Time Series Analysis In Meteorology

With the rapid development of modern science and technologies, system analysis theories are corresponding developing and improving. In particular, mathematically modeling for a complex system which has many uncertain factors has attracted lots of interests. In this field, probabilistic and statistical models are often proposed and established. Particularly, time series analysis, as a powerful method, has already taken a very important role in such modeling. This method has successfully applied to many fields such as financial analysis, signal processing and weather forecast.Nowadays, climatic issues such as global warming are becoming serious. Investigating and analyzing meteorological data is therefore making more sense. This dissertation will present the investigation and application of the time series theory and technology in analyzing meteorological data.In the first chapter of this dissertation, we will introduce the main content and contribution of our research. In addition,an introduction to the theory of time series analysis will be provided. In the second chapter, we will present the analysis of the data of the highest temperature each day in the season of summer which was collected at the Nanjing site. These analyses are based on the principles and techniques of stationary time series modeling. The results will show that the data in the years of 1980, 1982 and 1986 are stationary while in the other years are non-stationary. In addition, for the data in 1980, the former 88 points are used to establish an ARMA model which gives a prediction of the temperature of the latter four days. The consistency between the predicted values and the realistic records strongly supports the reliability of our modeling.In the third chapter, the investigation of the non-stationary data from 1951 to 2004 will be conducted. Our case study is based on the data of the highest temperature in the summer of 1998. Analogously, the former 88 points in this year are used for modeling, while the other four points are utilized to compare with the predicted data produced by the established model. The comparison will show the consistency between the predicted data and the recorded data, as well as the reliability of our modeling. Meanwhile, we also analyzed the data of other site. And in the final chapter, the time series predictability was discussed.Our research demonstrates the powerful application of the time series in the field of meteorological data analysis. The work presented in this dissertation will be very valuable and provide good suggestions in the investigations of any other meteorological data. collected in other places and time.