A Verification Approach for the Assessment of Extended-range Process Event Prediction
-
摘要: 基于延伸期过程性天气气候事件预报评估的特点,结合实际科研和业务工作的需求,提出了一种适用于延伸期过程预报的检验方法 (简称PPS方法)。该方法参考了常用的预报评分方法准则,借鉴了命中率、假警报率、欧式距离和动态时间弯曲距离等评估检验方法。利用命中率、假警报率和该方法对实际预报中可能出现的有漏报没有空报和既有漏报也有空报这两类情况的多个实例进行对比分析,表明该方法既能考虑大气随着时间的延长预报效果急剧降低的特性,也考虑了相似时间序列度量不精确匹配和形变的问题。利用该方法对1999—2010年冬季冷空气过程业务预报进行检验,结果表明:该方法能清晰表征延伸期预报时段内冷空气过程预报的准确程度,真实反映了目前延伸期预报准确率较低的现状,有较好的适用性。同时,该方法也适用于其他延伸期过程事件预报的评分,具有较好的应用前景。Abstract: Based on the features of forecast and assessment for extended-range weather and climate events, a verification approach named PPS (process-event prediction score) for process event forecast is proposed, which is combined with the actual requirements of extended-range forecast operation. This approach considers not only the criteria of event forecast scores including hit rate, false alarm rate commonly used in weather forecast operation, but also the advantages of other approaches such as Euclidean distance and dynamic time warping distance.As the forecast period is relatively long, it is very difficult to forecast a process event completely and accurately. Therefore, it is terrifically valuable for adjacent hit, denoting the forecast with one to two days lead or lag, in operational application. Based on the above-mentioned features, the periods of forecast and observation for process events are extended, respectively, and the virtual events are transformed into similar imaginary events. In terms of the accordance extent between forecast and observation, classified score table is constructed. Moreover, weight is used to show the influence of false alarm on forecast score.The features of PPS approach are assessed by couples of cases including "no false alarm but missing" and "missing and false alarm", and the relations of PPS to hit rate and false alarm rate are analyzed. Under the condition of "no false alarm but missing", scores of PPS and hit rate increase with the correct forecast number of days. The PPS score is generally higher than hit rate score, which indicates the increasing score effect from the expansion for process events of observation and forecast. In the case of missing and false alarm, PPS scores are higher than the hit rate score when false alarm rate is low. However, PPS scores will be lower than the hit rate score when false alarm rate significantly increases, which shows the influence of false alarm to PPS score. Combined with the features of process event forecast and the possible influence of false alarm on forecast skill, PPS score objectively reflects the actual skill of forecast. Compared with hit rate and false alarm rate, it is more efficient to represent the process event information involved in a forecast. Therefore, it is more applicable for assessing the skills of process event forecast.By this approach, skills of operational cold air process forecast are assessed during winters from 1999 to 2010. The results show that the PPS score reflects the accuracy of cold air process forecast well. Moreover, the verification actually indicates relatively low accuracy of extended-range forecast today. Above all, this approach can be used to assess extended-range process forecast and shows good prospect for operational application.
-
Key words:
- extended-range forecast;
- verification;
- PPS approach
-
表 1 双态分类联列
Table 1 Classification table for double state
预报 观测 出现 不出现 出现 a b 不出现 c d 表 2 评分分级表
Table 2 Classification table for prediction score
预测 实况 0 0.5 1 0 不参与评分 0 0 0.5 不参与评分 100 80 1 不参与评分 80 100 -
[1] 李崇银, 杨辉, 顾薇.中国南方雨雪冰冻异常天气原因的分析.气候与环境研究, 2008, 13(2):113-122. http://www.cnki.com.cn/Article/CJFDTOTAL-QHYH200802000.htm [2] 张勇.南方低温雨雪冰冻灾害历史罕见.气象, 2008, 34(4):132-135. doi: 10.7519/j.issn.1000-0526.2008.04.019 [3] 宋洁, 杨辉, 李崇银.2009/2010年冬季云南严重干旱原因的进一步分析.大气科学, 2011, 35(6):1009-1019. http://www.cnki.com.cn/Article/CJFDTOTAL-DQXK201106003.htm [4] 琚建华, 吕俊梅, 谢国清, 等.MJO和AO持续异常对云南干旱的影响研究.干旱气象, 2011, 29(4):401-406. http://www.cnki.com.cn/Article/CJFDTOTAL-GSQX201104000.htm [5] Saha S, Nadiga S, Thiaw C, et al.The NCEP Climate Forecast System.J Climate, 2006, 19(15):3483-3517. doi: 10.1175/JCLI3812.1 [6] Saha S, Moorthi S, Pan H-L, et al.The NCEP Climate Forecast System Reanalysis.Bull Amer Meteor Soc, 2010, 91(8):1015-1057. doi: 10.1175/2010BAMS3001.1 [7] Japan Meteorological Agency.Outline of Operational Numerical Weather Prediction at the Japan Meteorological Agency.Appendix to WMO Numerical Weather Prediction Progress Report.2007. [8] 陈丽娟, 李维京.月动力延伸预报产品的评估和解释应用.应用气象学报, 1999, 10(4):486-490. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=199904101&flag=1 [9] 孙国武, 信飞, 陈伯民.低频天气图预报方法.高原气象, 2008, 27(增刊):64-68. http://www.cnki.com.cn/Article/CJFDTOTAL-GSQX201303018.htm [10] 孙国武, 信飞, 孔春燕, 等.大气低频振荡与延伸期预报.高原气象, 2010, 29(5):1142-1147. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKZ201201008.htm [11] Flueck J A.A study of some measures of forecast verification.10th Conf Probability and Statistics in Atmospheric Sciences, Edmonton, AB, Canada, Amer MeteorSoc, 1987: 69-73. [12] 陈辉, 黄卓, 田华, 等.高温中暑气象等级评定方法.应用气象学报, 2009, 20(4):451-457. doi: 10.11898/1001-7313.20090409 [13] 周兵, 郭幼君, 何金海.降水概率预报评分方法分析.南京气象学院学报, 1999, 22(3):367-373. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX199903011.htm [14] 丁金才.天气预报评分方法评述.南京气象学院学报, 1995, 18(1):143-150. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX501.021.htm [15] 张强, 熊安元, 张金艳.晴雨 (雪) 和气温预报评分方法的初步研究.应用气象学报, 2009, 20(6):692-698. doi: 10.11898/1001-7313.20090606 [16] 罗阳, 赵伟, 翟景秋.两类天气预报评分问题研究及一种新评分方法.应用气象学报, 2009, 20(2):129-136. doi: 10.11898/1001-7313.20090201 [17] 王晨稀.短期集合降水概率预报试验.应用气象学报, 2005, 16(1):78-88. doi: 10.11898/1001-7313.20050110 [18] 刘懿, 鲍德沛, 杨泽红, 等.新型时间序列相似性度量方法研究.计算机应用研究, 2007, 24(5):112-114. http://cdmd.cnki.com.cn/Article/CDMD-10141-1013198523.htm [19] 吴学雁, 黄道平.基于事件的时间序列相似性度量方法.计算机应用, 2010, 30(7):1944-1946. http://www.cnki.com.cn/Article/CJFDTOTAL-JSJY201007071.htm