Research On Data Process Based On Automatic Weather Station

The development of science and technology promotes the advance of human society as well as strengthens the relations between them. The environment of our life has changed greatly as well as the people’s exploration. Many disasters happened in human history, such as earthquake, tornados and shower which have made catastrophy to many people. It is well know that the accurate weather forecast is one of the most difficult works in the world and lots of issues with it are still problematic. The monitoring of air on line can made great effort to improve the accuracy of weather forecast and lower greatly the loss of meteorological disaster. Especially in the process of surveying tornados, an accurate and valid data series about weather condition is vital to effectively predicate the bad weather.Based on the technology of wireless sensor net, automatic weather station becomes an effective way to survey the air condition of some place. During the last decade, more and more data set collected by the Automatic Weather Station (AWS) was employed meteorological service. In addition, the advance of high density monitoring both in temporal and space, unattended operation and normal collection made the AWS becomes a popular tool instead of traditional manual way, which can be seen frequently in the service of sport, agriculture and monitoring of extreme weather condition.However, the complicated weather system, the unstable ability of electronic instrument and unknown disturb of outer made part of those data series wrong, which become a bottleneck of improving the data validation. As a result, it is significant to monitor the quality of data series collect by AWS in the process of surveying the weather condition in extreme weather and other applications.The main works done in this dissertation are as follows.1. Based on the theory of chaos, the characteristics of multi-data source are analyzed and a new method for noise filtering of temperature series is developed; 2. By employing the method of dynamic surveying, nonlinear data series and the AR based Kalman filter method, we providea two stage Kalman filter to denoise the temperature series collected by AWS. Besides that, rainfall series have more stochastic than former, we present an effective method based on the Online Support Vector Machine and Unscented Kalman Filter method to filter a precipitation series. The empirical results reveal that the proposed method outperforms the other two models using traditional Kalman filter and support vector machine respectively.3. With the theory of Hiden Markov theory, the characteristic of rainfall series is studied and an important rule of a precipitation is obtained. Based on the introduced rule, an adaptive Hidden Markov Module (HMM) is been built to evaluate a session of precipitation series. In addition, with the theory of stochastic theory, we propose a Probability Finite State Automata based Algorithm (PFSAA) for detecting outliers of air temperature series data caused by sensor’s errors. The proposed method can be used effectively in other nonlinear system.4. GIS is good platform to support the analysis the relationship of an adjacent data series in space. We propose a new method for recovering of temperature series with the support of adjacent automatic weather stations (AWS) with a GIS platform. Besides that, by using the platform of GIS, which can be used to build relation between different data type, we provide a new method to value the precipitation series by the conference of synthesis radar, which is a good way to monitor the valid of daily rainfall.