The fourth report of the Intergovernmental Panel of Climate Change (IPCC) unambiguously indicates that the climate is changing. Climate change is induced by changes in atmospheric abundances of greenhouse gases and aerosols, solar radiation and land surface properties, which all alter the radiative balance of the Earth. Among them, greenhouse gases and aerosols are likely the compounds affecting the climate the most. While greenhouse gases warm the Earth, aerosols influence its radiative balance by scattering and absorbing solar radiation (the direct effect), and by modifying cloud amount and properties (the indirect effect). The radiative forcing of aerosols on the climate is however the less understood among the various forcings currently considered in the IPCC assessment. This is likely related to the lack of comprehensive and accurate observations of aerosols (and in particular of their vertical distribution) which thus makes global aerosol models difficult to constrain. Recent model intercomparisons have indicated that different assumptions regarding aerosol emissions, formation and growing properties, and removal processes generate large diversities within models. These large diversities are found in particular in the Arctic and the upper troposphere region where extreme surrounding conditions take place, limiting instrument sensitivity and accuracy. The objectives of this thesis are to better quantify the processes driving the aerosol distribution in regions where the uncertainties are the largest, including the Arctic region and the upper troposphere, and to assess the quality of CALIOP observations. For this purpose, we used the fully coupled global climate-aerosol-chemistry ECHAM5-HAMMOZ model conjointly with a suite of remote sensing observations including the CALIOP satellite instrument, which provides the vertical distribution of aerosols. The model predicts the size distribution and composition of aerosols as well as the number concentration of cloud droplets and ice crystals. We first investigated the processes that drive the transport of soluble and insoluble compounds toward the Arctic in the ECHAM5-HAMMOZ model. Recognizing that scavenging processes may be an issue in global models, we "re-visited" their properties in ECHAM5-HAMMOZ. We find that the model better reproduces the aerosol vertical distribution in the northern mid- and high-latitudes, especially in the free troposphere, by decreasing aerosol scavenging coefficients in the model. Using smaller aerosol scavenging coefficients results in an increase of aerosol burden and lifetime, especially in the Arctic. In addition, the Arctic haze phenomenon is better represented in the simulation using decreased aerosol scavenging coefficients. The relative contributions of different processes governing the transport of aerosols from the midlatitudes toward the Arctic together with the relative contributions of different geopolitical source regions were then quantified. We find that Europe and Siberia contribute the most to the Arctic pollution even though Asian anthropogenic emissions are much larger. The model suggests that BC emitted in Siberia and Europe is ten times more efficiently deposited onto the Arctic than BC emitted in Asia. We next investigated an aerosol layer that forms in the vicinity of the tropical tropopause layer (TTL) over the southern Asian and Indian Ocean region. Aerosols in this layer are observed with the CALIOP instrument even though ice clouds partly mask their signal. The aerosol layer follows the Inter Tropical Convergence Zone (ITCZ), which indicates that these aerosols are likely transported during convective processes. The ECHAM5.5-HAM2 model reproduces such an aerosol layer in the TTL over the same region but clearly overestimates CALIOP data. Model results suggest that aerosols in this layer are mostly sulfate particles transported during convective processes and originating from natural sources. The shortwave forcing of this aerosol layer is estimated. Finally, we assessed the accuracy of aerosol extinction retrievals from the CALIOP satellite instrument by comparing them with remote sensing observations such as AERONET, MODIS and MISR over the last four years. Overall, standard CALIOP products underestimate by 50% observed aerosol extinctions. However, this comparison is substantially improved by screening CALIOP data. The comparison of the CALIOP screened product with the ECHAM5.5-HAM2 model shows large disparities in model results. The model reproduces quite well the observed interannual variability and vertical distribution of aerosols in Europe and North Africa while it underestimates them in America and Asia. This is likely due to an underestimate of dust and anthropogenic emissions in North America and Asia.