Chinese Journal of Lasers, Volume. 49, Issue 17, 1710001(2022)

Estimation of Particulate Matter Mass Concentration Based on Generalized Regression Neural Network Model Combining Aerosol Extinction Coefficient and Meteorological Elements

Zusi Mo1, Lingbing Bu1、*, Qin Wang1, Xuefei Lin1, Samuel A. Berhane1, Bin Yang2, Chen Deng2, and Zhi Li2
Author Affiliations
  • 1Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters, School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing 210044, Jiangsu, China
  • 2Nanjing Mulei Laser Technology Co., Ltd., Nanjing 210038, Jiangsu, China
  • show less
    Figures & Tables(12)
    A systematic framework of lidar
    Flow chart for estimating PM2.5 and PM10 mass concentrations with GRNN based on extinction coefficient and meteorological elements
    GRNN structure
    Result of four fold cross validation
    Validation of GRNN model. (a) Comparison between estimated and measured mass concentrations of PM2.5 and PM10; (b) correlation between estimated and measured mass concentration of PM2.5; (c) correlation between estimated and measured mass concentration of PM10
    Lidar observation location distribution map (dot A is the locations of polarimetric lidar for horizontal scanning)
    Average PM2.5 mass concentration estimation under horizontal scanning in different periods on May 9, 2021
    Average PM10 mass concentration estimation under horizontal scanning in different periods on May 9, 2021
    Change in hourly averages of PM2.5 mass concentration, PM10 mass concentration, temperature, relative humidity, wind direction and speed, and surface pressure in Pukou District on May 9, 2021
    Comparison between GRNN model estimated mass concentrations of PM2.5 or PM10 and measured values
    • Table 1. Autocorrelation analysis (EX represents aerosol extinction coefficient, TE represents temperature, RH represents relative humidity, WS represents wind speed, and PS represents surface pressure)

      View table

      Table 1. Autocorrelation analysis (EX represents aerosol extinction coefficient, TE represents temperature, RH represents relative humidity, WS represents wind speed, and PS represents surface pressure)

      ParameterMass concentration of PM2.5Mass concentration of PM10EXTERHWSPS
      Mass concentration of PM2.510.7394***0.6339***-0.1877***-0.2894***-0.2067***0.1146**
      Mass concentration of PM10 10.5102***-0.1372***0.1005***-0.3139***0.0838**
      EX  1-0.1817***0.4662***-0.2241***0.0257**
      TE   1-0.3212***0.2272**-0.0177*
      RH    1-0.3284*-0.1311*
      WS     1-0.0717
      PS      1
    • Table 2. Performance indicators of GRNN model

      View table

      Table 2. Performance indicators of GRNN model

      Particulate matterQuantity of training sampleQuantity of validation sampleRMAE /(μg·m-3)RMSE /(μg·m-3)
      PM1012002510.861.5310.84
      PM2.512002510.850.812.58
    Tools

    Get Citation

    Copy Citation Text

    Zusi Mo, Lingbing Bu, Qin Wang, Xuefei Lin, Samuel A. Berhane, Bin Yang, Chen Deng, Zhi Li. Estimation of Particulate Matter Mass Concentration Based on Generalized Regression Neural Network Model Combining Aerosol Extinction Coefficient and Meteorological Elements[J]. Chinese Journal of Lasers, 2022, 49(17): 1710001

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: remote sensing and sensor

    Received: Nov. 15, 2021

    Accepted: Jan. 14, 2022

    Published Online: Jul. 28, 2022

    The Author Email: Bu Lingbing (lingbingbu@nuist.edu.cn)

    DOI:10.3788/CJL202249.1710001

    Topics