Satellite_Aerosol_Algorithms

Deriving aerosol properties (aerosol size and optical depth) from satellite requires that one know the aerosol type which is present. This is due to the changes in scattering and absorption properties for different aerosol types. When the particles are smaller than the observing wavelength then the aerosols scatter light in all directions. When the aerosols are relatively large compared to the observing wavelength then more light is scattered in the forward direction. The aerosol phase function describes how light is scattered at different angles and we have used a polar nephelometer to measure this aerosol property. Typically a limited set (10-20) of aerosol types (size and composition) are assumed and satellite measured radiances at several wavelengths are used to select the best aerosol type for each case. This typically involves the use of a "black box" lookup table approach but other pseudo-linearized approaches have been used as well (Porter, PhD, 1993). These remote sensing approaches depend heavily on having realistic aerosol models. For out work we have used the aerosol models of Porter and Clarke (1997) which we believe are fairly realistic in terms of their size distribution. 

In addition to aerosol uncertainties, the retrieval of accurate aerosol properties depends on correct calculation of the surface reflection. Over the ocean (away from sun glint) this requires accurate wind estimates. Over the open ocean large scale wind field weather models provide an accurate estimate of the wind speed and direction. Near coastal regions, large scale wind fields are not accurate and smaller meso-scale models are required. The figure below shows a MODIS image (at 250m resolution) which we have processed to obtain aerosol optical depths. For these calculations we have used meso-scale wind fields from various models. Although further tests are certainly needed, our initial tests suggest the meso-scale models are not sufficiently accurate near Hawaii. Based on varoius reasons (not discussed here), we believe the meso-scale wind models are not sufficiently accurate due to poor model initialization. We are now collaborating with various investigators to implement a novel approach to obtain improved wind fields which can be incorporated into better initialize the wind models.