I think it might be a bit tricky to get working, as understand it the soil moisture meters do indeed work by measuring the resistance, but this is based on the quantity of water in the soil, based on a low percentage, i.e mostly soil. I remember as a youngster there were oftern DIY water level detectors in the electronics mags based on the same idea, but they were all yes/no so gave an alarm if the probes got covered in water. The difference in resistance between just under and say 10cm is probably less than the change in resitance due to the type of water (ie. pure distilled water is not that good a conductor, saltwater is, chlorinated is different again, a 'wild' pond probably depends on the polutants), so the end result would probably more like the old bang-bang controllers.
Depending on the gap between the probes and base material you might also find that enough water remains on the surface once wet to trigger the sensor, so the boat might still be convinced its under water when its started flying! (wouldn't that be the ultimate flying boat?!)
I think to be a stable depth you would have to factor in the boats speed, as that would probably vary the speed at which it responds to any change in the 'planes, unless its done via the ballast tank which might be too slow.
It might be worth looking at spares for washing machines/dish washers since they often have sensors to detect a few cm of water, probably based on water pressure. There is usualy some part of their cycle that adds a small amount of water to the machine that is then squirted round whatever is being washed, but the accuracy of depth keeping is still very tight and you'd pobably need to keep control of pitch as well.
Another idea could be to have a float inside a periscope/snort attached to a rod that is then sensed. One linear sensor is a LVDT (linear voltage distance transducer) These work on a moveable magnetic core being moved between three fixed coils arranged in-line with the centre coil energised by an ac signal. as the core moves from its central position the ac signal induced in the end coils changes, one goes up as the core get nearer and couples more signal into it whilst the other reduces. The ratio of the signal in the two end coils thus gives the position. A bit more than basic electronics and not sure if you could buy one cheap enough, but have the advantage of being able to be water proof & friction free. Many years ago I made a very crude one by wrapping the coils on a biro tube with the core sliding up and down inside, but I no longer have the details. I was trying to make a gyro-compass from an old ex-WD gun-laying gyroscope from the local army surplus depot!
If you google you might find some circuits & ideas for LVDT, e.g. https://www.te.com/usa-en/industries/sensor-solutions/insights/lvdt-tutorial.html