Isopycnal floats as platforms for in-situ studies

By members of the RAFOS group

     It is now well established that isopycnal floats follow density surfaces very closely on all time scales, including those for internal waves. By isopycnal we mean that floats follow specific volume anomaly or delta-surfaces. We can use this property to study various processes: isopycnal stirring and mixing from in-situ changes in temperature, vorticity conservation from along-track variations in Lagrangian stretching vorticity, the internal wave sea state from pressure, and biologically important parameters such as oxygen.
     Lagrangian trajectories, particularly in fronts and eddy fields, reveal pathways of cross-frontal exchange (at various depths) and how fluid parcels move laterally and vertically in growing and decaying meanders. Changes in temperature signal incursions of water with different temperature-salinity properties and thereby indicate lateral stirring and mixing processes. These can occur quite suddenly in fronts, a clear indication that strong lateral shear rapidly increases the contact area between adjacent water masses.
     Stretching vorticity, defined as the distance between two specific volume anomaly surfaces , or static stability, , has been measured by numerous floats deployed in the North Atlantic Current (NAC) study. One float, caught in an anticyclonic lens, indicated a pycnostad with almost zero stratification implying recent exposure to the atmosphere. Generally speaking, floats evince greater N2 activity in energetic than in quiet regions.
     All floats in the ACCE study measured dissolved oxygen along their tracks. Floats that surface due to winter time convection show the oxygen levels rapidly approaching saturation. In spring, as the seasonal thermocline reestablishes itself, a gradual reduction in O2 levels takes places indicative of oxygen utilization. Near the surface changes in O2 levels can be quite strong and apparently rather patchy because they can quickly get erased by mixing events whereas in deeper more remote regions from the surface the O2 records are impressively quiet and stable at the ±0.1 ml/l level....

     We normally view the isopycnal RAFOS float as a marker of fluid motion (the Lagrangian approach), but one can also view it as a platform from which observations of in-situ processes can be conducted. While it has been shown elsewhere (Hebert ..., xxxx) that isopycnal floats can be designed to follow internal waves very closely, we focus on the idea of using floats as a framework or platform for in-situ studies. In this poster we present some examples of phenomena and processes that might be amenable to more systematic study. We begin with examples of thermal variability along isopycnal float tracks, continue with a discussion of thickness (stratification) variability, internal waves and tides and lastly studies of in-situ oxygen. The data come from two major studies: The North Atlantic Current study (NAC) between 1993-1995, and the Atlantic climate Change Experiment (ACCE) between 1997-1999.

Isopycnal operation
     The degree to which a float follows an isopycnal surface depends primarily upon how well its compressibility matches that of sea water. By adding a carefully designed compressible element (or compressee) to the float (which doubles as expendable ballast so the float can surface at the end of its mission) we can match the compressibility to within ±0.5 to ±1% of target value. In more practical terms this means that if a float deepens 600 m from one side of the Gulf Stream to the other, it can be within 10 m of the layer it originally floated on. If it were important enough, these numbers can be tightened further in our new ballasting facility by more deliberate matching of compressees and instrument housing (glass pipes). For isopycnal operation the floats must also have zero coefficient of thermal expansion. While this can be achieved in principle using a housing made of pure quartz, a simple and cost-effective approximation is to use borosilicate glass with which we obtain volume expansion coefficients about 1/20th to 1/10th that of sea water. All RAFOS floats, no matter what design, measure pressure and temperature with a resolution of <1 dbar and <0.02°C and an accuracy at O(5) dbar level and to within ±0.1°C.
     All NAC floats cycled between three density surfaces twice per day (by expanding and contracting its volume a prescribed amount (=1.5cc), thereby measuring a layer thickness centered about the center surface. The objective was to examine layer thickness variations (or stretching vorticity) in relation to their mesoscale activity. By cycling symmetrically above and below we minimize the effect of vertical shear on the Lagrangian tracking of the center layer. Internal wave activity leads to measurement uncertainty of about ±10 m, but with a four-day running average the layer thickness uncertainty is reduced to a few % such that long-term trends and changes emerge clearly.

In-situ temperature
     The most striking examples of temperature variability come from the floats deployed in the NAC study. By design these were deployed in and to both sides of the NAC, which is the current that separates two of the most strikingly different water masses in the world: subpolar Labrador Sea waters and tropical waters brought east and north by the Gulf Stream. Figure 1 shows a fairly simple track of a float (255) crossing the NAC (top) with the corresponding temperature (middle) and pressure records (bottom). Based on the track and the pressure record, we observe the float looping about on the cold side of the NAC brushing against the NAC around days 20-30, 105 and then crossing the NAC between days 140-170.
     Noteworthy: Note the tight correlation between temperature and pressure: as the float deepens (off-shore) the temperature increases. Temperature is a local observable while pressure tells us the float's position in relation to the baroclinic front of the NAC. This correlation might seem strange for one might expect the float to advect the temperature signal with it as it crosses the front, and only gradually through stirring and mixing lose memory of its origin and reflect its new environment. Instead, the rapid response of temperature seems to reflect a shearing of the entrained fluid by the strong advection of waters in the NAC. As a float crosses the NAC, the fluid parcel it was originally associated with becomes highly stretched in the downstream direction, increasing the contact area with the waters coming along the current. From other floats in the NAC project we find it rather striking how consistently this process of stirring into its new fluid environment takes place. But on closer inspection the temperature changes can exhibit a step-wise, irregular and sometimes delayed response. This steppiness appears to reflect a patchiness that can require weeks or more to attenuate. Other NAC floats that exhibit this rapid temperature adjustment crossing the NAC include 269, 272, 282, 309, but we should add that only a few floats cross the entire NAC from one side to the other in the south. Farther downstream (to the north), as a result of continued stirring across the front by the meandering of the NAC, the correlation between pressure and temperature weakens.
     We can use the vertical cycling of the float to sample the temperature of the adjacent upper and lower surfaces. We find that these also have a patchy character and that this patchiness changes more rapidly in time than on the center isopycnal. This can be expected due to the vertical shear between these surfaces. In this sense the floats sample the horizontal structure on these surfaces, but the corresponding spatial scales can only be surmised since we don't know the vertical shear with any accuracy. Figure 2 show examples of the temperature time history on the three surfaces for two floats. Float 282 (top) moves in and out of the NAC from the cold water side where the stratification is maintained by much colder and fresher water just below the surface. Float 335 (bottom) is well removed from any sharp water mass boundary.
     Noteworthy: Note the difference in small scale patchiness for the two cases. First, the frontal case exhibits sharp jumps of up to several degrees corresponding to the large differences between the Labrador Sea and North Atlantic Current waters compared to mid-ocean variability. Second, note the inversions in temperature reflecting interleaving of very different waters. This clearly indicates an interleaving character to the stirring between water masses. The stirring on the three surfaces takes place concurrently, but essentially independently even though the surfaces are separated only 50 to 100 m apart in the vertical. Third, note the fast time scales in the frontal region, reflecting no doubt the strong lateral shear compared to the open ocean. Since the floats follow specific volume anomaly or delta-surfaces, the temperature record will have a corresponding unique record of salinity (or Veronicity or spiciness).
     Question: As the floats move away from the frontal region, this small scale activity decreases in amplitude and rate. We find it striking that on the one hand the broad pattern of temperature change follows pressure (or position across the front) rather closely, but the following stirring and ultimate mixing takes much longer. Can these observations be used to make more specific inferences about the nature of isopycnal stirring and mixing in different regions?

Stretching vorticity or static stability
     The primary objective for cycling the floats between adjacent delta-surfaces was to examine how Lagrangian layer thickness varies in relation to their trajectories particularly in and out to either side of the strongly baroclinic front of the NAC. These observations have proven more difficult to work with than anticipated because layer thickness variations rarely correlate with other variables, in particular depth of the isopycnal as we originally had hoped. The vertical structure of the NAC can be characterized rather effectively in terms of a barotropic and gravest baroclinic mode. The higher order modes contain little kinetic energy although they are required for a more detailed characterization of the profile. One case where layer thickness did exhibit a qualitative correlation is for the same float discussed above, 255. Figure 3 shows the float's pressure and temperature record, and now also layer thickness. Curiously, the 30% decrease in layer thickness decreases as the float crosses the NAC exceeds by far a corresponding difference in thickness of the subthermocline layers (viewing the ocean as a two-layer system). Something else is going on here.
     Noteworthy: The fact that the temperature of this float adjusted so rapidly as it transited the NAC, and apparently layer thickness too, suggests that the potential vorticity of the layer also is modified by the same cyclonic/anticyclonic shear of the NAC during its cross-frontal transit. We emphasize again that we observed this type of side-to-side transit occurred only a few times. Dutkiewicz et al. (2001) have pointed out that such transits must, in some sense, occur only rarely else their be greater homogenization of adjacent water masses (such as for the sub-thermocline and deep waters of the Gulf Stream). Figure 4 illustrates a more typical case (float 282) where the layer thickness variations, although substantial, show little correlation with layer depth.
     Question: Note the rather slow variations in the low-pass filtered record of layer thickness. On the one hand this does not surprise since layer thickness contributes to the pressure field, which has a larger spatial scale associated with it than the velocity field. On the other hand, layer thickness is a component of potential vorticity, which includes second derivatives of pressure. If it were possible to obtain a Lagrangian frequency spectrum of relative (curvature and shear) vorticity and layer thickness, would they have the same shape? It seems likely that relative vorticity will exhibit faster time scales than layer thickness, yet PV must be conserved? A difficulty here is that floats are point observers and thus include all scales, whereas a statement about conservation of PV must be limited to a scale where dissipation can be ignored.
     The most striking example of variations in layer thickness was observed when a float got entrained into a large lens of warm, salty water in the northern Newfoundland Basin. The increase in layer thickness was at least a factor two, so large in fact that the float almost certainly didn't have time to reach equilibrium at the upper surface before changing volume to sink to the lower delta-surface. Prater and Rossby (2000) discuss this float in detail.

Internal wave and tide activity
     There have been some indications that the internal tide can exhibit very large amplitudes in the eastern North Atlantic. Deep drifters in meddy 'Sharon' revealed a conspicuous (80 m peak-to-peak) fortnightly modulation of the internal tide (Rossby, 1988). We use these isopycnal platforms to discern internal wave and tide activity including any amplitude modulation of these (with time).
     In another study using the COastal Ocean Lagrangian (COOL) float, it has been shown that isopycnal floats can track a density surface with considerable accuracy after they have reach in-situ equilibrium. Even on time scales of minutes, a float with a specially designed compressee will remain within a meter of its equilibrium surface (Hebert ref.). All NAC floats were equipped with a similar compressee. We think these floats tracked vertical displacements as small as 2-3 meters. However, these floats sampled pressure only twice a day such they give only a limited and aliased view of high-frequency activity. The ACCE floats, on the other hand, were programmed to sample pressure every four hours such that we can resolve the internal semidiurnal tide. Due to the longer missions in ACCE the floats carried a more corrosion-resistant compressee which has a higher level of stiction. This means that for perturbations less than O(5) meters the float will tend to undershoot the displacements by a factor 2-3. For larger excursions in the vertical the floats should track delta-surfaces reasonably well. To look at the pressure record a filter was constructed to catch and remove outliers and bad data. These were replaced by linear interpolation. The high-pass filter consisted of a Matlab 6-pole Butterworth filter with the high-pass frequency set at 0.3x1/8 hours = 1/26 hours.
     One of the quietest float records is that of float 576, which drifted into the southern Iceland Basin (but there are several others about as quiet). Figure 5 shows the high-pass filtered pressure and temperature records (top two). The bottom two show the full pressure record (left) and a scatter diagram of the pressure and temperature fluctuations (right). The last panel checks whether a correlation exists between the two. The plots show that the pressure variance remains rather limited at all times. The high-pass filtered standard deviations for pressure and temperature are 3.09 dbars and 0.043°C respectively. But one can also see larger fluctuations in the early part and a quiet section between roughly 300 and 400 days. The rather strong correlation between T' and P' suggests that the compressee is sticky at small these amplitudes.
     In contrast, float 448, Figure 6, exhibits some exceptionally strong and rapid fluctuations near the end of the record. A visual examination suggests a diurnal period. By running a local least squares fit to a sinusoid with 24 hour period we obtain an amplitude of 20 m. It is quite striking how suddenly this activity turns on. The rest of the record shows little difference from that of float 576. A third float, 475 in the southern Irminger Sea, Figure 7, shows for a period of time semidiurnal tidal activity very strong modulation with a period of about a month. It is evident from an examination of numerous ACCE floats that large tidal activity occurs comparatively rarely, yet when it does so, the signals can be quite large.
     Question: What causes the internal tide to have such an intermittent character?

In-situ observations of dissolved oxygen
     All URI ACCE floats measured dissolved oxygen using the ‘pulsed’ technique pioneered by Dr. Chris Langdon (LDEO). The sensors are only briefly activated once a day and thus can operate over the entire lifetime of the floats. From a Lagrangian viewpoint, interpreting the variability of dissolved oxygen presents the difficulty of separating the biological from the physical processes. However, there are times when the nature of the variability clearly indicates that one process dominates the other.
     Figures 8a and 8b show the track of a float (550) which outcrops in the Irminger Sea with the corresponding measurements of pressure, temperature, and dissolved oxygen, plus the satellite-derived values for significant-wave-height (SWH) and SeaWiFS Chl-a. The track and property plots are color-coded to indicate two distinct periods: red- when the float outcrops into the wintertime mixed layer; green- when either the oxygen record indicates significant removal due to respiration or the Chl-a record indicates high rates of production. When the float outcrops, the oxygen saturation level remains at 104% ± 2%. However, at the end of the first winter the oxygen shows momentary drops as large as 30% (2 ml/l). These occur at a time when the Chl-a indicates the onset of primary production and suggests that sinking of detrital matter may be concentrated to these newly-stratified density surfaces. Around day 340, late June, the oxygen record approaches a local minimum in dissolved oxygen. This takes place after the peak of spring-time production. We believe that this is another, although less intense, indication of removal from detritus. On day 360, the summer bloom peaks simultaneously with a slight, but noticeable, 2% (0.1 ml/l) momentary drop in dissolved oxygen. However, this bloom is accompanied by a O(40m) increase in depth, which could have also accounted for the oxygen drop.

     We have shown how the isopycnal float can be tuned to follow a specific volume anomaly or delta-surface quite closely. It can then be used to not only to identify Lagrangian pathways, but also to study local in-situ phenomena or processes, be they a dynamical signal or fluid property. Examples of layer thickness, internal wave and tide activity, temperature and oxygen variability in energetic and quiet regions were given. But we think this is just the beginning. One can readily imagine floats designed specifically to measure internal wave activity continuously and perform running spectral analyses to reduce the data volume before transmission. The intermittent nature of the internal tide is curious. Some day, when we are truly confident in the long-term stability of conductivity sensors, these could be added to the float to independently check the isopycnal behavior of the floats. Similarly, one can surmise an emerging interest in chemical/biological parameters such as pH, chlorophyll, down- and upwelling light (near the surface) and zooplankton biomass (measured acoustically) at depth.

     Hebert, D., M. Prater, J. Fontaine and T. Rossby. 1997: Results from the test deployments of the COastal Ocean Lagrangian (COOL) float, GSO Technical Report, 97-2
     Prater, M.D. and T. Rossby, 1999: An alternative hypothesis for the origin of the "Mediterranean" salt lens observed off the Bahamas in the fall of 1976. J. Phys. Oceanogr., 29, 2103--2109.

- - - - - - - - -
The RAFOS group at the Graduate School of Oceanography, University of Rhode Island, Kingston, RI 02881 includes

J. Fontaine: [Tel: 401-874-6946, Email:]
S. Fontana: [Tel: 401-874-6519, Email:]
D. Hebert: [Tel: 401-874-6610, Email:]
P. Lazarevich: [Tel: 401-874-6516, Email:]
Paula Perez-Brunius: [Tel: 401-874-6516, Email:]
M. Prater: [Tel: 401-874-6512, Email:]
T. Rossby: [Tel: 401-874-6521, Email:]
H. Zhang: [Tel: 401-874-6512, Email:]