GRAPES区域集合预报系统对登陆台风预报的检验评估

EVALUATION OF THE FORECAST FOR LANDED TYPHOONS BY GRAPES-REPS REGIONAL ENSEMBLE PREDICTION SYSTEM

针对2015年7—9月登陆中国大陆沿海的台风,利用GRAPES-REPS区域集合预报资料和集合统计诊断分析方法,对登陆台风的移动路径、时间、地点、强度和降水等进行检验评估,以期为预报员应用GRAPES登陆台风概率预报提供依据。检验结果表明,(1)集合平均移动路径要优于控制预报,集合预报各成员登陆地点存在20~340 km差异,但实况登陆地点均能落在集合成员登陆地点中。(2)对24 h和48 h登陆地点误差而言,集合平均较控制预报更接近实况。(3)随着预报时间的趋近,集合平均、控制预报和集合成员登陆地点距离误差逐渐缩小,登陆地点空间位置预报也没有明显的系统性误差。(4)集合成员对台风登陆时间预报偏早,平均提前2.3 h。(5)在强度预报中,尽管最低气压和近中心最大风速存在登陆前偏弱而登陆后偏强的趋势,但登陆点预报值区间包含了实况观测值,表明GRAPES-REPS集合预报能够较好展示多种可能信息。(6)不同量级降水AROC评分为0.56~0.76,具有预报参考价值;另外AROC评分的高低及台风暴雨落区的准确性与台风登陆点和登陆时间误差密切相关。可见,GRAPES-REPS区域集合预报可以在台风登陆地点、时间、强度和降水预报等方面提供更多的预报不确定性信息,有助于做出正确的预报决策。

Based on the data of GRAPES-REPS Regional Ensemble Prediction System, four typhoons which landed in the coastal regions of Chinese continent for the period from July to September 2015 are evaluated by the methods of ensemble verification and diagnosis. The contents of evaluation included the typhoon track, landfall time and position and typhoon intensity as well as precipitation in order to be used better by forecasters. The results show that the average track by ensemble forecast is better than that of control forecast. The landed position of all ensemble members is quite different, which ranges from 20 to 340km, but observed landed positions of these four typhoons strike within the forecasts of ensemble members.The landed position errors of 24h and 48h ensemble average prediction are lower than that of control forecasts （82km and 197km）.The results also showed that there are no obvious systematic errors of landed positions and the shorter the forecast time, the smaller error the landed position. For the landed time forecast, it is 2.5h ahead of ensemble members on average. For the intensity forecast of the typhoon, the minimum pressure and near-the-center maximum wind velocity are weak before landing and strong after landing with the observed landed position falling in the predicted values of ensemble members, which showed that GRAPES-REPS has the ability to make forecastsclose to the reality and exhibits the uncertainty of prediction. The AROC scores of precipitation category are from 0.59 to 0.76, which are positive for the forecaster. Thus, GRAPES-REPS can provide much more uncertainty information of typhoons track, landfall time, landfall position and intensity forecast than that of deterministic models, greatly improving the decision-making by forecaster.