Siliciclastic detritus recovered during Leg 168 was eroded from igneous, metamorphic, sedimentary, and metasedimentary rocks. With the exception of the sand layers with abundant intraformational mud clasts, the turbidite samples are remarkably consistent in composition (Table 2). Figure 9 shows that mean values for total-grain modes (Q = 35, F = 35, L = 30) plot within the generic dissected-arc provenance field, as defined by Dickinson et al. (1983). Qm-P-K data (mean of Qm = 46, P = 49, and K = 5) plot within the Circum-Pacific volcano-plutonic suite of Dickinson (1982). Polycrystalline modes agree with the arc-orogen field of Dickinson and Suczek (1979), with a mean of Qp = 16, Lv = 43, and Lsm = 41. The mean values of Lv = 52, Ls = 39, and Lm = 9 are consistent with a mixed magmatic-arc and rifted-continent source (Fig. 9).
Several specific sources and fluvial systems need to be considered in our interpretation of provenance. The Queen Charlotte Islands and Vancouver Island, for example, contain Cambrian to Tertiary volcanic, sedimentary, and granitic rocks (Muller, 1977; Clague, 1986). Common lithologies on southern Vancouver Island include metasedimentary and igneous rocks overlain by submarine volcanic rocks of Eocene age. The Coast Plutonic Complex of western British Columbia, Washington, and northern Oregon consists mainly of Cretaceous granitic intrusions, Jurassic to Tertiary sedimentary and volcanic rocks in the south, and Precambrian to Tertiary sedimentary and volcanic rocks in the north (Clague, 1986). The uplifted core of the Olympic Peninsula consists mostly of Tertiary metasedimentary rocks, surrounded by basalt of the Crescent Volcanics (Tabor and Cady, 1978; Brandon et al., 1998); sedimentary deposits are widespread along the west and north coasts of the peninsula. Farther east, the crystalline core of the Cascade Range is made up of Devonian to middle Tertiary metamorphic-plutonic complexes (Blakely and Jachens, 1990). The Western Cascades consist of basaltic, andesitic, and dacitic lava flows intercalated with pyroclastic rocks. Quaternary volcanic complexes of the High Cascades are basaltic to rhyolitic in composition, with andesite the most common lithology (McBirney, 1978). Collectively, mechanical weathering of these source regions should produce a diverse mixture of sand-sized grain types.
Differences in sample preparation and counting technique preclude reliable statistical comparisons among data sets, but several previous studies provide qualitative insights into the regional patterns of sediment dispersal. The Columbia River (Fig. 1) is the largest point source of fluvial discharge to Cascadia Basin, with headwaters extending eastward into the Rocky Mountains (Whetten et al., 1969). The upper Columbia sub-basin is underlain by sedimentary strata, plutonic, and coarse metamorphic bedrock (Knebel et al., 1968); widespread flood basalts contribute volcanic-lithic grains to the middle and lower basin (Beeson and Tolan, 1990), whereas andesitic debris from the High Cascades increases toward the river mouth (Whetten et al., 1969). Unfortunately, characterization of the Columbia River system in its natural state has been compromised by construction of numerous dams and reservoirs. Deposits in the westernmost reservoir (Bonneville) yield average Q-F-L values of Q = 32, F = 32, and L = 38, whereas samples between Bonneville reservoir and the river mouth average Q = 14, F = 28, and L = 58 (Whetten et al., 1969). Offshore, White (1970) showed that Holocene sands on the Washington-Oregon shelf are also enriched in lithic fragments. Average modes for the shelf deposits are Q = 16, F = 22, and L = 62, in reasonable agreement with the lower reaches of the Columbia River.
Farther north, the Fraser River empties into the Strait of Georgia (Fig. 1) near the city of Vancouver; this fluvial system drains over 250,000 km2 of geologically diverse terrain in south-central British Columbia (Pharo and Barnes, 1976; Clague et al., 1983). Garrison et al. (1969) showed that lithic grains and detrital quartz dominate the sand fraction of the Fraser River delta; average Q-F-L modes are Q = 42, F = 11, and L = 47. Discharge from the Fraser River mixes down-current with smaller streams that drain both southern Vancouver Island and the Olympic Peninsula (Mayers and Bennett, 1973). Collectively, this system creates a second major point source emanating from the Strait of Juan de Fuca, with sediments reflecting a mixture of many tributary sources and rock types.
Working offshore, Carson (1971) concluded that Holocene and Pleistocene near-surface sands in northwestern Cascadia Basin (including the northern part of Vancouver Valley) came from the west coast of Vancouver Island, and that influx from southern Vancouver Island and the Olympic Peninsula becomes increasingly important farther to the south. Chamov and Murdmaa (1995) studied samples from the northern edge of Nitinat Fan (ODP Site 888) and the Vancouver accretionary prism (ODP Site 889). Their average Q-F-L values for light-density grains (segregated using heavy liquids) are Q = 46, F = 18, and L = 36 (Site 888) and Q = 41, F = 19, and L = 40 (Site 889). These modes are similar to those documented at the Fraser River mouth (Garrison et al., 1969). Farther south, Scheiddeger et al. (1973) recognized two petrofacies of heavy minerals at Deep Sea Drilling Project (DSDP) Site 174 (Fig. 1). An amphibole-rich petrofacies (Pliocene) was derived from either the Klamath Mountains of southern Oregon or Vancouver Island, whereas the overlying Quaternary deposits match a Columbia River source. Near-surface sands from southern Cascadia Basin also contain lithic-rich light-mineral suites that probably were transported from the Columbia River (Duncan and Kulm, 1970).
Our results compare most favorably with the data of Gergen and Ingersoll (1986) and Marsaglia and Ingersoll (1992), who analyzed sands from DSDP sites using similar techniques. Sands from DSDP Site 177, which is located well to the north of our study area (50°28´N), contain much higher percentages of feldspar (Q = 29, F = 58, and L = 13) and more sedimentary-rock and metamorphic-rock fragments (Lv = 25, Ls = 48, and Lm = 27) as compared to the samples from Leg 168 (Fig. 9). Based largely on these negative comparisons, we conclude that the flux of sand into the Leg 168 corridor from detrital sources north of Vancouver Island was not significant. The average detrital modes for Quaternary sands at DSDP Site 174 (distal Astoria Fan) are Q = 36, F = 40, and L = 24; proportions of volcanic-rock to sedimentary + metamorphic-rock fragments are roughly equal, and monocrystalline modes are Qm = 46, P = 40, and K = 14 (Fig. 9). This petrofacies typifies Columbia River output during the Quaternary and comes fairly close to matching the modal character of Leg 168 samples (Fig. 9). The Pliocene abyssal-plain facies beneath Astoria Fan shows an interesting shift to more lithic-rich composition (Q = 21, F = 27, and L = 52), with a significant increase in volcanic-rock fragments (Lv = 69, Ls = 6, and Lm = 25). Although a Vancouver Island provenance was suggested for the Pliocene lithofacies by Scheidegger et al. (1973), its petrographic character differs significantly from our Leg 168 results. We believe that additional criteria are required to discriminate more effectively between potential point sources at the mouths of the Columbia River and Strait of Juan de Fuca.
Under circumstances typical of siliciclastic margins, interbeds of turbidite sand and hemipelagic mud usually share common detrital sources (e.g., Underwood et al., 1993), but this expectation is not always realized. Transport directions can be different for downslope-seeking density currents, surface-water plumes, and (or) isobath-parallel bottom currents (Underwood, 1986; Hathon and Underwood, 1991). Selective partitioning of size fractions by transporting agents can also cause contrasts in mineral abundance and clay crystallinity (Gibbs, 1977; Carson and Acaro, 1983). Sluggish nepheloid-layer transport down submarine canyons might be particularly susceptible to size partitioning (Baker, 1976). Smectite tends to increase in the smaller size fractions, and chlorite tends to increase in coarser size fractions. Thus, one might expect grain-size partitioning to reduce smectite in turbidite sands relative to muddy interbeds.
In our study, relative abundance of clay minerals within adjacent sand/mud pairs are similar enough to argue for a shared detrital source. More significantly, we did not document any systematic shifts in clay mineralogy in consort with the type of host lithology. Some of the erratic variations probably represent the natural heterogeneity of the clay-sized sediment budget. Another contributing factor might be intermingling of nepheloid clouds, mass flows, and surface currents converging on the study area from several directions. Finally, glacial-interglacial climate change during the Pleistocene probably played a role in the zonation of weathering behavior; the resolution of our nannofossil age control is not high enough, however, to define or recognize such climate-driven cycles in the depositional record.
As with our petrographic data, quantitative comparisons with and among previously published data sets are inhibited by differences among the sample types, methods of sample preparation, and X-ray diffraction systems. As one qualitative comparison, Holocene clay-mineral suites from the lowermost reservoir of the Columbia River contain relatively high contents of smectite, as might be expected with widespread chemical weathering of volcanic source materials; the average smectite content in Bonneville reservoir is 55%, with individual values as high as 86% (Knebel et al., 1968). Artificial compartmentalization of this fluvial system by dam construction probably has less effect on downstream changes in suspended sediment composition than it does on bedload composition, so the mean value of 55% may be a reasonable estimate for the integrated basin-wide output. Tributaries in the lower Columbia River basin contain an even higher average of 89% smectite (Knebel et al., 1968). More significantly, perhaps, offshore studies by Duncan et al. (1970) and Karlin (1980) showed that smectite content decreases across the continental margin, from over 50% to less than 20%, with increasing distance from the Columbia River mouth. Griggs and Kulm (1970) also found that the Holocene clays funneled south through Cascadia Channel are enriched in smectite (average = 52%) relative to nearby Pleistocene deposits in southern Cascadia Basin (average = 37%). These differences may have occurred because of shifts in the contributions from different Columbia River sub-basins during glacial-interglacial cycles (Griggs and Kulm, 1970). Another contributing factor might have been differences in the balance between mechanical and chemical weathering during glacial-interglacial cycles.
One of our unexpected discoveries is the low content of detrital smectite throughout the study area. Working with surface cores from nearby sites, Carson and Acaro (1983) also documented low contents of smectite (typically <20% in size fractions <0.5 µm), but the average for our Pleistocene and Pliocene mud samples is a scant 8% smectite (Table 2). This shift in clay mineralogy, relative to Columbia River output, indicates that the extent of chemical alteration of volcanic rocks and pyroclastic deposits in the detrital source areas was minor. The predominance of chlorite-rich muds (mean value of chlorite + kaolinite is 52%) indicates, instead, that the source regions were subjected to rapid mechanical weathering, particularly glacial erosion. Source terrains probably included polymictic mixtures of sedimentary, metamorphic, and igneous lithologies. Clay-mineral suites are nearly identical in the intensely glaciated fjords, inlets, and shelf environments of southern Alaska (Molnia and Hein, 1982; Naidu and Mowatt, 1983), as well as along the continental margin of southern Oregon and northern California, where volcanic input is also lower (Griggs and Hein, 1980; Karlin, 1980).
Quaternary sand samples from the Astoria Fan system are broadly similar to those recovered during ODP Leg 168 (Fig. 9). On the other hand, clay-mineral suites from the Columbia River are significantly enriched in smectite relative to the mud samples from the Leg 168 sites. Evidently, surface currents spread the muddy Columbia River discharge toward the north, but the plume becomes diluted or redirected as it reaches the Washington slope and moves downslope into the northwest portion of Cascadia Basin (Carson and Acaro, 1983). Most of the Columbia River sediment is funneled through Quinault, Willapa, and Astoria Canyons, then directed south across the Astoria Fan system (Duncan and Kulm, 1970; Baker, 1976). Accordingly, despite similar detrital modes, we conclude that the Columbia River has not been the principle source for Pliocene and Pleistocene sediments in the Leg 168 study area. A mixed source is more likely, with potential contributions from Vancouver Island, the Olympic Peninsula, and perhaps western British Columbia via the Fraser River and Straight of Juan de Fuca. Lithologic similarities among those terrains, especially when combined with the likely effects of glacial homogenization and postglacial recycling, precludes additional refinement of provenance. As a final point, we acknowledge that our interpretation is difficult to reconcile with previous explanations for the more lithic-rich Pliocene petrofacies beneath Astoria Fan (Scheidegger et al., 1973; Marsaglia and Ingersoll, 1992). Vancouver Island probably was not the source for the Pliocene petrofacies at Site 174; the Klamath Mountains may be a more viable alternative, as suggested by Scheidegger et al. (1973).
Judging from regional bathymetry, several overlapping pathways probably carried individual turbidity currents into the Leg 168 corridor of Cascadia Basin. The first path begins with small submarine canyons that are incised into the western margin of Vancouver Island and continues via Vancouver Valley (Fig. 1). Given the corridor's north- to northeast-striking fabric of basement structures, it seems likely that some of the sandy sediments were transported southward through the Vancouver Valley and Juan de Fuca Channel system. This interpretation is also supported by the seismic-reflection expression of small channel-levee features within the Hydrothermal Transition Transect (Shipboard Scientific Party, 1997a). The second main pathway is through Barkley, Nitinat, and Juan de Fuca Canyons, all of which discharge near the apex of Nitinat Fan (Fig. 1). These canyons connect upstream to the Strait of Juan de Fuca, and their discharge flows around the northern margin of the fan. Nitinat Valley directs turbidity currents westward into Vancouver Valley just north of Sites 1026 and 1027 (Fig. 1). Another contribution, indistinguishable from the Nitinat Valley input, might be unconfined turbidity currents that spread across the northern part of Nitinat Fan. At the northwestern fan fringe, such sheet flows should either spill into Vancouver Valley or deflect toward the south in response to the basement fabric.
Smectite-group minerals play a crucial role in the fluid budgets and mechanical behavior of subduction margins (Vrolijk, 1990). The dehydration reaction of smectite to illite, for example, may be largely responsible for the freshening of pore waters within such systems as Nankai Trough and Barbados Ridge (Kastner et al., 1991, 1993; Vrolijk et al., 1991; Bekins et al., 1995). In addition, clay minerals of the smectite group are mechanically weak (e.g., Wang, 1980; Morrow et al., 1982; Bird, 1984; Logan and Rauenzahn, 1987). As diagenesis progresses with increasing temperature and depth, clay-mineral suites should become increasingly depleted in smectite (Hower et al., 1976; Bruce, 1984). The sedimentary section, consequently, should strengthen with depth because of concomitant increases in the coefficient of internal friction. At the same time, decreases in shear strength may occur locally because of the buildup of excess pore-water pressure and migration of fluids into fault zones (Moore and Vrolijk, 1992; Shipley et al., 1994; Moore et al., 1995). In the cases of both Cascadia and Nankai Trough, the transition from stable-sliding to stick-slip behavior along the basal décollement of the accretionary prism may be governed by the smectite-to-illite diagenetic front (Hyndman and Wang, 1993; Hyndman et al., 1995). This important hypothesis needs to be evaluated thoroughly within the context of strike-parallel zonation of clay mineralogy, particularly as documented outboard of the Cascadia subduction front. Based on the data discussed above, sediments near the Columbia River mouth probably start off with enough detrital smectite (>50% of the <2-µm-size fraction) to affect fluid budgets and physical properties substantially as they pass through the subduction front and diagenesis progresses in the deeper subsurface. On the other hand, smectite content decreases markedly toward the Strait of Juan de Fuca, reaching an average value of only 8% within the Leg 168 study area. Diagenetic changes of smectite-poor clay-mineral suites should be much more subtle, and such small shifts in mineralogy should not alter material properties significantly within (or below) the Vancouver corridor of the accretionary prism.