This dissertation is a collection of studies focusing on improving our understanding of atmospheric aerosols using both observational data and model simulations. EOF analysis of Aerosol Index (AI) product from Total Ozone Mapping Spectrometer (TOMS) and Ozone Monitoring Instrument (OMI) reveals global distribution of absorbing aerosols, with major sources lying in Sahara deserts, the Sahel region, South America and South Africa. Analysis of aerosol Single Scattering Albedo (SSA) data from AErosol RObotic NETwork (AERONET) further indicate trends in SSA over a number of globally distributed stations, which might be associated with changes in aerosol composition and thus their optical properties. More importantly, the changes in SSA alter the radiative forcing of aerosols. They may also potentially impact satellite retrievals of aerosol properties as generally a constant SSA is assumed in the retrieval algorithms. In order to assess satellite retrieved aerosol optical properties, collocated pixel level Aerosol Optical Depth (AOD) and Ångström Exponent (AE) data from MODerate resolution Imaging Spectroradiometer (MODIS) are compared with AERONET measurements over 10 stations representing typical aerosol regimes. The results show that while MODIS AOD well agrees with AERONET in both the magnitude and seasonal variability for all stations, comparatively large discrepancies are found in the AE, especially for over land. Further investigation reveals that the dependence of the AE on AOD for MODIS data are quite different from AERONET data, which suggest problems in the aerosol models used in MODIS retrieval. MODIS ocean data are generally reliable. Focusing on ocean data, a strong correlation between the AE and ENSO index has been found, and the roles of relevant physical mechanisms are discussed. While the exact cause of the correlation is still unclear, the results indicate aerosol properties can be influenced by major climate modes such as ENSO. The sensitivity of aerosol Direct Radiative Forcing (DRF) to perturbations of major aerosol parameters are tested using the GISS GCM. Among the three perturbed parameters, AOD, SSA and asymmetry parameter g, DRF appears to be most sensitive to SSA. Moreover, changing aerosol dry sizes result in larger fluctuation in DRF than the previous three parameters. Based on the sensitivity studies, an optimal fitting technique based on AERONET data is developed to better constrain aerosol dry size parameterization in the GCM. Model results for AOD and SSA are also improved by adjusting the size and applying “uncertainty parameters”. The fitting results indicate an overall underestimate in GCM aerosol loading. In particular, aerosol absorption has been underestimated by approximately a factor of 2. The low bias might be attributed to insufficient aerosol mass loading, lack of internal mixing of black carbon with other species, etc. After incorporating the optimized sizes and uncertainty parameters into the GCM, estimated global mean DRF is significantly larger than the original aerosol field. Regionally the changes in DRF are more diverse due to the relative fraction of absorbing and non-absorbing aerosols. The method still has limitations. Further improvements are required including examining the fine/coarse aerosol fraction, better identifying the absorbing species, and using advanced observations with global coverage.