Scholarlink

<p>Accurate estimation of atmospheric aerosols is essential for understanding air quality and climate interactions, particularly in complex mountainous regions. This study evaluates black carbon (BC) and aerosol optical depth (AOD) over the Western Ghats using ground-based observations and MERRA-2 reanalysis data, enhanced through a Deep Bayesian Filtering (DBF) framework integrating Long Short-Term Memory (LSTM) and Sequential Monte Carlo (SMC) methods. Baseline results show strong agreement between MERRA-2 and observations (BC RMSE: 0.30&#xa0;μg/m<sup>3</sup>, AOD RMSE: 0.02, R up to 0.98), but performance improves significantly after DBF application (BC RMSE: 0.19&#xa0;μg/m<sup>3</sup>, AOD RMSE: 0.01, R<sup>2</sup> up to 0.97). Seasonal analysis indicates higher uncertainty during the pre-monsoon period due to dust and biomass burning, while post-monsoon conditions show improved stability. Altitude-based results reveal increasing aerosol concentrations and uncertainty with elevation. Cross-validation confirms strong generalization ability with minimal variance across folds, while uncertainty quantification demonstrates reliable confidence intervals across sites. Ablation studies confirm the importance of all DBF components in achieving optimal performance. Compared with existing reanalysis and machine learning approaches, the proposed model exhibits superior predictive accuracy. Future research should integrate multi-satellite hyperspectral observations and expand DBF applicability to global regions, enabling real-time aerosol forecasting and climate impact assessment under diverse atmospheric conditions.</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

Assessing Aerosol Optical Depth and Black Carbon Levels in High Altitude Regions Validation Against MERRA-2 Data

  • 期刊论文

  • 发表日期：

    2026年6月27日

查看全文
  • R. M. Jayabalakrishnan,
  • K. Boomiraj,
  • P. Jothimani,
  • R. Murugaragavan,
  • P. Raja

摘要

Accurate estimation of atmospheric aerosols is essential for understanding air quality and climate interactions, particularly in complex mountainous regions. This study evaluates black carbon (BC) and aerosol optical depth (AOD) over the Western Ghats using ground-based observations and MERRA-2 reanalysis data, enhanced through a Deep Bayesian Filtering (DBF) framework integrating Long Short-Term Memory (LSTM) and Sequential Monte Carlo (SMC) methods. Baseline results show strong agreement between MERRA-2 and observations (BC RMSE: 0.30 μg/m3, AOD RMSE: 0.02, R up to 0.98), but performance improves significantly after DBF application (BC RMSE: 0.19 μg/m3, AOD RMSE: 0.01, R2 up to 0.97). Seasonal analysis indicates higher uncertainty during the pre-monsoon period due to dust and biomass burning, while post-monsoon conditions show improved stability. Altitude-based results reveal increasing aerosol concentrations and uncertainty with elevation. Cross-validation confirms strong generalization ability with minimal variance across folds, while uncertainty quantification demonstrates reliable confidence intervals across sites. Ablation studies confirm the importance of all DBF components in achieving optimal performance. Compared with existing reanalysis and machine learning approaches, the proposed model exhibits superior predictive accuracy. Future research should integrate multi-satellite hyperspectral observations and expand DBF applicability to global regions, enabling real-time aerosol forecasting and climate impact assessment under diverse atmospheric conditions.