SAMPLE OPTIMIZATION OF ENSEMBLE FORECAST IN SIMULATING TROPICAL CYCLONES BASED ON OBSERVED TRACKS
-
摘要: 集合预报是从一定误差范围内的一组初值出发，这组初值(样本)代表了大气状态的概率分布，集合预报中集合样本的好坏严重影响分析质量。质量较差样本进入集合预报中难免会降低集合预报的整体质量。由于集合样本是模拟大气可能状态的概率分布，因此样本的优选是提高分析质量的关键。通过对集合样本优胜劣汰来分析样本优选对模拟效果的影响。由于台风预报中台风路径的模拟至关重要，因此样本优选的方案为将样本模拟的路径信息与观测的台风报文路径相比较后，保留误差较低的样本，剔除误差较高的样本，从而提升样本的整体质量。但过多的样本被替换将导致集合离散度的大幅下降，因此替换样本的数量要适度。研究结果表明样本优选极可能有利于热带气旋路径和强度模拟的改进，其中对“妮妲”路径误差的改进为4% ~13%，对“鲇鱼”路径误差的改进为11%~28%，对“妮妲”的强度误差改进为5%~37%，“鲇鱼”的强度误差改进为1%~27%。Abstract: Ensemble forecasting is widely used in numerical weather prediction (NWP). However, the ensemble may not satisfy perfect Gaussian probability distribution owing to limited members, and some members significantly deviate from the true atmospheric state. Such impossible samples (belonging to small probability events) may downgrade the accuracy of ensemble forecast. In this study, the observed tropical cyclone (TC) track is used to restrict the probability distribution of samples by investigating the evolution of TCs. The method employed uses observation data rather than data assimilation. By restricting the probability distribution, ensemble spread could be decreased through sample optimization. In addition, the prediction results showed that track and intensity errors could be reduced by sample optimization. Comparing vertical structures of TCs considered in this study, different thermal structures was found to exist. This difference may have been caused by sample optimization, and it may affect intensity and track. Nevertheless, it should be noted that the replacement of a large number of inferior samples may inhibit the improvement of the simulated results.
-
Key words:
- ensemble forecast /
- sample optimization /
- tropical cyclone /
- observed track
-
图 3 热带气旋“妮妲”的模拟和观测中心分布图
黑点代表观测中心，红点代表好样本，蓝点代表坏样本，绿点代表一般样本。四列代表的时间为2016年8月1日06、07、08、09时，七行从上至下表示SOno、SO5、SO10、SO15、SO20、SO25、SO30试验。
图 4 热带气旋“鲇鱼”的模拟和观测中心分布图
黑点代表观测中心，红点代表好样本，蓝点代表坏样本，绿点代表一般样本。四列代表的时间为2016年9月26日06、07、08、09时，七行从上至下表示SOno、SO5、SO10、SO15、SO20、SO25、SO30试验。
图 7 观测和模拟的平均绝对路径误差
SO5、SO10、SO15、SO20、SO25、SO30所对应的热带气旋“妮妲”时间为2016年8月1日09时—2日18时，“鲇鱼”时间为2016年9月26日09时—28日21时。
图 8 台风“鲇鱼”中心的经向剖面图
X轴为经度(°E)，Y轴为气压(hPa)。第一、二、三列分别为12 h (2016年9月26日12时)、24 h(2016年9月27日00时)、48 h(2016年9月28日00时)预报结果。第1~第7行则分别代表SOno、SO5、SO10、SO15、SO20、SO25、SO30试验。
表 1 观测和模拟试验的路径、最低气压、最大风速误差 “妮妲”对应的时间为2016年8月1日09时—2日18时，“鲇鱼”对应的时间为2016年9月26日09时—28日21时。
气象因素 SOno SO5 SO10 SO15 SO20 SO25 SO30 路径/km “妮妲” 52.76 49.88 45.88 53.83 50.49 59.62 49.91 “鲇鱼” 56.99 45.28 41.14 46.06 47.67 45.23 50.53 最低气压/hPa “妮妲” 10.29 10.37 9.52 9.11 9.70 9.59 9.10 “鲇鱼” 6.97 6.16 6.39 6.53 7.29 6.86 6.75 最大风速/(m/s) “妮妲” 5.08 4.41 3.41 3.20 4.37 4.12 4.07 “鲇鱼” 3.14 2.51 2.29 2.55 3.36 2.57 2.77 
下载: 导出CSV
-
[1] SAETRA Ø, HERSBACH H, BIDLOT J R, et al. Effects of observation errors on the statistics for ensemble spread and reliability. European Centre for Medium-Range Weather Forecasts, Reading, Berkshire, United Kingdom[J]. Mon Wea Rev, 2004, 132(6): 1487-1501. doi: 10.1175/1520-0493(2004)132<1487:EOOEOT>2.0.CO;2 [2] RICHARDSON D S. Skill and relative economic value of the ECMWF Ensemble prediction system[J]. Q J Roy Meteor Soc, 2000, 126(563):649-667. doi: 10.1002/qj.v126:563 [3] PALMER T N. The economic value of ensemble forecasts asa tool for risk assessment: From days to decades[J]. Q J Roy Meteor Soc, 2002, 128(581): 747-774. doi: 10.1256/0035900021643593 [4] TAILLARDAT M, MESTRE O, ZAMO M, et al. Calibrated ensemble forecasts using quantile regression forests andensemble model output statistics[J]. Mon Wea Rev, 2016, 144(6): 2375-2393. doi: 10.1175/MWR-D-15-0260.1 [5] HAMILL T M, COLUCCI S J. Evaluation of Eta-RSM ensemble probabilisticprecipitation forecasts[J]. Mon Wea Rev, 1998, 126(3): 711-724. doi: 10.1175/1520-0493(1998)126<0711:EOEREP>2.0.CO;2 [6] STENSRUD D J, BAO J W, WARNER T T. Using initial conditionand model physicsperturbations in short-range ensemble simulations of mesoscale convective systems[J]. Mon Wea Rev, 2000, 128(7): 2077-2107. doi: 10.1175/1520-0493(2000)128<2077:UICAMP>2.0.CO;2 [7] VUKICEVIC T, JANKOV I, MCGINLEY J. Diagnosis and optimization of ensemble forecasts[J]. Mon Wea Rev, 2008, 136(3): 1054-1074. doi: 10.1175/2007MWR2153.1 [8] JANKOV I, SCHULTZ P J, ANDERSON C J, et al. Theimpact of different physical parameterizations and their interactionson cold season QPF in the American river basin[J]. J Hydrometeorol, 2007, 8(5): 1141-1151. doi: 10.1175/JHM630.1 [9] JANKOV I, GALLUS JR W A, SEGAL M, et al. Influenceof initial conditions on the WRF-ARW model QPF responseto physical parameterization changes[J]. Wea Forecasting, 2007, 22(3): 501-519. doi: 10.1175/WAF998.1 [10] HAMILL T M, WHITAKER J S. Global ensemble predictions of 2009's tropical cyclones initializedwith an ensemble Kalman filter[J]. Mon Wea Rev, 2011, 139(2): 668-688. doi: 10.1175/2010MWR3456.1 [11] CANGIALOSI J P, FRANKLIN J L. 2013 National Hurricane Center Forecast verification report[M]. NOAA/National Hurricane Center, 2014: 84. [12] LEE T C, WONG M S. The use of multiple-model ensemble techniques for tropical cyclone track forecast at the Hong Kong Observatory[C]. Preprints, Tech Conf on Data Processing and Forecasting Systems, Cairns, QLD, Australia, WMO, Topic 2. 2002. [13] YAMAGUCHI M, SAKAI R, KYODA M, et al. Typhoon ensemble prediction system developed at the Japan Meteorological Agency[J]. Mon Wea Rev, 2009, 1378): 2592-2604. http://cn.bing.com/academic/profile?id=1b99bc86957859a184d5c3e2bd19cbff&encoded=0&v=paper_preview&mkt=zh-cn [14] QI L, YU H, CHEN P. Selective ensemble-mean techniquefor tropical cyclone track forecast by using ensembleprediction systems[J]. Q J Roy Meteor Soc, 2014, 140(680): 805-813. doi: 10.1002/qj.2196 [15] DONG L, ZHANG F. OBEST: An observation-based ensemble subsetting technique for tropical cyclone track prediction[J]. Wea Forecasting, 2016, 31(1): 57-70. doi: 10.1175/WAF-D-15-0056.1 [16] HOUTEKAMER P L. Global and local skill forecasts[J]. Mon Wea Rev, 1993, 121(6): 1834-1846. doi: 10.1175/1520-0493(1993)121<1834:GALSF>2.0.CO;2 [17] WHITAKER J S, LOUGHE A F. The relationship between ensemble spread and ensemble mean skill[J]. Mon Wea Rev, 1998, 126(12): 3292-3302. doi: 10.1175/1520-0493(1998)126<3292:TRBESA>2.0.CO;2 [18] GRIMIT E P, MASS C F. Measuring the ensemblespread-error relationship with a probabilistic approach: Stochasticensemble results[J]. Mon Wea Rev, 2007, 135(1): 203-221. doi: 10.1175/MWR3262.1 [19] HOPSON T M. Assessing the ensemble spread-error relationship[J]. Mon Wea Rev, 2014, 142(3): 1125-1142. doi: 10.1175/MWR-D-12-00111.1 [20] 麻素红, 陈德辉.国家气象中心区域台风模式预报性能分析[J].热带气象学报, 2018, 34(4): 451-459. http://www.itmm.org.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20180402&flag=1 [21] 黄燕燕, 薛纪善, 陈子通, 等. 2011年"洛克"和"桑卡"双台风预报效果的初值试验研究[J].热带气象学报, 2017, 33(1): 30-42. http://www.itmm.org.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20170104&flag=1 [22] 张进, 麻素红, 陈德辉, 等. GRAPES_TYM改进及其在2013在西北太平洋和南海台风预报中的表现[J].热带气象学报, 2017, 33(1): 64-73. http://www.itmm.org.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20170107&flag=1 [23] 朱磊, 万齐林, 刘靓珂, 等.雷达径向风资料EnKF同化应用对Vicente(2012)登陆台风强度变化过程预报的影响试验研究[J].热带气象学报, 2017, 33(3): 345-356. http://www.itmm.org.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20170306&flag=1 [24] 王晓峰, 许晓林, 杨续超, 等.数值模式对强台风"菲特"登陆期间预报能力评述[J].大气科学学报, 2017, 40(5): 609-618. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX201705004.htm [25] 昌磊, 余锦华.西北太平洋TC移动速度异常及预报误差特征的分析[J].大气科学学报, 2017, 40(1): 71-80. http://d.old.wanfangdata.com.cn/Periodical/njqxxyxb201701007 [26] ZHU L, WAN Q, SHEN X, et al. Prediction and predictability of high-Impact western pacific landfalling tropical cyclone Vicente (2012) through convection-permitting ensemble assimilation of Doppler radar velocity[J]. Mon Wea Rev, 2016, 144(1): 21-43. doi: 10.1175/MWR-D-14-00403.1 [27] MENG Z, ZHANG F. Test of an ensemble Kalman filter for mesoscaleand regional-scale data assimilation. Part Ⅲ: Comparisonwith 3DVAR in a real-data case study[J]. Mon Wea Rev, 2008, 136(2): 522-540. doi: 10.1175/2007MWR2106.1 [28] MENG Z, ZHANG F. Tests of an Ensemble Kalman filter for mesoscale and regional-scale data assimilation. Part Ⅳ: comparison with 3DVAR in a month-long experiment[J]. Mon Wea Rev, 2008, 136(10): 3671-3682. doi: 10.1175/2008MWR2270.1 [29] BARKER D M, HUANG W, GUO Y R, et al. A three-dimensional variational data assimilation system for MM5: Implementation and initial results[J]. Mon Wea Rev, 2004, 132(4): 897-914. doi: 10.1175/1520-0493(2004)132<0897:ATVDAS>2.0.CO;2 [30] WHITAKER J S, HAMILL T M. Ensemble data assimilation without perturbed observations[J]. Mon Wea Rev, 2002, 130(7): 1913-1924. doi: 10.1175/1520-0493(2002)130<1913:EDAWPO>2.0.CO;2 -
点击查看大图
图(8) / 表(1)
计量
- 文章访问数: 445
- HTML全文浏览量: 7
- PDF下载量: 241
- 被引次数: 0
