28 6 2012 12 JOURNAL OF M ETEOROLOGY AND ENVIRONM ENT Vol. 28 No. 6 December 2012. J. 2012 28 6 50-57. LI Jiao REN Guo-yu REN Yu-yu et al. Prediction models of summer extreme high temperature in Liaoning province J. Journal of Meteorology and Environment 2012 28 6 50-57. 1,2,3 李娇 2 任国玉 2 任玉玉 1,2 沈志超 1,2 孙秀宝 1. 210044 2. 100081 3. 112000 利用 1957 2006 年辽宁地区夏季 23 个站极端最高气温资料和国家气候中心气候监测室的 74 项环流特征量资料, 应用 EOF 方法对高温极值样本进行分解, 研究辽宁极端高温的时空分布规律结果表明 : 第一特征向量表现为区域整体一 致的特征, 中心区位于辽西北 辽北, 第二 三特征向量空间分布表现为东西部反位相和南北反位相的特征普查前 3 个时间系 数与前期环流指数的相关关系, 认为前 3 个时间系数的显著影响因子是不同的采用 CSC 准则确定最优预测因子, 分别建立 各时间系数的回归统计模型, 并对高温极值历史拟合序列进行回报检验和预测检验回报结果表明, 各站历史拟合率均保持在 一定水平, 但拟合率在辽西地区较差各年历史拟合率极不均衡, 多数年份较为稳定, 但个别年份拟合率较低未来 3 a 试验 性预测效果逐年下降, 模型对未来 1 a 预测能力较好, 可以作为业务预测的参考 高温极值 ; EOF; 环流指数 ; 最优子集回归 ; 统计模型 ; 气候预测 ; 辽宁 P468. 0 + 21 A 1673-503X 2012 06-0050 - 08 15-17 18-22 1-5 1 6-9 1. 1 756 10-14 10 EHE 500 hpa 850 hpa 1957 12 4 3 500 hpa 23 14 1 1. 2 1957 2006 23 EOF CSC 3 2007 2009 2012-01 - 06 2012-02 - 13 2007BAC29B02 201206024 1986 E-mail lijiaostu@ 163. com E-mail guoyoo@ cma. gov. cn
6 51 lnn i + I j = 1 n j lnn j ] 4 4 I n ij i j S 1 = nr 2 3 R 3 2 EOF 1957 2006 23 EOF 63. 3% 8% 6% 3 77. 8% 1 1 23 1 3 EOF Fig. 1 The distributions of selected 23 weather stations Table 1 Variance contribution of first three EOF in Liaoning province components of extreme high temperature 1. 2. 1 23 1 14. 56615 0. 63331 0. 63331 2 1. 94054 0. 08437 0. 71768 3 1. 40868 0. 06125 0. 77893 m n X m n p 3 X m n = V m p T p n 1 2a 1 T V 1. 2. 2 CSC PC1 23 2b - 7 ~ 11 PC1 CSC CSC 0. 3 /10 a 20 60 80 90 20 90 k CSC k CSC k = S 1 + S 2 2 3a 4a 2 S 1 24 2 S 2 EOF S 2 = 2I = 2 I I n ij lnn ij + nlnn - I [ ni i = 1 j = 1 i = 1 PC2 3b
52 28-3. 708 ~ 2. 99 PC2 PC3 4b 0. 13 /10 a - 2. 542 ~ 1. 871 0. 326 /10 a Fig. 2 2 EOF a b The spatial distribution of the first EOF vector of summer extreme high temperature a and its time coefficient b in Liaoning province Fig. 3 3 EOF a b The spatial distribution of the second EOF vector of summer extreme high temperature a and its time coefficient b in Liaoning province Fig. 4 4 EOF a b The spatial distribution of the third EOF vector of summer extreme high temperature a and its time coefficient b in Liaoning province 3 1 2 3 74
6 53 0. 451 3. 1 0. 4 α = 0. 01 1 12 22 1 2 74 74 14 3 = 18 2 3 3108 α = 0. 01 0. 361 α = 0. 001 Table 2 2 3 The first three circulation indices influencing significantly time coefficient of summer extreme high temperature in Liaoning province α PC1 PC2 PC3 7 47 2 150 E 120 W - 0. 424 7 55 5 0 360-0. 433 7 52 2 150 E 120 W - 0. 438 8 51 1 60 150 E - 0. 412 10 25 110 W 60 E 0. 464 0. 01 10 24 20 W 60 E 0. 404 12 67 25 35 N 80 100 E 0. 459 12 68 30 40 N 75 105 E 0. 502 12 46 1 60 150 E - 0. 473 2 24 20 W 60 E 0. 462 9 67 25 35 N 80 100 E 0. 425 9 68 30 40 N 75 105 E 0. 453 0. 01 11 67 25 35 N 80 100 E 0. 415 2 41 55 W 25 W - 0. 402 3 34 5 E 360-0. 443 7 38 110 150 E - 0. 429 0. 01 1 28 175 115 W - 0. 405 1 39 175 115 W - 0. 461 PC1 5 3 2 10 2 12 500 hpa - 0. 473 α = 0. 001 2 30 40 N 75 105 E 12 PC1 26 PC2 25 2 500 hpa
54 28 RMSE R F 3 PC3 3 F 0. 01 F 3 3 3 3 RMSE R F Table 3 RMSE multiple correlation coefficients and F test for three optimum subset regression equations 3. 2 PC1 PC2 PC3 RMES 2. 73 1. 21 0. 84 27 R 0. 70 0. 50 0. 71 F 11. 53 25. 00 22. 53 10 3. 3. 1 5a 2 0. 82 0. 98 0. 903 1 0. 40 0. 68 0. 553 5b 2 0. 8 0. 9 43 a 36 a 4 a 1 0. 3 3. 3. 2 6 2007 2009 3 a 3. 3 2 3. 0 CSC 2007 3 14 2 2 y 1 = 20. 455-0. 2x 3-0. 206x 4-0. 097x 9 + 1. 077x 10 5 1 a y 2 = - 50. 269 + 0. 039x 1 + 0. 032x 3 6 y 3 = 12. 479-0. 064x 1-0. 134x 2-0. 187x 3 + EOF 0. 125x 4-0. 318x 5 7 PC1 5 x 3 x 4 x 9 x 10 1957 2009 23 4 t 53 t 5 Sig 0. 478 0. 995 0. 05 3 95%
6 55 Fig. 5 5 a b The fitting rate of past records of summer extreme high temperature for the 23 weather stations a and each year b in Liaoning province Fig. 6 6 2007 a 2008 b 2009 c 3 a d The absolute errors of prediction values of summer extreme high temperture in the 23 weather stations of Liaoning province in 2007 a 2008 b 2009 c and their average d
56 28 J. 4 2011 16 2 199-208. 1 EOF 11. 50 J. 2009 64 3 289-302. 12. J. 2009 28 3 653-662. 2 PC1 13 Guan Z Yamagata T. The unusual summer of 1994 in East Asia IOD teleconnections J. Geophysical Research PC2 PC3 Letters 2003 30 10 1541-1544. 14. J. 2004 32 3 182-3 186. 15. 10 J. 2010 27 3 4 621-627. 16. J. 2009 25 1 19-22. 4 3 a 17. 1 a J. 2005 20 1 36-39. 18. J. 2003 29 4 44-1. 1951 1990 19. J. 1998 22 2 217-227. 2. 1956 2008 2009. J. 2010 15 4 405-417. J. 1999 10 4 503-508. 3. 21. J. 2005 10 2 701-716. 4. 50 2001 21 2 221-229. J. 2003 58 1-10. 5. J. 2010 46 6 J. 2008 47 3 112-116. 6. 21 J. 2011 33 1 120-23. M. 127. 2007 106-107 244-245. 7. 21 J. 2008 30 7 1084 - J. 2008 36 11 4589-4590 1092. 4620. 8. 25. J. J. 2008 63 3 227-2008 17 2 156-160. 236. 26. D. 9. 2005. J. 2008 19 6 655-660. 10. 2006 49 3 662-671. 47. C / / 20. EOF J. 22. 54-58. 24. EOF 27. J.
6 57 Prediction models of summer extreme high temperature in Liaoning province LI Jiao 1 2 3 REN Guo-yu 2 REN Yu-yu 2 SHEN Zhi-chao 1 2 SUN Xiu-bao 1 2 1. College of Atmospheric Sciences Nanjing University of Information Science and Technology Nanjing 210044 China 2. The Laboratory for Climate Studies of China Meteorological Administration National Climate Center Beijing 100081 China 3. Tieling Meteorological Service Tieling 112000 China Abstract Based on the summer extreme high temperature data from the 23 weather stations in Liaoning province and 74 circulation indices information from the climate monitoring department of the National Climate Center NCC China Meteorological Administration CMA the temporal and spatial distribution features of summer extreme high temperature in Liaoning province were analyzed by a Empirical Orthogonal Function EOF decomposition method. The results show that the first EOF vector is characterized by a uniform anomaly over the whole area and the centers are in the northern and northwestern Liaoning province while the second and third EOF vectors are the reversed phase patterns in the east and west areas and in the south and north areas respectively. The correlation coefficient betw een first three time coefficient series and preceding circulation indices are calculated. It is found that the influencing factors are different for the three time coefficients. The optimum subset regressions are chosen as prediction equations using a CSC evaluation method and the fitting rate of past records in the 23 w eather stations and each year are tested. It shows that the fitting rate of past records in the 23 weather stations is generally stable except for in the western Liaoning province. In addition the fitting rate is unbalanced each year and it is stable in most years while it is low in few individual years. While the prediction effect is good for the first year in the future it declines yearly in the follow ing tw o years. The results can be used as references in climatic prediction. Key words Extreme high temperature Empirical orthogonal function EOF Circulation index Optimum subset regression. TIF Statistic models Climate prediction Liaoning province