Many sophisticated general circulation models which include chemistry and transport mechanisms for trace gases and aerosols have been developed in the past several years. At the same time, the methods available for assimilating observed data into numerical weather prediction models of all types has grown immensely. The lack of growth has been in the intersection of the two areas. The most sophisticated forecasts of chemical trace species typically rely on initial conditions which are no more than the previous forecast. This forecast is most often initialized from a climatological estimate of trace species emission. A three-dimensional variational data assimilation system has been developed to address the issue of initial state estimation for aerosols. The assimilation seeks to find a maximum-likelihood analysis based on MODIS level 2 aerosol optical depth and model background from the Weather Research and Forecast model with MADE/SORGAM aerosol chemistry. The resulting analyses verify more favorably with aerosol optical depth from AERONET and AVHRR PATMOS-x than do WRF/CHEM forecasts without assimilation.