共查询到20条相似文献,搜索用时 46 毫秒
1.
C. Serra G. Fernández Mills M. C. Periago X. Lana 《Theoretical and Applied Climatology》1996,54(3-4):175-186
Summary A principal component analysis (PCA), based on a network consisting of 60 pluviometric gauges and their daily precipitation data, is attempted in order to describe the main winter and autumn patterns governing precipitation in Catalonia (NE Spain). This PCA procedure is applied to the interstation correlation matrix and rotated component loadings are then deduced and extensively interpreted. The PCA results are then used in a clustering process (Average Linkage), leading to two rainfall divisions, one for each season, which are then compared.With 5 Figures 相似文献
2.
Principal component analysis (PCA) with oblique rotation is applied to Large Eddy Simulation (LES) results to discern and quantify coherent structures within the convective boundary layer (CBL). Sensitivity tests are first conducted on a moderately convective LES run. Once the ability of PCA to generate robust results is verified, the method is applied to LES runs spanning a range of stability regimes. Interregime similarities and differences in the coherent structures are discussed. For the moderately convective LES run, three-dimensional convective cells are arrayed in two-dimensional bands aligned with the geostrophic wind. The resulting gravity waves in the free atmosphere and convective inflow and outflow in the boundary layer are also captured by the PCA. Convective modes are more sensitive to the ratio of w
* to u
* than are the dynamic modes.PCA has demonstrated advantages over previous analysis methods. PCA score maps provide information on the spatial distribution of phenomena that has not been available from traditional conditional sampling studies. Principal components provide information on the vertical structures of phenomena that would be obscured by life-cycle effects or erratic tilts from the vertical in the conventional approaches to either conditional sampling or composite analysis. Future work includes application of this technique to multi-level observational time series from a surface-layer tower for the Risø Air/Sea Experiment (RASEX). 相似文献
3.
In an attempt to clearly separate the volcanic signal,we use a mixture of principal componentanalysis(PCA)and superposed epoch analysis to identify volcanic signal in the global surface tem-perature field.In this way,the spatial and temporal pattern of volcanic signals is identified in theglobal surface temperature records.Our results show that the strongest ENSO and volcanic signalsare related with the first and the third principal components respectively.Both ENSO and volcanicsignals have responses in the second principal component. 相似文献
4.
5.
基于主分量的神经网络水位预报模型应用研究 总被引:5,自引:0,他引:5
根据气象和水文资料,以上游面雨量、水位值为预报因子,以西江流域的梧州水位为预报量,发现预报因子与预报量有很好的相关性。采用人工神经网络与主分量分析相结合的方法,建立了梧州水位的预报模型。结果表明,该预报模型对历史样本拟合精度高,试报效果及预报稳定性明显好于传统的神经网络预报模型,可在预报业务中应用。 相似文献
6.
华南热带气旋相关物理量场的线性及非线性统计分析研究 总被引:1,自引:0,他引:1
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。 相似文献
7.
8.
Correlations between sea surface temperature in eastern equatorial Pacific and rain days over China in summer 总被引:1,自引:0,他引:1
Huang Jiayou 《大气科学进展》1990,7(2):127-136
Sea-surface temperature (SST) in the eastern, equatorial Pacific and rain days over China in summer are ana-lysed using correlation moments, that is proposed by author and principal component analysis (PCA). Occurrences of the strong rain-day anomalies over China are associated with extreme SSTs in some years. Areas significantly affect-ed by the phenomena include North and Northeast China. 相似文献
9.
JIN Long KUANG Xueyuan HUANG Haihong QIN Zhinian WANG Yehong 《Acta Meteorologica Sinica》2005,19(2):216-225
Because of overfitting and the improvement of generalization capability (GC) available in the construction of forecasting models using artificial neural network (ANN), a new method is proposed for model establishment by means of making a low-dimension ANN learning matrix through principal component analysis (PCA). The results show that the PCA is able to construct an ANN model without the need of finding an optimal structure with the appropriate number of hidden-layer nodes, thus avoids overfitting by condensing forecasting information, reducing dimension and removing noise, and GC is greatly raised compared to the traditional ANN and stepwise regression techniques for model establishment. 相似文献
10.
R. Huth 《Theoretical and Applied Climatology》2007,89(1-2):1-8
Summary The Arctic Oscillation (AO) appears as the leading unrotated mode of principal component analysis (PCA) of monthly mean sea
level pressure anomalies, whereas the North Atlantic Oscillation (NAO) results from rotated PCA, regardless of the number
of PCs rotated. Three criteria are employed to decide whether the interpretation in terms of the NAO or AO should be preferred:
the degree of simple structure, the similarity between the PC loadings and correlation/covariance maps, and the sensitivity
to spatial subsampling. All these criteria favour, to a different extent, the interpretation in terms of the NAO. This is
further supported by more general arguments. Therefore, the statistical arguments suggest that in interpreting the Northern
Hemisphere circulation variability, the sectorial view, i.e. the NAO, should be preferred to the hemispheric view, i.e. the
AO. Our analysis supports the idea expressed in other studies that the AO is rather a statistical artifact. 相似文献
11.
利用主成分分析和相关分析表明,近40 a来福建春雨量场的主要信息集中在前2个主分量上,可解释方差的83.8%,同时第1、第2主成分与福州、厦门春雨的相关显著性水平都超过0.01,因此,可用近百年福州、厦门3—4月降水量构造新序列代表福建春雨的前2个主分量,使分析资料延长至百年。采用连续小波分析和正交小波分析方法以及功率谱方法,研究福建春雨空间分布型的多时间尺度特征,研究结果表明:第1主分量表示的全省一致性旱涝具有显著的准4 a和24~26a的周期振荡;第2主分量表示的春雨分布南北差异主要表现为准两年和准11 a的周期振荡;春旱年南部的降水少于北部;从年代际角度看,20世纪50—60年代显著偏少,80年代显著偏多。福建目前处于春雨正常—偏少期,且南部的春旱比北部严重。 相似文献
12.
基于MATLAB的主成分RBF神经网络降水预报模型 总被引:13,自引:3,他引:10
以前期500 hPa高度场、海温场为预报因子,采用径向基函数(RBF)神经网络与主成分分析相结合的方法,建立了广西中部5月平均降水预报模型。在5年独立样本的预测检验中,预测的平均相对误差、均方误差及平均绝对误差分别为18.12%、50.52和34.23。对比分析RBF神经网络与BP(Back Propagation)神经网络的预测结果,表明RBF神经网络预测结果更准确、精度更高。 相似文献
13.
14.
A spectral analysis of Iberian Peninsula monthly rainfall 总被引:2,自引:0,他引:2
J. A. García A. Serrano M. de la Cruz Gallego 《Theoretical and Applied Climatology》2002,71(1-2):77-95
Summary A spectral analysis of Iberian Peninsula monthly rainfall series was carried out. The data set consists of monthly precipitation
records from 40 meteorological observatories over 74 years (1919–1992). The stations are representative of most of the Iberian
Peninsula. The rainfall series were analyzed spatially by means of Principal Component Analysis (PCA) and temporally by means
of the Multi-Taper Method (MTM) of spectral analysis of by Monte-Carlo Singular Spectrum Analysis (MCSSA). The PCA gave six
dominant modes of variation which explain 75% of the variance with each component affecting a different region of the Peninsula.
The spectral analysis showed 7 year oscillations for the dominant pattern and 2.7 and 16 years for the third pattern. The
7-year oscillation seems to be related to other climatic oscillations recorded in the northern hemisphere while the 2.7-year
oscillation could be related to the ENSO phenomenon.
Received July 18, 2000 Revised April 19, 2001 相似文献
15.
采集大连某机场2004—2007年大雾、轻雾和无雾天气事件共186例,选取雾天气事件前期(前一日08:00,14:00,20:00(北京时)实测资料)的温、压、湿、风等要素指标为预报因子,基于学习向量量化神经网络(learning vector quantization,LVQ),采用逐级预报思想建立起某机场雾天气事件的预报模型。在网络训练过程中,动态调整网络神经元比例参数,提高模型的预报能力;采用根据检验准确率适时终止训练的"先停止"技术,有效提高了模型的泛化能力。预报试验表明:无论是拟合率还是独立预报准确率,模型均已达到较高水准,具有实际应用意义。 相似文献
16.
17.
运用一种基于神经网络的非线性主成分分析法(nonlinear principal component analysis,NLP-CA)对中国1951—2003年53 a四季气温距平场(surface air temperature anomaly,SATA)进行分析,NLPCA第一模态结果显示中国四季气温异常具有一定的非线性特征,并且具有显著的季节性差异,即春、夏两季的非线性较强,秋、冬两季较弱。一维NLPCA对原始气温距平场的近似比一维PCA(principal component analysis)更好地反映了气温场的实际分布情况。 相似文献
18.
Pattern characteristics of Indian monsoon rainfall using principal component analysis (PCA) 总被引:1,自引:0,他引:1
C.V. Singh 《Atmospheric Research》2006,79(3-4):317-326
In the present study the Principal Component Analysis (PCA) is used to determine the dominant rainfall patterns from rainfall records over India. Pattern characteristics of seasonal monsoon rainfall (June–September) over India for the period 1940 to 1990 are studied for 68 stations. The stations have been chosen on the basis of their correlation with all India seasonal rainfall after taking the ‘t’ Student distribution test (5% level). The PCA is carried out on the rainfall data to find out the nature of rainfall distribution and percentage of variance is estimated. The first principal component explains 55.50% of the variance and exhibits factor of one positive value throughout the Indian subcontinent. It is characterized by an area of large positive variation between 10°N and 20°N extending through west coast of India. These types of patterns mostly occur due to the monsoon depression in the head Bay of Bengal and mid-tropospheric low over west coast of India. The analysis identifies the spatial and temporal characteristics of possible physical significance. The first eight principal component patterns explain for 96.70% of the total variance. 相似文献
19.
The aim of this work is to obtain an index for predicting the probability of occurrence of zonda event at surface level from sounding data at Mendoza city, Argentine. To accomplish this goal, surface zonda wind events were previously found with an objective classification method (OCM) only considering the surface station values. Once obtained the dates and the onset time of each event, the prior closest sounding for each event was taken to realize a principal component analysis (PCA) that is used to identify the leading patterns of the vertical structure of the atmosphere previously to a zonda wind event. These components were used to construct the index model. For the PCA an entry matrix of temperature (T) and dew point temperature (Td) anomalies for the standard levels between 850 and 300 hPa was build. The analysis yielded six significant components with a 94 % of the variance explained and the leading patterns of favorable weather conditions for the development of the phenomenon were obtained. A zonda/non-zonda indicator c can be estimated by a logistic multiple regressions depending on the PCA component loadings, determining a zonda probability index \( \widehat{c} \) calculable from T and Td profiles and it depends on the climatological features of the region. The index showed 74.7 % efficiency. The same analysis was performed by adding surface values of T and Td from Mendoza Aero station increasing the index efficiency to 87.8 %. The results revealed four significantly correlated PCs with a major improvement in differentiating zonda cases and a reducing of the uncertainty interval. 相似文献
20.
Properties of the circulation classification scheme based on the rotated principal component analysis 总被引:3,自引:0,他引:3
Dr. R. Huth 《Meteorology and Atmospheric Physics》1996,59(3-4):217-233
Summary The study addresses some methodological issues of application of principal component analysis (PCA) to the classification of circulation patterns. The obliquely rotated PCA in T-mode (i.e. with time observations corresponding to variables and grid points to realizations) is applied to 500 hPa geopotential heights over Europe and adjacent parts of Atlantic Ocean. The solutions are examined for various numbers of principal components rotated, and for both raw and anomaly data, with the aim to find the way of determining the optimum number of circulation types. This is done, among others, by examining temporal and spatial stability of solutions, their compliance with simple structure requirements, and temporal behaviour of classifications. Some of the solutions that are pre-selected according to the rule based upon the separation between successive eigenvalues prove to perform considerably better than unselected ones; some of them do not. Which pre-selected solutions should be given preference is impossible to decide in advance, without a detailed scrutiny. Nevertheless, even after such a scrutiny is done, more than a single classification are acceptable. The final choice of the optimum solution depends on the aims of the intended study: It should balance the demands on statistical stability of types and on resemblance between types and daily patterns classified with them.With 6 Figures 相似文献