(2007), show good agreement between the Fluidity-ICOM mixing bin values and those from Özgökmen et al. (2007), Fig. 12. The value p=2p=2 for the MpMp metric is found to work well. The successful use of M2M2 demonstrated here builds on the good results obtained with M2M2 in Loseille and Alauzet (2011b) by extension to a turbulent KU-60019 research buy and time-varying flow. A smaller value of p   would lead to a more equal weighting between the smaller- and larger-scale fluctuations and a more uniform mesh would be expected ( Loseille and Alauzet, 2011b).

Conversely, as p   increases, the larger-scale fluctuations will become increasingly dominant and the meshes produced will become more like those for the M∞M∞ case ( Loseille and Alauzet, 2011b). The ability to capture a range of scales will also be useful for modelling of the lock-exchange in three dimensions, where the lobe and cleft instability adds to the complexity of the flow. The extension to three dimensions offers an important and tractable avenue for future investigation which also presents the opportunity for more extensive comparison to published results from other types of model e.g. Özgökmen selleck chemicals llc et al., 2009a and Özgökmen et al., 2009b. Whilst there are many other factors which

will affect the efficiency of the individual models, such as the discretisation method, the adaptive meshes are able to produce flow characteristics that are equivalent to fixed meshes with one to two orders of magnitude more vertices (or degrees of freedom). This reduction in the number of vertices presents a significant improvement in the efficiency of the simulation for the finite-element discretisation method and numerical configuration used here. Such decreases

in computational demand are not limited to the discretisation method and mesh structure considered here with, for example, 80% efficiency gains for the lock-exchange Clomifene problem using a quadtree finite-volume discretisation reported in O’Callaghan et al. (2010). In addition, the reduction in computational demand with the use of adaptive meshes can provide an offset against the inherent increased cost of, for example, a finite-element discretisation method on an unstructured mesh compared to a finite-difference model on a structured mesh. The performance of the adaptive mesh is highly dependent on the choice of metric. Changing the adaptive mesh settings can and will change the solution, particularly for a turbulent system such as the lock-exchange. However, the impact is not necessarily any greater than changing the discretisation method or the resolution of a fixed mesh. The effective use of an adaptive mesh with the simple metrics used here demands consideration of the problem to which it is applied and preliminary test simulations to obtain an appropriate set of solution field weights.