Magnetic clouds (MCs), described as large-scale structures with a magnetic flux-rope topology, interact with the surrounding interplanetary medium during their propagation. McComas et al. (1988), and later Dasso et al. (2006), were the first to propose that magnetic reconnection might peel off the outer magnetic structure of ICMEs. This erosion process is the core subject of this thesis. First, we confirm the occurrence of MC erosion, thanks to a multi-spacecraft study combining a set of key signatures expected from this process. Based on careful determination of the MC main axis, we estimate the amount of magnetic flux eroded by analyzing the azimuthal flux imbalance during the spacecraft sampling of the flux rope. Consistent with this process, we also find clear magnetic reconnection signatures at the MC front boundary. Finally, we investigate the characteristics of suprathermal electrons in the back region of the MC. Those electrons are shown to signal large-scale topological changes also expected from the erosion process. In a second part, a statistical analysis of all MCs of solar cycle 23 is performed in order to quantify this process. We show that the process is very frequent. It often occurs in large amounts, and at both the front and back boundaries of MCs. The statistics of reconnection jets at these boundaries confirms the significance and high recurrence of the process even up to 1 AU. In a last part, since the erosion mechanism can lead to the removal of part of the southward oriented magnetic field at the front or the rear of MCs, we study the potential impact of this mechanism on their geo-effectiveness by using a standard MC model and an empirical model of the ring current strength. We also estimate the radial evolution of this process based on simple models. We conclude that most of the erosion is expected to occur within Mercury's orbit.
