六个气候系统模式对西南地区2 m温度的预报检验分析

通过分析BCC_CSM1.1m、ECMWF_SYSTEM5、NCEP_CFSv2、FGOALS_f、FGOALS_s2、PCCSM4气候模式的模式初始误差、空间相关系数和时间距平相关系数,评估国内外6种气候预测模式在西南地区的预测能力。结果表明：FGOALS_s2和ECMWF_SYSTEM5模式的初始误差最大,FGOALS_f模式的最小。模式在西南地区西南部模拟2 m温度的初始误差较小,ECMWF_SYSTEM5模式在四川盆地东部的预测场均方根误差较小。ECMWF_SYSTEM5模式预报的空间相关系数最大（小）值随时间有所下降（上升）,尤其是2016~2018年变化趋势较为明显,其产生原因可能与欧洲中心模式升级有关。采用ECMWF_SYSTEM5模式（西南和东部区域）和BCC_CSM1.1m模式（西北区域）预报的空间相关系数较高。随着预报时效的延长,各模式预报西南子区域的空间相关系数均呈先减后增的变化趋势;就东部子区域而言,BCC_CSM1.1m、NCEP_CFSv2和ECMWF_SYSTEM5模式预报的空间相关系数变化较小,其余各模式呈现单峰或双峰的变化特征;各模式对西北子区域的预报稳定度较高。从预报时效为1个月的时间距平相关系数水平分布看,各模式对西南区域东部的预报效果优于西部,这可能是西部地势较高且天气物理过程具有特殊性导致的。PCCSM4模式对高原区域预报效果较好,但对于地形起伏较大的区域,其预报效果仍然较差。采用FGOALS_s2和PCCSM4模式对西南子区域的预报效果较好,FGOALS_s2和ECMWF_SYSTEM5模式对西北子区域的预报效果较好,PCCSM4和NCEP_CFSv2模式对东部子区域的预报效果较好。从预报时效为1~3个月的时间距平相关系数变化趋势看出,各模式在预报时效为1个月的时间距平相关系数较高（低）的区域,其数值随预报时效的延长有下降（上升）的变化趋势。

Seasonal Prediction Performance of Temperature in Southwest China by Six Climate Prediction Models

The Model initial error, Temporal Correlation Coefficient, and Anomaly Correlation Coefficient are investigated by using the hindcasts from the six climate modes, including BCC_CSM1.1m, ECMWF_SYSTEM5, NCEP_CFSv2, FGOALS_f, FGOALS_s2, PCCSM4. The result shows that FGOALS_s2 and ECMWF_SYSTEM5 have the maximum initial error among six models, and FGOALS_f has the minimum error. The southwestern part of evaluated area has the smaller initial error compared with other parts. For the temperature forecast in eastern Sichuan Basin, ECMWF_SYSTEM5 has the highest sensitivity to terrain changes in Southwest China, and the root mean squares of the analytical field are small. The maximum (minimum) ACCs of the ECMWF_SYSTEM5 prediction decreases (increase) over time, especially from 2016 to 2018, which may be related to the upgrading of the European Center mode. The ACCs of ECMWF _ SYSTEM5 model ( southwest and east regions ) and BCC _ CSM1.1m model ( northwest region ) are higher. The ACCs of BCC_CSM1.1m, NCEP_CFSv2, and ECMWF_SYSTEM5 are little. The reasons for this difference still need to be further analyzed. From the perspective of the horizontal distribution of the TCCs predicted by six climate model for the temperature of southwest China in one month lead, the forecast effect of each model for the eastern part is better than that of the west. Due to its particularity, the climate models have poor treatment effects. PCCSM4 has higher TCCs on the plateau area, but its forecast effect is still poor for the transitional area, with large terrain undulations, between the basin and the plateau. Judging from the regional average TCCs in 1~6 months lead, FGOALS_f has the worst forecasting effect, and it shows a single-peak change trend with the increase of the lead months, and the other models show a multi-peak change trend. It can be seen from the trend of TCCs changes with the forecast time of 1~3 months, for each model, the area with a higher (low) TCCs of one-month forecast time, its value decreases (increases) with the increases of forecast time.