Visual Analysis Of Time Series Environmental Data Of Shandong Province

With the rapid development of science and information technologies, collecting data using intelligent devices are ubiquitous in the process of manufacture and will accumulate massive high-dimensional time series data. It is almost impossible to manually display or analyze these massive high-dimensional data. Therefore, using computer program to display and analyze these massive data has become a hot and difficult problem in the field of computer science in recent years. Air quality data as a type of time series has received more attention because of the increasingly serious of environmental pollution problem. In order to help people perceive the pattern of air quality time series data of Shandong province, we has invented some novel data visualization and data analytics methods in this thesis.There are 17 cities and 144 air quality monitoring stations in Shandong province and the monitoring station will record six pollutants,for example,SO₂,PM2.5,etc, per hour. For each pollutant, there will be 144 time series, and how to visualize these large-scale data becomes the first challenging problem. Due to traditional visualization method can’t display these massive data well, in this thesis we propose a visualization method that combine ThemeRiver method and time series clustering method. In this method, we can not only display individual time series clearly but also can discern some similarity pattern and overview in the time series datasets. There will be a multivariate time series for each station, and how to find the linear relationships between these pollutants is our second challenging issue, we use scatterplot matrix visualization to visualize the multivariate time series in this thesis. Due to the traditional scatterplot matrix can’t automatic identify the outliers and the temporal characteristic can’t present, we use Mahalanobis distance based method to automatically identify the outliers and use animation techniques to present the temporal in the datasets. How to present the primary pollutant for each station and compare primary pollutant of monitoring stations or cities as time went on is also an important topic in air quality research. In order to achieve these goals, we should use a visualization method that combines temporal visualization and hierarchical visualization, we propose a visualization method that combines Circle packing and pie visualization. Through the three methods that we proposed, the users can have a better understanding of the air quality of Shandong province and our methods provide better decision support for environmental management.