Forecasting the Future With Yesterday's Climate: Temperature Bias in AI Weather and Climate Models

Abstract AI‐based climate and weather models provide fast, skillful forecasts yet face a key challenge: predicting future climates while being trained with historical data. We investigate this issue by analyzing boreal winter land temperature biases in AI weather (FourCastNet V2 Small and Pangu Weather) and climate (Ai2 Climate Emulator version 2) models. We evaluate these models during time periods that are significantly more recent than the bulk of their training data, allowing us to assess how well they generalize to more modern conditions. We find that all models produce cold‐biased mean temperatures, resembling climates from 15 to 20 years earlier than their prediction period. Furthermore, FourCastNet's and Pangu's cold bias is strongest for the hottest predicted temperatures, indicating limited training exposure to modern extreme heat events. In contrast, ACE2's bias is more evenly distributed but largest in regions, seasons, and parts of the temperature distribution where historic global warming is most pronounced.