《海洋预报》| 盘古气象模型在南极区域的预报技巧评估

摘要：采用南极站点观测数据、GFS模式全球预报数据和ERA5再分析数据，对盘古气象模型在南极区域的预报技巧进行了系统评估。结果表明：无论以ERA5再分析资料还是南极地面观测数据作为真值，盘古气象模型在南极区域的预报技巧均优于美国GFS全球数值预报产品；对于初始误差较大的变量，盘古气象模型的误差发展呈现先减小后增大的趋势。进一步分析表明，在风速较大条件下，盘古气象模型和GFS的预报风速均偏弱；在温度较低的条件下，盘古气象模型和GFS的预报温度均偏低。相比于GFS预报产品，盘古气象模型对中国南极长城站（中山站）冬季（8月）的海平面气压预报误差偏大约0.5 hPa(1.0 hPa)，且对长城站冬季（8月）的10 m风速预报误差偏大约0.4 m/s。然而，盘古气象模型在长城站冬季（8月）的2 m温度预报，以及中山站的10 m风速和2 m温度平均预报误差，均明显小于GFS预报产品。

关键词：盘古气象模型 南极 预报技巧 深度学习模型 

Abstract：We verify the forecast skill of Pangu-Weather model over the Antarctic region against in-situ meteorological station observations, the Global Forecast System（GFS） forecasts, and the ECMWF Reanalysis v5（ERA5） data. The results show that the prediction skill of Pangu-Weather model is better than the GFS forecasts,with respect to the ERA5 reanalysis or the surface observations at Antarctic meteorological stations. The forecast error growth feature of Pangu-Weather model is decreasing initially, followed by increasing, for those variables with larger initial errors. Further analysis shows that the wind speed predicted by the Pangu-Weather model and GFS is relatively weak under conditions of high wind speed, and the air temperature predicted by the PanguWeather model and GFS is relatively low under conditions of low temperature. Compared to the GFS forecasts in August, the Pangu-Weather model has a sea level pressure forecast error of about 0.5 hPa（1.0 hPa） for the Great Wall Station（Zhongshan Station）, and a forecast error of about 0.4 m/s for the 10 m wind speed at the Great Wall Station. However, the August forecast errors of the Pangu-Weather model for 2 m air temperature at Great Wall Station, and 10 m wind speed and 2 m air temperature at Zhongshan Station are significantly smaller than those of the GFS forecasts.

Key words：Pangu-Weather model; Antarctic; prediction skill; deep-learning model

读者也可以通过公众号菜单栏

“相关知识”板块进入《当期目录》和《过刊浏览》

查阅详细期刊内容

未来两天海洋环境预报

35个美丽海岛环境预报