Xin Huang, Aijun Ding

Weather prediction is essential to human daily life. Current numerical weather prediction (NWP) models are still subject to substantial forecast bias and rarely consider the impact of atmospheric aerosols, despite of the consensus of aerosols as the most important sources of uncertainty in predicting climate change. Here we show aerosols as an important driver biasing daily temperature prediction. By using observation minus forecast analysis based on the Global Forecast System and sounding observations, we found that monthly-averaged bias in 24-hour temperature forecast varies between ±1.5℃ in regions dominated by different types of aerosols. The biases depend on the properties of aerosols over different underlying land surface and on aerosol-cloud interactions over oceans. We also revealed that forecast errors are rapidly magnified with time over regions featuring higher aerosol loadings. Our study provides “direct” evidence of aerosols’ impacts on daily weather forecasts and bridges the gaps between weather forecast and climate science regarding the understanding of the impact of atmospheric aerosols.