High spatial resolution aerosol and surface reflectance retrieval and validation using Sentinel-2

Aerosols may vary spatially quite rapidly in an urban environment, but present aerosol products fail to detect them due to limitations in coarse spatial resolution. Aerosol dispersal can be mapped at local scales using Sentinel-2 with relatively high spatial (10, 20, and 60 m) and temporal (5 days) resolutions. A new high-resolution (60 m) coupled aerosol-surface reflectance retrieval algorithm Modified Sensor Invariant Atmospheric Correction (MSAIC) has been developed to address the urban air pollution problem from Sentinel-2. Sentinel-2 retrieved AOD products validated against Aerosol Robotic Network (AERONET) (R²= 0.830, and RMSE= 0.156) and MODIS (R²= 0.655, and RMSE= 0.240) AOD products. For light to medium aerosol loading (AOD < 0.2), it was largely successful in extracting AOD with uncertainties <0.10. Additionally, MSIAC produces precise surface reflectance estimation at 60 m resolution across the 13 band of Sentinel-2. This is important as the accuracy of satellite-retrieved AOD is determined by surface reflectance correctness. Thus, it is crucial also to create an accurate estimation of surface reflectance. In the absence of in-situ observations and a Radiometric Calibration Network (RadCalNet) over India, two indirect approaches were used to verify Sentinel-2 retrieved surface reflectance products: (a) Sensitivity analysis , and (b) the use of invariant targets. Little or no change in surface reflectance was observed for different aerosol concentrations, and insignificant change in surface reflectance was observed for invariant targets. Additionally, Sentinel-2 retrieved surface reflectance was validated using observations obtained from the radiometric calibration network RadCalNet over La Crau (France) with uniform landscape and low AOD. Results for this validation will be presented to demonstrate the quality of this Sentinel-2 analysis compared to previous results. For further improvement of the algorithm, more investigation is required over the in-situ sites with varying (high) AOD concentration, less uniform and low reflectance landscapes, and not just desert sites like Gobabeb and Psedo-Invariant Calibration Sites (PICS). We argue that co-located global networks of continuous ground monitoring stations are required simultaneously characterising surface reflectance and aerosol over a range of surface and atmospheric conditions. Such a network would allow thorough quality evaluation of satellite retrieved products conducted over land.