EVALUATION OF THE FORECAST FOR LANDED TYPHOONS BY GRAPES-REPS REGIONAL ENSEMBLE PREDICTION SYSTEM
-
摘要: 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。
-
关键词:
- GRAPES-REPS /
- 区域集合预报 /
- 登陆台风 /
- 预报评估
Abstract: 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.-
Key words:
- GRAPES-REPS /
- regional ensemble prediction system /
- landed typhoon, evaluation /
-
图 6 同图 5,但为登陆地点
图 7 三个台风的连续5次集合成员预报台风登陆时间
说明同图 5。
表 1 2015年三个台风登陆实况概况
台风名 强度 近中心最大风速/(m/s) 最低气压/hPa 沿海登陆地点 登陆时间 降水强中心 “莲花”(1510) TY 35 970 广东省陆丰市甲东镇 7月9日04 UTC 登陆点北侧 “苏迪罗”(1513) STY 45 950 福建省福田 8月8日14 UTC 登陆点北侧 “杜鹃”(1521) STY 48 944 福建省莆田市 9月29日01 UTC 登陆点北侧 表 2 双态分类联列表
双态分类 预报出现 预报不出现 观测相加 观测出现 X Y X+Y 观测不出现 Z W Z+W 预报相加 X+Z Y+W 表 3 降水观测统计
分级(i) 概率范围/% 未出现次数 出现次数 0~9 0~9 a1 b1 10~19 10~19 a2 b2 20~29 20~29 a3 b3 30~39 30~39 a4 b4 40~49 40~49 a5 b5 50~59 50~59 a6 b6 60~69 60~69 a7 b7 70~79 70~79 a8 b8 80~89 80~89 a9 b9 90~99 90~99 a10 b10 -
[1] 朱乾根, 林锦瑞, 寿绍文, 等.天气学原理和方法[J].北京:气象出版社, 2007. [2] 薛建军, 李佳英, 张立生, 等.我国台风灾害特征及风险防范策略[J].气象与减灾研究, 2012, 35(1):59-64. [3] 钱传海, 董林. 我国台风业务现状及其关键技术[C]//第十七届全国热带气旋科学研讨会论文集. 北京: 中国气象学会, 2015. [4] Annual Tropical Cyclone Report 2010, 2011, 2012, 2013, 2014[R].http://www.usno.navy.mil/JTWC/annual-tropical-cyclone-reports. [5] Annual Report on Activities of the RSMC Tokyo Typhoon Center 2010, 2011, 2012, 2013, 2014[R].http://www.jma.go.jp/jma/jma-eng/jma-center/rsmc-hp-pub-eg/annualreport.html. [6] 端义宏, 余晖, 伍荣升.热带气象强度变化研究进展[J].气象学报, 2005, 63(5):636-645. [7] 陈联寿, 罗哲贤, 李英.登陆热带气旋研究进展[J].气象学报, 2004, 62(5):542-549. [8] 陈玉林, 周军, 马奋华.登陆我国台风研究概述[J].气象科学, 2005, 25(3):319-329. [9] 李江南, 王安宇, 杨兆礼, 等.台风暴雨的研究进展[J].热带气象学报, 2003, 19(s1):152-159. [10] 马雷鸣.国内台风数值预报模式及其关键技术研究进展[J].地球物理学进展, 2014, 29(3): 1 013-1 022. [11] EPSTEIN E S. Stochastic dynamic prediction[J]. Tellus, 1969, 21(6): 739-759. [12] LEITH C E. Theoretical skill of Monte Carlo forecasts[J]. Mon Wea Rev, 1974, 102(6): 409-417. [13] ZHANG Z, KRISHNAMURTI T N. Ensemble forecasting of hurricane tracks[J]. Bull Amer Meteor Soc, 1997, 78(12): 2 785-2 795. [14] CHEUNG K K W, CHAN J C L. Ensemble forecasting of tropical cyclone motion using a barotropicmodel, Part Ⅰ:Perturbations of the environment[J]. Mon Wea Rec, 1999, 127(6):1 229-1 243. [15] KUMAR T S V, KRISHNAMURTI T N. Michael fiorinomultimodelsuperensemble forecasting of tropicalcyclones in the Pacific[J].Mon Wea Rec, 2003, 131(3):574-583. [16] 周霞琼, 端义宏, 朱永禔.热带气旋路径集合预报方法研究Ⅰ——正压模式结果的初步分析[J].热带气象学报, 2003, 19(1):1-8. [17] 周霞琼, 端义宏, 朱永褆.应用集合预报产品开展热带气旋路径概率预报试验[J].气象学报, 2002, 60(1): 111-118. [18] 王晨稀, 梁旭东.热带气旋路径集合预报试验[J].应用气象学报, 2007, 18(5):586-593. [19] 黄小刚, 费建芳, 陆汉城, 等.基于集合Kalman滤波数据同化的热带气旋路径集合预报研究[J].大气科学, 2007, 31(3): 468-478. [20] 黄小刚, 费建芳, 陆汉城, 等.基于集合Kalm滤波数据同化与偏差修正的热带气旋强度集合预报研究[J].气象学报, 2010, 68(1):79-87. [21] 陈静, 佟华, 徐枝芳, 等.NMC强天气模式诊断变量和概率预报产品及初步应用检验[J].气象, 2010, 36(12):42-49. [22] 杨昌贤, 郑艳, 林建兴, 等.数值预报产品检验和评估[J].气象研究与应用, 2008, 29(2):32-37. [23] 张涵斌, 陈静, 智协飞, 等.GRAPES区域集合预报系统应用研究[J].气象, 2014, 40(9): 1 076-1 087. [24] 潘旸, 沈艳, 宇婧婧, 等.基于贝叶斯融合方法的高分辨率地面-卫星-雷达三源降水融合试验[J].气象学报, 2015, 73(1):177-186.