**4. Discussion**

Our results o ffer estimates of northern Gulf of Mexico sediment delivery and oceanic transport conditions, including locations for gravity flow; and routing of riverine and shelf sediment into submarine canyons. With e fforts such as these, that treat multiple time- and space-scales, modeling tools can be developed to deepen our understanding of how sediment is carried from riverine sources to various oceanic sinks. The challenges of integrating various modeling approaches across di fferent spatial and temporal scales are substantial and require further research and code development. Both physical aspects (the implementation of erosion, resuspension of complex sediments into large-scale simulations), as well as numerical challenges (two-way coupling, temporal and spatial interpolation at the boundaries between models) require an additional community e ffort. The treatment of physical phase-transitions, such as between wave-supported suspended sediment flows and actual turbidity currents, requires more fundamental research.

The models developed for the workflow operated over a broad range of spatial and temporal scales. For example, the *RANS*/*TURBINS* model represented relatively thin (tens of meters) turbidity currents at higher temporal (<1 s) and spatial (~3 m) resolution than a fforded by *ROMS*' hydrodynamic and suspended sediment transport model. As a first step toward multi-scale modeling at the spatial level, our workflow follows sediment routing from the watershed scale via *WBMsed*, to the continental shelf scale via *ROMS*, to specific sediment gravity flows via the *HurriSlip* modules and *RANS*/*TURBINS*. Likewise, the processes encompassed in our workflow operate over a range of temporal scales, from that of hours for the *TURBINS* model, to the timescale of storm fluctuations for riverine delivery, flow ignition, and suspended transport. Changes in sediment transport that operate at seasonal and interannual timescales are likewise built into our workflow by using forcing functions for weather that represent variations in winds, precipitation, and air temperatures that operate at these timescales. Barriers in applying our methods to longer timescales (i.e., longer than decadal) include both computational limits, and di fficulties in assuring that subtle biases in the models and their parameterizations do not cause the calculations to drift from realistic conditions.

The model workflow presented here is sequential, with limited two-way coupling. A fairly straightforward step is to link the riverine discharge model (*WBMsed*) to the oceanic *ROMS* and *CSTMS* models. It would facilitate studies aimed at quantifying oceanic dispersal of fluvial sediments for poorly gauged river systems [83]. Future e fforts should explore a more direct model coupling between the suspended sediment transport and gravity flow mechanisms. Within this workflow, *ROMS* estimates the bed shear stresses, which can be used for the flow ignition model (*HurriSlip*). Locations of a slope failure can trigger simulation of a gravity current (e.g., Figure 11), which moves sediment downslope. More direct coupling between these modules would account for sedimentation via suspended sediment transport within the slope failure module, and for net erosion and deposition via gravity currents within the regional scale (*ROMS*) resuspension model.

Regional modeling in the northern Gulf of Mexico is not trivial. Sediment transport modeling requires high-spatial-resolution models to resolve the complex and steep bathymetry. The intense coastal circulation, eddy shedding from the Gulf Loop Current [84], and sporadic strong forcing from storms and hurricanes can a ffect sediment transport pathways across the continental shelf and slope. Therefore, a telescoping grid approach, from coarse (kilometers) to fine (10s of meters) horizontal scales, is required to obtain viable long-term (1–10 years) and a ffordable computations. Within our implementation, this was realized by using a low-resolution model for the entire Gulf, telescoping to finer-resolution for the region surrounding the bird-foot delta (Figure 3A). A similar approach has been employed to represent decadal-scale sediment transport in the northern Gulf of Mexico [22].

Joint modeling and field experiments are needed to develop reliable sediment transport models for the Gulf of Mexico continental shelf and slope. Sediment depositional data with which to compare the model calculations are severely lacking, especially at the spatial scales considered here. Recent observational e fforts, some motivated by the response to the Deepwater Horizon event, have shown that sediment can be mobilized in deep Gulf of Mexico locations [30,85]. Many of our workflow's sediment transport routines were based on parameterizations for other continental shelf systems, or on laboratory measurements. To improve and gain confidence in the models developed for this workflow requires allied field and modeling studies of sediment processes for the northern Gulf of Mexico continental shelf and slope. Because field sampling during and immediately after storm events is inherently challenging, coupled models that are consistent with observed transport processes and sedimentation are needed to characterize conditions during the extreme events most likely to lead to large sediment fluxes in the deep Gulf of Mexico, and which can damage o ffshore infrastructure (e.g., [1]).
