The main objective of the thesis is to identify the optimal set-up for future satellite gravimetry missions aimed at monitoring mass transport in the Earth’s system.The recent variability of climatic patterns, the spread of arid regions and associ- ated changes in the hydrological cycle, and vigorous modifications in the ice coverage at polar regions have been attributed to anthropogenic influence. As such, it is important to continue monitoring the Earth system in order to properly constrain and improve the geophysical and climatic models and to better interpret the causes and consequences of climate change. Satellite gravimetric data are also exploited to further the knowledge on other geophysical processes with high societal and scientific impact, such as megathrust earthquakes, drought monitoring and Glacial Isostatic Adjustment (GIA). The primary focus of the study is to properly quantify the errors in the gravimetric data to be collected by future gravimetric satellites, in particular those related to the measurement of the temporal gravitational field variations. One source of errors comes from the background force models describing rapid mass transport processes; another error source is related to the background static gravity field model. These models are used to complement geophysical signals that are missing or improperly represented in the gathered satellite gravity data. However, they are built on the basis of in situ data that lack global coverage and, therefore, suffer from a limited accuracy (particularly in remote areas). Although the fidelity of these models is constantly improving, the satellite data accuracy is also increasing with the on-going technological and methodological advances. etermining the net effect of these conflicting trends is the main driver to study the propagation of errors in background models into the estimated models. Other sources of errors arise from imperfections of the on-board sensors, such as the ranging sensor or the Global Navigation Satellite System (GNSS) receiver. The influence of the sensors errors is divided into the major independent contributions, with the corresponding frequency description, and assembled into a detailed noise model. The model predicts the effects of i) the inaccurately known orbital positions, ii) the noise in the inter-satellite metrology system, iii) the noise in the on-board accelerometers, iv) the wrongly-estimated Line of Sight (LoS) frame accelerations resulting from errors in the radial orbital velocities, and v) errors in the orientation of the LoS vector. The model has been validated with the help of actual Gravity Recov- ery And Climate Experiment (GRACE) a posteriori residuals, which are compared to the output of the noise model considering a simulated GRACE mission. Therefore, once the assumptions describing sensor and model accuracies are modified to reflect those predicted for future gravimetric missions, it is reasonable to expected that this noise model reproduces realistic errors for those missions. Also relevant is the analysis of the sensitivity of the data in terms of isotropy. As learned from the GRACE mission, the nearly-constant North-South alignment of the measurement direction makes the data less sensitive to gravitational changes along the East-West direction. Although formally not an error itself, the anisotropic data sensitivity amplifies the errors in the data. The sensor and model errors are propagated firstly to the gravimetric data and further to the gravitational field, in full-scale simulations of the cartwheel, trailing and pendulum satellite formations. The results are analysed in terms of i) the observation error in the frequency domain and ii) the estimated gravity field model error in the frequency and spatial domains. The error budgets for these formations are also quantified. The results indicate that the pendulum formation with no along-track displacement is least sensitive to model and sensors errors, in particular to temporal aliasing. The conducted study reveals serious limitations in the cartwheel mission concept, since the orbit errors are considerably amplified by the diagonal components of the gravity gradient tensor, while the pendulum and trailing formations are only affected by (small) off-diagonal components. The spatial error patterns provide valuable clues on how to best combine the different formation geometries in order to produce minimum anisotropy in the sensitivity of collected data. The data from the pendulum formation show some anisotropic sensitivity but the combination of such data with those from a trailing formation, such as the GRACE Follow On (GFO), would eliminate this disadvantage (as well as the low accuracy near the poles of the pendulum formation). Unlike alternative proposals for dual-pair satellite missions, such as the Bender constellation, the dual trailing/pendulum constellation would provide global coverage in case of failure of one satellite pair and dense temporal sampling at high latitudes. Furthermore, the data from gravimetric missions are shown to benefit greatly from the data gathered by numerous non-dedicated satellites. From the conducted simulations, it is predicted that the achievable temporal resolution is increased to a few days for the degrees below 10 and, crucially, with no significant level of temporal aliasing. Longer estimation periods allow for higher degrees to be estimated, with greatly reduced effects of temporal aliasing in the resulting gravity field models.