Duan Mingkeng, Wang Panxing. A new weighted average method on ensmble mean forecasting. J Appl Meteor Sci, 2006, 17(4): 488-493.
Citation: Duan Mingkeng, Wang Panxing. A new weighted average method on ensmble mean forecasting. J Appl Meteor Sci, 2006, 17(4): 488-493.

A New Weighted Average Method on Ensmble Mean Forecasting

  • Received Date: 2005-09-05
  • Rev Recd Date: 2006-06-19
  • Publish Date: 2006-08-31
  • The ensemble mean forecasting is one of the major forecasting products in an ensemble prediction system. Compared with the results from a single deterministic forecast, the ensemble average forecast has higher forecasting skills. For the forecasts starting at the same initial time, the ordinary ensemble mean forecasting is the arithmetic mean of all ensemble members. In the ensemble, each of ensemble members has the same weight. A new ensemble average method considering the different weight of ensemble members (known as "the weighted ensemble average method") is presented. In this method, the ensemble average is calculated from all the ensemble members, which has the same initial time and lead time, with the different weight. Through this method, it is expected to get the better forecasting performance than the ordinary ensemble mean forecast (known as "the common ensemble average method").Some research conclusions indicate that the results of the ensemble forecasting are more inclined to the value of major ensemble members; at the same lead time, the higher success rate of ensemble forecasting always correspond to the more consistent members of the forecasting results (the prerequisite of this result is that the ensemble prediction system has better reliability and the initial ensemble perturbations have enough spread so that the forecasting results can more comprehensively cover the possibilities of forecasting errors).The starting point of the weighted ensemble average method is to adjust the weight of ensemble members in order that the members with consistent forecast results have larger weights. By this process, the consistence among the ensemble members is emphasized. Here, the weighted ensemble average method uses the climatologically likely intervals to group the ensemble members according to the proximity extent each other; then according to the size of the climatologically likely interval and the ensemble members in it, the corresponding weight is adjusted; eventually, based on the adjusted weight, the calibrated ensemble mean forecasts are gotten.Based on the NCEP EPS forecasting dataset, the verification methods of ACC and RMSE are used to prove the performance of the weighted ensemble average method. The test results at the various regions show that the method can upgrade the ensemble mean performance in a way. Even in the worst situation, the effects of the weighted ensemble average method do not appear below the ordinary ensemble average method. Compared to the resolution improvement of the numerical weather prediction or development of enormous ensemble prediction system further, the enhancement of the weighted ensemble average method is significant.In addition to the medium-range ensemble average forecasting, this method can also be used in the ensemble mean of the continuous variables during the other types of forecasts, predictions to improve the forecast performance.
  • [1]
    Toth Z, Kalney E. Ensemble forecasting at NMC: the generation of perturbations. Bull Amer Meteor Soc, 1993, 74 (12): 2317-2330. doi:  10.1175/1520-0477(1993)074<2317:EFANTG>2.0.CO;2
    [2]
    Tracton M S, Kalnay E. Operational ensemble prediction at the National Meteorological Center: practical aspects. Wea Forecasting, 1993, 8 (3): 379-398. doi:  10.1175/1520-0434(1993)008<0379:OEPATN>2.0.CO;2
    [3]
    Molteni F, Buizza R, Palmer T N, et al. The ECMWF ensemble prediction system: methodology and validation. Q J R Meteorol Soc, 1996, 122 (529): 73-119. doi:  10.1002/(ISSN)1477-870X
    [4]
    Houterkamer P L, Lefaiver L, Derome J, et al. A system simulation approach to ensemble prediction. Mon Wea Rev, 1996, 124 (6): 1225-1242. doi:  10.1175/1520-0493(1996)124<1225:ASSATE>2.0.CO;2
    [5]
    李泽椿, 陈德辉.国家气象中心集合数值预报业务系统的发展及应用.应用气象学报, 2002, 13 (1): 1-15. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20020101&flag=1
    [6]
    杨学胜, 陈德辉, 冷亭波, 等.时间滞后与奇异向量初值生成方法的比较试验.应用气象学报, 2002, 13 (1): 62-66. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20020107&flag=1
    [7]
    皇甫雪官.国家气象中心集合数值预报检验评价.应用气象学报, 2002, 13 (1): 29-36. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20020103&flag=1
    [8]
    陈静, 陈德辉, 颜宏.集合数值预报发展与研究进展.应用气象学报, 2002, 13 (4): 497-507. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20020465&flag=1
    [9]
    Toth Z, Zhu Y J, Marchok T. The use of ensemble to identify fo recasts with small and large uncertainty. Wea Forecasting, 2001, 16 (8): 463-477. doi:  10.1175/1520-0434%282001%29016%3C0463%3ATUOETI%3E2.0.CO%3B2
    [10]
    Kalnay E, Kanamitsu M, Kistler R, et al. The NCEP/NCAR 40-year reanalysis project. Bull Amer Meteor Soc, 1996, 77 (1): 437-471. doi:  10.1175/1520-0477%281996%29077%3C0437%3ATNYRP%3E2.0.CO%3B2
    [11]
    刘还珠, 张绍晴.中期数值预报的统计检验分析.气象, 1992, 18 (9): 50-54. http://www.cnki.com.cn/Article/CJFDTOTAL-QXXX199209012.htm
  • 加载中
  • -->

Catalog

    Figures(5)

    Article views (3427) PDF downloads(1348) Cited by()
    • Received : 2005-09-05
    • Accepted : 2006-06-19
    • Published : 2006-08-31

    /

    DownLoad:  Full-Size Img  PowerPoint