THE LINEAR AND NONLINEAR STATISTICAL STUDY OF WESTERN PACIFIC PHYSICAL FIELD WITH TROPICAL CYCLONES IN SOUTH CHINA
-
摘要: 选取西太平洋海平面温度场、海平面气压场、500 hPa位势高度场作为物理因子场，并普查计算1949—2013年影响我国华南地区的热带气旋的年均频数、年均最低气压、年均最大风速作为相关物理量场。对因子场分别应用主成分分析 (PCA) 与核主成分分析 (KPCA) 算法进行主成分提取，在此基础上，对因子场的前六主成分进行功率谱估计与凝聚谱分析。最后，利用典型相关分析 (CCA) 与核典型相关分析 (KCCA) 算法对因子场与物理量场进行典型相关分析。结果表明，基于非线性的KPCA算法提取出西太平洋物理因子场前六成分的解释方差贡献率，均高于PCA算法；海温场、海压场、高度场的第一成分各自存在大概18年的周期性振荡变化，同时，在周期为2~3年的范围内，这三者的振荡频率的互相关性最强；而因子场与物理量场的非线性典型相关系数，明显高于线性典型相关系数。Abstract: Western Pacific sea surface temperatures (SST), sea level pressure (SLP), 500 hPa geopotential height (HGT) were selected into a physical factors field, and the average frequency of each TC in the south of China, values of average annual lowest air pressure, and annual maximum wind speed were used as a physical field that were surveyed and calculated for the time from 1949 to 2013. Principal component analysis (PCA) and kernel principal component analysis (KPCA) algorithms were used to extract principal component factor fields. On this basis, the first six games of the principal components were taken for power spectrum estimation and cohesion spectrum analysis. Finally, canonical correlation analysis (CCA) and kernel canonical correlation analysis (KCCA) algorithms were used for canonical correlation analysis on the factor field and physical field. The results show that based on KPCA the explained variance contribution rates of the first six components are higher than that with PCA. The first components of SST, SLP and HGT fields have periodic fluctuations of about 18 years. Meanwhile, for the fluctuation of 2 to 3 years, the cross correlation of oscillation frequency is the strongest. The nonlinear canonical correlation coefficient is higher than the linear coefficient.
-
Key words:
- tropical cyclone /
- principal component /
- power spectrum estimation /
- period /
- canonical correlation
-
表 1 SST、SLP、HGT的前六个主成分解释方差贡献率 (%)
主成分数 1 2 3 4 5 6 SST KPCA 73.55 12.44 8.08 3.44 1.34 1.16 PCA 70.97 12.00 7.84 3.42 1.25 1.08 SLP KPCA 86.32 6.61 4.24 1.07 0.97 0.79 PCA 85.05 5.80 4.26 0.98 0.86 0.64 HGT KPCA 83.14 8.64 3.87 2.33 1.33 0.68 PCA 82.93 8.48 3.81 2.29 1.30 0.56 
下载: 导出CSV
表 2 SST、SLP、HGT的功率谱估计
L
T0
∞1
182
93
64
4.55
3.66
37
2.68
2.39
2Pk 0.203 8 0.274 0 0.110 4 0.070 7 0.046 3 0.039 4 0.057 4 0.071 6 0.081 9 0.044 1 SST P 0.203 7 0.274 0 0.110 4 0.070 7 0.046 3 0.039 4 0.057 4 0.071 6 0.082 0 0.044 1 ${\tilde P}$ 0.372 1 0.333 9 0.257 7 0.190 9 0.144 9 0.115 3 0.096 7 0.085 5 0.079 4 0.077 5 Pk 0.175 6 0.230 3 0.094 0 0.097 1 0.1174 0.106 3 0.068 8 0.042 6 0.044 0 0.023 4 SLP P 0.175 5 0.230 2 0.094 0 0.097 3 0.117 4 0.106 3 0.068 8 0.042 6 0.044 0 0.023 4 ${\tilde P}$ 0.381 1 0.339 8 0.259 0 0.189 8 0.143 0 0.113 3 0.094 7 0.083 6 0.077 6 0.075 8 Pk 0.166 4 0.217 6 0.094 2 0.090 8 0.099 5 0.094 1 0.078 8 0.075 1 0.060 8 0.022 1 HGT P 0.172 9 0.220 3 0.093 3 0.100 7 0.110 2 0.095 5 0.070 2 0.062 9 0.053 2 0.020 3 ${\tilde P}$ 0.404 3 0.354 5 0.261 7 0.186 7 0.138 2 0.108 3 0.089 9 0.079 0 0.073 3 0.071 4
下载: 导出CSV
表 3 因子场凝聚谱估计
T 18 9 6 4.5 3.6 3 2.6 2.3 2 Rk 0.223 8 0.036 2 0.037 7 0.003 8 0.304 4 0.510 1 0.6119 0.474 1 0.446 2 SST-SLP R 0.223 6 0.035 9 0.037 4 0.003 7 0.304 0 0.510 4 0.611 5 0.474 0 0.446 9 F 3.031 7 0.587 5 0.271 5 0.166 4 5.646 0 12.76 4 20.728 13.502 12.26 5 Rk 0.490 6 0.170 6 0.232 4 0.133 3 0.199 5 0.219 9 0.2939 0.107 9 0.001 0 SST-HGT R 0.529 0 0.149 6 0.236 5 0.146 2 0.266 8 0.340 4 0.415 8 0.180 4 0.025 0 F 15.218 2.810 0 5.409 9 3.652 3 6.120 1 10.290 15.406 6.188 0 2.016 4 Rk 0.292 5 0.285 1 0.496 6 0.718 8 0.793 3 0.565 0 0.397 6 0.304 0 0.243 4 SLP-HGT R 0.374 1 0.359 2 0.571 7 0.759 9 0.8120 0.646 6 0.479 8 0.324 9 0.251 5 F 7.385 2 6.736 7 14.627 32.701 51.304 25.460 14.958 5.269 8 1.310 2
下载: 导出CSV
表 4 KCCA与CCA的典型相关系数及检验
主成分 1 2 3 CCA r 0.602 6 0.199 5 0.052 2 χ2 29.427 2.579 5 0.162 4 KCCA r 0.638 2 0.456, 2 0.367 1 χ2 53.621 22.495 8.613 0
下载: 导出CSV
-
[1] 于玉斌, 赵大军, 陈联寿.干冷空气活动对超强台风"桑美"(2006) 近海突然增强影响的数值模拟研究[J].热带气象学报, 2015, 31(1):21-31. http://www.itmm.gov.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20150103&flag=1 [2] 王晨稀.基于随机全倾向扰动的台风路径集合预报试验[J].热带气象学报, 2015, 31(1):32-42. http://www.itmm.gov.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20150104&flag=1 [3] 吴俞, 薛谌彬, 郝丽清, 等.强台风"山神"外围超级单体引发的龙卷分析[J].热带气象学报, 2015, 31(2):213-222. http://www.itmm.gov.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=20150208&flag=1 [4] LIU H X, ZHANG D H, CHEN J W, et al. Prediction of tropical cyclone frequency with a wavelet neural network model incorporating natural orthogonal expansion and combined weights[J]. Natural Hazards, 2013, 65(1):63-78. doi: 10.1007/s11069-012-0343-x [5] WU A, HSIEH W W. Nonlinear canonical correlation analysis of the tropical Pacific wind stress and sea surface temperature[J]. Clim Dyn, 2002, 19(8):713-722. doi: 10.1007/s00382-002-0262-8 [6] PROHASKA, JOHN T. A technique for analyzing the linear relationships between two meteorological fields[J]. Mon Wea Rev, 1976, 104(1):1 345-1 353. doi: 10.1175/1520-0493%281976%29104%3C1345%3AATFATL%3E2.0.CO%3B2 [7] WEARE B C. A statistical study of the relationships between ocean surface temperatures and the Indian monsoon[J]. J Atmos Sci, 1979, 36(12):2 279-2 291. doi: 10.1175/1520-0469(1979)036<2279:ASSOTR>2.0.CO;2 [8] 梁建茵. 6月西太平洋副高脊线的年际变化及其对华南降水的影响[J].热带气象学报, 1994, 10(3):274-279. http://www.itmm.gov.cn/rdqxxb/ch/reader/view_abstract.aspx?file_no=19940310&flag=1 [9] 姚才, 金龙, 黄明策, 等.遗传算法与神经网络相结合的热带气旋强度预报方法试验[J].海洋学报, 2007(4):11-19. http://www.cnki.com.cn/Article/CJFDTOTAL-SEAC200704001.htm [10] HOTELLING H. Analysis of a complex of statistical variables into principal components[J] Journal of Educational Psychology, 1933(24): 417-441, 498-520. http://www.worldcat.org/title/analysis-of-a-complex-of-statistical-variables-into-principal-components/oclc/6537284 [11] JOLLIFFE I T. Principal component analysis, second edition[M]. New York, Springer, 2002. [12] SCHÖ B, LKOPF, MÜ S K R, et al. Kernel principal component analysis[J]. Lecture Notes in Computer Science, 1997:583-588. [13] BERNHARD S, SMOLA A. Nonlinear component analysis as a kernel eigenvalue problem[J]. Neural Computation, 1998, 10(5):1 299-1 319. doi: 10.1162/089976698300017467 [14] VAPNIK V N. The nature of statistical learning theory[J]. Neural Networks IEEE Transactions on, 1995, 10(5):988-999. http://www.springer.com/us/book/9780387987804 [15] SHAWE T J, HARDOON D R, SZEDMAK S. Canonical correlation analysis: an overview with application to learning methods[J]. Neural Computation, 2004, 16(12): 2 639-2 664. doi: 10.1162/0899766042321814 [16] SHOTARO A. A kernel method for canonical correlation analysis[J]. In Proceedings of the International Meeting of the Psychometric Society (IMPS2001, 2006, 40(2):263-269. https://www.researchgate.net/publication/2555798_A_Kernel_Method_For_Canonical_Correlation_Analysis [17] VÍA J, SANTAMARÍA I, PÉREZ J. Canonical correlation analysis (CCA) algorithms for multiple data sets: Application to blind SIMO equalization[C]//Signal Processing Conference, 2005 13th European. IEEE, 2005: 1-4. [18] 中国气象局.台风年鉴1949—1988年[Z].北京:气象出版社. [19] 中国气象局.热带气旋年鉴1989—2013年[Z].北京:气象出版社. [20] 中国台风网. http://www.typhoon.gov.cn/, 2015/07/14. [21] SATHIYA K S. Asymptotic Behaviors of Support Vector Machines with Gaussian Kernel[J]. Neural Computation, 2003, 15(7): 1 667-1 689. doi: 10.1162/089976603321891855 [22] 黄嘉佑.气象统计分析与预报方法 (第三版)[M].北京:气象出版社, 2004. [23] HANCOCK D J, YARGER D N. Cross-spectral analysis of sunspots and monthly mean temperature and precipitation for the contiguous United States[J]. J Atmos Sci, 1979, 36(4):746-753. doi: 10.1175/1520-0469(1979)036<0746:CSAOSA>2.0.CO;2 [24] THOMSON D J. Spectrum estimation and harmonic analysis[J]. Proceedings of the IEEE, 1982, 70(9):1 055-1 096. doi: 10.1109/PROC.1982.12433 -
点击查看大图
图(3) / 表(4)
计量
- 文章访问数: 699
- HTML全文浏览量: 5
- PDF下载量: 507
- 被引次数: 0
