【作者】 王静；

【导师】 高炜； 施润和；

【作者基本信息】 华东师范大学 ， 地图学与地理信息系统， 2016， 硕士

【摘要】 近些年来,细颗粒物PM2.5是影响我国大多数城市空气质量的首要污染物,一方面,PM2.5影响着大气环境的清洁度,严重时可引起灾害性的重雾霾天气,直接导致城市能见度的下降,对人们的生产和生活造成不利的影响；另一方面,PM2.5严重威胁着人体健康,因此,开展PM2.5的监测和预测研究具有重要的现实意义。卫星遥感具有获取范围广、速度快、周期短、受条件限制少等特点,可以在PM2.5的监测当中发挥重要作用。本文以长江三角洲主要9个城市：上海市、南京市、杭州市、苏州市、无锡市、常州市、南通市、宁波市、绍兴市为研究区域,借助2013年6月1日至2015年5月30日研究区内68个地面监测站监测得到的PM2.5数据,探究了长三角主要城市的PM2.5的小时、曰、月和季节变化特征；提取与监测站相对应的MODIS AOD数据,探究PM2.5与AOD的相关性；利用再分析资料提供的气象要素数据,包括温度、气压、相对湿度、纬向风速、经向风速、表面粗糙度和边界层高度数据,探究各个气象要素对PM2.5的影响,最后,基于多元逐步回归分析和人工神经网络的方法分城市和分季节估算PM2.5。结果表明,长江三角洲主要9个城市的PM2.5具有明显的小时、日、月和季节变化特征：中午12时至13时是PM2.5浓度比较高的时段,PM2.5的日变化波动性很强；各个城市12月的PM2.5浓度值远远大于其他的月份,且12月PM2.5的波动程度比较大；PM2.5在冬季的浓度值大于其他季节,且波动性最大,其次是春季。夏季的浓度最小,且波动性最小；各个城市PM2.5受不同的气象要素控制和影响,其中温度、相对湿度和风是最显著的三个要素：基于多元逐步回归分析的方法估算PM2.5,各个城市全年尺度和不同季节的估算模型通过了显著性检验,每个城市的最优估算模型R2均在0.6以上,其中无锡市冬季PM2.5的估算模型R2最高,为0.884,且平均相对误差绝对值最小,为0.244,杭州市夏季的估算模型疋最低,为0.606,RMSE为56.999。基于人工神经网络估算的PM2.5,与真实测量值的基本走势大致一致,但是当实测的PM2.5浓度值偏高或者偏低时,预测值存在偏差。对比分析两种模型,基于神经网络建立的估算模型的精度高于基于多元线性逐步回归建立的估算模型。总的来说,基于卫星遥感获取的AOD监测地面空气污染状况具有可行性。基于以上的分析研究发现,PM2.5的浓度高低和多种气象要素具有复杂且紧密的联系,为了进一步探究,本文以2013年上海市入冬以来一次重雾霾事件为例,借助空气质量数据、地面气象要素、卫星遥感数据并结合后向轨迹模式追溯了此次重污染过程的污染过程以及其成因。更多还原

【Abstract】 Recent years, haze happens more and more frequently in China. PM2.5 is primary pollutant. Haze seriously threatened the life and health of local citizen, declines city’s visibility. PM2.5 draws more attention from public. Therefore, it has great significance to monitor the concentration of PM2.5.This paper chose nine cities in Yangtze River Delta as study area. Data sets include ground level PM2.5 concentration from June 1,2013 to May 30,2015; MODIS aerosol products and meteorological factors (temperature, relative humility, surface pressure, u wind, v wind, surface roughness, boundary layer height). This paper utilizes them to set up annual and seasonal estimation models of PM2.5 concentrations.PM2.5 in nine cities has obvious features. The hourly average PM2.5 concentration in Ningbo is lower than other cities. PM2.5 in Nanjing is highest. The diurnal variation of PM2.5 is not stable. The average concentration of PM2.5 is highest in November. Compare to other seasons, seasonal averaged concentration is higher in winter. From the analysis of meteorological factor with PM2.5, we can conclude that, temperature, relative humility and wind are the most significant three elements.This paper utilizes two methods to estimate PM2.5:step wise regression and single-hidden layer feedforward. Stepwise regression models involve ground level PM2.5, MODIS AOD and meteorological factors, R2 of every model is higher than 0.6. The highest R2 is 0.884. The lowest MAPE is 0.244. Single-hidden layer feedforward based estimation model shows the same PM2.5 trend between measured value and estimated value. This study shows that satellite remote sensing effectively estimates the ground level PM2.5 concentration. The air pollution was the heaviest and most long-lasting pollution event in the historical record of Shanghai from 30 November to 9 December,2013. The aim of this work was to analyze the characteristics and formation mechanisms of this episode. For this purpose, the air quality data, ground meteorological data and satellite remote sensing data were used.更多还原