基于遥感技术和机器学习的城市街区PM_(2.5)空间分布特征研究——以合肥市滨湖新区为例

Spatial Distribution Characteristics of PM_(2.5) in Urban Block Based on Remote Sensing Technology and Machine Learning: Taking Binhu New District of Hefei City as an Example

为获取合肥市滨湖新区地面PM_(2.5)浓度,利用地面风速对2022年7月31日-2023年7月31日MODIS传感器MCD19A2数据进行风速订正,并与同期气象数据和合肥市10个国控监测站点的PM_(2.5)数据建立AOD-PM_(2.5)反演模型。结果表明:BP网络训练到第10代时训练结果最为理想,均方误差MSE值约为0.027。训练集、验证集、测试集和总体结果回归值r均在0.88以上,总体结果高达0.93。通过BP预测工具箱建立模型,拟合优度达到0.807。选择滨湖新区7个典型测点,测量2023年7月12-16日每日平均PM_(2.5)浓度、地面风速和温湿度,结合气象数据代入反演模型对比发现,估算值和实测值曲线高度吻合,证明该模型可用于反演滨湖新区地面PM_(2.5)浓度。对PM_(2.5)浓度进行估算和分析,结果表明:1)滨湖新区西南侧PM_(2.5)浓度最高,东南侧最低;2)功能区PM_(2.5)浓度呈现为工业区>商业区>居住区>生态园区;3)工业园区废弃物的排放会导致PM_(2.5)浓度升高,对工厂设备的优化使用以及合理处理生产过程中的废弃物可降低PM_(2.5)浓度;4)商业区和居住区中的尾气排放、扬尘、工业排风等均会导致PM_(2.5)浓度升高,优化交通组织,推动新能源汽车的发展可降低PM_(2.5)浓度;5)产业园区的发展模式对城市PM_(2.5)浓度有显著影响,加快产业结构调整,推动产业绿色技术创新可降低PM_(2.5)浓度;6)城市绿地空间对PM_(2.5)有显著调节作用,增加绿地空间连通度、聚集度,加强大型绿地建设可发挥较强的PM_(2.5)消减作用。该研究所建立的AOD-PM_(2.5)反演模型旨在为城市街区PM_(2.5)空间分布特征研究提供可靠方法,具有重要的实践意义。

In order to obtain the ground-level PM_(2.5) concentration in Binhu New District of Hefei City,the ground wind speed was used to correct the wind speed of the MCD19A2 data of the MODIS sensor from July 31,2022,to July 31,2023,and the AOD-PM_(2.5) inversion model was established with the same period of meteorological data and the PM_(2.5) data from 10 state-controlled monitoring stations in Hefei City.The test results were most satisfactory when the BP network was set up to the 10th generation,and the Mean Square Error(MSE)value was 0.027.The regression values of R for the training,validation,and test sets were greater than 0.88,and the overall result was as high as 0.93.The model was built using the BP prediction toolbox and the goodness-of-fit was 0.807.Seven typical measuring points were selected in the Binhu New District to measure the daily average PM_(2.5),ground wind speed,temperature,and humidity from July 12,2023,to July 16,2023.After combining the meteorological data into the inversion model,it was found that the estimated and measured value curves were highly consistent,which proves that the model can be used to reverse the ground PM_(2.5) concentration in Binhu New District.The results of the estimation and analysis of the PM_(2.5) concentration showed that:1)the highest PM_(2.5) concentrations were found on the southwest side of the Binhu New District and the lowest on the southeast side.2)In terms of land use,PM_(2.5) concentration was presented as industrial area>commercial area>residential area>ecological park.3)Waste emissions in industrial zones led to an increase in PM_(2.5).Optimizing the use of factory equipment and proper treatment of waste in the production process can reduce PM_(2.5).4)Tailpipe emissions,dust,and industrial exhaust in commercial and residential areas led to higher PM_(2.5).Optimizing traffic organization and promoting the development of energy vehicles can reduce PM_(2.5).5)The development mode of industrial parks has a significant impact on the concentration of PM_(2.5);accelerating the adjustment of industrial structure and promoting the innovation of industrial green technology can reduce the concentration of PM_(2.5).6)Urban green spaces had a significant regulatory effect on PM_(2.5).Increasing the connectivity and aggregation of green spaces and strengthening the construction of large-scale green spaces can play a stronger role in PM_(2.5) reduction.The AOD-PM_(2.5) inversion model provides a reliable method for the study of the spatial distribution characteristics of PM_(2.5) in urban neighborhoods,which has important practical significance.