Sediment transport

Discharge of sediments from catchments into the GBR Lagoon has increased many-fold over the last 150 years as a result of human activities (Furnas, 2003; Kroon et al., 2012). Most of these sediments are delivered to GBR during high flood events associated with tropical cyclones and monsoonal rainfall (Brodie and Furnas, 2001). Because of their size and geomorphology, dry catchments (Fitzroy and Burdekin) deliver the majority of sediment loads to the GBR coast (Brodie et al., 2013). River flows and sediment loads exhibit high variability over the range of time scales (Kuhnert et al., 2012; Darnell et al., 2012, Kroon et al., 2012) including diurnal, seasonal, annual and decadal scales.

Estuaries along the GBR coast are generally ranked as tidally dominated (Webster et al., 2008). Strong tidal currents in a constrained channel can maintain high levels of the suspended sediment in the mid-estuary region even during periods of low river flow. During extreme flood events a considerable amount of sediments can be deposited on the coastal floodplain, but this amount is not well quantified. Under moderate flood conditions a general consensus is that estuaries along the Australian GBR coast are mainly too small to have much impact on the magnitude of the fine sediment delivery to the lagoon (Neil et al., 2002). During non-flood conditions these estuaries, generally, trap significant fraction of catchment sediment. Estimates of sediment loads to GBR are often based on measurements above the estuary, missing inputs and losses of sediment occurring downstream from the measurement point.

Turbidity plumes from the major rivers may extend tenths and even hundreds of kilometres along the shelf, but tend to be constrained to within 20 km of the coasts by buoyancy and wind stress (Webster et al., 2008; Brodie et al., 2012). Most of the suspended sediment settle out of the flood plumes and deposit on the seabed within 10 km of the coast (Orpin et al, 1999; Lewis et al., 2014). A sedimentary pool created by these sediments can be partly buried into deeper sediments and partly resuspended by wind-driven waves and tides and then dispersed further by currents. Coastal currents, under the dominant southeast (SE) trade winds, tend to carry sediments northward along the coast. A fraction of the fine sediments can be ultimately trapped in northward facing embayments sheltered from the exposure to south-easterly swells by coastal features (e.g. Broad Sound, Bowling Green Bay, Princess Charlotte Bay) (Larcombe and Woolfe 1999). The amount of modern sediments trapped in these embayments is not well known. Recent measurements, for example, indicate that inorganic sediments delivered from the Burdekin River are trapped within 50 km from the mouth and very little of this sediment reaches north-west into the Bowling Green Bay and Cleveland Bay (Lewis et al., 2014). In Princess Charlotte Bay (Northern GBR) the terrestrial silt-clay component is dominated (more than 80%) by sediment derived from the coastal plain and input from local catchments (Olley et al., 2013).

Sediment processes on the GBR region have led to the development of a strongly sediment partitioned shelf, with modern mud-rich sediments almost exclusively restricted to the inner and inner-middle shelf of 0 to 20 m depth (Neil et al., 2002). Further off-shore, at depths of 20 to 40m, a middle shelf zone is marked by a thin veneer of mixed terrigenous-carbonate ribbons and sand dunes. The passage of intermittent cyclones creates strong northward along-shelf currents, which cause erosion of the middle shelf seabed and transport of mobile bedload. At depths of 40 to 80 m, an outer shelf zone of reef perimeter is dominated by carbonate sediment (Larcombe and Carter, 2004).

Because of the high natural variability and the wide range of spatial and temporal scales of the sediment processes on GBR, inferring marine manifestations of the altered land use practices (e.g. persistent changes in the turbidity levels or changes in depositional rates of coastal sediments, instrumental to coral reef health) is a challenge, particularly over the whole GBR scale. According to one school of thought, chronic turbidity at coral reefs due to suspension of fine sediments by tides and winds will not be significantly affected by changes in sediment inputs since the sedimentary pool in the Lagoon is already large (Larcombe and Woolfe, 1999). On the other hand, recent measurements suggest that the geological deposits together with newly imported materials additively determine water clarity inshore as well as mid-shelf (Wolanski et al., 2008; Fabricius et al., 2013; Thompson et al., 2014; Fabricius et al., 2014). This new evidence shows significant correlation between catchment loads and turbidity on the shelf. The processes connecting catchment loads to the observed changes of turbidity across the GBR, however, are still not well understood. Transport of fine inorganic sediment and biogeochemical cycling of organic particulates have been considered recently as potential drivers of these changes (Lewis et al., 2014; Fabricius et al., 2014).

 

Numerical model

 

The fine sediment transport model adds a multilayer sediment bed to the hydrodynamic model grid and simulates sinking, deposition and resuspension of multiply size-classes of suspended sediment (Margvelashvili et al., 2008). The model solves advection-diffusion equations of the mass conservation of suspended and bottom sediments and is particularly suitable for representing fine sediment dynamics, including resuspension and transport of biogeochemical particles. Sediment particles settle on the seabed due to the gravity force and resuspend into the water column whenever the bottom shear stress, exerted by waves and currents, exceeds the critical shear stress of erosion. The resuspension and deposition fluxes are parameterised with the Ariathurai and Krone (1976) formula. Estimates of the bottom shear stress, required by this formula, are derived through the Madsen boundary layer model (Madsen, 1994). Bottom roughness is scaled by ripple dimensions (Grant and Madsen, 1982) which are considered the model input parameters and must be specified through observations or calibration study.

Sediments in benthic layers undergo vertical mixing due to bioturbation, represented by local diffusion. The corresponding diffusion coefficient is scaled with the sediment depth so that the bioturbation ceases to operate beneath the biologically active layer. The resistance of sediments to resuspension increases with the sediment depth.

The numerical grid for sediment variables in the water column coincides with the numerical grid for the hydrodynamic model. Within the bottom sediments, the model utilises a time-varying sediment-thickness-adapted grid, where the thickness of sediment layers varies with time to accommodate the deposited sediment. There are 4 benthic layers in the GBR model grid. Horizontal resolution within sediments follows the resolution of the water column grid.

The sediment transport model can be fully coupled to the hydrodynamic model implying that both models run in parallel and have the same time step. The alternative option is to simulate sediment transport in off-line mode. In this case currents and diffusion coefficients saved from the hydrodynamic model run provide inputs into the stand-alone sediment transport model. The simulation time-step is much larger than that of the coupled model and there is no feedback from the sediment processes to the hydrodynamics, i.e. the impact of sediments on flow, density and turbulence are not simulated.  This decoupling of the sediment and hydrodynamic models provided substantial benefits in computational efficiency, and is implemented for the eReefs sediment transport models.

The sediment model is initialised with the observed distribution of gravel, sand and mud in the seabed of the shelf region.  Catchment sediments discharged into the GBR over the simulation period are represented in the model by two classes of particles having varying settling velocities. The model is intended as a decision support tool to estimate GBR-wide distribution of fine sediments; it also supports biogeochemical model simulations and provides input to nested fine-resolution relocatable model RECOM. The sediment transport model produces stable, realistic simulation, even when subject to extreme forcing events such as cyclone Yasi. The model is now routinely running in near real-time, using WaveWatch III outputs (AUSWAVE) as forcing fields.

 

Calibration and validation study

 

The sediment transport model was calibrated in two stages. First, the model parameters (initial conditions and spatially varying bottom roughness) have been refined through the ensemble assimilation of the 6 month remote sensing data for TSS. Ensemble of models, produced through the assimilation step, has been reduced then to a single model, which was subsequently validated against time-series of the observed turbidity (i.e. data from Reef Rescue coastal sensors and GBROOS shelf moorings). The validation of the 4-year run of the model revealed a long-term drift of the solution which was handled through the manual adjustment of the model parameters. The quality of the calibrated model varies across the GBR region and with time (as was expected). The distribution of the simulated suspended sediment on GBR is generally consistent with observations. The model tends to overestimate suspended sediment levels in northern Queensland and underestimate TSS levels in Torres Strait. The model is capable of predicting the broad patterns of suspended sediment in the GBR, and variability at tidal and seasonal time-scales, and is suitable for use in scenario analyses. Due to the inherently stochastic nature of the sediment processes on the shelf and resolution and process limitations of the model itself, there is uncertainty in forecasts of sediment concentrations at specific locations in space and time.

A number of preliminary scenarios have been simulated with the calibrated model. Numerical experiments highlight the role of very fine fraction of catchment sediments (representing either tails of the sediment size-distribution or flocs of fine particles characterised with a low settling velocity) as a carrier of sediment signals propagating from catchments to the GBR region. Scenarios with varying loads of sediments from catchments illustrate spatial and temporal variability of changes of suspended sediment levels on GBR. The response of TSS to varying loads from catchments in these scenarios is expressed in terms of changes of the annual mean probability for TSS to exceed 2 mg/L. This threshold (2 mg/L) represents GBRMPA (2009) guideline trigger value for TSS effects on marine ecosystems in open coastal and midshelf waters of GBR. Preliminary analysis of these simulations suggests a relatively short term-response of the GBR system to changes in catchment loads. Scenario with a 4-year run of the model having elevated loads from catchments does not show an incremental, multi-year build-up of the excessive TSS levels in the region. Instead, the TSS response to the increased load from catchments is most pronounced during wet years (and, according to some scenarios, during the first dry year following wet years). During the subsequent dry years changes in catchments have a much smaller impact on the probability for TSS to exceed 2 mg/L.