I am a great believer in considering the theoretical limits of RF penetration but also looking for the real world experiences of others. Theory is often a good base line but practice is proof positive of what is possible. Many good antennas have been invented through experimentation rather than a purely theoretical approach. RF is a very unpredictable energy, especially when it comes to materials penetration.
With that in mind I did some Googling on Radio Controlled model submarines. I have always stuck to fixed and rotary wing aircraft but know that the RC submarine chaps have exactly the same issues as your project.....i.e. which frequency is best in water and what range can be expected. The TX and RX positions are reversed in your case but with careful design that should not be a major obstacle.
As already suggested, think low frequencies in the HF bands and not GHz microwave frequencies
Take a look here and do your own research on RC submarines
http://www.rcgroups.com/forums/showthread.php?t=853144Be aware that I know of no RC electronics that will operate in a 100C environment. As such you would either need to use bespoke transmitters or protect the transmitter from heat in a Vacuum Dewar. RC telemetry products used in the Quadcopters can provide excellent data communications but sadly most (all?) use GHz freqeuncies.
You could set up a water (fluid) filled test tank with a Transmitter antenna waterproofed and submerged in it and driven by an RF signal generator. Monitor the signal level at the desired distance with a sensitive wideband receiver and see which frequencies dos and do not make the distance. A practical experiment may tell you much about the challenges you face with penetration through water or other fluids. Once you know which frequencies are, and are not usable you can begin your research in to available transmitter modules and data rates. In your position I would buy some cheap data modules and see how they behave at 100C by placing them in a domestic Oven set at 100C
They may go bananas, but they may also work, albeit at a frequency different to the spec ! Regarding MCU controlled oscillator correction.... nice idea but if you want to apply the KISS principle, go with AFC in the receiver.
All this take no account of using transmitter frequencies within legal allocations. That is another topic but should be considered if this is to be a commercial product. Short range data links do have approved frequencies such as 433MHz for car key fobs, wireless alrams, telemetry etc. Wireless computer keyboards used to operate at 27MHz HF which may be an option ? It is worth checking your local licence free transmitter regulations if that is a concern for you.
Some food for thought anyway
Aurora