RAFOS Float Data Report of the North Atlantic Current Study 1993-1995
by S.Anderson-Fontana, M.Prater, & H.T. Rossby
1. Introduction
This is the final data report of all RAFOS float data collected during the 1993-1995
study of the North Atlantic Current (NAC) and adjacent waters in the Newfoundland
Basin. The objective of the program, jointly supported by the Office of Naval Re-
search and the National Science Foundation, was to study the structure of the cur-
rents in the NAC region and the exchange of waters between the subtropical and sub-
polar gyres in the Newfoundland Basin. One hundred floats were deployed on two den-
sity surfaces corresponding to Sigma-t = 27.2 and 27.5, respectively. All floats
were designed to cycle once or twice a day to density surfaces 0.1 Sigma-t units
above and below their nominal level to determine changes in stratification and temp-
erature along the trajectories. Three separate float deployments took place: July-
August 1993, November-December 1993 and October-November 1994. The first cruise was
on the R/V Oceanus, the other two on the CSS Hudson. CTD casts were taken at nearly
all deployments. Most float missions had a duration of ten months. The floats were
tracked using four moored sound sources developed by Sparton of Canada, energized by
power modules from Webb Research Corp. of Falmouth, Massachusetts (Rossby et al.,
1993). The sources were deployed on the R/V Oceanus cruise (July 1993) and recovered
on a CSS Hudson cruise in June 1995.
2. Description of RAFOS f/h floats
The float used in this project is based on the design developed for the Anatomy of
Gulf Stream Meanders project (Rossby et al., 1994). That paper should be consulted
for a detailed discussion of the principle of operation & some first test results
in the Gulf Stream. The f/h float is basically a standard isopycnal RAFOS float to
which a small pump, or volume changer (VOCHA), has been added. The pump changes
the volume or, equivalently, the density (since the mass is constant) of the float
+/- 0.01% (1.5 cm3 in 15 liters), causing it to move up or down to the correspond-
ing isopycnal. Once the float is at equilibrium (which was assumed to be attained
within a given time delay), the pressure and temperature are recorded. The differ-
ence in pressure between the two neighboring surfaces yields an estimate of strat-
ification. The floats are made very nearly isopycnal in seawater by adding a comp-
ressible element (the compressee) so that the complete float has very nearly the
same compressibility (+/- l%) as that of the seawater. See Rossby et al. (1985) or
Goodman and Levine (1990) for a detailed discussion.
3. Float Ballasting
All floats were ballasted at the Graduate School of Oceanography. Ballasting con-
sists of several stages, all to ensure that the float will descend to the desired
level (or density surface). We will summarize the ballasting procedure here. First
the assembled float is weighed on a digital precision scale, then "preballasted"
in an unpressurized tank filled with room temperature tap water. The float is ini-
tially positively buoyant in the tank, and weights (approximately 2 kg) are added
until the float is almost neutrally buoyant. This extra weight is released when
the float mission is over, thus allowing the float to rise to the surface and ex-
pose the glass-enclosed Argos antenna. Free-hanging chains are taped to the bot-
tom of the float before the float is placed in the high-pressure tank. Additional
trim weights are added so that the float has a slight negative buoyancy, with some
of the chain now resting on the floor of the pressure tank. The float is position-
ed so that a video camera, aimed through a porthole in the side of the tank, is
able to view a graduated paper scale inside the float. The tank is then closed
and pressurized. The float itself is significantly less compressible than water,
and increases in buoyancy (and thus rises) with increased pressure, lifting more
of the suspended chain off the tank floor. The f/h floats, however, have a comp-
ressee, whereby the compressibility of water is nearly matched, & therefore have
little or no change in buoyancy with pressure. The scale height versus tank pres-
sure is recorded, and this scale reading indicates the length of chain suspended.
The tank's pressure is slowly cycled up and down, and stopped at intervals to re-
cord the pressure level and the scale reading. When we feel that the pressure ver-
sus scale reading relationship is well defined, the tank is depressurized, the
water temperature is measured again, the trim weights are removed and weighed, &
the float is removed from the tank. Since individual floats vary slightly in pro-
perties (due to irregularities in the glass and stiffness of the compressee), in-
dividual pressure-scale relationships (or more precisely, compressibility) must
be computed, and thus this ballasting process must be repeated for each float.
The pressure-scale relationship allows us to determine the correct weight to add
to the float to achieve neutral buoyancy at the target pressure, albeit with tank
water at tank temperature. To correct for the target temperatures and salinities,
the target density and the density of the tank water are computed from the seawa-
ter equation-of-state (Fofonoff and Millard, 1983). The difference between these
densities multiplied by the float's target volume (accounting for the float's
compressibility and thermal expansion) gives the additional weight required to
assure neutral buoyancy at the specified level in the ocean. This seawater weight
correction is approximately 500 grams. For a float to reach a target isopycnal to
within 0.1 sigma-t, the float's equation-of-state must be known to (and the mass
of the float corrected to) 1.5 grams out of a total float mass of 15000 grams.
The uncertainty of the ballasting procedure itself is less than 0.03 sigma-t units
(0.5 gram), but this is for fresh water in the tank. In practice the accuracy is
closer to +/- 0.1 sigma-t in the three groups of floats that were ballasted, as
can be seen from the differences in nominal and estimated sigma-t in Tables 2a-c.
Knowledge of the ballasting errors comes from a detailed comparison with the CTD
casts made at the time of deployment and the time history of temperature during
their missions. The reason for this larger error comes from the additional step
of adding a nearly 500 gram weight to make a float neutrally buoyant on an iso-
pycnal surface in the ocean. In principle this is very straightforward. In prac-
tice, errors creep in, primarily due to variations in apparent density resulting
from the type of anti-corrosion treatment or paint used to protect the different
groups of drop weights. We know from more recent experience ballasting 12 floats
for a Gulf Stream project in September 1995 that these errors or uncertainties
can be reduced significantly. Those 12 floats, whose add-on weights were not
painted or given any surface treatment, went to the target density +/- 0.03
sigma-t units.
In discussing the density surfaces, we often refer to a certain sigma-t surface.
In reality we think the floats behave much more like specific volume anomaly
surface followers. A very useful discussion of RAFOS float & ballasting issues
can be found in a report edited by A. Bower (1994).
4. Deployment of Sound Sources
The sound sources were developed by Sparton of Canada to provide efficient wide-
area insonification. The power and electronic packs were supplied by Webb Re-
search Corp., with energy sufficient for well over two years of service at two
transmissions/day. The signaling system itself is the conventional SOFAR signal
and consists of a frequency-modulated sweep of 80-s duration, where the frequency
is incremented linearly from 259.375 to 260.898 Hz in 2-s steps, with phase cont-
inuity preserved at each step (Webb, 1977). The source has a dipole radiation pat-
tern with a horizontal source level of 195.5 dB re 1 microPascals at 1 m. (See
Rossby et al.,1993 for a complete description of the system.) Table 1
shows the locations of the four sound source moorings. The sources were at a depth
of about 1400m. All sources functioned reliably for the duration of the study, but
SS#2 (mooring #M3) had a significantly reduced source level due to a spurious mech-
anical resonance close to the operating frequency, resulting in a degraded trans-
mitting voltage response. The power packs were kept separate from the resonant pipe
projectors via a 100 foot long umbilical cable in order to minimize possible damage
due to vibration from the source. This may have been fortunate, for after recovery
considerable internal chafing around the shock mounts for the electronics was dis-
covered.
5. Deployment of Floats
The deployment strategy of the project was as to release floats (1) from three dif-
ferent cruises so as to sample the NAC at different times; and (2) on both sides of
and within the NAC to explore cross-frontal pathways and exchange processes. The two
density surfaces chosen were sigma-t = 27.2 and 27.5. The deeper surface is the
shallowest one that does not make local contact with the atmosphere in winter, where-
as the shallower surface was chosen because it was expected to outcrop, at least
north of the subpolar front (the eastward extension of the NAC). Mission lengths of
ten months ensured that the floats would experience all four seasons. The last group
of floats had only eight month missions due to the scheduled recovery of the sound
sources in June 1995. Tables 2a, 2b and notes to 2b and 2c summarize in detail all
float deployments: date, position, sigma-t surface (target and estimated) & mission
length, with footnotes as necessary.
6.Float Tracking
A total of 100 floats were released on the three cruises. At the end of its mission,
each float releases ballast, returns to the surface, and telemeters data to Argos, a
French satellite-based data collection and platform location system. Once all the
float's data are transferred to our computer system and processed, the trajectory of
the float can be reconstructed from the time series of acoustic travel times. The
pressure and temperature time series are also included in the telemetered data. Most
floats transmitted data for five to six weeks before their batteries wore out,though
each float data set was complete, or nearly so, in less than two weeks. There were
only four floats that failed to transmit any information: 250, 251, 278 and 319.
Three floats ended their missions early when CPU activity, as monitored by their
watchdog timers, ceased (253 and 322 after 57 days, and 338 after one day); two lost
their ballast weights very early, one (302) during deployment & the other (271) soon
after; one had no acoustic data (281); one (317) transmitted only 10% of its data; &
one (290) failed to transmit any reliable data. A number of floats in the 1st group
and a few later ones (297, 309 and 323) had a subtle problem in their temperature
circuits. The thermistor resistance controls the frequency of an oscillator circuit.
If the frequency was very close to certain values related to the computer clock
cycle, the frequency would,in some floats, "lock" onto it. This was due to the omis-
sion of a by-pass capacitor in the counter chip, which was added in the later float
releases. Tests were unable to find any bias or errors in temperature at frequencies
in between these locked frequencies.
The floats were able to hear strong signals from sources 1, 3 and 4 until the end of
each listening window (25 minutes). In most cases, the travel time series from the
three sources were sufficient for accurate float tracking, with very few source
geometry problems due to the absence of usable signals from source 2.
Float and sound source clock drifts are factored into the float trajectory calcula-
tions. The float clock drift is obtained from the time of the first transmission
from the float to the Argos system compared to the expected time of the first trans-
mission. The sound source drifts were first estimated based on the arrival times at
a few selected floats just prior to surfacing, compared to the expected arrival
times at the surface positions (assuming little surface drift). We assume a linear
sound source clock drift. Errors are introduced in these calculations, however, due
to (1) the elapsed time between the recording of the final arrival times by the
float prior to surfacing and the transmission of the first Argos position (several
hours); and (2) the chosen speed of sound used in the calculations. Fortunately when
the sound sources were recovered in June 1995, we were able to obtain an accurate
time of transmission from sound source 1 during the recovery process. Based on this
information and the earlier calculations showing sources 1, 2 and 3 to have very
similar drifts, we've determined the drifts as listed in Table 1. The drift for
source 4 is based on its relationship to the other 3 in the earlier calculations.
URI North Atlantic Current Floats Fall 1995 Addendum
Flt. # Launch Surface Launch Mission
(1st ARGOS) yearday length
(1995) (days)
-----------------------------------------------------------------------
359 37.613 -72.810 33.876 -51.476 252 180
360 37.198 -72.998 36.980 -52.494 243 180
361 37.158 -72.970 34.570 -64.493 243 180
362 37.198 -72.998 50.079 -33.253 243 180
363* 37.158 -72.970 47.087 -30.939 243 180
364 37.717 -72.772 31.017 -46.635 252 180
365 37.618 -72.807 39.157 -50.660 252 180
366 37.722 -72.770 39.529 -41.114 252 180
367* 37.610 -72.813 39.258 -35.983 252 180
368 37.112 -72.962 35.318 -68.108 243 180
369 37.725 -72.768 35.151 -50.265 252 180
370 37.112 -72.962 44.944 -40.672 243 180
------------------------------------------------------------------------
Note:
Only two sound sources were available for tracking these floats. The
geometry of the sources prevented tracking of the floats generally
beyond 30 to 60 days. Temperature and pressure data only are available
after that time.
* Records have been cut off when the floats surfaced prematurely.