Abstract : Assessment and comprehension of gully erosion dynamics at their temporal and spatial scale in Badlands (marl's geology, south of French Alps need fast and easy reproductible methods of characterization of these phenomena. We tested two approaches of spatial observation in order to accomplish this goal: an aerial approach by stereophotogrammetry from unmanned aerial vehicle (drone) images, and a terrestrial approach by ground laser scanner. For the aerial approach an adequate combination of ground preparation, systematic errors compensation (resulting from image acquisition) and an image matching strategy enabled us to obtain a drone DTM with resolution, accuracy, and hydrologic coherence within the limit of the data (average image ground resolution 3 cm). In regard to the terrestrial approach, results of LiDAR measurement test on different geometric surfaces show that the noise is of 1 cm from a measure distance of 30 m. The application of a simple average filter in a regular grid of 1 cm makes it possible to reduce the noise and also to recreate geometric forms of more than 3 cm large. The "advantage-constraints" ratio between these two approaches shows us that the aerial approach produces DTM with better extent and continuity, while the terrestrial approach produces more precise and detailed DTM. These results reveal the enormous potential of drone images for the development of inexpensive DTM. In addition, ground LiDAR arises as an interesting topographic surveying system for the monitoring of elementary processes of gully erosion, allowing for very detailed measurements in space and time.