The ADCP for this site is at a fixed depth relative to the bottom. It doesn’t move vertically to profile, rather it sends out pulses of sound to profile the water column. You might want to look at the ADCP Principles of Operation Primer produced by Teledyne RDI. In particular, take at look at Chapter 5.
Any variation you are seeing in the depth record would be due to the tides, slight differences in deployment depth from deployment to deployment and/or mooring tilt. As a general rule (for all upward facing OOI ADCPs), the instruments are programmed to sample N number of depth bins of size X with N set large enough to ensure the ADCP samples up to and through the surface, where values <0 are above the surface in the air. Note, that is a general rule, not 100% applicable!
You might also find this existing conversation helpful: https://discourse.oceanobservatories.org/t/adcp-range-bins-contaminated-by-sea-surface-a-robust-qc-option/209
You are mostly correct in removing any negative bins (sampled past the surface and into the air), however there are other effects on the data near the surface that you may also want to address. Make sure to review the references linked in that conversation above about removing side lobes. First pass, though, yes filter out all negative bins.
As far as the large negative value goes (-396 m), delete it. Sometimes the unit is running in air either before or after a deployment (operators haven’t had a chance to power it down yet), and you can get some errant values in the data record. We do try to trap these time periods out when ingesting the data into the system, but the occasional bad data point can leak through.
Let me know if you need more help/guidance after you get a chance to look at those links.
Chris