The Prediction And Spatial Statistics Analysis Of PM_2.5

With the rapid development of our economy, people pay more attention to high quality living environmental issues. PM_2.5 enters the public sight and becomes the air pollution that people care nowadays. The thesis researches PM_2.5 from following aspects :Firstly, this paper sets up the prediction model for PM_2.5 level based on the discrete param-eter Markov chain. Daily average concentration data of PM_2.5 are graded with the atmospheric pollutant concentration limits, and then we establish the discrete parameter Markov chain pre-diction model. Next, this paper uses the one-step state transition probability matrix and C-K equation to test the validity of the model. Finally, according to the ergodic properties of the discrete parameter Markov chain, the thesis obtains the steady-state distribution and return pe-riod. This paper forecast level of PM_2.5 that based on Nanjing's data. The result shows that the discrete parameter Markov chain model for grade prediction of PM_2.5 is effective.Secondly, because the traditional global Moran's I can not make the best of the informa-tion contained in data, this paper improves traditional global Moran's I. Through calculating the relationship between traditional global Moran's I and improved global Moran's I, we find improved global Moran's I that contains more information. For doing significance test, we use matrix algorithm of Lee to calculate the expectation and variance. The thesis analysis PM_2.5 per hour data of the 7 air quality monitors in Chengdu to explore spatial autocorrelation in the over-all region. The result shows that improved global Moran's I can mining spatial autocorrelation effectively.Thirdly, because of traditional local Moran's I's localization in large samples , this paper improves traditional local Moran's I. And we calculate the relationship between traditional local Moran's I and improved local Moran's I. For doing significance test, this thesis uses matrix algorithm of Lee to calculate the expectation and variance. We analysis PM_2.5 per hour data of the 7 air quality monitors in Chengdu to explore spatial autocorrelation in the local region. The result shows that improved local Moran's I can mining spatial autocorrelation effectively.Finally, in order to describe the interaction between regions, this paper combines geographic distance and economic variable to structure the geography — economic dynamic spatial weight matrix. Based on the geographic data and economic data of 21 cities in Sichuan Province,we combine the spatial weight matrix and the improved Moran's I to analysis spatial autocorrelation and spatial distribution of PM_2.5 pollution in Sichuan province. The result shows that combining the geography—economic dynamic spatial weight matrix with improved Moran's I can explore spatial autocorrelation effectively.