Scholarlink

<p>High spatiotemporal resolution data on carbon monoxide (CO) concentration distribution play a crucial role in the detection and management of atmospheric CO pollution. In this study, we integrated CO column concentration data from the Sentinel-5P satellite, ERA5 reanalysis meteorological data, and ground-based observation data to fully account for the complex spatiotemporal heterogeneity of CO and its nonlinear relationships with predictor variables. A hybrid model, Bo-LightGBM-BiGRU, was developed to estimate daily CO concentrations across Shandong Province. Bayesian optimization was employed to construct a Gaussian process surrogate model, enabling adaptive modeling of the complex nonlinear relationships between CO concentrations and multi-source environmental variables by iteratively minimizing the mean squared error (MSE) objective function. Ten-fold cross-validation results demonstrated that the Bo-LightGBM-BiGRU model achieved high estimation accuracy, with a coefficient of determination (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({R}^{2}\)</EquationSource> </InlineEquation>) of 0.913, mean absolute percentage error (MAPE) of 10.054%, mean absolute error (MAE) of 61.379 <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\;\mu g\bullet m^{-3}\)</EquationSource> </InlineEquation>, and root mean square error (RMSE) of 77.747 <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\;\mu g\bullet m^{-3}\)</EquationSource> </InlineEquation>. Temporally, CO concentrations in Shandong Province in 2021 exhibited a “V”-shaped trend, first declining and then rising, with the highest concentration occurring in January and the lowest in July. Seasonally, CO concentrations varied significantly, being substantially higher than the average in winter and lowest in summer. Spatially, CO concentrations displayed a general pattern of being higher in the central region and lower in the eastern and western parts. The eastern coastal areas had lower concentrations than the provincial average, likely due to the dispersal effects of the temperate monsoon climate. According to the results of the bivariate Moran’s I index, there was a weak positive correlation between CO concentrations and population/traffic density in inland areas of Shandong Province.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Estimation of near-surface carbon monoxide concentration in Shandong based on Bo-LightGBM-BiGRU modeling

  • 期刊论文

  • 发表日期：

    2026年3月9日

查看全文
  • Yajing Kang,
  • Xuming Yang,
  • Chunkang Zhang

摘要

High spatiotemporal resolution data on carbon monoxide (CO) concentration distribution play a crucial role in the detection and management of atmospheric CO pollution. In this study, we integrated CO column concentration data from the Sentinel-5P satellite, ERA5 reanalysis meteorological data, and ground-based observation data to fully account for the complex spatiotemporal heterogeneity of CO and its nonlinear relationships with predictor variables. A hybrid model, Bo-LightGBM-BiGRU, was developed to estimate daily CO concentrations across Shandong Province. Bayesian optimization was employed to construct a Gaussian process surrogate model, enabling adaptive modeling of the complex nonlinear relationships between CO concentrations and multi-source environmental variables by iteratively minimizing the mean squared error (MSE) objective function. Ten-fold cross-validation results demonstrated that the Bo-LightGBM-BiGRU model achieved high estimation accuracy, with a coefficient of determination ( \({R}^{2}\) ) of 0.913, mean absolute percentage error (MAPE) of 10.054%, mean absolute error (MAE) of 61.379 \(\;\mu g\bullet m^{-3}\) , and root mean square error (RMSE) of 77.747 \(\;\mu g\bullet m^{-3}\) . Temporally, CO concentrations in Shandong Province in 2021 exhibited a “V”-shaped trend, first declining and then rising, with the highest concentration occurring in January and the lowest in July. Seasonally, CO concentrations varied significantly, being substantially higher than the average in winter and lowest in summer. Spatially, CO concentrations displayed a general pattern of being higher in the central region and lower in the eastern and western parts. The eastern coastal areas had lower concentrations than the provincial average, likely due to the dispersal effects of the temperate monsoon climate. According to the results of the bivariate Moran’s I index, there was a weak positive correlation between CO concentrations and population/traffic density in inland areas of Shandong Province.