**3. Results**

The sections below describe model calculations from components of the workflow to demonstrate their capabilities.

#### *3.1. Suspended Sediment Transport*

The *ROMS*/*CSTMS* ocean model calculated current velocities, bed stresses, suspended sediment fluxes, and erosion/deposition from October 1, 2007, through September 30, 2008. *CSTMS* results for 2007–2008 indicated that the overall signature of sedimentation calculated from suspended sediment was deposition near fluvial sources, with patchy erosion and deposition elsewhere (Figure 8A). Sediment delivered by the Mobile River was largely retained within Mobile Bay. The Atchafalaya River sediment was deposited near the delta, but resuspension events on the inner shelf (depths < 30 m) created westward sediment transport along the coast. Mississippi River plumes more widely dispersed sediment around its bird-foot delta with some of the river load deposited in deeper water (>200 m). The model indicated that the deep sea experienced strong intermittent currents capable of mobilizing sediment, termed benthic storms [80].

Analysis of suspended sediment delivery to the continental slope indicated that about 70% resulted from delivery during low-intensity storms such as frontal systems, and fallout from the Mississippi River plume. The remaining 30% of the year-long delivery of suspended sediment to the continental slope occurred rapidly, during the days surrounding the passage of Hurricanes Gustav and Ike. This supports our first hypothesis, that episodic sediment transport down a submarine canyon is fed by sediment input from wave and current resuspension. Shelf erosion during non-hurricane times accounted for a small fraction of the cumulative erosion seen for the year (Figure 8B). The patchiness of erosion seen in the deep sea (Figure 8B) corresponded to the sediment texture assumed by the model (see Figure 6A). Hurricanes Gustav and Ike created widespread erosion on the shelf, and this material contributed disproportionately to sediment delivery from the shelf to the slope, compared to other resuspension events during the preceding eight months when elevated Mississippi River discharge also occurred (Figure 8). Bed shear stresses during the hurricanes were sufficient to suspend fine-grained sediment across the shelf break. Hurricanes Gustav and Ike produced distinct patterns of erosion and deposition (Figure 8C,D), mainly due to their differences in strength, duration, and storm track (see Figure 4). In general, Ike created higher bed stresses, sediment concentrations, and erosion; but in some locations, Gustav had more impact.

Suspended sediment fluxes along three cross-slope transects (locations shown on Figure 5D) were analyzed to evaluate the phasing and magnitude of hurricane-driven sediment delivery to the continental slope and beyond. The size of sediment flux generally decreased with water depth across the continental slope (Figure 9). While peak fluxes on the continental shelf coincided with the passage of hurricanes, there was often a lag of several days before suspended sediment reached deeper waters. Along the western continental slope, suspended sediment fluxes were larger for Hurricane Ike than Gustav and decreased in the deeper waters (Figure 9A). The peak suspended sediment fluxes at depth (~1300 m) occurred several days after the peak fluxes calculated for the shelf–slope break (~128 m). For the De Soto Canyon, the model estimated net downslope flux towards the south to southeast during

the hurricanes (Figure 9F–I). Suspended sediment flux actually increased offshore there between depths of ~650–1115 m during and after the passage of the hurricanes, suggesting that suspended sediment transported over the sides of the canyon settled to the bottom boundary layer and contributed to the suspended sediment fluxes calculated within the canyon (Figure 9F–I).

**Figure 8.** Net erosion (<0) and deposition (>0) calculated for suspended sediment transport for (**A**) entire model run (1 October 2007–20 September 2008); (**B**) prior to hurricanes: time-integrated from 1 October 2007 up until 25 August 2008; (**C**) during Hurricane Gustav; (**D**) during Hurricane Ike. Bathymetric contours (in black) drawn for depths of 10, and every 100 m up to 1500 m.

Sediment fluxes down the Mississippi Canyon lagged behind the passage of the storms, being delayed by 1–5 days relative to the occurrence of peak wave energy on the shelf (Figure 9B–D). The model results showed that these lags corresponded to the time needed for nepheloid layers generated by cross-shelf transport of storm resuspension to be carried to, and settle into, continental slope depths. For example, Hurricane Gustav made landfall on 1 September 2008. For the Mississippi Canyon transect, the model indicated that Gustav created peak sediment fluxes on the outer continental shelf (water depth 98 m) around 2–3 September; while offshore, sediment fluxes did not peak until 4 September (688 m depth) and 8 September (1008 m depth) (Figure 9B,C). The distance along the Mississippi Canyon transect from the 98 m deep site to the 1008 m deep site is about 66 km, so the ~4.5 day lag in delivery to the 1008 m site can be explained by an average horizontal transport velocity of about 0.16 m/s. Similarly, vertical settling delays a storm's impact on deep-sea locations. The fine sediment classes used in the model would settle about 10 or 100 m per day, so that fall out from nepheloid layers would require days to weeks to reach the near-bed continental slope and deeper. This process is illustrated using modeled suspended sediment concentrations along the Mississippi Canyon when Gustav was centered over the Lousiana shelf (Figure 10A), and five days later, the nepheloid layer was delivered to, and settled into, continental slope depths (Figure 10B).

**Figure 9.** Time-series of depth-integrated suspended sediment flux (kg m<sup>−</sup><sup>1</sup> s<sup>−</sup>1) calculated along three cross-shelf transects for summer, 2008. Suspended sediment flux across the (**A**) western slope, (**B–E**) Mississippi Canyon (red vectors), and (**F–I**) De Soto Canyon (green vectors). Vector angles correspond to flux direction in accordance with map conventions (down means southward flux). Water depth, latitude, and longitude of each calculation provided as text on figure panels. Locations of transects shown in Figure 5D. Dashed lines mark landfall times of Hurricanes Gustav (9/1) and Ike (9/13).

**Figure 10.** Suspended sediment concentrations calculated along the Mississippi Canyon transect during and after Hurricane Gustav show that sediment delivery to the mid-Canyon lagged several days behind peak storm conditions on the shelf. Model estimates for (**A**) 1 September 2008, and (**B**) 6 September 2008.

#### *3.2. Density Flow Ignitions*

Results from the *HurriSlip* model sugges<sup>t</sup> that during extreme storms, bed stresses are large enough to create conditions suitable for the ignition of turbidity currents from near-bottom layers of suspended sediment, especially in areas near the shelf break (Figure 6A). The modeling also suggests that small-thickness sediment mass-failure events, which may evolve into turbidity currents, are widespread around the shelf-slope transition under hurricane conditions (Figure 6B). There is some association between predicted ignitions' locations, and the geomorphic channelizations of the upper continental slope (Figure 6B).

Sediment resuspension sources: The results on the resuspension of sediments into bottom waters (*SuspendiSlip*) indicated suspended sediment concentrations (*SSC*) during times of wave activity averaged ~300 ppm v/v, up to ~5000 ppm v/v (5% v/v) at levels 1 m above the bottom. During the storm events, in shoreface areas including at the delta front, some wave-induced bottom orbital velocities >4 m/s were indicated. At depths of 20–40 m this was reduced to >2 m/s. As modeled, wave-induced resuspensions occurred down to water depths of 189 m (at surface wave periods >13 s) in areas not sheltered from the storm wave e ffects.

Numerous density-flow ignition events were indicated. They were overwhelmingly in the bottom 1–2 m of the water column, but occasionally occupied water masses as thick as 8 m or more. Bulk densimetric Richardson Number values for the bottom flows ranged widely, but during the storm events were <<1.0 near-bottom i.e., were supercritical states susceptible to the onset of density flow [64]. The Knapp–Bagnold criterion discriminated events more closely and with the gravity influence of slope, identified locations of plausible density flow ignition (Figure 6A). There is some indication that suspension events in the waxing and waning of a storm are more likely to ignite because of the balance between densities and velocities.

Mass failure sources: In agreemen<sup>t</sup> with the extensive evidence of mass sediment failures in the region [11,81], the modeling indicated a potential for seafloor failures due to the combined e ffects of intense storm wave activity, shallow depth, and significant slope. The present prediction with *WaveSlip*, however, also extends over sandy areas not only the mudslide province at the Mississippi Delta front. There seems to be an increased potential for the failure to transform into turbidity current in sandy sediments [82].

Wave-induced liquefaction was predicted in the modeling for conditions of <30 m water depth, somewhat sandy sediments, surface wave wavelengths of >150 m, and significant wave heights of 10 m. Developed (residual and momentary) normal pore pressure increases to exceed normal overburden pressure were modeled down to subbottom depths of 10 m and more at some locations. In those circumstances, e ffective shear strength was reduced to near zero. The possibility of cyclic strain reduction of shear strengths was also investigated. However, the cumulative strains induced by waves, even during extreme events, were insu fficient to produce significantly lowered (remoulded) shear strengths, the strains being at most of order 10−<sup>2</sup> cumulative (10−<sup>4</sup> to 10−<sup>6</sup> per cycle).

The circular-slip analyses indicated mass-failure instabilities (*FoS* <<1.0) over broad areas of sloping seafloor in the top 0.5 m of the seabed (Figure 6B). More deeply-seated failures, down to 20 m sub-bottom, were predicted at a small number of locations at about 30 m water depth. Still, all had a *FoS* >>2 and, therefore, apparently limited potential for actual failure. (They are also at the limits of the analysis in terms of wave-breaking and infinite sediment column.) *HurriSlip* results appear to sugges<sup>t</sup> that without liquefaction or remoulding, probably very few significant wave-induced mass failures would occur in the region. However, the smaller occurrences which are also predicted, remain as candidates to release turbidity flows. They include particularly, many locales with a high likelihood of failure (*FoS* <<1) during storms, in seabed areas down to 100 m water depth, with a significant slope, and often near to the shelf edge.
