[b]For fluid flow, I was planning on using a well BC, with the pumping rate divided by 2*pi as recommended in the online help. However, the help says that in the case of axisymmetric models a specified flux boundary is preferred. What is the reason for this?[/b]
NeumannBC's are usually applied to have an "areal" inflow or outflow. The value for the assignment can be calculated according to the pumping rate Q, filter section B, radius R: q=Q/(2*pi*R*B).
[b]Could someone with experience modelling density-dependent flow provide some insight as to whether these BCs are best suited to this particular problem?[/b]
The selection of these boundary condition is not related with density-dependent flow. The principle is the same for non-density-dependent flow.
[b]And potentially comment on whether the divergence or convective form of the transport equation should be used?[/b]
If you solve the convective formulation of the transport equation and if you use NeumannBC’s or WellBC's for flow plus NeumannBC's/Nodal mass source sink for mass you specify only the dispersive transport component. The advective transport component remains unspecified. Accordingly, you cannot control the advective component and most likely you observe larger budgets to what you expected.
In contrast, if you apply the divergent form you specify all mass drivers (advective, dispersive, diffusive). As a result, calculated budget quantities will satisfy your expectations. The disadvantage of the divergent form is a more difficult handling at outflowing boundaries. Artificial accumulation of the concentration distribution may also occur in particular if the advective flow field is heterogeneous.