Most air quality assessments have been traditionally conducted within a framework that simulates only criteria air pollutants (CAPs). In recent years, increasing attention has been given to the development and application of a multi-pollutant (MP) framework, including both CAPs and air toxics that are classified as hazardous air pollutants (HAPs). An air quality modeling system with MP treatments and interactions provides an essential tool to replicate the complex atmosphere. To apply and evaluate a 2002 MP modeling platform, several annual simulations have recently been conducted by the U.S. EPA's Office of Air Quality Planning and Standards using a MP version of the U.S. EPA's Community Multi-scale Air Quality (CMAQ) modeling system v4.6.1 with horizontal grid resolutions of 12-km over the Eastern U.S. (EUS) and 36-km over the continental U.S. (CONUS). Model evaluation for concentrations and depositions of ozone (O3) and its precursors, fine particulate matter (PM2.5) and its components, mercury, and 34 other HAPs has been performed in terms of spatial distributions, scatter correlations, and temporal variations with available surface monitoring networks. In this study, the model performance will be further examined by evaluating CMAQ column predictions against available satellite measurements for 2002 simulations at 12-km and 36-km grid resolutions. The satellite data will be taken from the Measurements of Pollution in the Troposphere (MOPITT), the Total Ozone Mapping Spectrometer (TOMS)/the Solar Backscattered Ultraviolet (SBUV), the Global Ozone Monitoring Experiment (GOME), the SCanning Imaging Absorption SpectroMeter for Atmospheric CHartographY (SCIAMACHY), and the MODerate-resolution Imaging Spectroradiometer (MODIS). The discrepancies in the spatial distributions as well as seasonalities between the observed and simulated column abundances will be analyzed. The model biases due to the uncertainties in the model inputs, formulations, and satellite data retrievals, the possible sources of such biases, as well as potential model improvements will also be discussed.