The primary drawback to physical models is that it typically very difficult or expensive to get the same behavior of the phenomena you are studying at different scales. Depending on the study, you need to try your best to match the dimensionless quantity that is involved in the phenomenon you are studying. Fluids have many properties (surface tension, viscocity, expansion coeffs, etc..) that typically don't scale well: http://en.wikipedia.org/wiki/Dimensionless_quantity
Often you use a fluid other than water and a different timescale and sometimes you can adjust the properties of the fluid by changing it's temperature and try to dial in the dimensionless quantity of interest. But I bet it is difficult to swap out all that water in the Bay Model or drastically change the temperature..