一种基于可预报性的暴雨预报评分新方法Ⅱ:暴雨检验评分模型及评估试验
|
摘要
针对当前暴雨预报检验采用二分类事件检验方法存在较严重的"空报""漏报"双重惩罚,没有考虑暴雨时空分布不均和预报评分可比性不够等问题,在分析预报员对暴雨预报评分期望值基础上,设计了一种基于可预报性的暴雨预报检验评分新方法和计算模型,分析了理想评分,并对2015—2016年4—10月中国中央气象台5 km×5 km定量降水格点预报和降水落区等级暴雨预报进行评分试验,获得了以下结果和结论:(1)预报员对暴雨预报评分期望值呈现梯级下降特征,与传统的TS评分存在显著差异;(2)设计了一种基于可预报性的暴雨预报检验新方法,通过引入e指数函数构建暴雨预报评分基函数,进而构建暴雨评分模型,该模型可以较好地拟合预报员对暴雨预报评分的期望值,同时改善了评分在不同量级阈值处的断崖式突变情况;(3)提出了预报与观测的邻域匹配方法,即一个预报点与所定义邻域中的一组观测相匹配,并利用距离加权最大值法确定暴雨评分值权重系数,预报与观测距离越近,距离权重系数越大,评分值权重越大,提高了评分的合理性,避免了距离较远的匹配站点得高分不利于鼓励预报员提高预报精度的问题;(4)对中国中央气象台逐日5 km×5 km水平分辨率的定量降水格点预报产品和中央气象台定量降水落区等级预报产品进行了评分试验,暴雨预报准确率全国平均值大于60分。基于可预报性的暴雨预报检验新评分与传统暴雨预报TS评分逐日演变特征相似,但可以较好地解析TS为0的预报评分,解析后的新评分与预报员和公众的心理预期更为接近。
In order to solve the problem that the current verification score(Threat Score, TS) for heavy rainfall forecasting severely suffers from the double punishment due to the relatively high level of missing rate and false alarm rate as well as the ignorance of the uneven temporal and spatial distribution of heavy rainfall in China, the present study designs a new verification method and computational model for heavy rainfall forecasts based on the predictability of heavy rainfall and analysis of the expectation scores of forecasters. A new score model is designed and tested using the 5 km×5 km gridded quantitative precipitation forecast and precipitation location forecast issued by the China Central Meteorological Observatory from April to October during 2015-2016. The results and conclusions are as follow.(1) Forecaster′s expectation scores for heavy rainfall forecast show a staircase-like descending characteristic, which is different from the traditional TS score.(2) A new forecast verification method based on the predictability of heavy rainfall is designed, which constructs the heavy rainfall forecast score basic function by first introducing an exponential function and then constructing the heavy rainfall grading model. The model can well fit the expectation of the forecaster's score for the heavy rainfall, and improve the score by reducing its cliff-like mutation at different levels of the threshold.(3) A neighborhood method of matching forecasts and observations is proposed, that is, a forecasting point is matched with a set of observations in a defined neighborhood, and a distance-weighted maximum score method to define the weighting coefficient of the rainstorm score is used. Thereby, the closer the distance between the forecast point and the observation point is, the greater the distance-weighted coefficient is and the higher the contribution of this point to the score value is.This method improves the rationality of the score and avoids the problem that a high score from a distant matching station is not encouraging for forecasters to improve the accuracy of forecasts.(4)Quantitative gridded precipitation forecasts at the5 km×5 km horizontal resolution and quantitative precipitation location forecasts of the China Central Meteorological Observatory are verified using this new method.The accuracy of heavy rainfall forecasts is over 60 paints on average over entire China,and the daily evolution characteristic of the new score is similar to that of the traditional TS score.However,for the forecasts whose TS scores are 0,this new score is more consistent with the psychological expectations of both the forecasters and the public.
In order to solve the problem that the current verification score(Threat Score, TS) for heavy rainfall forecasting severely suffers from the double punishment due to the relatively high level of missing rate and false alarm rate as well as the ignorance of the uneven temporal and spatial distribution of heavy rainfall in China, the present study designs a new verification method and computational model for heavy rainfall forecasts based on the predictability of heavy rainfall and analysis of the expectation scores of forecasters. A new score model is designed and tested using the 5 km×5 km gridded quantitative precipitation forecast and precipitation location forecast issued by the China Central Meteorological Observatory from April to October during 2015-2016. The results and conclusions are as follow.(1) Forecaster′s expectation scores for heavy rainfall forecast show a staircase-like descending characteristic, which is different from the traditional TS score.(2) A new forecast verification method based on the predictability of heavy rainfall is designed, which constructs the heavy rainfall forecast score basic function by first introducing an exponential function and then constructing the heavy rainfall grading model. The model can well fit the expectation of the forecaster's score for the heavy rainfall, and improve the score by reducing its cliff-like mutation at different levels of the threshold.(3) A neighborhood method of matching forecasts and observations is proposed, that is, a forecasting point is matched with a set of observations in a defined neighborhood, and a distance-weighted maximum score method to define the weighting coefficient of the rainstorm score is used. Thereby, the closer the distance between the forecast point and the observation point is, the greater the distance-weighted coefficient is and the higher the contribution of this point to the score value is.This method improves the rationality of the score and avoids the problem that a high score from a distant matching station is not encouraging for forecasters to improve the accuracy of forecasts.(4)Quantitative gridded precipitation forecasts at the5 km×5 km horizontal resolution and quantitative precipitation location forecasts of the China Central Meteorological Observatory are verified using this new method.The accuracy of heavy rainfall forecasts is over 60 paints on average over entire China,and the daily evolution characteristic of the new score is similar to that of the traditional TS score.However,for the forecasts whose TS scores are 0,this new score is more consistent with the psychological expectations of both the forecasters and the public.
引文
陈法敬, 陈静. 2015. “SEEPS”降水预报检验评分方法在我国降水预报中的应用试验. 气象科技进展, 5(5): 6-13. Chen F J, Chen J. 2015. The application experiment of a new score for precipitation verification based on the SEEPS principle. Adv Meteor Sci Technol, 5(5): 6-13 (in Chinese)
陈静,刘凑华,陈法敬等. 2019. 一种基于可预报性的暴雨预报评分新方法Ⅰ:中国暴雨可预报性综合指数. 气象学报,77(1):15-27. Chen J, Liu C H, Chen F J, et al. 2019. A new verification method for heavy rainfall forecast based on predictability I:Synthetic predictability index of heavy rainfall in China. Acta Meteor Sinica, 77(1):15-27 (in Chinese)
矫梅燕. 2010. 现代数值预报业务. 北京: 气象出版社, 162-176. Jiao M Y. 2010. Modern Numerical Weather Prediction. Beijing: China Meteorological Press, 162-176 (in Chinese)
Jolliffe I T, Stephenson D B. 2016. 预报检验——大气科学从业者指南. 李应林, 译. 北京: 气象出版社, 1-10. Jolliffe I T, Stephenson D B. Forecast Verification: A Practitioner's Guide in Atmospheric Science. Li Y L, trans. Beijing: China Meteorological Press, 1-10 (in Chinese)
唐文苑, 周庆亮, 刘鑫华等. 2017. 国家级强对流天气分类预报检验分析. 气象, 43(1): 67-76. Tang W Y, Zhou Q L, Liu X H, et al. 2017. Analysis on verification of national severe convective weather categorical forecasts. Meteor Mon, 43(1): 67-76 (in Chinese)
陶诗言. 1980. 中国之暴雨. 北京: 科学出版社, 1-10. Tao S Y. 1980. Heavy Rainstorm in China. Beijing: Science Press, 1-10 (in Chinese)
王雨, 公颖, 陈法敬等. 2013. 区域业务模式6 h降水预报检验方案比较. 应用气象学报, 24(2): 171-178. Wang Y, Gong Y, Chen F J, et al. 2013. Comparison of two verification methods for 6 h precipitation forecasts of regional models. J Appl Meteor Sci, 24(2): 171-178 (in Chinese)
中国科学院大气物理研究所. 1998. 东亚季风和中国暴雨——庆贺陶诗言院士八十华诞. 北京: 气象出版社. Institute of Atmospheric Physics, Chinese Academy of Science. 1998. East Asian Monsoon and Chinese Rainstorm: Celebrating the Eighty Birthday of Tao Shiyan. Beijing: China Meteorological Press (in Chinese)
Casati B, Wilson L J, Stephenson D B, et al. 2008. Forecast verification: Current status and future directions. Meteor Appl, 15(1): 3-18
Cherubini T, Ghelli A, Lalaurette F. 2002. Verification of precipitation forecasts over the Alpine region using a high-density observing network. Wea Forecasting, 17(2): 238-249
Stephenson D B, Casati B, Ferro C A T, et al. 2008. The extreme dependency score: A non-vanishing measure for forecasts of rare events. Meteor Appl, 15(1): 41-50
WMO. 2009. Recommendations for the verification and intercomparison of QPFs and PQPFs from operational NWP Models, revision 2 (WMO TD No. 1485). https://www.wmo.int/pages/prog/arep/wwrp/new/documents/WWRP2009_1.pdf, 2008.10
WMO. 2015. Standardized Verification of deterministic NWP Products. http://www.wmo.int/pages/prog/www/DPS/Manual/Table-of-content_Manual-gdpfs.html, 2015.05.12
陈静,刘凑华,陈法敬等. 2019. 一种基于可预报性的暴雨预报评分新方法Ⅰ:中国暴雨可预报性综合指数. 气象学报,77(1):15-27. Chen J, Liu C H, Chen F J, et al. 2019. A new verification method for heavy rainfall forecast based on predictability I:Synthetic predictability index of heavy rainfall in China. Acta Meteor Sinica, 77(1):15-27 (in Chinese)
矫梅燕. 2010. 现代数值预报业务. 北京: 气象出版社, 162-176. Jiao M Y. 2010. Modern Numerical Weather Prediction. Beijing: China Meteorological Press, 162-176 (in Chinese)
Jolliffe I T, Stephenson D B. 2016. 预报检验——大气科学从业者指南. 李应林, 译. 北京: 气象出版社, 1-10. Jolliffe I T, Stephenson D B. Forecast Verification: A Practitioner's Guide in Atmospheric Science. Li Y L, trans. Beijing: China Meteorological Press, 1-10 (in Chinese)
唐文苑, 周庆亮, 刘鑫华等. 2017. 国家级强对流天气分类预报检验分析. 气象, 43(1): 67-76. Tang W Y, Zhou Q L, Liu X H, et al. 2017. Analysis on verification of national severe convective weather categorical forecasts. Meteor Mon, 43(1): 67-76 (in Chinese)
陶诗言. 1980. 中国之暴雨. 北京: 科学出版社, 1-10. Tao S Y. 1980. Heavy Rainstorm in China. Beijing: Science Press, 1-10 (in Chinese)
王雨, 公颖, 陈法敬等. 2013. 区域业务模式6 h降水预报检验方案比较. 应用气象学报, 24(2): 171-178. Wang Y, Gong Y, Chen F J, et al. 2013. Comparison of two verification methods for 6 h precipitation forecasts of regional models. J Appl Meteor Sci, 24(2): 171-178 (in Chinese)
中国科学院大气物理研究所. 1998. 东亚季风和中国暴雨——庆贺陶诗言院士八十华诞. 北京: 气象出版社. Institute of Atmospheric Physics, Chinese Academy of Science. 1998. East Asian Monsoon and Chinese Rainstorm: Celebrating the Eighty Birthday of Tao Shiyan. Beijing: China Meteorological Press (in Chinese)
Casati B, Wilson L J, Stephenson D B, et al. 2008. Forecast verification: Current status and future directions. Meteor Appl, 15(1): 3-18
Cherubini T, Ghelli A, Lalaurette F. 2002. Verification of precipitation forecasts over the Alpine region using a high-density observing network. Wea Forecasting, 17(2): 238-249
Stephenson D B, Casati B, Ferro C A T, et al. 2008. The extreme dependency score: A non-vanishing measure for forecasts of rare events. Meteor Appl, 15(1): 41-50
WMO. 2009. Recommendations for the verification and intercomparison of QPFs and PQPFs from operational NWP Models, revision 2 (WMO TD No. 1485). https://www.wmo.int/pages/prog/arep/wwrp/new/documents/WWRP2009_1.pdf, 2008.10
WMO. 2015. Standardized Verification of deterministic NWP Products. http://www.wmo.int/pages/prog/www/DPS/Manual/Table-of-content_Manual-gdpfs.html, 2015.05.12