For begining, try this:
http://subsim.questions.cz/sopro/sopro.1.9.zip
Doc here:
http://subsim.questions.cz/sopro/doc/sopro.html
This is simple tool I made which tries to simulate propagation of sound in sea water. It does not include bottom though.
Bottom can block rays bending up to the surface, which is called bottom shadow, and which is similar to surface shadow, which can be simulated in my tool (see examples).
Shallow seabed will also limit creation of convergence zones, as they need rather large depth to occur.
All that is also affected by how well seabed reflects the sound. Rock reflects best, sand worse, mud almost nothing.
If the water is shallow, let's say 100m, there are new group of phenomenons, and raytracing approach I use is no longer much useful. It's called 'shallow water propagation'. Generally we talk about sound channel which only passes some frequencies, and blocks other frequencies, and bottom characteristics have even larger effect.
With active sonar, all above applies too, but bottom will also show on the sonar (which it wont on passive). So sub can be masked by bottom reflections. They you can use Doppler sonar to differentiate sub from the bottom. Active sonar is also somewhat more affected by multipath propagation. That means mainly reflected paths (bottom, surface), including multiple reflections.
This is just rough sketch of main problems involved, and I probably even forgot something really important. Feel free to ask for more specific details.
I'd like to know more about that project of yours anyway. Maybe I could suggest some level of realism, as I thought about most of these problem from implementation view. But it should correspond to the realism level of the rest of the simulation.