The above summary of the characteristic features of subaerial rock-avalanche deposits resulting from giant landslides, clearly indicates many similarities between the Site 899 serpentinite breccias and such deposits. These similarities include the presence in both of (1) angular fragments with jigsaw/crackle textures, (2) a general lack of sorting, compatible with very rapid, essentially instantaneous formation, but with larger boulders occurring near the upper surface of the units, (3) a matrix generated by fragmentation of the clasts and a size continuum between the clasts and matrix, and (4) internal deformation zones or slip zones. In addition, rock-avalanche deposits are generally tabular units that are surprisingly extensive. Observation at Site 899 and the general stratigraphic framework for Unit IV show the serpentinite breccias to be interbedded with sedimentary units. Biostratigraphic ages suggest a normal stratigraphic succession and therefore the breccias are most simply interpreted as normal bedded units. Overall, the similarities are sufficiently striking to warrant interpreting the serpentinite breccias as rock-avalanche deposits resulting from giant landslides. This interpretation raises a number of issues and these are considered next.
Large subaerial landslides are generated in areas of significant topographic relief. Many have been reported from fold-mountain belts, particularly in regions such as the South Island of New Zealand and the European Alps where glacial erosion has generated over-steepened slopes along the sides of glacial valleys (Craw et al., 1987). However, large landslides are also particularly common in regions of extensional tectonics, slope failure often being associated with active fault scarps. There are many accounts of such deposits from the Basin and Range province in the southwestern United States although only in some cases (e.g., Woodford and Harriss, 1928; Longwell, 1951; Burchfiel, 1966; James et al., 1993) can megabreccia deposits be directly demonstrated to have a landslide origin.
The Iberia Margin was a region of extensional tectonics at the time of emplacement of the serpentinite breccia units, and, by analogy, their generation in this area at that time was to be expected. The presence of single rock-avalanche deposits as thick as 90 m requires considerable relief with perhaps 500-m- or even 1-km-high fault scarps comparable to those found in the Basin and Range province at the present time. Basement relief on this general scale exists at present beneath the sediments of the Iberia Abyssal Plain. However, the present basement relief cannot have generated the serpentinite breccia units as the units now form the top of one of the basement highs, whereas they must have formed at the foot of a major escarpment. This in turn implies significant deformation in the area after the formation of the breccia units and prior to the start of the deposition of the sediments that mantle the basement. A similar problem arises in the interpretation of the mass flow deposits of Unit IV recovered in Holes 897C and 897D. One possibility at both Site 897 and Site 899 is that an extensive (possibly tabular but unconstrained) deposit on a fault block terrace was subsequently back-rotated, raising the leading edge to create the existing topographic highs. Unfortunately, we cannot accurately document the bedding and show it to be tilted. A more fundamental problem is whether the breccia units were formed in a subaerial or submarine environment. All the published accounts of giant avalanche deposits cited above describe rock units that were thought to have been emplaced in subaerial conditions. Indeed, some authors consider that entrapped air was an important element contributing to the fluidity of the rock avalanches, thus accounting for their tabular nature. However, the presence of fine-grained sediments between the Site 899 serpentinite breccias, and the occurrence in the sediments of marine microfossils strongly suggests that the breccias were emplaced in a submarine environment. Certainly, submarine landslides do occur (Lipman et al., 1988; Moore et al., 1989; Cochonat et al., 1990; Holcomb and Searle, 1991) and possibly pore waters or entrapped seawater contribute to the fluidity of the rock avalanches and explain in some cases the long distances traveled. Tucholke (1992) has presented persuasive geomorphological evidence for the occurrence of a massive submarine rockslide in the rift-valley wall of the Mid-Atlantic Ridge. However this unit has not been drilled and the internal textures and fabric of this submarine rockslide are unknown. It is noteworthy that the region is again one of extensional deformation. We agree with Tucholke who argued that "there is no a priori reason why flowslide-like phenomena should not occur in the submarine realm."
The striking similarity of the fabrics developed in the submarine serpentinite breccias to fabrics developed in demonstrably subaerial deposits is perhaps surprising. The high fluidity of the deposits suggests that, in both environments, they were emplaced in a fluidized state. An explanation of the similarity may be that the fundamental process involved in the fluidization is independent of the properties of the medium, water replacing air in the submarine environment. Melosh (1979) has proposed that large landslide deposits may suffer "acoustic fluidization" induced by a transient strong acoustic wave field, and shows that acoustically fluidized debris may behave as a newtonian fluid with a viscosity in the range of 105-107 P. Such an explanation can also account for the long runout landslides on the moon (Howard, 1973), where neither water nor air can have been effective in generating the fluidized mass.
The source for the serpentinite breccias was presumably a fault scarp exposing massive serpentinite. Such a source would naturally explain the serpentinite bulk composition of the breccia composition and the absence of sedimentary, pelitic-metamorphic, or granitic clasts. Serpentinites are structurally weak and particularly susceptible to slope failure along escarpments. The trace-element data presented in the Site Report (Shipboard Scientific Party, 1994, table 8) indicate that the bulk composition of the serpentinite was similar to that of the serpentinite exposed at Site 897. The similar bulk composition of source and breccia indicates that probably little material was entrained during the flow and emplacement of the breccia units—another characteristic feature of rock avalanche deposits. More importantly, drilling at Site 899 suggests that during the early extensional history of the Iberia Margin, serpentinite did not just form a narrow ridge close to the ocean/continent transition, but outcropped along a zone perhaps as wide as 20 km—the approximate distance between Sites 897 and 899.
Large serpentinite escarpments also occur in transform fault zones and serpentinite breccias, somewhat similar to those being considered here, have been recovered from some of these zones. Bonatti et al. (1974) described a dredged suite of ultramafic-carbonate breccias from the Romanche and Vema fracture zones in the equatorial Atlantic. The samples are quite varied in texture and, in most, the volume occupied by calcareous cement is roughly equal to that occupied by serpentine. The original authors argued that the textures indicated that the samples were brecciated prior to calcitization and that the brecciation was probably tectonic in origin. This is certainly possible, but we note that in situ brecciation of serpentinites at Site 897, an area selected as not being close to a fracture zone, is interpreted as generating rocks with more than 50% calcium carbonate. Some of the extensively calcitized rocks there show no evidence for prealteration brecciation of the serpentinite; the brecciation is interpreted as an integral part of the calcitization process (Morgan and Milliken, this volume).
In contrast, in a wide-ranging review, Lockwood (1971b) suggested that many fragmental serpentinite units were deposited very rapidly by submarine landslides, mudflows or turbidity currents. Later, Bonatti et al. (1973) described some "sedimentary serpentinites" from the Mid-Atlantic Ridge and concluded that they were indeed of sedimentary origin and that they were probably emplaced by gravity sliding, slumping, and turbidity current transport of serpentinite debris originating from the upper parts of ultramafic transverse ridges at the Romanche Fracture Zone. Most of the dredged samples described are finer grained than the very coarse breccias at Site 899.
Bonatti et al. (1974) also stressed the similarity of rocks described from the Romanche and Vema fracture zones to ultramafic breccias found in ophiolite occurrences in the Apennines and the South Pennine nappes of eastern Switzerland. The description by Bernoulli and Weissert (1985) of relations near Davos, Switzerland are very reminiscent of relations at Site 897 with sediments overlying brown calcitized, brecciated serpentinites, in turn overlying green massive serpentinites. The Davos breccias are described as polyphase in origin and forming in a transform zone. Barrett and Spooner (1977), describing similar ophiolitic breccias from the East Ligurian Apennines, considered that most of the breccias represent talus accumulations at the base of major submarine fault scarps. In both of these areas, the ophiolites may indeed include landslide deposits, but in situ and tectonic brecciation were probably also active processes.
As noted above, the Upper Breccia Unit in Hole 899B also contains a small proportion of clasts of metamorphosed magnesium-rich igneous rocks (Shipboard Scientific Party, 1994, Fig. 17). These clasts are described in detail by Cornen et al. (this volume). The rocks can be interpreted as medium- to fine-grained metamorphosed basalts. Some large landslide deposits preserve the original stratigraphy in the sense that relative position of the source lithologies may be preserved during transport and deposition. However, an examination of the data on 158 clasts in the Upper Breccia Unit (Shipboard Scientific Party, 1994, table 6) suggests that the metamorphic clasts are randomly distributed throughout the unit. The metamorphic clasts tend to be small in size, perhaps because the source material was highly fractured. Thus, although a "ghost stratigraphy" cannot be demonstrated, the most likely explanation is that the altered basalts formed a thin veneer on top of the serpentinite escarpment and that the rocks became mixed with the dominant serpentinite lithology during the emplacement of the breccia unit. Elsewhere in the breccia source region, this veneer of volcanic rocks may have been missing, as the source serpentinite appears to have been variably oxidized and calcitized before becoming incorporated into the breccia.
Interestingly, no gabbro or mafic cumulate rock fragments have been found in the serpentinite. Thus, at the time of formation of the serpentinite breccia, the crustal section in the source region was already highly attenuated. The extension had generated an escarpment that exposed highly metamorphosed basaltic rocks, probably from at least one kilometer below the original eruptive surface. These apparently rested as a thin veneer directly on mantle rocks, without any intervening gabbros. Whereas this might have been a depositional contact, this condensed stratigraphy was more likely also the result of extension eliminating any intervening gabbros and cumulate rocks.