Actually the ultrasonic approach should work better with water than it does with snow. Snow works fine except when it is very dry and fluffy. The ultrasonic approach isn’t cheap, but using the sensor described in the following link might be a cheaper alternative.
http://www.parallax.com/dl/docs/prod/audiovis/Distance28015.pdf
All my experimentation has been with the Senscorp sensors which cost about $75 plus some kind of an interface board for 1-wire readings. A relatively cheap approach to the interface board is the HobbyBoards solar sensor and feed the output of the ultrasonic sensor to the Vad pin of the DS2438 chip.
Interesting project, but I don’t have any close by ponds or streams.
Cheers
MikeyM