TL;DR
This paper introduces a novel hybrid method for retrieving aerosol optical depth (AOD) over land using multi-angle data from the GF-5 02 Directional Polarimetric Camera (DPC). The method combines Dark Dense Vegetation (DDV) and multi-angle techniques, achieving high accuracy (R² > 0.9, RMSE < 0.1) and improved computational efficiency compared to existing methods.
What Problem Does This Paper Address?
Accurately monitoring aerosols and particulate matter (PM) is crucial for understanding air quality and its impact on climate and human health. Satellite-based remote sensing offers a wide spatial coverage, but retrieving aerosol information over land is challenging due to the significant influence of surface reflectance. Traditional methods often struggle to disentangle surface and atmospheric signals, especially with multi-angle data which can provide richer information but are computationally intensive. This paper aims to develop a more efficient and accurate method for AOD retrieval over land, specifically for the densely populated and industrialized Jing–Jin–Ji (JJJ) region in China, which suffers from significant air pollution.
Methods
The core of this study is a novel hybrid retrieval methodology that synergistically integrates the Dark Dense Vegetation (DDV) approach with multi-angle observations. The process involves three main steps:
- Nadir AOD Estimation: Initial AOD is derived from DPC nadir-angle data. This step leverages linear parameters of land surface reflectance between blue and red bands, obtained from the MODIS MOD09 surface product. This helps to establish a baseline AOD and account for surface characteristics.
- Atmospheric Correction: After estimating the initial AOD, atmospheric correction is performed. This step aims to remove the atmospheric effects from the observed top-of-atmosphere reflectance.
- Multi-Angle Variance Analysis: The normalized reflectance at all available observation angles is then analyzed. The variance of these normalized reflectances is calculated. This variance is then used to refine the AOD retrieval, effectively exploiting the multi-angle information to improve accuracy and reduce surface influence. The method iteratively searches for the best-fitting aerosol model and AOD by minimizing this variance.
The authors also developed a specific set of 16 aerosol models tailored to the JJJ region, considering combinations of water-soluble aerosols (WASO), black carbon (BC), and dust, based on AERONET data.
Datasets
- GF-5 02 DPC Data: Multi-angle and polarized intensity data from the Directional Polarimetric Camera aboard the Chinese GaoFen-5 02 satellite were used for AOD retrieval. Data from January to June 2022 over the Jing–Jin–Ji (JJJ) region were utilized.
- MODIS MOD09 Surface Reflectance Product: This dataset was used to derive land surface reflectance parameters (specifically, linear relationships between blue and red bands) for the initial AOD estimation and for surface reflectance analysis. The 8-Day L3 Global 500 m MOD09 product was selected.
- AERONET Beijing-RADI Site Data: Ground-based aerosol optical depth measurements from the AERONET network were used for validation.
Key Insights and Ideas
- Synergistic Hybrid Approach: Combining DDV with multi-angle methods offers a powerful way to overcome the limitations of each individual technique, leading to more robust aerosol retrieval over land.
- Exploiting Multi-Angle Variance: The variance of normalized reflectance across multiple viewing angles is a sensitive indicator of atmospheric conditions and can be effectively used to refine AOD retrieval, especially in reducing surface contamination.
- Region-Specific Aerosol Models: Developing aerosol models tailored to the specific composition of the JJJ region (WASO, BC, dust) improves the accuracy of the retrieval process.
- Computational Efficiency: The proposed method achieves a significant reduction in computation time (approximately 30% faster than GRASP/MODEL), making it more practical for operational use.
Key Results
- The developed method successfully retrieved AOD images over the JJJ region.
- Validation: The retrieved AOD showed high accuracy when compared to AERONET data, with a correlation coefficient (R²) greater than 0.9 and a root mean square error (RMSE) below 0.1.
- The performance was comparable to the GRASP algorithm, a well-established method.
- The method demonstrated a computational efficiency improvement of approximately 30% compared to GRASP/MODEL.
- MODIS DDV and DB products showed less accuracy than the proposed method and GRASP.
Limitations and Future Work
- Low AOD Conditions: The paper acknowledges that the method exhibits suboptimal performance at low AOD, particularly for certain aerosol models (Models 2-4 and 8). Future work will focus on optimizing performance in these conditions.
- Surface Reflectance Uncertainty: Uncertainties in MODIS surface reflectance might affect retrieval accuracy, especially at low AOD.
- Regional Specificity: The current aerosol models are specific to the JJJ region. Future work aims to validate the method over longer time spans and in more diverse geographical regions.
- Calibration Uncertainties: The paper highlights that calibration uncertainties in DPC data can impact retrieval accuracy, suggesting that minimizing these uncertainties is crucial for future improvements.
- Future Enhancements: The authors plan to incorporate adaptability to polarized data, built-in calibration optimization, and simultaneous retrieval of aerosol component proportions into future versions of the method.
Summary updated