研究报告

  • 赵雪,侯丽丽,王鑫龙,武高峰,梁爽,赵文吉.基于LUR模型的2019年北京地区PM2.5与PM10浓度空间分异模拟[J].环境科学学报,2020,40(11):4060-4069

  • 基于LUR模型的2019年北京地区PM2.5与PM10浓度空间分异模拟
  • Simulation of spatial distribution of PM2.5 and PM10 concentrations in Beijing in 2019 based on LUR model
  • 基金项目:国家重点研发计划(No.2018YFC0706004)
  • 作者
  • 单位
  • 赵雪
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 侯丽丽
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 王鑫龙
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 武高峰
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 梁爽
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 赵文吉
  • 首都师范大学资源环境与旅游学院, 北京 100048
  • 摘要:在城市区域内,空气污染物的浓度在小范围内存在显著差异,而离散的地面监测点分布不均匀,且监测范围有限,无法满足污染物暴露评估等研究的需求.本研究基于GIS空间分析和多元逐步回归的模型构建的方法,建立了土地利用回归(LUR)模型,并模拟了北京市2019年PM2.5和PM10浓度的空间分布特征.选择土地覆盖数据、气象数据(风速、降水、温度)和植被覆盖度数据等预测变量,以研究区34个监测站点为中心建立0.1~5 km共7个系列缓冲区,表征不同尺度下各变量对PM2.5和PM10浓度的影响.研究结果表明:①进入PM2.5回归模型中的变量有:年均风速、温度、降水量和周围中等植被覆盖、耕地和不透水面的面积;进入PM10回归模型中的变量有:年均风速和周围中等植被覆盖的面积.两个模型的调整R2分别为0.829和0.677,模型精度较高.②抑制污染物浓度的变量,影响力随着空间范围扩大而增强;使污染物浓度增加的变量,影响力随着空间范围缩小而增强.③浓度模拟结果显示,PM2.5和PM10在西北部山区浓度较低,南偏东的城区浓度较高,并且向南有逐渐增加趋势.4植被覆盖度这一变量不仅进入了上述两个方程,且影响力都强于其他土地利用类型,故以后的模型改进应该考虑植被覆盖度这一因素.
  • Abstract:In urban areas, the concentration of air pollutants can be significantly different in a small region. However, ground monitoring sites are discrete and often unevenly distributed, with limited monitoring scope covered by each site. Such distributed monitoring data is difficult to meet the needs of pollutant exposure assessment and related research. In this study, a land use regression (LUR) model was established based on GIS spatial analysis and multiple stepwise regression method, and the spatial distributions of PM2.5 and PM10 concentrations in Beijing in 2019 were simulated. Predictive variables were tested and selected, such as land cover, meteorological data (wind speed, precipitation, temperature) and vegetation coverage. Based on data from 34 monitoring sites, a total of 7 series of buffer zones with sizes ranging from 0.1 km to 5 km were established to characterize the influence of various variables on PM2.5 and PM10 concentrations on different scales. Results show that: ① Six predictive variables were introduced into the LUR model of PM2.5, including the annual average wind speed, temperature, precipitation and the areas of surrounding medium vegetation coverage, farmland and impervious surface. Two predictive variables were introduced into the LUR model of PM10, including the annual average wind speed and the area of surrounding medium vegetation cover. The adjusted R2 were 0.829 and 0.677, respectively, meaning a satisfying precision. ② For variables that lower the pollutants, their impacts on pollutant concentration were enhanced with an increasing of space scope; while for those variables enhancing the pollution, their impacts were boosted by a smaller space.③ Simulation results of concentration distribution showed that the concentrations of PM2.5 and PM10 were lower in the Northwestern Mountainous Area, and higher in the Southeastern urban area, with a gradually increasing trend toward the south. 4The impact of vegetation coverage, which was newly introduced into the above two equations, was observed greater than the impact of other land use types on the simulations. It is recommended that factor of vegetation coverage should be considered in future model improvement.

  • 摘要点击次数: 82 全文下载次数: 134