BCC-CPSv3-S2Sv2模式对中国区域降水和气温预测的性能评价

Performance evaluation of BCC-CPSv3-S2Sv2 for precipitation and temperature prediction over China

利用北京气候中心第3代气候模式预测业务系统的第2代次季节至季节预测子系统(Beijing Climate Centerclimate prediction system version 3-subseasonal to seasonal version 2,BCC-CPSv3-S2Sv2)模式发布的次季节至季节(sub-seasonal to seasonal,S2S)回算数据,从确定性预报和概率性预报这2方面对该模式降水和气温的预测性能进行系统评价,并与上一代北京气候中心第2代气候模式预测业务系统的第1代次季节至季节预测子系统(Beijing Climate Center-climate prediction system version 2-subseasonal to seasonal version 1,BCC-CPSv2-S2Sv1)模式和欧洲中期天气预报中心综合预报系统(European Centre for Medium-Range Weather Forecasts-integrated forecasting system,ECMWF-IFS)模式预测结果的表现对比,研究现有模式较上一代在预测表现方面是否有所提升,且其是否达到国际先进水平。结果表明:现有模式整体预测效果优于上一代,但差于ECMWF-IFS模式。在提前1周时,现有模式对降水和气温的预测技巧有明显改进,但在提前3、4周时,预测效果仍较差,降水预测相关系数的平均值在提前4周时为0.11。BCC-CPSv3-S2Sv2春季降水预测对青藏高原中部地区在各个起报时间始终有较好的预测效果,相关系数均大于0.4。此外,现有模式冬季降水预测表现与ECMWF-IFS模式相近。对于气温预测,提前4周时其在西北、华北部分区域的预测技巧比提前3周、甚至2周更高,表明模式能较好把握S2S尺度上气温预测的可预报性来源。从概率预报的表现来看,现有模式降水和气温预测的欠离散问题更突出,但均值误差变小。此外,现有模式降水预测对负异常事件的预测表现更好。

This study evaluates the performance of Beijing Climate Center‑climate prediction system version 3‑subseasonal to seasonal version 2(BCC‑CPSv3‑S2Sv2)in sub‑seasonal to seasonal(S2S)precipitation and temperature prediction over China by using both deterministic and probabilistic metrics.Apart from this,forecasts from Beijing Climate Center‑climate prediction system version 2‑subseasonal to seasonal version 1(BCC‑CPSv2‑S2Sv1),the previous forecast system of China Meteorological Administration,the BCC‑CPSv3‑S2Sv2,and European Centre for Medium‑Range Weather Forecasts‑integrated forecastingsystem(ECMWF‑IFS)are also evaluated so as to compare the performance of three models to investigate whether the current model has improved its forecast performance compared with the previous generation and whether it has reached the international advanced level.Results show that the overall prediction performance of BCC‑CPSv3‑S2Sv2 is better than that of the BCC‑CPSv2‑S2Sv1,but worse than that of the ECMWF‑IFS.When predicting precipitation and temperature one week in advance,the prediction accuracy of precipitation and temperature of BCC‑CPSv3‑S2Sv2 is improved obviously,but when predicting precipitation three or four weeks in advance,it is still poor,and the average correlation coefficient of precipitation forecast four weeks in advance was 0.11.Specifically,the precipitation prediction of BCC‑CPSv3‑S2Sv2 shows a persistent good prediction result over the central Qinghai‑Tibet Plateau at each onset time in spring,and the correlation coefficient is greater than 0.4.Moreover,the winter precipitation prediction of BCC‑CPSv3‑S2Sv2 performs similarly with that of ECMWF‑IFS.For temperature prediction,the performance for 4‑week ahead is more accurate than 3‑week ahead,even 2‑week ahead in Northwest and North China,which means that BCC‑CPSv3‑S2Sv2 can better grasp the source of the predictability of temperature prediction at S2S scale.From a probabilistic forecast perspective,the underdispersion problem of precipitation and temperature forecast is more prominent in the existing model,but the mean error is smaller.In addition,existing model precipitation forecasts are better at predicting negative abnormal events.