Analysis of different aerosol plume products in order to identify a useful Smoke Indicator (SI) and applying the SI and Satellite AOD from VIIRS as inputs of a Neural Network to predict surface level PM2.5

Particulate Matters, or PM, are small particles that are found in the air, including inorganic sulfates, nitrates as well as biomass products such as smoke. Fine Particulate Matter with diameters < 2.5 microns are called PM2.5 and they are believed to pose significant health risks. Surface sampling of PM2.5 can be quite expensive and limited number of existing ground networks makes it difficult to monitor PM2.5 on a 24 hours basis. Satellite detected column integrated Aerosol Optical Depth (AOD) can be useful in predicting PM2.5 concentration. Studies have found linear relationship existing between AOD and PM2.5. However, this strongly depends on aerosol properties, PBL height, meteorological conditions, absence of aloft plumes etc. Studies have found that this relationship does not work well in the presence of aloft plume. Identifying and quantifying smoke plumes is important for better interpretation of the linkage of passive satellite observations of aerosol optical depth (AOD) and surface aerosol concentration (PM2.5). In this research work, we study smoke plume products such as VIIRS Aerosol Model Index (AMI) and dust/smoke mask product, GOES ASDTA smoke product, NAAPS smoke plume products for summer of 2010-2012 and apply them to filter out aloft smoke contaminated cases to improve relationship between VIIRS AOD and surface level PM2.5 over New York State. We also identify a useful smoke indicator (SI) from these smoke tools and apply the SI and satellite AOD as inputs of a Neural Network in order to predict surface PM2.5.