I was asked to help wade through the processing of a short cruise in mid-2007 on the R/V Atlantic Explorer. The data were collected with an OS75 using VmDAS. Various problems exist with the data, resulting in my prolonged exposure. After a day of examination, I need to quit, but leave this record of what I found. I conclude that at all levels, the Atlantic Explorer ADCP installation should be examined by knowledgable experts, and fixed. The data from the BATS224 cruise I looked at are barely recoverable, and only provide a crude sense of the ocean currents.

Problems with Instrument Configuration

Data were collected using

Beam Orientation

Beam 3 is approximately 45deg starboard of forward. When the ship is on station, all 4 beams should be close to zero. When the ship gets underway, the forward beams should register positive beam velocities and the aft beams should register negative beam velocities of the same magnitude, i.e. forward and aft should be symmetric about zero.

Beam Orientation
Figure: beam_orientation

Problems with single-ping data in beam coordinates

On the Atlantic Explorer, this symmetry is only present in shallow bins. In deeper bins, values that would normally be "bad" by the instrument (missin) are instead replaced by a single value (about 1.6), in all beams. I call this a background "tone" below.

The following figure shows beam1 velocity on station (upper panel) and amplitude (lower panel). Velocities are centered around zero, the background "tone" is reddish). Velocity values degrade starting around bin 30, with two distinct holes around bin 35 and 60. Below bin 75 all data are replaced by the background "tone".

Addional note - The white rectangles are regions of higher noise, possibly low-level acoustic interference by some other instrument, selected from the amplitude panel. In the velocity panel, the shallower region of possible acoustic interference is coincident with a loss of velocity data, filled in by the background "tone" - The vertical black bars were selected as regions of lower amplitude. The same bars, drawn in the velocity panel, are coincident with holes in the deeler velocities (also filled with the background "tone")

beam 1 velocity and amplitude
Figure: on-station beam 1 velocity and amplitude

The following figures show the beam velocity in 4 different bins, going from shallow to deep. The bottom figure is a closeup of the beam values at bin 100. Note the transition from on-station (before decimal day 172.82) to underway, and the increasing presence of the background "tone" in deeper bins.

bin 10 beam velocity
Figure: bin 10 beam velocity
bin 45 beam velocity
Figure: bin 45 beam velocity
bin 70 beam velocity
Figure: bin 70 beam velocity
bin 100 beam velocity
Figure: bin 100 beam velocity
bin 100 beam velocity
Figure: bin 100 beam velocity, zoom in

In this figure, note also that the background "tone" is changing over time.

In underway data (next figure), the background noise level is increased and bubbles (?) causing blockage and increased noise are evident (lower panel, labeled in figure). Shallow velocity (upper panel) has holes where the noise is high, and is still filled with the background "tone" in the deeper water. Some other instrument is als pinging at a regular, but longer interval, and those "hits" are also visible in the amplitude.

beam 1 velocity and amplitude
Figure: beam 1 velocity and amplitude, underway

The transition from on-station to underway is shown in the next two figures in beam velocity, amplitude, and correlation.

One question: "Why do beams 2 and 4 lose the "tone" when the ship is underway, whereas beams 1 and 3 hear the tone where the value should be bad"?

Figure: velocity by beam
Figure: amplitude by beam
Figure: correlation by beam

Problems with Single-ping Data in Earth Coordinates

If all beams have the same beam velocity (eg. zero, or the background "tone") then measured horizontal velocities are zero and error velocity is zero. This has two major implications for processing.

(1) "bad" values are now zero measured velocity, so when ship velocity is added (to obtain ocean velocity) the bad values are now the negative of the ship velocity. If these are not edited out, but are accepted into the averaged data, they will seriously bias the final data in the diretion of motion. (see an example below)

(2) "bad" velocities now have near zero error velocity, so this cannot be used to edit out bad data.

The following figure shows measured velocity on station (east and north components are in upper and lower panels, respectively). Note the ubiquitous near-zero values where we previously saw "bad" data.

measured velocity
Figure: on-station measured velocity

In comparison, underway data has holes where the amplitude was noisy, and a few deep profiles of background "tone" that still make it through. These brightly colored velocity values will have zero error velocity, but will seriously bias any average of the data.

measured velocity
Figure: underway measured velocity

Problems with Averaged Data

Before looking at the single-ping data, I had run the ENX data through CODAS processing (to edit on a single-ping basis, average, rotate, scale). Only the data above bin 35 were salvagable, because values below that were tainted by the background "tone". In addition, the heading device was changing, navigation was sporadic, and times were shifting.

The result was a crude picture of the local currents, but I would put pretty large (10cm/s??) errors on this. Because of the reciprocal track, we can be sure that the flow near the island is NE at about 20cm/s, the flow offshore (64W, 31.5N) is S/SW at about 20cm/s, and we may or may not trust what is farther away:

processed data vector plot
Figure: processed data vector plot

An example of the edited velocities from which the vector plot are draw, is below:

processed data
Figure: processed data u,v

Compare the edited (averaged) data to the original data (below). This is a good example of a dataset where judicious editing is necessary before and after averaging, and there are still problems.

processed data
Figure: processed data u,v


There is something very wrong with the deck unit or transducer system on the Atlantic Explorer. There are setup problems that make processing difficult or impossible.


April 18, 2008.

Unsolicited opinions from Jules Hummon at the Unversity of Hawaii