A couple reasons, I'd bet - first, a hole wide enough to fit the probe is pretty unusual, and would be an expensive research project all on its own. Second, the probe near the surface will emit the largest signal, so you get some really good SNR data to calibrate your algorithms against. If you don't understand the signals you're receiving when it's shallow, you won't understand them any better when it's deep - a regime where we have more theories than data. Or maybe something else.