The past existence of a giant ice stream off the northern end of the Antarctic Peninsula is suggested by a convex-upward, elongated sediment body now at a mean water depth of 1000 m. Because of its morphological characteristics, i.e., a set of parallel to subparallel ridges and grooves to 100 km long, with an overall width of 25 km, we call this depositional body a bundle structure. We hypothesize that bundle structures form by accumulation of basal deformation till under their parent ice streams. From its location, size, and overall characteristics the bundle structure described here constitutes the best preserved, largest, deepest, and relatively low latitude evidence of giant ice streams that flowed offshore Antarctica during glacial maxima. Bundle structures reveal the very dynamic behavior of ice caps in the northern Antarctic Peninsula during the last glacial maximum, with catchment areas draining rapidly under marine-based critical subglacial conditions. Bundle structures represent a new megascale streamlined glacial landform.

[1] Submarine landslides are ubiquitous along Mediterranean continental margins. With the aim of understanding mass-wasting processes and related hazard at the scale of a large marine basin encompassing multiple geological settings, we have compiled data on their geometry, age, and trigger mechanism with a geographic information system. The distribution of submarine landslides in the Mediterranean reveals that major deltaic wedges have a higher density of large submarine landslides, while tectonically active margins are characterized by relatively small failures. In all areas, landslide size distributions display power law scaling for landslides > 1 km³. We find consistent differences on the exponent of the power law (θ) depending on the tectonic setting. Available age information suggests that failures exceeding 1000 km³ are infrequent and may recur every ~40 kyr. Smaller failures that can still cause significant damage might be relatively frequent (failures > 1 km³ may recur every 40 years). The database highlights that our knowledge of submarine landslide activity with time is limited to a few tens of thousands of years. Available data suggest that submarine landslides may preferentially occur during lowstand periods, but no firm conclusion can be made on this respect, as only 70 landslides (out of 696 in the database) have relatively accurate age determinations. The temporal pattern and changes in frequency-magnitude distribution suggest that sedimentation patterns and pore pressure development have had a major role in triggering slope failures and control the sediment flux from mass wasting to the deep basin.