Paper - Caris

Figure 4: Image of 3 REMUS AUV CTD cast sites in Solomons, MD. Data were used to test Velocwincodethe cast was approximately 30 feet. Multiple copies of this downward cast were duplicated within the textle to create a test AUV CTD data le for which we could start writing code to pick the beginning andend of each sound speed prole cast. Here we assumed that each time the AUV began its decent it wouldrst come up to the surface (where it would have at some point a depth reading of zero), get a positionx, decend vertically down the water column to the seaoor, then travel to the next cast location andcome up again to the surface to repeat the process again. Once the Velocwin algorithm was manipulatedto account for these changes to split out the SSPs, we collected more data in May to test its durability.2.2 ExperimentFor the May 28 experiment in Solomons, MD on the NOAA S/V BAY HYDROGRAPHER, our majortask (after moding the Velocwin program to be able to accept the CTD sensor data from the AUV) wasto test this code with a simple, yet more typical, cast-like situation. We deployed the REMUS 100 toacquire three simultaneous sound speed casts within close proximity to one another (between 9-29 meters)on the BAY HYDROGRAPHER in Solomons (Figure 4). The CTD data output from the REMUS wasagain inserted in Velocwin. However, we noticed that our original assumption in which the depth readingwould be zero just before the downcast was incorrect. We had unfortunately used this assumption todetermine how to split up the text le of back-to-back CTD casts. Figure 5 illustrates also that as theAUV goes from one cast location to another, it oscillates along the seaoor. Our solution was to adjustthe Velocwin algorithm to incorporate box car averaging.Although this worked for the present dataset we were not certain whether it would work for others.7