Soils play an important role in the carbon (C) cycle. They constitute an important part of the global C pool, by storing C in de forms of organic matter and carbonates. The soil micro-organisms respiration transforms organic carbon (OC) into CO2. Two considerable concerns exist in relation with this: the global climate change due to the greenhouse gas warming potential of CO2; and the decrease of soil quality. This thesis aims to consider the hillslope as elementary unit to estimate CO2 emissions from soils, and to provide a detailed mechanistic understanding of the processes which regulate it at the landscape scale. Our study was carried out on a cropland of the Belgian loess belt. In-situ measurements of respiration and of its controlling variables were conducted during 3 years along a 150 meters eroding hillslope, both at the surface and along soil profiles. We demonstrate that, accounting for soil OC dynamics only at flat landscapes would under-estimate annual soil-atmosphere CO2 exchanges by c. 20 %, as the respiration rates are ca. 50% higher at the footslope relative to the summit. We show that downslope, while OC is only weakly stabilized by organo-mineral associations, the high soil moisture strongly control OC stabilization. Also, we highlight that soil microorganisms in these stabilizing wet conditions have a higher temperature sensitivity. Hence, decomposition rates of this downslope buried OC can potentially increase under future dryer and/or warmer environmental conditions, making this large store become an increasing C source to the atmosphere. Given the predicted climate changes, this is a topic at which scientists and policy makers should look closely and carefully.