资源环境科技发展态势分析平台

Twenty-first-century climate change projections are uncertain, especially on regional scales. An important source of uncertainty is that climate models exhibit biases, which limits their ability to predict climate. One of the largest biases is the too warm sea surface temperature (SST) in the eastern tropical Atlantic (TA), reflecting deficient atmospheric and oceanic circulation. Here we show that CO₂-forced TA-sector climate changes simulated by state-of-the-art climate models exhibit a strong mean-state dependence. In particular, models simulating largest SST warming in the eastern TA, consistent with the warming observed since the mid-20^th century, typically exhibit a more realistic mean state than models simulating largest warming in the western TA. The former models exhibit a larger climate sensitivity, and predict stronger and in part qualitatively different climate changes over the TA sector, for example in precipitation. These findings may help to reducing uncertainty in TA-climate change projections.