课程名称：数据的物理分析
地 点：西康路1号威尼斯5994电气馆507
授课时间：5月28日—6月1日；6月4日—6月8日（周一到周五，为期两周）
下午2:00—5:00（每天授课3小时）
授课对象：青年科研人员和研究生
课程简介：
数据分析是任何科学和工程学科研究的基础。科学研究中的数据分析最终目的是为了解析隐藏在数据中的物理系统的演变特征，但大多数的数据分析方法更侧重于数学方面而不是物理方面。这种数据分析方法给理解物理系统带来了许多副作用，这主要是因为，分析方法作为黑箱，盲目地应用于数据，不考虑在某一具体问题上的适用性而作机械的应用。本门课程将介绍数据分析中物理观点，通过对传统分析方法，特别是对气候和海洋研究中的常用方法仔细分析，揭示其优缺点和可适用性。由此进一步讨论数据分析的物理观念和约束数据分析的一些基本物理定律，着重介绍满足这些物理约束的黄变换（希尔伯特-黄变换）的一系列方法，实现从数据的数学分析向物理分析的跨越。
授课专家：
吴召华副教授于1988年在南京大学大气科学系获得学士学位，1988-1991年在中科院大气物理研究所获得硕士学位，1998年在美国华盛顿大学大气科学系获得博士学位，2000-2001年在美国马里兰大学海洋-陆地-大气研究中心做博士后研究，2001-2005年在美国华盛顿东南大学计算机科学系担任讲师，2002-2008年在美国马里兰大学海洋-陆地-大气研究中心工作，2009至今在美国弗罗里达州立大学气象学系以及弗罗里达州立大学海洋-大气预测研究中心担任副教授。吴召华副教授的主要研究领域为大气和气候动力学，尤其擅长研究地球气候系统变化的基础理论。他对气象和气候资料的分析方法研究也有一定的贡献，他提出了一个改进的数据分析方法：经验模式分解（EMD）和噪声数据分析方法（NADA），为人们加深对气候变率的理解和全面认识全球气候变化提供了依据。
Objective:
The course discusses advantages and drawbacks of various types of the data analysis methods used to interpret data sets in the Earth, ocean, and atmospheric science. This is a tools class: the objective is to provide a working knowledge of the analysis tools. Emphasis is placed on the application of the tools discussed in class to the analysis of climate data.
Course Content Summary:
The data analysis has been used in every scientific or engineering field. Although the ultimate goal of data analysis in scientific research is to understand physical systems hidden in data, most of data analysis methods emphasize more on the mathematical aspects than on the physical aspects. This approach has brought many side-effects to understanding physical systems through data analysis, e.g., an analyzer taking an analysis method as a black box and applying it blindly to data without considering the deficiencies of the method and questioning its applicability.
In this course, I will present physical perspectives of data analysis. By introducing and disseminating the traditional data analysis methods, especially those widely used in climate sciences, I will introduce the popular methods based on well-established mathematical rules that are used widely in climate literatures; discuss the applicability of these methods to climate data; and expose their limitations. From there, I will introduce the physical perspective of data analysis in which a few most fundamental physical principles serve as the constraints of data analysis. In this part, we will take advantages of the recently developed Huang Transform (HT) (formerly Hilbert-Huang Transform, HHT), a novel and rapidly spreading adaptive and local data analysis method that has already been widely used with great success in many scientific and engineering fields, to illustrate the physical aspects of data analysis, especially the temporal locality of analysis and the "physicapability" of an analysis method: the capability of the method of isolating physically meaningful signals in data.
The course will cover both fundamental and advance concepts and techniques of data analysis, at a level comprehensible to graduate students. Through the class, home works and projects will be assigned to help to understand class content as well as to facilitate research in one’s own interests. For this reason, it is welcome that students bring in their own research project into the class.
Course Contents (tentative):
1.Data, data processing, and data analysis: An overview
Part I: Global domain analysis
2.Statistical descriptions of data
3.Methods with a priori determined basis (Fourier transform and wavelets)
Part II: Adaptive but global domain data analysis
4.Adaptive basis
5.Traditional analysis of temporal-spatial structures of data
Part III: Adaptive and local domain data analysis: Huang transform
6.Nonlinearity and non-stationarity of data
7.Adaptive and local analysis of data
8.Empirical Mode Decomposition
9.Ensemble Empirical Mode Decomposition
10.Multidimensional Ensemble Empirical Mode Decomposition
11.Analysis the evolution of temporal-spatial structures
12.Holo-spectral Analysis
Part IV: Optional topics
13.Trend and detrending
14.Reference frames for climate anomaly
