This AI mimics important climate patterns like daily variations and long-term trends. It can make reliable predictions for up to several months ahead, using just past temperature data.
What this means in practice
- •For climate modeling teams: Run extended daily temperature predictions at lower computational cost by using AI emulators trained on existing climate model data.
- •For weather forecast modelers: Integrate AI emulators to generate longer-range daily temperature fields that retain climate variability and trends without full physical model runs.
Authors
Kirien Whan, Nikolaj T. Mücke, Karin van der Wiel
Abstract
We introduce EC-EarthFlow, a generative flow matching model that emulates simulations from the physical climate model EC-Earth3. The model is trained on transient simulations from EC-Earth3 (1950-2166, SSP2-4.5) to predict the day ahead temperature field from the previous days temperature as well as annual mean temperature. Predictions are made auto-regressively with rollout periods of between a month and an extended season. Using only this variable of interest, we are able to reproduce the daily variability, spatial patterns, annual cycle and long-term trend from EC-Earth3 at a substantially lower computational cost than the physical model. We demonstrate that EC-EarthFlow is stable for long inference periods, and that it can learn the physical relationships as simulated in EC-Earth3.