甘肃省2016年夏季大气首要污染物PM10的空间分析

发布时间：2018-05-14 13:42

本文选题：PM10 + 空间分布　；参考：《兰州大学》2017年硕士论文

【摘要】：随着各种工业过程中超细颗粒物的排放以及大气中二次形成的超细颗粒物与气溶胶等原因,大气中的可吸入颗粒物(PM10)逐渐成为成为全国大中城市的首要空气污染物,并严重影响城市生态环境及居民身体健康。探讨区域内PM10浓度在地理上的空间分布以及空间点模式具有重要的价值与现实意义。本文以甘肃省为研究区,以PM10为研究对象,利用甘肃省2016年夏季33个空气污染监测站点的六项空气指标(PM10, PM2.5, CO, NO2, 03,SO2)的75天的日数据进行建模研究。首先利用提出的基于Elman神经网络的平均影响值(MIV)的机器学习算法来确定其它五个空气指标对PM10的影响重要程度,筛选出其中对PM10影响最大的指标PM2.5,并且利用统计方法相关系数和散点图验证了该结论的正确性。然后分别利用普通克里金方法和以筛选出来的PM2.5为辅助变量的协同克里金方法对研究区每天的PM10进行插值,通过交叉验证的方式比较两种插值模型的精度。实验对比结果表明,协同克里金插值方法在插值精度上优于普通克里金,并且以协同克里金方法对研究区进行插值,以了解PM10的空间分布情况。最后,本文利用PM10浓度极大值点进行了空间点模式分析。利用协同克里金插值出来的75天的PM10空间分布,找出每天具有PM10浓度极大值的地理位置,将这些地理位置全部绘制在研究区内进行空间点模式分析。分析表明研究区夏季PM10的分布情况主要集中在城市人口密度大的区域,具有强烈的聚集性。
[Abstract]:With the emission of ultrafine particles in various industrial processes and the secondary formation of superfine particles and aerosols in the atmosphere, PM10 in the atmosphere has gradually become the main air pollutant in large and medium-sized cities in China. And seriously affect the urban ecological environment and the health of residents. It is of great value and practical significance to study the geographical distribution of PM10 concentration and the spatial point model. In this paper, taking Gansu Province as the research area and PM10 as the research object, the 75-day daily data of PM10, PM2.5, CO, NO2, and SO2) from 33 air pollution monitoring stations in Gansu Province in the summer of 2016 are used to model the model. Firstly, the machine learning algorithm based on the average influence value of Elman neural network is proposed to determine the importance of the other five air indexes to the PM10. The index PM2.5, which has the greatest influence on PM10, is selected, and the validity of the conclusion is verified by using the correlation coefficient and scatter plot of statistical method. Then the common Kriging method and the cooperative Kriging method with the filtered PM2.5 as the auxiliary variable are used to interpolate the daily PM10 in the study area, and the accuracy of the two interpolation models is compared by cross-validation. The experimental results show that the interpolation accuracy of cooperative Kriging interpolation method is better than that of common Kriging method, and the cooperative Kriging method is used to interpolate the study area to understand the spatial distribution of PM10. Finally, the spatial point model is analyzed by using the maximum of PM10 concentration. By using the spatial distribution of PM10 which is interpolated by co-Kriging for 75 days, the geographical location with maximum PM10 concentration per day is found out, and all these geographical positions are plotted in the study area for spatial point pattern analysis. The analysis shows that the distribution of PM10 in summer in the study area is mainly concentrated in the area with high urban population density and has a strong agglomeration.
【学位授予单位】：兰州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X513

【参考文献】