A potentially universal algorithm for estimating aerosol scattering reflectance from satellite remote sensing data

- A universal algorithm of atmospheric correction for ocean color satellite data
- Based on assumption of aerosol scattering reflectance following the Angstrom law
- Use the epsilon spectrum to match the two closest aerosol models
- It removes the black ocean assumption of the standard atmospheric correction.
- It can be applied for satellite data over Case 1 and Case 2 waters.

Accuracy of aerosol scattering reflectance in atmospheric correction is crucial for satellite ocean color remote sensing. A new approach is developed to estimate aerosol scattering reflectance from satellite data based on a look-up table of in situ water-leaving reflectance. A decision rule is set to determine the water-leaving component of the total satellite-measured radiance from the table under an assumption that aerosol scattering reflectance follows the Angstrom law. The epsilon spectrum, obtained from the ratio of aerosol scattering reflectance, is used to select the two closest aerosol models for determining the corrected epsilon values, which is adjusted to new values according to the reference wavelength. The actual aerosol scattering reflectance is obtained from the mean value of the reflectance and the new epsilon spectrum. The performance of the model is evaluated using the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) imagery, showing that the component of aerosol scattering reflectance can be completely separated from the total satellite-measured radiance, and effects of non-zero water-leaving reflectance can be eliminated over highly turbid waters. The accuracy is validated by in situ measured aerosol optical thickness (AOT) using a handheld multi-band sun photometer and the Aerosol Robotic Network (AERONET) measured AOT data during the 2006 winter cruise and 2007 autumn cruise over the East China Sea (ECS). The analysis gives the mean relative errors of 23.73% and 32.04%, respectively. Our analysis shows that aerosol scattering reflectance can be significantly overestimated over turbid waters when using the atmospheric correction method based on the black ocean assumption (BOA). It may also have some errors over oceanic waters because a small 1% bias of satellite-measured reflectance in the near-infrared (NIR) bands will lead to a relatively large error of ~ 10% in aerosol scattering reflectance in Band 1. This error will spread to water-leaving reflectance, possibly leading to negative values of water-leaving reflectance. This amplifying effect of aerosol scattering reflectance can be limited by matching aerosol models using epsilon spectrum in our model. Because the assumption of aerosol scattering reflectance following the Angstrom law can eliminate effects of different types of waters including open ocean, coastal regions, lakes and rivers, our model offers a potentially universal algorithm for estimating aerosol scattering reflectance from satellite remote sensing data.