两类天气预报评分问题研究及一种新评分方法
Dichotomous Weather Forecasts Score Research and a New Measure of Score
-
摘要: 探讨了预报评价的意义及应遵循的原则, 对常用的几种两类预报评分方法进行分析, 指出其应用的局限性, 得到一个判定所作预报水平是否高于随机预报、具有预报技巧的简易判别式; 提出评分权重的概念, 指出以往评分存在问题的根源是评分权重分配不当, 使评分结果的真实性受到影响, 评分无可比性, 进而提出一种考虑了评分权重的新评分方法。新评分方法满足预报评价的原则, 侧重于对两类事件中事件概率较小一方预报效果的评估, 评分结果不受事件概率影响, 具有可比性。对比分析表明:新方法比其他方法优越, 能更准确地反映预报水平, 使不同季节、不同地域的预报评分可进行比较, 是一个通用的评分方法。Abstract: The significance of forecast estim ate and the principles are discussed. It is assumed that the scores are objective; and moreover they can objectively reflect the forecast level.The scores should be comparable, guiding forecast in the right direction. Several usual methods of dichotomous forecasts score are analyzed, revealing that accuracy and critical success index (CSI) in different areas are incomparable due to the influence of event possibility. It shows that the true skill statistic (TSS) is approach to the probability of detection (POD) when forecasting rare events.When events do not appear but false alarms are made, TSS can't be calculated. Heidke skill score and Girbet skill score make up for the above weaknesses. The three skill scores are all obtained by comparing forecast with random ones, hence there are (n11n22-n12n21) in the three formulas. It can be used as the discriminant for forecast skill. When (n11n22-n12n21)> 0, it indicates that the forecast level is better than that of random forecast, otherwise, it will be worse than it. On the basis of the relationship between event probability and the difficulty of forecast, a new method of score weight is considered and proposed. The essay points out the exiting problems are resulted from improper score weight, leading the score result unreliable and not comparable. The new method of score is based on CSI, and combines CSI of the two event forecasts. The focus of score is laid upon estimating the one with smaller probability in the two events.The principles of forecast score are fulfilled. By comparative analysis, the new method is proved to be superior to other methods, especially on estimating rare events. They can reflect the forecast level and changes more accurately.The advantages are as follows :With the increase of samples, the new score tends to be more stable than other scores in the rare events fo recast, thus leading to a rapid judgement for forecast level. When forecast level is improved, the new score will be able to reflect it correctly and distinctly. The new score is objective, just and real, and is compatible for different seasons and regions. So it is a uniform standard in forecast score.
-
Key words:
- forecast estimate;
- skill score;
- score weight;
- probability;
- comparable
-
表 1 两类预报列联表
Table 1 Contingency table for dichotomous (yes/no) forecasts
表 2 两类预报评价方法
Table 2 Verification methods for dichotomous (yes/no) forecasts
表 3 Finley预报列联表
Table 3 Contingency table for Finley's forecasts
表 4 预报实例及评分研究
Table 4 Examples of forecast and score research
表 5 预报水平变化时各种评对比分析
Table 5 Comparative analysis of scores at changed forecast level
表 6 各类评分特性分析
Table 6 Characteristic analysis of scores
-
[1] Hughes P.The great leap forward.Weatherwise, 1994, 47: 22-27. [2] Whitnah D R.A History of the United States Weather Bureau. Illinois:University of Illinois Press, 1961: 1-267. [3] Burton J.Robert Fitzroy and the early history of the Meteorological Office.BrJ Hist Sci, 1986, 19: 147-176. doi: 10.1017/S0007087400022949 [4] Finley J P. Tornado predictions. Amer Meteor J, 1884, 1: 85-88. http://science.sciencemag.org/content/ns-4/80/126 [5] Murphy A H.The Finley affair : A signal event in the history of forecast verification. Wea Forecasting, 1996, 11: 3-20. doi: 10.1175/1520-0434(1996)011<0003:TFAASE>2.0.CO;2 [6] Gilbert G K.Finley's tornado predictions.Amer Meteor J, 1884, 1: 166-172. [7] Palmer W C, Allen R A. Note on the Accuracy of Forecasts Concerning the Rain Problem.US Weather Bureau, 1949:1-4. [8] Donaldson R J, Dyer R M, Krauss M J. An Objective Evaluator of Techniques for Predicting Severe Weather Events. 9th Conf Severe Local Storms, Norman, Oklahoma, Amer Meteor Soc, 1975:321-326. http://citeseerx.ist.psu.edu/showciting?cid=3237109 [9] Schaefer J T. The critical success index as an indicator of warning skill. Wea Forecasting, 1990, 5:570-575. doi: 10.1175/1520-0434(1990)005<0570:TCSIAA>2.0.CO;2 [10] Black T L.The new NMC mesoscale eta model:Description and forecast examples. Wea Forecasting, 1994, 9: 265-278. doi: 10.1175/1520-0434(1994)009<0265:TNNMEM>2.0.CO;2 [11] Peirce C S.The numerical measure of the success of prediction. Science, 1884, 4:453-454. http://citeseerx.ist.psu.edu/showciting?cid=1558742 [12] Hanssen A W, Kuipers W J A. On the relationship between the frequency of rain and various meteorological parameters. Mededeelingen en Verhandelingen, 1965, 81:2-15. http://www.worldcat.org/title/on-the-relationship-between-the-frequency-of-rain-and-various-meteorological-parameters-with-reference-to-the-problem-of-objective-forecasting/oclc/37281102 [13] Murphy A H, Daan H.Forecast Evaluation∥Murphy A H, Katz R W.Probability, Statistics, and Decision Making in the Atmospheric Sciences. Westview : Westview Press, 1985: 379-437. [14] Flueck J A. A Study of Some Measures of Forecast Verification. 10th Conf Probability and Statistics in Atmospheric Science, Edmonton, Alberta, Amer Meteor Soc, 1987: 69-73. [15] Doolittle M H.The verification of predictions.Amer Meteor J, 1885, 2:327-329. [16] Doolittle M H.Association ratios. Bull Philos Soc Washington, 1888, 10:83-87;94-96. [17] Heidke P.Berechnung des Erfolges und der Gǜte der Windstärkevorhersagen im Sturmwarnungdienst (Calculation of the success and goodness of strong wind forecasts in the storm warning service). Geogr Ann Stockholm, 1926, 8: 301-349. [18] Murphy A H, Winkler R L.A general framework for forecast verification.Mon Wea Rev, 1987, 115:1330-1338. doi: 10.1175/1520-0493(1987)115<1330:AGFFFV>2.0.CO;2 [19] Murphy A H. What is a good forecast? An essay on the nature of goodness in weather forecasting.Wea Forecasting, 1993, 8:281-293. doi: 10.1175/1520-0434(1993)008<0281:WIAGFA>2.0.CO;2 [20] 张军, 葛军, 田俊杰, 等.概率天气预报及其应用.北京:气象出版社, 1998: 127-137. [21] 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 [22] 新田尚, 立平良三, 市桥英辅.宁松, 译.最新天气预报技术.北京:气象出版社, 1997: 145-146. [23] Brownlee K A.Statistical Theory and Methodology in Science and Engineering (2nd ed). John Wiley and Sons, 1965. [24] Doswell C A Ⅲ, Davies-Jones R P, Keller D L.On summary measures of skill in rare event forecasting based on contingency tables. Wea Forecasting, 1990, 5:576-585. doi: 10.1175/1520-0434(1990)005<0576:OSMOSI>2.0.CO;2 [25] 赵有娟, 赵康林, 王小虎.涿州地区冬半年辐射雾的预报.军事气象水文, 2005, (6) : 32-33. [26] 宋伟, 雷显飞, 宋作义.东北平原夏季雷雨预报.航空气象, 2005, (2): 31-34.