A new MFRSR data analysis algorithm is presented. Our earlier algorithm assumed a monomodal aerosol size distribution, while the new algorithm allows us to partition the aerosol optical thickness into fine and coarse aerosol modes. In addition, we retrieve the fine mode effective radius and Angstrom exponent. A bimodal gamma distribution is used to describe the aerosol particle size distribution. The algorithm has been tested using a multi-year data set from the local MFRSR network at the DOE Atmospheric Radiation Measurement (ARM) Program site in Southern Great Plains (SGP). Our retrieved aerosol optical thicknesses (total, fine, and coarse) are compared with the corresponding AERONET almucantar retrieval results derived from a CIMEL sunphotometer co-located with the MFRSR at the SGP Central Facility. A constrained variant of the algorithm (zero NO2 column values) has been used to define the range of physically justified values of the fine mode effective radius, and for comparison with AERONET particle size retrievals. We use the multiple MFRSR measurements obtained during the year 2000 at the SGP Extended Facilities to examine geographical and seasonal variability of aerosol properties. A correspondence has been found between the geographical variation in the fine mode particle size and aerosol composition (nitrates versus sulfates) as measured by National Atmospheric Deposition Program. We similarly find good correspondence between our retrieved aerosol sizes and the PM2.5 to PM10 ratios obtained from EPA AirData monitoring. Examination of the data from the SGP Central Facility obtained for the period 1993--1997 reveals a decreasing trend in coarse mode aerosol optical thickness during the 1993--1995 period, consistent with the decay of the stratospheric aerosol following the 1991 eruption of Mt. Pinatubo. In contrast, fine mode optical thickness exhibits only seasonal variability with summer maxima during this period.