The German CHAMP, US/German GRACE, and European Space Agency (ESA) GOCE and Swarm Earth Explorer satellites have provided a data set of accelerometer observations allowing the derivation of thermospheric density and wind products for a period spanning more than 15 years. With the advent of highly accurate satellite accelerometer measurements, the neutral density and wind characterization has been significantly improved. These observations provided detailed information on the thermospheric forcing by Solar Extreme Ultraviolet radiation and charged particles, and revealed for the first time the extent of forcing by processes in lower layers of the atmosphere. Because the focus of most of previous research was on relative changes in density, the scale differences between the CHAMP, GRACE, GOCE and Swarm data sets, so far, have been largely ignored. These scale differences originate from errors in the aerodynamic modelling, specifically in the modelling of the gas-surface interactions (GSI) of the satellite. Once detailed 3D geometry models of these satellites are available, the key parameters to describe the satellite aerodynamics can be estimated by cleverly making use of variations in satellite orientation and simultaneous observations by multiple satellites. The first step for obtaining more consistent density and wind data sets consisted of meticulously modelling the satellite outer surface. For this dissertation work, this was done by collecting information from technical drawings and pre-launch pictures, and generating a CAD model of the selected satellites. In the following phase, these geometries were given as input to a rarefied gas-dynamics simulator. The Direct Simulation Monte Carlo approach was used with the SPARTA software to compute the force coefficients under different conditions of satellite speed, atmospheric temperature and local chemical composition. Once all the mission scenarios had been simulated, an aerodynamic data set was generated and applied in the processing of satellite accelerations into thermospheric density and wind data products. To this aim, the Near Real-Time Density Model (NRTDM) software, developed at TU Delft, was used. The data were generated from accelerometer observations and, when necessary, with the help of GPS-based accelerations estimated by a Precise Orbit Determination (POD) technique. Multiple comparisons were performed with empirical and physics-based models. This helped in determining for which conditions the models are performing better, and also which models’ features would need further development. In the second step, the interaction between atmospheric particles and satellite surfaces was investigated. The way in which atmospheric particles collide with the satellite surfaces have a large influence on the satellite aerodynamic forces and, if proper assumptions are not implemented, can produce large discrepancies in the final thermospheric products. Initially, the GSI assumptions were selected in agreement with the fully diffusive reflection mode. This assumption was adopted to exclusively investigate the geometry modelling influence on thermospheric products. Later, to cover also this research area, multiple simulations described different reflection modes. A wide range of GSI parameters was investigated, and more optimal values were found allowing the derivation of new consistent thermospheric products. Within this study, the energy accommodation coefficient, which describes the energy exchange between particles and satellite surfaces, played a crucial role. Although the value of 0.93 is used commonly in the literature, in this study lower values were identified as optimal. Indeed, a value of 0.82 for the GOCE satellite, and a value of 0.85 for the Swarm and CHAMP satellites have been found to provide more consistent thermospheric data. This resulted in new improved thermospheric density and wind data sets, which have been made available to the scientific community. Among the possible applications, these data can be used for data assimilation for improving current atmospheric models. Resolving the problem of deriving the true absolute thermosphere density scale from satellite dynamics measurements improves orbit predictions for the space debris population and its long-term evolution. Moreover, the new capabilities for computing more consistent drag, density and wind, can also be exploited for future missions that are currently in the design phase.