Our study focuses on the avulsion process, the key for the depocenter migration of stacked channel-levee systems in the deep-sea. To present, the avulsion process has been explained as the abandonment of a channel (parent channel) and the birth of a new one (new channel). A true indice of avulsion has been given by the detection of the HARPs, a seismic term known since the 80s for high-amplitude reflection packets at the base of new channel-levee systems. The HARPs are suggested, in analogy to the studies from ODP Leg 155 of the Amazon Fan, to be sandy deposits in lows adjacent to breached parent channels. Our present work, by re-visiting the Amazon Fan, highlights the great complexity of the HARPs architecture (at the scale of multi-channel high-resolution seismic data) and their lithological heterogeneity. The study of some specific zones within the Zaïre Fan allowed to reveal both the earliest stages of an avulsion process related to levee breaches and first splays linked to unconfined flows on the levee, and the latest stage of this process, with the first levee deposition and evolution following the HARPs deposition. These results provide new insights for the understanding of the processes involved during an avulsion (levee breach, HARPs accumulation, levee aggradation) and clues to refine the previous conceptual models and to reach a higher resolution knowledge of this process, by describing the inter-relationship of the different architectural sub-bodies of the HARPs and giving details on both the earliest and latest stages of the avulsion.
