Acta Optica Sinica, Volume. 45, Issue 12, 1201008(2025)

Atmospheric Model Construction Method Based on K‑means Clustering and Random Forest Regression

Haosen Wang1,2, Chen Cheng2、*, Hailiang Shi2, Xianhua Wang2, Hanhan Ye2, Shichao Wu2, and Erchang Sun2
Author Affiliations
  • 1School of Electronic and Information Engineering, Anhui Jianzhu University, Hefei 230601, Anhui , China
  • 2Anhui Institute of Optics and Fine Mechanics, Hefei Institute of Physical Science, Chinese Academy of Sciences, Hefei 230031, Anhui , China
  • show less
    Figures & Tables(24)
    Flowchart of atmospheric radiative transfer model construction
    Flowchart of random forest regression model construction
    Tropospheric height of eastern China region in different months
    Classification of temperature and pressure over eastern China in different months
    Temperature patterns of different categories in the eastern China. (a) January; (b) April; (c) July; (d) October
    Pressure profiles of different ategories in the eastern China region in January
    Regional mean pressure profiles for representative months. (a) 1–20 km; (b) 7–13 km
    Classification results of water vapor profiles in the eastern China region for different months
    Water vapor patterns of different categories in the eastern China. (a) January; (b) April; (c) July; (d) October
    Classification results of ozone profiles in the eastern China region for different months
    Ozone patterns of different categories in the eastern China. (a) January; (b) April; (c) July; (d) October
    Total volume fraction of CH4 in the eastern China region for different months
    Methane patterns in the eastern China region for representative months
    CO2 volume fraction in the eastern China region for different months
    Carbon dioxide patterns in the eastern China region
    Comparison of radiation transfer simulation results from 1976 US standard atmosphere model and self-built atmospheric model with HIRAS measured spectra in July
    Deviation between simulated spectra and measured spectra in July
    Comparison of radiation transfer simulation results from 1976 US standard atmosphere model and self-built atmospheric model with HIRAS measured spectra in January
    Deviation between simulated spectra and measured spectra in January
    • Table 1. Atmosphere profile data sources

      View table

      Table 1. Atmosphere profile data sources

      ParameterData sourceSpatial resolutionTemporal resolutionVertical coverage height
      H2OERA50.25°×0.25°1 month1000‒1 hPa
      TemperatureERA50.25°×0.25°1 month1000‒1 hPa
      PressureWACCM0.9°×1.25°6 h0‒150 km
      CO2CarbonTracker3°×2°1 d0‒47 km
      O3WACCM0.9°×1.25°6 h0‒150 km
      CH4WACCM0.9°×1.25°6 h0‒150 km
    • Table 2. Characteristic distribution of water vapor profiles for different categories in January

      View table

      Table 2. Characteristic distribution of water vapor profiles for different categories in January

      LabelIntegral value of volume fraction /10-6Mean square error /10-6Tropospheric height /kmLand type
      16716.32‒8512.83306.71‒388.199.8‒10.2Land: 98%; sea: 2%
      25198.31‒6865.46225.85‒312.589.8‒10.2Land: 94%; sea: 6%
      36660.71‒8913.93285.30‒364.049.8‒10.2Land: 67%; sea: 33%
      48511.71‒9820.21356.41‒455.6815.9‒16.0Land: 4%; sea: 96%
    • Table 3. Characteristic distribution of water vapor profiles for different categories in July

      View table

      Table 3. Characteristic distribution of water vapor profiles for different categories in July

      LabelIntegral value of volume fraction /10-6Mean square error /10-6Tropospheric height /kmLand type
      167447.57‒77530.172700.56‒3035.5416.5‒16.6Land: 9%; sea: 91%
      261678.44‒67805.202602.73‒2766.6716.5‒16.6Land: 1%; sea: 99%
      359871.98‒66801.972450.88‒3149.9516.5‒16.6Land: 94%; sea: 6%
      457860.39‒63252.412402.92‒2696.6415.2‒15.3Land: 100%; sea: 0%
    • Table 4. Characteristic distribution of ozone profiles for different categories in January

      View table

      Table 4. Characteristic distribution of ozone profiles for different categories in January

      LabelIntegral value of volume fraction /10-6Mean square error /10-6Tropospheric height /kmLand type
      16716.32‒8512.83306.71‒388.199.8‒10.2Land: 98%; sea: 2%
      25198.31‒6865.46225.85‒312.589.8‒10.2Land: 94%; sea: 6%
      36660.71‒8913.93285.30‒364.049.8‒10.2Land: 67%; sea: 33%
      48511.71‒9820.21356.41‒455.6815.9‒16.0Land: 4%; sea: 96%
    • Table 5. Characteristic distribution of ozone profiles for different categories in July

      View table

      Table 5. Characteristic distribution of ozone profiles for different categories in July

      LabelIntegral value of volume fraction /10-6Mean square error /10-6Tropospheric height /kmLand type
      16716.32‒8512.83306.71‒388.199.8‒10.2Land: 98%; sea: 2%
      25198.31‒6865.46225.85‒312.589.8‒10.2Land: 94%; sea: 6%
      36660.71‒8913.93285.30‒364.049.8‒10.2Land: 67%; sea: 33%
      48511.71‒9820.21356.41‒455.6815.9‒16.0Land: 4%; sea: 96%
    Tools

    Get Citation

    Copy Citation Text

    Haosen Wang, Chen Cheng, Hailiang Shi, Xianhua Wang, Hanhan Ye, Shichao Wu, Erchang Sun. Atmospheric Model Construction Method Based on K‑means Clustering and Random Forest Regression[J]. Acta Optica Sinica, 2025, 45(12): 1201008

    Download Citation

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

    Category: Atmospheric Optics and Oceanic Optics

    Received: Jan. 20, 2025

    Accepted: Feb. 17, 2025

    Published Online: Apr. 27, 2025

    The Author Email: Chen Cheng (chengchen@aiofm.ac.cn)

    DOI:10.3788/AOS250520

    CSTR:32393.14.AOS250520

    Topics