平稳随机序列分析在水温预报中的应用

目前国内外进行海水温度预报主要是应用数理统计方法,平稳随机序列分析法就是其中之一,本文根据海水温度自身前后之间的关系,采用平稳随机序列分析法,用小麦岛海洋站３６ａ（１９６０～１９９５年）逐旬表层水温观测资料,建立多元线性回归方程,从而算出未来时刻的水温预报值。     

表层海水温度场趋势预报的几种方法简介

本文主要介绍了用于表层水温场趋势预报的几种方法:方差分析法、周期分析法和综合时间序列分析法,并用这些预报方法对东海及外缘海域、[Ⅰ,Ⅱ区]进行了水温场趋势预报试验,选出了预报海区表层海水温度场的各种显著周期。本文还给出了1982年的预报结果分析及方差分析法和周期分析法的拟合结果。分析得出预报海区海水温度场存在6～7年的长周期振动和2～3年的短周期振动,为表层海水温度场趋势预报提供了重要依据。     

周期外推方法在海水水温预报中的应用

针对海水温度的周期性变化,利用周期外推方法进行海水水温预报,并用平稳过程对周期变化以外的部分进行拟合,最后得到预报值。与实际观测资料比较,检验了预报的结果。     

黄东海近海区表层水温纵向预报试验——黄东海温盐预报方法研究之二

海水温度和盐度的变化与鱼类的行动和分布息息相关。当前在海洋渔业生产上广泛开展渔情预报工作,对发展海洋渔业生产,进行科学捕鱼作出了一定的贡献,而海水温、盐度的预报,往往是分析渔情,预测渔况和发布渔情预报的重要依据。本研究课题是直接为海洋渔业生产服务的,因而,我们主要选择渔汛期和渔场区进行了水温纵向预报试验。所谓纵向预报,这里是指多年同月水温序列的预报,借以同横向时间序列预报相区别。     

表层海水温度场的正交综合因子场分解预报方法——东海及其外缘海域月平均表层水温预报

表层海水温度预报的途径目前在我国仍以统计预报为主,数值预报和数值动力统计预报尚在探索中.表层海温统计预报一般又分为单要素时间序列预报和多要素相关预报两类.这些方法各有长处和短处,因此在预报实施中,应根据各种方法的特点,互相补充,综合发布预报.我们在应用经验正交函数分析预报表层海水温度场的基础上,也曾对表层海温场的多因子相关预报方法进行了多种试验.     

东海及外缘海域表层海水温度场的分析与预报

东海及外缘海域月平均表层海水温度场变化的持续性较好,正温度距平场最长可持续18个月,负温度距平场可持续14个月。能谱分析结果表明,由于该区受黑潮暖流系统的控制,存在明显的6、7年周期。 文中给出了几种以数理统计为基础的表层水温场预报方法,即单要素和多因子的经验正交场分解时间序列预报方法,对东海及外缘海域月平均表层海水温度场进行了连续6年(1978-1983)的实时预报,其效果比较稳定,预报误差较小,从1978～1983年的平均预报误差为0.53℃。     

应用经验正交函数分析预报表层海水温度的方法——东海及外缘海域月平均表层水温预报

东海及其外缘海域表层海水温度场的热状况直接影响着我国海洋资源的开发与利用,尤其是近年来随着外海远洋渔场的不断开发,对海洋环境预报的要求日益迫切,为此开展了大面积表层海水温度预报的研究.大面积表层海水温度预报,实际上是一个温度场的预报.若用热动力学数值模式的计算方法直接求解这个多维的非线性问题,目前还是比较困难的.     

海温预报知识讲座 第二讲 数理统计方法在海温预报中的应用

概率统计方法被引用到水文气象预报中来,至今已有五十余年。到60 年代后期,统计预报已成为要素预报的重要手段,尤其是70 年代以后“模式输出统计预报”在预报业务中使用以来,使统计预报出现了新的局面。现在水文、天气预报的概率统计方法已形成为一个独立的分支,它已成为自动化预报系统中不可缺少的组成部分,是使海温预报走向客观化、定量化的有效手段之一。 海水运动的规律是不确定的,是随机的,即使观测资料足够多,也不可能把海洋运动规律完全描述出来,而应用概率论、数理统计的方法预报未来海水温度变化是可能的,这就是概率统计海温预报…     

平稳时间序列分析在海温预报中的应用

本文根据老虎滩站１９６３－１９９４年逐年５月份海洋表层水温距平值，采用平稳时间序列分析的方法，建立的回归预报方程，从而给出时间序列未来时刻的预报值。     

南通洋口港海洋气象要素分析和预报

本文主要内容是南通洋口港海洋气象几种环境要素的分析和预报。选取国内外数值预报产品为预报因子,采用0,1回归方法,建立了四个关于海洋表面上的风力等级预报方程;采用三个简明的数学表达式反映海水表层温度的变化规律;采用投票法作出近海最大波高预报;采用统计分析方法,做出全年逐日、逐月高(低)潮时客观预报;采用误差分析法预测潮水到达紫菜养殖区或港口施工区的时间;实现了洋口港海洋气象4个要素的定量预报,并投入到紫菜养殖生产和洋口港海上施工气象科技服务工作中应用,其预报可信度受到了用户和海事等相关部门的认可与肯定。     

Statistics of wave forces on large horizontal cylinders

To the first order in a Stokes expansion, the pressure force exerted by a sea state on a large horizontal cylinder represents a stationary random Gaussian process. A relationship is obtained between the spectrum of this process and the wave spectrum. As a consequence, the basic statistical properties of the height and period of the individual waves of the force-process are also obtained. It is proven that these statistical properties agree very well with the data from a small scale field experiment.     

Stochastic Modeling of Short Term Variations of Sea Level in Eastern Canada

Hourly sea level records from three stations in eastern Canada (Québec-Lauzon, Harrington-Harbour and Halifax) are analyzed both in frequency domain from 1970 to 1979 and in time domain during 1973. At the three stations, the deterministic model explains 90 to 96% of the total variability of sea level. The semidiurnal and diurnal tides contribute largely to its variations. The residual series, less than 10% of the initial variations of sea level, contain irregular values including extreme values of seiches and storm surges. Such random variations are analyzed and modeled following the method described by Box and Jenkins (1976). The long period variations (2 to 30 days) can be attributed to meteorological forcing (atmospheric pressure and winds). The short period variations (some hours to one day) can be attributed to longitudinal seiches, semidiurnal and diurnal atmospheric tides, and inertial oscillations. The water discharge from the St. Lawrence River contributes 29% of the monthly residual sea level at Qué     

聚类分析在浅海变性水团分析中的试验

The statistical cluster analysis method is applied to the modified water masses in the shallow sea and a series of experiments are carried out. A simple and convenient method for defining the quantity and boundary of water masses and mixing zone from the hierarchical tree is given. Two cluster methods of gradual examination are proposed and compared with the hierarchical method.     

Variability of the Baltic Sea level and floods in the Gulf of Finland

The statistical analysis of the long-term data on the variability of the Baltic Sea level has revealed the complicated character of the wave field structure. The wave field formed by the variable winds and the disturbances of the atmospheric pressure in the Baltic Sea is a superposition of standing oscillations with random phases. The cross spectral analysis of the synchronous observation series of the level in the Gulf of Finland has shown that the nodal lines of the standing dilatational waves are clearly traced with frequencies corresponding to the distance from the nodal line to the top of the gulf (a quarter of the wave length). Several areas of the water basin with clearly expressed resonant properties may be distinguished: the Gulfs of Finland, Riga, and Bothnia, Neva Bay, etc. The estimations of the statistical correlation of the sea level oscillations with the variation of the wind and atmospheric pressure indicate the dominant role of the zonal wind component during the formation of the floods in the Gulf of Finland. The probable reason for the extreme floods in St. Petersburg may be the resonance rocking of the eigenmode oscillations corresponding to the basic fundamental seiche mode of the Gulf of Finland with a period of 27 h when the repeated atmospheric disturbances in the Baltic Sea occur with a period of 1–2 days.     

基于支持向量回归的海面高度统计预报方法研究

本文提出一种基于支持向量回归的统计预报方法,通过经验正交分解对原始数据矩阵进行时空分解,提取出空间模态和时间系数。由于海面高度变化具有非线性、大惯性的特点,对时间系数进行小波分析,能有效滤除其中的高频信号,得到表征海面高度变化的低频信号。利用支持向量回归方法对小波分解后的低频信号构建预报模型。最后,进行小波重构,还原时间序列长度,实现未来7天的海面高度预报。通过黑潮附近海域的海面高度预报结果验证,该预报方法的预报效果优于整合滑动平均自回归预报方法。本文通过机器学习的算法实现了海面高度的预报,为海洋预报方法提供了新的思路。     

Non-linear effects of the statistical model of shallow-water wind waves

This study deals with the non-linear effects of shallow-water wind waves. It is assumed that wind waves are an ergodic, random process, quasi-normal, stationary in time and homogeneous in space. The probability density function of sea surface oscillations is approximated by Gram-Charlier series in a modified form due to Edgeworth. Based on the above assumptions and on the fundamental statistical and stochastical laws the author has obtained the following characteristics: the probability density functions of wave heights, wave phases and velocity of the point mean wave height and mean wave length. The derived formulae are compared by the experimental data of the international expedition ‘Lubiatowo-74’. The analysis carried out has shown that real waves and the Gaussian model contradict considerably.     

Observation of the abyssal current in the West Mariana Basin

In order to investigate long-term variations of deep sea currents and temperature in the western North Pacific, a direct current measurement was made at 12.5°N, 137°E from July 1985 to July 1986. The current meter was moored at a depth of about 4,000 m (bottom depth 4,604 m) in the West Mariana Basin, very close to the deep water passage to the North Pacific Basin. Throughout the observed period, the current flows southward with an average speed of 0.8 cm sec^–1. There are significant variations of both currents and temperature within the period of 60 days. For the shorter time scales, in addition to the tidal oscillations with one day and half day periods, there is a notable spectral peak of the current with a clockwise rotation at a period of 2.2 days, which is slightly shorter than the local inertial period of 2.3 days.The observed southward current seems to indicate that the deep sea water in the West Mariana Basin flows out through the sill which is deeper than 4,000 m and is located about 200 km southeast of the mooring point. A simple analysis of the linear plane wave indicates that the medium time-scale variation with a period of 60 days is associated with the barotropic Rossby wave whose wavelength is 390 km and whose trough direction is 30° clockwise from the north.     

Research on the spectral analysis and test method of period signals in monthly mean sea level

Ｒｅｓｅａｒｃｈｏｎｔｈｅｓｐｅｃｔｒａｌａｎａｌｙｓｉｓａｎｄｔｅｓｔｍｅｔｈｏｄｏｆｐｅｒｉｏｄｓｉｇｎａｌｓｉｎｍｏｎｔｈｌｙｍｅａｎｓｅａｌｅｖｅｌ￥ＭａＪｉｒｕｉ；ＴｉａｎＳｕｚｈｅｎ；ＺｈｅｎｇＷｅｎｚｈｅｎａｎｄＣｈａｉＸｉｎｍｉｎ（Ｒ...     

Warm water intrusion from the Kuroshio into the coastal areas south of Japan

Sea surface temperature (SST) has been measured in the south of Japan using a thermometer set up in the ferry boat to investigate the characteristics of the warm water intrudes into the coastal areas from the Kuroshio. Time series analysis was applied to the SST data with satellite images and hydrographic observation data from April 1987 to September 1989. The results indicate that the warm Kuroshio water intruded into the coastal areas on the Enshu-nada and the Kumano-nada Seas intermittently with periods of about 50 and 20 days associated with the fluctuation of the Kuroshio path and the Kuroshio frontal disturbance respectively. The intrusion with a 50-day period was dominant when the Kuroshio took a stationary small meander path (B- and C-types). The warm water spread to the west at 20 cm s^–1, and was estimated to have a depth of 150 m at least and supply enough heat to make up the loss due to the evaporation in the coastal area. During the straight path of the Kuroshio, it was detected that the warm water intruded into coastal areas only with a 20-day period. The warm water that intrudes with a period of 20 days spreads to the west at 25 cm s^–1 in a small scale.     

Red noise and regime shifts

The analysis of interdecadal physical and biological variability is made challenging by the relative shortness of available time series. It has been suggested that rapid temporal changes of the most energetic empirical orthogonal function of North Pacific sea surface temperature (sometimes called the Pacific Decadal Oscillation or PDO) represents a “regime shift” between states with otherwise stable statistics. Using random independent time series generated to have the same frequency content as the PDO, we show that a composite analysis of climatic records recently used to identify regime shifts is likely to find them in Gaussian, red noise with stationary statistics. Detection of a shift by this procedure is not evidence of nonlinear processes leading to bi-stable behavior or any other meaningful regime shift.