Wednesday, 13 January 2016
Algorithms to derive Aerosol Optical Depth(AOD)from satellites rely on several assumptions, one of which is an assumption about the type of aerosol present in the atmosphere and its optical properties. In the Suomi National Polar Orbiting (SNPP) Visible Infrared Imaging Radiometer Suite (VIIRS) AOD retrieval algorithm, five aerosol models (dust, high absorbing smoke, low absorbing smoke, clean urban, polluted urban aerosols) are used to generate Look-Up Tables (LUTs) to invert satellite-measured radiances into AODs. The algorithm dynamically chooses a particular aerosol model based on a minimization of the multispectral reflectances based on direct radiative transfer using mainly the M3 (488 nm) and M5 (675 nm) bands. To investigate the ability of the algorithm to chose an aerosol model accurately, it is useful to compare the VIIRS Aerosol Model Index (AMI)against an alternative VIIRS based approach. Using the same satellite is useful to remove many satellite specific issues such as geolocation and observational time. In particular, a very simple band ratio approach has been developed as a way to qualitatively differentiate between clear sky, dust, and smoke episodes. The two complementary indices used are Dust Aerosol Index and Non-Dust Aerosol Index computed using reflectances measured at 412, 445, and 2250 nm. Spectral differences in aerosol absorption of dust and smoke, Rayleigh scattering, and surface reflectance allows the use measurements at these wavelengths to provide information on the presence of dust and smoke in the atmosphere. In this study, we assessed the performance of the VIIRS AMI by comparing its spatial structures to Dust/Smoke Index. Since the AMI classification is based on the optical properties of the aerosols which are Optical depth specific, it is also useful to consider the VIIRS derived AOD as a separate parameter when making these comparisons.
We used one year of VIIRS data products for standard statistical analysis including metrics based on a suitable truth table for each class including dust, clear, thick and thin smoke. The analysis focuses on treating the dust/smoke mask as a truth data set to evaluate the AMI using metrics such as Probability of Correct Detection (POCD), Accuracy, and False Alarm Ratio(FAR). Spatial and temporal analysis of the results based on one year of VIIRS data, including time series and global/regional frequency maps, will be presented.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner