First the velocity profiles are differentiated with respect to depth to eliminate the CTD-package's motion. Then a depth record is obtained by either using the CTD-pressure record or integrating the vertical velocity in time. Now the shear profiles are averaged together within depth bins. Here one of the crucial steps is to carefully edit shear estimates to reject outliers. The average shear profile is then integrated vertically to obtain a baroclinic velocity profile. The missing barotropic correction can be calculated if precise position of the start and end point of the cast is known (from GPS, for example) and the velocity record is continuous in time (Fischer and Visbeck, 1993a). Several problems are known for the LADCP system. First, some levels close to the bottom of the profile, but still several hundreds of meters above the sea floor, will not have any useful data if the ping rate is constant. At those levels the bottom reflection of the previous ping arrives at the same time as the backscatter from the current ping and results in ambiguous Doppler shifts. This problem can be reduced by employing asynchronous ping rates (T. Chereskin, pers. comm. 1996). Furthermore, if the CTD package approaches the sea floor the velocity recording fails altogether since the minimum distance for velocity profiling is about 20--30~m. Both of those data gaps can significantly degrade the final velocity estimate. Finally, the range of the individual profiles in the deep ocean is significantly reduced due to dramatically decreased abundance of scatterer. As a consequence the number of shear estimates per depth cell drop and problems arise in the the detection of statistical outliers.

Note that version 5 and later of the MATLAB software uses a new approach without ever computing velocity shears. It also allows to incorporate bottom track data (Visbeck, 2002).