西太平洋副热带高压和南亚高压对东亚区域天气气候影响显著，运用数值模式集合系统提升其预报准确率对我国天气预报意义重大。采用国家级气象业务规范指标，系统地评估了中国气象局数值预报中心自主研发的GRAPES全球集合预报业务模式系统对2019年西太平洋副热带高压和南亚高压的集合预报技巧，并考察了不同集合方法对预报效果的影响，从而为东亚天气特别是极端事件的预报提供参考。结果显示，GRAPES全球集合预报系统对西太平洋副热带高压脊线的预报技巧最高，强度和面积次之，表现为偏弱的估计，西伸脊点的预报效果相对较差，表现为较观测偏东；对南亚高压强度和中心纬度指数的预报技巧较高，而对中心经度指数预报技巧相对较低。采用最大（小）值法可以有效降低该模式对西太平洋副热带高压强度和面积（西伸脊点）指数的预报偏差。而在南亚高压预报中，集合平均法比最值法具有略高技巧。对于极端性预报，最大值法较集合平均法可以显著提升对西太平洋副热带高压和南亚高压指数极端情形的预报性能，这从个例分析中也得到证实。从而表明集合最值法比平均法可能更适用于该模式的极端事件预报，应在业务应用中加以重视。

Based on the operational standard indices, the prediction skills of the Western-Pacific Subtropical High (WPSH) and South-Asian High (SAH) using 2019 real-time forecasts derived from the Global Ensemble Prediction System of GRAPES (GRAPES-GEPS) in China Meteorological Administration (CMA) Numerical Prediction Center were evaluated and the effects of different ensemble approaches on the prediction skills of WPSH and SAH indices were further investigated in this study. The results show that for WPSH, the GRAPES-GEPS has its highest prediction skill for the ridge line index, considerably high skill for the intensity and area indices, but relatively low skill for the western boundary index, and for SAH, it has comparatively high skill for the intensity and center latitude indices, but relatively lower skill for the center longitude index. Prediction errors of the GRAPES-GEPS for the WPSH forecasts are featured by the weaker intensity and area and the more eastward center position, compared with the observation, which can be effectively reduced by employing the maximum/minimum approach from ensemble members, relative to the ensemble mean approach. By direct comparison, prediction errors of the GRAPES-GEPS for the SAH forecasts are featured by the weaker intensity and the more southward center position, which tends to be slightly reduced using the ensemble mean approach. Finally, for the extreme forecast, the maximum approach provides superior performance for both WPSH and SAH than the ensemble mean approach, which can be validated in terms of the two extreme cases. These results clearly indicate that the maximum approach could better improve the GRAPES-GEPS performance for the extreme forecasting of the two primary circulation patterns than the traditional ensemble mean approach.