we have discussed this and shown that acoustic elements can be used (it can be hard to identify modes, but they work).
So if it is not anchored you can let it sit on top of the ground and have frictional contact between the tank and the ground, and then run it with nonlinear transient dynamics say typically with seismic loads (you cannot use modal since it is a linear analysis and does not account for the complex contact between the two structures).
Now I have never done anything like that (measurements or FEA), so I cannot comment if the modes you calculate (probably free suspended tank), are close to the real modes when the structure rests on the ground - obviously the BC are quite different.
In any way seismic transient nonlinear transient dynamic analysis uses implicit time integration since it needs to account for all nonlinearities (contact, material geometry), and thus a mode superposition is not used, but an implicit scheme, so modes are not used.