Main
page
Forum
South-western Atlantic Ocean during 1993-94
Data and Methods
Buoy data
The drifting buoys used in this work were built at INPE (Instituto Nacional de Pesquisas Espaciais - National Institute for Space Research), Brazil, for use in the COROAS project. Like all the surface drifters in use within the WOCE Surface Velocity Program (SVP), the WOCE drifters were fabricated following the design proposed by Sybrandy and Niiler (1991). Owing to their simplicity and low cost, the buoys are also known as Low Cost Drifters (LCDs).
Souza (2000) describes in detail the LCDs design and data transmission process. The LCDs were equipped with a thermistor able to measure the sea surface temperature to a precision of about 0.05 oC. Temperature data and buoy positioning were assessed through the ARGOS system.
The LCDs were deployed in the BC in a position close to the shelf break near 25 oS, 44 oW off the São Paulo State coast, Brazil. The deployment campaign was conducted by the COROAS project, onboard the R.V. Prof. W. Besnard. The buoys were deployed following a box pattern, five at a time, in February 1993, July 1993 and January 1994. For more information about the LCDs’ launching procedure refer to Stevenson and Castro Filho (1996).
Following the general flow of the BC, the 15 buoys launched in this current drifted southwestward covering the COROAS region of interest off the São Paulo coast in Brazil (~23 oS to 27 oS, 42 oW to 47 oW). From the second group of buoys launched in the BC, 2 buoys stopped transmitting data immediately after the launching. All the other 13 buoys traversed the COROAS study area. After a variable time which depended on each particular buoy trajectory, 12 buoys out of these 13 reached the vicinity of the BMC region at about 30 oS. Their trajectories in the SWA are seen in Figure 1. The buoys drifted in the study area from March 1993 to July 1994, having individual lifetimes that varied from 2 weeks to about 11 months.
Time series of position and temperature for particular eddies or rings found in the buoys’ trajectories were manually extracted from the original latitude and longitude time series. The eddies or rings were identified as closed or almost closed tracks found in the total trajectory of a particular buoy. The eddies were also separated according to the surface current in which the buoy was drifting: BC, SAC or BCC. From the positional time series, the displacement time series were computed and, having the start date (di) and finish date (dn) of occurrence of a particular eddy, the period of rotation (TR) is simply dn - di. Using TR together with the measurement of the buoy displacement along the eddy perimeter (P), it was possible to compute the tangential velocity (VT) for each eddy as follow: VT = P/TR.
From the time series it was also possible to compute each eddy’s mean position, mean temperature, temperature standard deviation and to report its direction of rotation (cyclonic or anticyclonic). An empirical estimate of the eddies’ typical diameter was made assuming that all the eddies were circular in shape, which is approximately correct for the biggest eddies. It was also assumed, although not necessarily correct, that the buoys were always surrounding the edge of the eddies. Hence, the eddy diameter (D) is simply the eddy perimeter (P) divided by p. This estimate of the eddy diameter from the buoys’ trajectories is useful for future comparisons with eddies found in the satellite images.
Using the methodology of extracting the eddies as closed tracks directly from the original drifters’ time series, it was possible to resolve eddies in the range of a couple of hours to a maximum rotational period of about 40 days. A total number of 126 eddies were detected during the period of this study. However, in the case when the eddies were tracked for periods longer than their rotational period they were counted more than once.
As expected, there is a direct relation between the eddy rotational period (TR) and its perimeter (P). This relation, however, changes from small scale eddies to mesoscale ones. Small scale eddies of TR lower than 5 days and P lower than 100 km were found to dominate the spectrum of eddies sampled by the buoys (about 65 % of the frequency), although they are more difficult to be spotted directly in the trajectories.
Knowing that the majority of the eddies identified through this methodology were small scale ones, subsequent analysis separated them from the mesoscale ones. We defined two classes of eddies: (a) Class 1: small-scale eddies, TR < 5 days and P < 100 km; (b) Class 2: mesoscale eddies, TR >= 5 days and P >=100 km.
According to Knauss (1997), the scaling of the non-linear and Coriolis (f) terms of the equation of motion is useful when trying to find out the relative importance of these terms in different types of motion in the sea. For instance, the Rossby number (Ro) is the non-dimensional ratio between the non-linear (acceleration) and the Coriolis terms of the equations of motion. Ro is expressed in the form Ro = U / fL, where U is the typical speed of the motion and L is the length scale often associated with the radius of curvature. Considering eddies or rings in the ocean, a common practice is to scale them in relation to their maximum rotational velocities and radius (Olson, 1980; cited by Chassignet, Olson and Boudra, 1990).
From the measurements of the radius (radius = D / 2), tangential velocity (VT) and average latitude of each eddy found in the buoys trajectories, the Rossby number was computed to indicate the relative importance of the aceleration and Coriolis forces in the particular eddies. Statistics for Ro were computed for the eddies in the particular currents BC, BCC and SAC and also in the classes 1 and 2 defined above.
Considering an idealised two layer density ocean in the front between the BC and MC (the western subtropical front in the BMC region) and in the BC/BCC front, the internal Rossby radius of deformation (Rd) was also computed for the range of latitudes where the eddies were present. This property gives a length scale at which the rotational (f) forces become comparable to the buoyancy forces (pressure gradient) in the equation of motion (Richards and Gould, 1996). Rd is defined as Rd = (g' Ho)1/2 / f (Pond and Pickard, 1983) where the reduced gravity (g') is the gravity (g) times the density difference between layers (g' = g (dr/r)) and Ho is the upper layer depth.
In the western STF, BC was considered to carry tropical waters with density (r) of 1025 Kg/m3 in a 200 m water column above SACW, whose typical density was assumed to be 1027 kg/m3. In the BC/BCC front, the BCC density was assumed to be 1023 kg/m3 (coastal waters) extending in a water column of 100 m above TW carried by BC with density of 1026 kg/m3. These numbers were based on the T-S diagrams and vertical profiles of temperature and salinity presented by Castro and Miranda (1998) and Ciotti, Odebrecht, Fillmann and Moller Jr. (1995) for the SBCS.
According to Richards and Gould (1996), wavelengths of about 4 to 6 times the Rossby radius of deformation dominate a fully developed eddy flow. The eddies present in the buoy trajectories will be also characterised in this paper by the statistics of the ratio D / Rd for the particular currents and for classes 1 and 2. Other analyses made of the eddy properties included the linear correlation between eddy size (represented, for instance, by P or D) and eddy rotational period (TR) or tangential velocity (VT) for classes 1 and 2. The linear relationships can be used as empirical models for the prediction of TR or VT of eddies present in satellite images of the study region, where only eddy dimensions can be assessed in particular images.
Satellite data
The high-resolution Advanced Very High Resolution Radiometer (AVHRR) images used in this work were provided by INPE, which operates an antenna in Cachoeira Paulista, Brazil (22 o 41 ’S, 45 oW). Although this antenna is able to record four AVHRR images per day (one image at approximately 12 h for each of the two operational NOAA satellites), problems of storage space and lack of personnel involved in the acquisition phase consistently caused loss of acquisition.
Because of the perceived importance of the COROAS project, however, INPE managed to guarantee the consistency of AVHRR image recording since 1992. Because of a shortage of recording media (CCT tapes), just one single image was recorded at INPE per day although this was adequate for the COROAS objectives. During the months of Austral winter (June to August), the cloud coverage in the SWA tends to increase dramatically, leaving sometimes the entire COROAS area of study without any useful information. In this case, the COROAS AVHRR scenes were simply not recorded and stored at Cachoeira Paulista.
After the selection of the interesting scenes (mainly the ones corresponding to the same period of time when the LCDs were in the water, March 1993 to July 1994), the AVHRR scenes stored in the CCT tapes were extracted and the sea surface temperature (SST) images were generated. The SST images were generated according to the National Oceanic and Atmospheric Administration (NOAA) algorithms described by Kidwell (1995).
A total number of 81 SST images were generated for this work. They cover the period between 10 March 1993 and 11 July 1994. All the images were resampled to 4 km x 4 km pixel size and geolocated to the Mercator projection over a coherent area in the SWA (26.4 oS to 42.7 oS; 38.8 oW to 58.8 oW).
Individual eddies were visually located in the high-resolution AVHRR images. Typically, the eddies present in the satellite SST images are identified as closed elements with the borders delimited by strong horizontal thermal gradients in relation to adjacent waters. The gradients, however, change in intensity according to the stage of formation or coalescence of a particular eddy. Besides, the absolute temperatures inside and outside the eddies are also not constant, which make very difficult the establishment of a single palette of colours to reveal eddies in a temporal sequence of SST images. Cloud coverage is another crucial problem whenever a particular feature needs to be tracked in satellite images.
In this work, each AVHRR image was processed independently to enhance the presence of eddies. Each time one of these features was located, it was treated as an ellipse, and its minor and major axis were measured. From the ellipse’s equation, the eddy perimeter (P) was estimated. Moreover, the mean latitudinal and longitudinal position of the eddy was assessed, and the temperature profiles for the eddy’s minor and major axes were generated. From these profiles (and discarding the cloud covered pixels when detected), an average temperature was computed for each particular eddy.
Using this procedure, 78 eddies were located in the overall set of high-resolution AVHRR images. From this total a few were double counted when they persisted from one image to the subsequent one. Mainly because of cloud coverage but also advection, unfortunately, the location and tracking of a particular eddy in a sequence of images was rare. This made the estimation of the eddies’ lifetime extremely difficult.
The majority of
the eddies found in the satellite images have perimeters bigger than 100
km. This contradicts what was found in the buoy trajectories, but can be
explained by the visualisation technique employed here (in which, of course,
bigger eddies are visually easier to detect than smaller ones). The warm
core and cold core eddies present in the AVHRR images were subject to simple
statistical analysis in order to assess their typical length scales. Their
statistics were compared to the ones found for the eddies present in the
buoys’ trajectories. The ratio between the eddies’ average diameter and
the Rossby radius of deformation was also computed.
rbds@soc.soton.ac.uk (c) April 2001, all rights reserved.