f Turbidity Current Modelling

Algorithms for modelling of turbidity currents in three dimensions, to assist in the exploration of poorly imaged deposits, are described and examples given of their use. Key aspects of algorithm design are:-<br>1. Algorithm efficiency. Program must be capable of accurately modelling currents and their deposits with good resolution and on small computers. This is achieved by appropriate simplifications of the full 3D modelling equations.<br>2. Ease of use. Program must be robust and should not require large numbers of unknown parameters. Appropriate simplification is again important here.<br>3. Accuracy. The program has been validated by comparison to both flume tank and real world datasets.<br>4. Invertibility. In practice, inverse modelling is needed rather than forward modelling. The approach adopted is genetic searching based upon an objective function comparing observed to modelled net-to-gross ratio. The results of this can be refined using symmetry concepts.<br><br>In a little more detail, the governing equations are depth-averaged versions of the Cauchy equations of motion. Newtonian rheology is not assumed and so these do not reduce to the Navier-Stokes equations. Instead, the equations bring out the importance of the relationship between depth-averaged flow-speed and basal shear-stress as the key factor governing overall flow behaviour. These equations still leave a number of unknown factors to be determined but these can be estimated from the behaviour of real-world and flume-tank flows.<br><br>Initial testing shows, not surprisingly, that the geometry of the seafloor over which a flow is modelled plays a central role in the deposit geometry and deposit net-to-gross distribution. Structural reconstruction of the seafloor at the time of the flow is therefore usually vital. This suggests the possibility of a promising workflow link between seismic interpretation, providing input into reconstructed seafloor scenarios, and forward modelling to understand the consequences for probable sand distribution.<br><br>Non-uniqueness (i.e. many different model parameter sets giving equally good fits of model to data) is usually considered to be a serious problem in any stratigraphic inversion procedure. However, if the links between good solutions are known (i.e. the symmetries are understood), then a range of equally good solutions can be constructed from any single good solution. Unfortunately, the symmetries are rarely known a priori for any but the simplest of models. Instead, symmetries can be investigated using large numbers of simulations and these empirical-symmetries used to guide refinement of solutions found by, for example, genetic search algorithms. Output from these methods can also be used to enhance seismic interpretation in deep-water systems.<br><br>