JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, C10029, doi:10.1029/2009JC006087, 2010
Coastal and mesoscale dynamics characterization using altimetry
and gliders: A case study in the Balearic Sea
Jérôme Bouffard,1 Ananda Pascual,1 Simón Ruiz,1 Yannice Faugère,2
and Joaquín Tintoré1,3
Received 29 December 2009; revised 9 June 2010; accepted 21 June 2010; published 13 October 2010.
[1] Dynamics along the continental slopes are difficult to observe given the wide
spectrum of temporal and spatial variability of physical processes which occur (coastal
currents, meanders, eddies, etc.). Studying such complex dynamics requires the
development of synergic approaches that use integrated observing systems. In this context,
we present the results of an observational program conducted in the Balearic Sea
combining coastal gliders and altimetry. The objectives of this experiment are to study
regional dynamics using new technologies, such as gliders, in synergy with satellite
altimetry and to investigate the limitations and potential improvement to altimetric data
sets in the coastal zone. In this regard, new methodologies have been developed to
compute consistent altimetric and glider velocities, and a novel technique to estimate
absolute glider velocities, combining surface glider geostrophic velocities with integrated
currents estimated from the glider GPS positioning, has been applied. In addition, the
altimetric velocity computation has been improved, especially in the coastal zone, using
high‐frequency along‐track sampling associated with new filtering and editing techniques.
This approach proves efficient for homogenizing the physical contents of altimetry and
glider surface currents (percentage of standard deviation explained is >40) and
characterizing regional dynamics in the Balearic Sea through a combined analysis of a
high‐resolution observing system, such as the appearance of anomalous intense mesoscale
features missing in the classical circulation scheme of the Balearic Sea.
Citation: Bouffard, J., A. Pascual, S. Ruiz, Y. Faugère, and J. Tintoré (2010), Coastal and mesoscale dynamics characterization
using altimetry and gliders: A case study in the Balearic Sea, J. Geophys. Res., 115, C10029, doi:10.1029/2009JC006087.
1. Introduction
[2] New monitoring technologies are being progressively
implemented in coastal ocean observatories, increasing our
understanding of coastal and nearshore processes and contributing to a more science based and sustainable management of the coastal area.
[3] By autonomously collecting high‐quality observations
in three dimensions, gliders allow high‐resolution oceanographic monitoring and provide useful contributions to the
understanding of mesoscale dynamics [e.g., Hodges and
Fratantoni, 2009; Ruiz et al., 2009a] and multidisciplinary
interactions that significantly affect upper ocean biogeochemical exchanges, an issue of worldwide relevance in the
context of climate change. However, isolated measurements
from fleets of gliders are not sufficient as, for many processes, glider measurements remain scarce, both in space
1
Marine Technologies, Operational Oceanography, and Sustainability
Department, Mediterranean Institute for Advanced Studies, Mallorca,
Spain.
2
CLS Space Oceanography Division, Ramonville, France.
3
Also at Balearic Islands Coastal Observatory System, Mallorca, Spain.
Copyright 2010 by the American Geophysical Union.
0148‐0227/10/2009JC006087
and time. Instead, a multisensor approach that combines in
situ and remote‐sensing measurements should provide a
better understanding of observed features, especially through
an increase of space‐time coverage. This is one of the key
conclusions and recommendations reached during the recent
OceanObs09 Conference.
[4] To properly address the new scientific challenges associated with the coastal marine variability (including physical,
biogeochemical, and ecosystem variations), an intensive
observational program has been conducted in the Balearic
Sea (western Mediterranean), in particular aiming at combining coastal glider and satellite altimetry data. The physical content of the two data sets and their potential synergy
and limitations in the coastal domain have been explored.
Four glider missions have been used, carried out from July
2007 to June 2008. These missions were approximately
simultaneous and well colocalized with Envisat satellite passages whose measurements are almost perpendicular to the
main Balearic oceanographic features (see Figure 1). The
Balearic subbasin is a region of worldwide interest characterized by both permanent and variable signals, covering a
wide range of dynamical scales from intense mesoscale (in
terms of filaments, eddies, or shelf‐slope flow modifications)
to seasonal and interannual variability. The main Balearic
C10029
1 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
C10029
Figure 1. Location of Envisat track 773, bathymetry and main surface circulation characteristics of the
Balearic subbasin (western Mediterranean) [modified from Pascual et al., 2004].
oceanographic features share the results of process interactions at basin, subbasin, and local scales. Thermohaline
adjustments take place over the western Mediterranean Sea
and govern the eastward and northeastward spreading of
Atlantic Water (AW) inflowing through Gibraltar Strait.
This flow can reach the Balearic Islands, therefore influencing north and south heat transport and exchanges that
appear to have key relevance to upper level ecosystem variability and fisheries.
[5] The general surface circulation of the Balearic Sea is
controlled by the presence of two fronts and their associated
currents [Font et al., 1988; Font, 1990; Pinot et al., 1994;
Onken et al., 2008]. The Catalan front is a shelf‐slope front
that separates old AW, in the centre of the Balearic subbasin, from the less dense water transported by the Northern
Current (NC), which is also old AW but is fed into the Gulf
of Lions and the Catalan shelves by fresh continental water.
The NC flows southwestward along the continental slope
until it either exits the basin through the Ibiza Channel, or
retroflects cyclonically over the insular slope forming the
Balearic Current (BC; see Figure 1). The BC is also fed by
recent warm and fresh AW waters coming from the Algerian
Basin through the Mallorca and Ibiza channels. Both the
Northern and Balearic currents have widths of the order of
50 km and are in good geostrophic balance, as winds only
seem to produce transient perturbations in near‐inertial
oscillations [Font, 1990]. In addition to the general basin
scale circulation, the Balearic subbasin is also characterized
by frontal dynamics near the slope areas: Mesoscale eddies
[Tintoré et al., 1990; Pinot et al., 2002; Rubio et al., 2009]
as well as filaments and shelf‐slope flow modifications [La
Violette et al., 1990] have been found to modify, not only
the local dynamics (significant vertical motion associated
[Pascual et al., 2004]), but also the large scale patterns, as
shown by Pascual et al. [2002] in a detailed study of the
blocking effect of a large anti‐cyclonic eddy, as well as
showing a clear influence of basin circulation on phytoplankton biomass [Jordi et al., 2009]. Seasonal frequency
observations of the NC [Béthoux, 1980; Font et al., 1988]
reveal higher transports in winter than in summer (1.5–2 Sv
and 1 Sv, respectively), while the opposite is found in the
BC (0.3 Sv in winter compared with 0.6 Sv in summer).
Regarding interannual variability, a recent study with along‐
track satellite altimetry [Birol et al., 2010], has shown large
year‐to‐year differences, suggesting complex non‐linear
interactions between basin scale and mesoscale circulations.
[6] Several previous studies at the Balearic subbasin scale
have been done, however most were from observational
cruises carried out on the mainland side and focused on the
NC or on the exchanges of water through the Balearic
channels; few experiments have been devoted to study the BC
and its associated front. In a recent work, Ruiz et al. [2009b]
provided the first positive insights concerning the use of
gliders in synergy with altimetry in order to monitor dynamics
in semi‐enclosed basins such as the Balearic area. That paper
showed reasonable qualitative agreements between absolute
dynamic topography (ADT) from altimetry and dynamic
height (DH) derived from glider measurements. However,
although the first results were encouraging, the study was
limited to only two glider missions over a relatively short
distance (<90 km) that did not allow robust altimetry‐glider
velocity comparisons (especially in terms of velocity gra-
2 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
C10029
Table 1. Altimetric Product Characteristics
Products
c
Along‐track 1 Hz SLA Envisat track 773
Along‐track 20 Hz SLA Envisat track 773
Gridded field ((M)SLA)d Multimissions
Resolution
Spatial Filtering
a
∼7 km
∼0.350 km
1/8°
20 km
20 km
>42 km
Editing
b
New
Newb
Standard
Time Sampling
35 days Instantaneous
35 days Instantaneous
7 days Time‐averaged
a
Forty‐two kilometers in Ruiz et al. [2009a].
See section 3.1.1.
c
Sea level anomaly, SLA.
d
Map of sea level anomaly, (M)SLA.
b
dients in the coastal zone). Moreover, the use of a reference
level at 180 m to compute glider DHz180 does not appear to
totally satisfy the dynamical processes marked by a deep
thermocline, such as eddies or coastal currents which might
be expected in this area. In addition, some altimetric data
near the coast was missing, especially due to land contamination in the altimeter footprint and associated data eliminations. This last issue suggested the need for specific
algorithms dedicated to coastal zone applications (e.g., altimetric waveform retracking, dedicated quality control procedure, etc.).
[7] On the basis of these issues, new strategies have been
developed within the framework of this study in order to
more precisely characterize coastal and mesoscale dynamics
in terms of current velocity. The altimetric velocity computation has been improved, especially in the coastal zone,
by associating high‐frequency along‐track sampling to new
filtering and editing techniques, and a new methodology has
been applied to estimate absolute glider velocities by using
velocities derived from glider GPS positions, a complementary and relevant variable not fully exploited in previous
studies. Thus, beside the main scientific objective of our
study, which consists at characterizing dynamical processes
at regional and coastal scales, the two complementary
objectives are to investigate limitations and potential improvements to altimetric velocity computation in the coastal
area; and to develop and assess methods for future combining of glider and altimetry data sets by homogenizing
their physical content.
[8] To give an answer to the questions addressed, this
study is organized as follows: After a detailed presentation
of the data sets used, we explain the strategies developed to
compute homogeneous altimetric and glider surface absolute
geostrophic currents (SAGC) with a special emphasis on
error budget evaluation. We then provide qualitative and
quantitative cross validations between altimetric and glider
current velocity through four examples corresponding to
characteristic dynamical events associated with the Balearic
current dynamical system. Finally, our conclusions are
summarized.
2. Data Sets and Variables Used
2.1. Sea Surface Height From Altimetry
[9] Altimetry allows a direct computation of geostrophic
velocity anomalies [e.g., Pascual et al., 2009, and references
therein]; by adding the geostrophic mean currents derived
from a mean dynamic topography (MDT) [Rio et al., 2007],
it is then possible to build SAGC. However, conventional
altimetry measurements remain largely unusable in the
coastal zone [Anzenhofer et al., 1999; Volkov et al., 2007;
Bouffard et al., 2010] due to several factors such as inaccurate geophysical corrections (e.g., atmospheric and tidal
signals) as well as environmental issues (e.g., land contamination in altimetric and radiometric footprints). At
present, new coastal altimeter products are under development (COASTALT, PISTACH, X‐TRACK). Several studies show that we can now be confident of such experimental
data for scales greater than 20 km and up to 15 km far from
the coast [Vignudelli et al., 2005; Bouffard et al., 2008,
2010; Durand et al., 2008, 2010; Cipollini et al., 2010]. The
main developments involve the application of coastal‐
oriented corrections and the review of the data recovery
strategies near the coast.
[10] In this paper, we will specifically focus on the impact
of filtering and new editing strategies, combined with high‐
frequency along‐track sampling. In this regard, we will use
three different altimetric data sets (see Table 1 for altimetric
product characteristics): Two along‐track data (1 and 20 Hz)
and a multisatellite gridded field, the map of sea level
anomalies (M)SLA.
2.1.1. Along‐Track Data Sets
[11] We used Envisat data sets provided by CLS
(Y. Faugère et al., personal communication, 2009) for track
773, cycles 59, 63, 67, and 69 being associated with the dates
of passage of the satellite of 8 July 2007, 25 November 2007,
13 April 2008, and 22 June 2008 respectively. These are
almost simultaneous with the glider missions; in each case
there is a temporal lag of less than 1 week.
[12] The processing of altimeter data used here is similar
to the one described in the “Ssalto/Duacs User Handbook”
[Ssalto/Duacs, 2006]. The only differences were the sampling, editing, and filtering of the along‐track data. The sea
surface height (SSH) built from ocean retracking (see
Envisat RA2‐MWR Handbook, 2007, http://envisat.esa.int/
handbooks/ra2‐mwr/) is corrected for path delay effects
(wet and dry troposphere, ionosphere, etc.) (see Le Traon
and Ogor [1998] and Le Traon and Ogor [2003] for
details) and for geophysical effects. The tide model used is
GOT00.2 (based on spectral analysis of tide gauge and altimetric signals) [Ray, 1998]. The barotropic response of the
ocean to atmospheric pressure forcing and wind effects is
modeled by a correction based on a combination of the
classical inverse barometer (IB) correction for low frequencies (lower than 1/20 day−1) and the Modèle aux Ondes de
Gravité 2‐dimensions (MOG2D) [Carrère and Lyard, 2003]
for higher frequencies.
[13] The data are sampled every 350 m (for 20 Hz data) or
resampled every 7 km (for 1 Hz data) along the tracks using
cubic splines. A mean profile, <SSH>, is removed from the
individual SSH measurements, yielding sea level anomaly
(SLA) values. The mean profile contains the geoid signal
3 of 17
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Table 2. Glider Cruise Characteristics: Dates, Total Length (Lineal and Real Distances) and Number of Profiles
Jul 2007
Nov 2007
Apr 2008
Jun 2008
Go Date
Return Date
Total Cruise Lineal Distance
(Go + Return, km)
Total Cruise Real Distance
(Go + Return, km)
Mean Distance Between
Profiles (km)
Total Profilesa
06072007
23112007
07042008
20062008
13072007
30112007
23042008
24062008
176
88
318
84
188
107
388
97
0.71
0.54
0.30
0.29
269
201
1321
335
a
During the 2007 missions, only downcasts were collected, whereas for the missions in 2008, both downcasts and upcasts were collected.
and the mean dynamic topography over a 7 year averaging
period (1993–1999).
[14] A key aspect of our data processing is the editing,
that is to say, the method of selecting good altimetric data
over corrupted data (see section 3.1.1). In official standard
products, for example that of Ruiz et al. [2009b], the
remaining measurement noise is reduced by applying
Lanczos cutoff and median filters for a 42 km window to the
1 Hz SLA. That 42 km filter could, however, generate too
much smoothing of the SLA, given the spatial scales associated with coastal dynamics (first Rossby radius of ∼14 km
in the Mediterranean Sea) [Robinson et al., 2001]. Moreover, the small length of glider cruises associated with this
window size could entail predominant edge effects along the
whole track (see section 3.1.3).
2.1.2. Gridded Field Product
[15] Multisatellite AVISO (M)SLA are also used. The
corrected along‐track SSH obtained for each mission
(Jason‐1 and Envisat for the period analyzed in this study)
have been inter‐calibrated with a global crossover adjustment of the Envisat data using Jason‐1 data as a reference
[Le Traon and Ogor, 1998]. The mapping method to produce
(M)SLA from along‐track data is detailed in Le Traon et al.
[2003] and consists of a suboptimal space‐time objective
analysis that takes into account along‐track correlated errors.
The products used in this study are specific for the Mediterranean Sea. For information on resolution, correlation
scales, and measurement noise see Pujol and Larnicol [2005].
[16] For each set of SLA (1 Hz, 20 Hz, and (M)SLA), a
mean dynamic topography (MDT) has been added in order
to obtain the ADT. Here, we use a regional MDT of the
Mediterranean Sea, as described by Rio et al. [2007], resulting
from a combination of model outputs, drifting buoys, and
altimeter data. This MDT was built to be compatible with
altimetry, i.e., it represents a MDT averaged over the same
period (1993–1999) as the temporal mean that is removed to
compute SLA.
2.2. Data From Gliders
[17] We used data from repeated glider surveys in the
Balearic Sea conducted between July 2007 and June 2008
(see Table 2 for details). Gliders are autonomous underwater
vehicles providing high‐resolution hydrographic and biogeochemical measurements. These vehicles control their
buoyancy to allow vertical motion in the water column and
make use of their hydrodynamic shape and small fins to
induce horizontal motions. The platform used in this study is
a Teledyne Webb Research Slocum Electric glider for
shallow water (200 m maximum depth) with a net horizontal
speed of ∼25 km/d, which takes into account data transmission when the glider is at the surface, once every 6 hours
in this type of survey.
[18] In all of the missions, the glider operated between
surface and 180 m. During the first surveys in 2007, only
downcasts were collected whereas for the missions in 2008
both downcasts and upcasts were registered, obtaining spatial
resolutions of ∼600 and ∼300 m, respectively. However, as
shown in Table 2, spatial resolution is not a constant value
as it depends on how the glider is ballasted as well as on the
presence of intense currents. Glider conductivity‐temperature‐
depth (CTD) profiles were processed and calibrated against
independent CTD casts from the SeaBird‐19 probe installed
on the Mediterranean Institute for Advanced Studies
(IMEDEA) ship. Profiles extracted from glider data were
corrected for thermal lag using the recursive filter introduced
by Lueck and Picklo [1990] and Morison et al. [1994]. The
final profiles were vertically averaged into 1 m bins.
[19] Surface geostrophic currents can be estimated from
the glider surface DH (obtained from temperature and
salinity fields) with respect to an arbitrary reference level
which is in general considered to be at the maximum depth
of the glider measurements (here 180 m). This implies that
geostrophic velocities at this reference level are negligible,
which is not always a correct assumption as many dynamical features usually have a deeper extension. However,
depth averaged absolute currents can also be retrieved from
GPS glider positioning, which gives one reference velocity
vector for every ∼7 km at every glider surfacing. The attitude sensor of the platform and the angle of attack can
introduce an associated error in the cross‐ and along‐track
estimates of depth averaged currents of about 2–3 cm/s
[Merckelbach et al., 2008].
3. Absolute Sea Surface Geostrophic Current
Computation
3.1. Absolute Surface Geostrophic Current From
Altimetry
3.1.1. Editing Procedure and Impact on the Coastal
Data Retrieval
[20] The use of high‐frequency along‐track sampling
(20 Hz, ∼350 m) allows small scale dynamics, as present in
the northwestern Mediterranean, to be captured. However,
this high‐resolution altimetric data is extremely noisy and
must be edited [see Bouffard et al., 2008]. In this study, the
determination of altimetric outliers is based on the assumption that deviations of measurements from the mean value
should vary spatially smoothly and follow a uniform distribution. Outliers can be detected when the deviations exceed a
pre‐determined range in the ranked deviation series. We
propose here to use three times the standard deviation (3s) of
along‐track SLA as the upper and lower limits to remove
main residual outliers related to land contaminations. This 3s
selection procedure is repeated 10 times (iterative process).
4 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
C10029
Figure 2. Altimetric surface absolute geostrophic current (SAGC) processing scheme.
The edited along‐track SLA are then spatially low‐pass (LP)
filtered using a 20 km Loess filter [Cleveland and Devlin,
1988]. This quality control procedure is applied to the
20 Hz (or 1 Hz) fully corrected along‐track data (see Figure 2).
[21] Our editing strategy has no impact on the 1 Hz data
(data are pre‐edited, for more details refer to section 4 of
the Envisat RA2/MWR ocean data validation and cross‐
calibration activities, yearly report 2008, http://www.aviso.
oceanobs.com/fileadmin/documents/calval/validation_
report/EN/annual_report_en_2008.pdf). The data selection,
however, impacts on the 20 Hz along‐track altimetric SLA
eliminating efficiently land‐contaminated data from 2%–
15% of the 20 Hz original data set, especially in the 20 km
coastal band (see triangles in Figure 3, right).
[22] In general, in the 50 km coastal band, data close to
the coast are more likely to be eliminated by the editing
procedure because altimeter and on‐board radiometer footprints may encounter land. Apart from November 2007,
where only few data points are eliminated along the whole
track (<2%), more than 25% (65% in July 2007) of the
original data are excluded at 21–28 km distances from the
coast (see Figure 3, left). Despite data eliminations by our
editing procedure, Table 3 shows that the 20 Hz sampling
allows us to recover more data than the 1 Hz data in close
proximity to the coast; between 6 and 13 km closer to the
coast by recovering 10 points measurements between the
last 1 Hz available data and the coastline and by applying
the 3 s statistical criterion to more data (20 times more than
at 1 Hz).
3.1.2. From Sea Level Anomaly to Surface Absolute
Geostrophic Current
[23] The across‐track SAGC is calculated by adding the
interpolated MDT to the edited and filtered SLA (see
section 3.1.1 and Figure 2). By construction, this current is
perpendicular to the satellite track, so in our case it is almost
parallel to the Balearic and Iberic shelf break (see Figure 1)
which should allow us to intercept the main components of
dynamics related to BC and NC.
[24] The across‐track altimetric SAGC is given by
Vgabs
0
g @ SLA þ MDT
¼
;
f
@x
ð1Þ
where g is the gravitational acceleration, f is the Coriolis
parameter, x is the axis set along the track direction, SLA′ is
the residual filtered sea level anomaly (from both edited
1 and 20 Hz data) and MDT is the mean dynamic topography [from Rio et al., 2007].
[25] The along‐track altimetric gradient is estimated by
using the optimal filter developed by Powell and Leben
[2004] with a spatial frame of 20 km. In the next sections,
even if not precise, the computed altimetric SAGC corresponds to components that are perpendicular to Envisat
track 773.
3.1.3. Sensitivity to Altimetric Noise
[26] Very few studies have analyzed the noise content of
20 Hz altimetric SLA. Recent comparisons between 20 Hz
altimetric SLA and tide gauges in the western Mediterranean
Sea indicate that 20 Hz altimeter measurement errors range
from 2 to 5 cm, mainly depending on the data editing and
smoothing process [Bouffard et al., 2008, 2010].
[27] Here, a Monte Carlo procedure is employed in order
to test the sensitivity of the SAGC computations to any
remaining noise in the 20 Hz edited SLA. To perform the
Monte Carlo simulation, an artificial data set was created by
adding Gaussian random noise with zero mean to the original SLA. Here, for 20 Hz altimetric signals (before the
spatial filtering), the spatial standard deviation (STD) noise
level is set at 5 cm (which approximately corresponds to the
spatial STD of the edited 20 Hz SLA for the missions used).
The resulting corrupted SLA was used to generate new estimates of the SAGC following the methodology described in
Figure 2. A total of 10,000 Monte Carlo simulations were
carried out for each of the four missions.
[28] For the missions of July 2007, November 2007, April
2008, and June 2008 the spatial mean of the STD difference
5 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Figure 3. (left) Percentage of raw 20 Hz altimetric data eliminated (black histograms) and number of
original data (shaded curve) as functions of distance to the coast (by 7 km class). (right) Sea level anomaly
(SLA; 20 Hz, cm) as a function of latitude (°N) and corresponding impact of editing. Triangle, eliminated
SLA; shaded curve, edited SLA; black curve, edited and 20 km low‐pass (LP)‐filtered SLA.
6 of 17
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Table 3. Distance From the Coast to the Closest Valid Altimetric
Measurements for Each Missiona
1 Hz
20 Hz
20 Hz + editing
Jul 2007
Nov 2007
Apr 2008
Jun 2008
34.2
20.6
21.4
27.2
20.0
20.0
20.3
12.8
14.1
26.8
14.7
16.4
a
Distances in km.
between signals a and b (STD(a‐b)) between uncorrupted
and corrupted SAGC are 6.3, 9.5, 4.7, and 8.5 cm/s,
respectively, while the mean differences are close to 0 cm/s
along the whole transect (see Figure 4, shaded curve). This
means that the SLA noise impacts the SAGC gradient but
not its spatial average. The differences of mean STD errors
as a function of missions are mainly due to the length of
track used (see Table 2) and the relative weight of edge
effects used to determine the averaged errors along the whole
track. Thus, the longer the mission, the lower the averaged
STD error. Figure 4 shows an example of the Monte Carlo
simulation for the mission of April 2008.
[29] As shown in Figure 4 (black curve), outside of the
20 km edge of the track (i.e., the window used both for the
LP Loess filtering and slope calculations), the Monte Carlo
STD are less than 5 cm/s whereas in the first and last 20 km
of the track, the associated errors increase and become
higher than 10 cm/s because of edge effects. This suggests
that only missions of sufficient length (i.e., >40 km as in the
July 2007 and April 2008 missions) can be used to efficiently assess the gradient of across‐track SAGC with
respect to the geostrophic velocities computed from glider.
C10029
The other missions, November 2007 and June 2008, have
too short a length for gradient SAGC comparisons; however, they can be used to assess the mean SAGC.
[30] In summary, when we look at the whole mission, the
averaged error due to the altimetric noise is about 0 cm/s
along the whole track, whereas the associated STD errors
are between 4 cm/s (>20 km from the coast) and 10 cm/s
(within 20 km of coast). These error values are of the same
order of magnitude as the STD differences that Vignudelli
et al. [2005] (respectively, Bouffard et al. [2008]) found by
performing a direct comparison in the Corsica Channel
between improved along‐track TOPEX/Poseidon‐derived
(respectively, improved multisatellite‐derived) velocities and
mooring velocity anomalies.
3.1.4. Impact of Sampling and Editing on the Velocity
Computation
[31] As shown previously, the geostrophic slope calculation is sensitive to the SLA noise, especially in the coastal
zone. This indicates the crucial importance of SLA editing
before SAGC computation. Figure 5 clearly shows the impact
of this editing on altimetric velocities for the missions of
July 2007 and April 2008.
[32] From Figure 5 it appears that the SLA editing has a
strong influence on the SAGC computation, especially within
a 50 km coastal zone (Figure 5, dashed boxes). Indeed when
the editing procedure is not applied, the resulting coastal
SAGC values are unrealistically large in July 2007 (>80 cm/s).
In April 2008, the editing also entails deep modifications in
the coastal area where the SAGC values derived from edited
SLA are about 15 cm/s. Beyond a 50 km coastal band, the
editing has less influence as less data are eliminated (see
section 3.1.1).
Figure 4. Errors (cm/s) obtained from Monte Carlo simulation as a function of latitude (°N), for the mission of April 2008. Shaded curve, average error; black curve, standard deviation (STD) error. The dashed
orange rectangles indicate areas where the STD error is greater.
7 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
C10029
Figure 5. Altimetric across‐track surface absolute geostrophic current (SAGC; cm/s) as a function of
latitude (°N) derived from 1 Hz (pink thin curve), raw 20 Hz (shaded thick curve), and edited 20 Hz
SLA (red thick curve) in (left) July 2007 and (right) April 2008. Areas with major differences are located
inside the orange dashed boxes.
[33] When comparing with velocities computed from 1 Hz
SLA, significant differences are also observed. In July 2007,
the 1 Hz SAGC are very low in the 50 km coastal zone
whereas a relative strong current (between 10 and 20 cm/s)
is observed by using the 20 Hz edited data. In April 2008,
differences are also observed in the neighborhood of the BC
where the 1 Hz SAGC exhibits a current 5 cm/s less than
that from edited 20 Hz data. This is also the case between
40.1°N and 40.4°N.
[34] In section 3.2, a direct comparison with glider calculated velocities provides robust insights about the respective
performance of these different along‐track altimetric data
sets.
3.2. Absolute Surface Geostrophic Velocity From
Glider Data
3.2.1. General Methodology and Reference Level Issue
[35] Ruiz et al. [2009b] used a reference depth level of
180 m to estimate DHz180 from glider data and obtained
coherent glider geostrophic velocities flowing northeastward
along the north Mallorca coast. For specific events, velocity
estimates from 42 km smoothed DHz180 appeared to be not
very sensitive to the test reference level, seeming to indicate
that the layer between 200 and the bottom does not play a
key role in the dynamics of the upper layer. However, on
closer consideration, small differences in DHz180 gradients
entail significant modifications in the geostrophic velocity
pattern and magnitude. For this reason, a more robust
strategy has been developed in this study. This strategy aims
at solving the reference level issue by combining glider
CTD current (Vgz180: cross‐track geostrophic component of
the baroclinic current relative to 180 m), glider GPS current
(Vabs : 180 m depth average absolute current), and model
outputs. Figure 6 shows the scheme used to process glider
SAGC.
[36] At each vertical level z, Vgz180 is derived from the
DHz180 using an optimal filter as described by Powell and
Leben [2004] (as used in the altimetry data). Then, the
difference between the 180 m depth average Vgz180 and Vabs
should correspond to the absolute velocity at 180 m. By
adding a reference level correction (RLC, Vabs − Vgz180 ) to
Vgz = 0180 at the surface, we should therefore be able to
compute the glider SAGC.
[37] However, these two averaged velocities do not have
exactly the same physical content. Whereas Vabs is influenced by the whole depth averaged dynamical components,
including ageostrophy, high‐frequency barotropic signals,
cyclostrophy, and inertial current, Vgz180 is only the result of
the baroclinic geostrophy contribution. Therefore, Vabs
needs to be partially corrected by using modeled high frequency geophysical corrections (HFGC, MOG2D + Ekman
currents).
3.2.2. High‐Frequency Geophysical Corrections
(HFGC)
3.2.2.1. Ageostrophic Ekman Effects
[38] The Ekman current is the motion induced by the wind
according to the theory of Ekman [1905] and should represent one of the main ageostrophic contributions included
in Vabs but which is missing in Vgz180. In this study, an Ekman
component has been estimated following the method of
Poulain et al. [2009]. The wind data used are from SeaWinds on QuikSCAT Level 2B Ocean Wind Vectors in
25 km Swath Grid (http://cersat.ifremer.fr/fr/data/discovery/
by_parameter/ocean_wind/quikscat_l2b). To remove the very
high frequency signals, which have no impact on the Ekman
oceanic motion, the data have been filtered with a low‐pass
filter at a 36 hours cutoff frequency. Given that the angle
between the wind and surface motion (in theory p/4) and
Ekman depth D strongly depend on location and time [Rio
and Hernandez, 2003], we therefore evaluate those parameters by using three drifters launched over the Balearic Sea.
We found values of = 24° and averaged depth D = 36 m,
which are in agreement with past studies of the Mediterranean
Sea [see Ursella et al., 2006; Poulain et al., 2009]. The
Ekman current was then validated with independent drifter
velocities [Escudier, 2009], 180 m depth averaged (assuming
8 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
C10029
Figure 6. Glider surface absolute geostrophic current (SAGC) processing scheme.
the theoretical spiral in the Ekman layer and approximating to
0 cm/s below), space‐time interpolated at glider positions,
and removed from Vabs .
3.2.2.2. Barotropic High‐Frequency Current
[39] Barotropic currents induced by high‐frequency
atmospheric forcing are not included in Vgz180 and, in theory,
have also to be removed from Vabs (see Figure 6). For this,
we use a correction of the ocean response to atmospheric
wind and pressure forcing from the MOG2D finite element
barotropic model [Carrère and Lyard, 2003] for high frequencies (i.e., <20 days), and an IB correction for lower
frequencies. The model is forced by surface atmospheric
pressure and wind from European Centre for Medium‐
Range Weather Forecasts analysis. Barotropic currents are
provided on a regular grid of 0.25° × 0.25° every 6 hours
and space‐time interpolated at glider positions.
4. Dynamical Structures Observed
4.1. Surface Patterns and Associated Vertical
Structures
[40] In this section, surface patterns (from remote sensing)
are qualitatively compared to the vertical structures (from
glider). We use both Vabs , sea surface temperature (SST),
SAGC derived from (M)SLA, and glider baroclinic geostrophic velocities derived from single CTD measurements
(Vgz180) projected onto the Envisat‐altimetric track 773
position. Figure 7 shows an example of the combined use of
remote sensing and a glider when observing dynamics in our
study area during the “go” glider transect (northward). For
the missions of July 2007, November 2007, and June 2008
(respectively, April 2008), the glider cruise began approximately 3 days (respectively, 7 days) before the Envisat
passage and crossed several dynamical patterns where significant across‐track surface currents were observed (see
Figure 7).
[41] In July 2007, the glider intercepts the BC in the
coastal Balearic zone (Figure 7a). Farther north, the remote
sensed SST data shows a marked thermal front corresponding
to the Catalan front and an associated south border of the NC
is seen. In addition, between these two major dynamical
patterns an eddy is also observed in the velocities. The
southwestern branch of this eddy is almost parallel to the
glider position which means that the associated across‐track
velocity would not be clearly seen by the glider. This is
confirmed when we look at the glider transect where, out of
the BC, the across‐track velocity is low. It also appears that
the BC, which borders the coast of Mallorca, is enlarged by
about 40 km large and marked by a significant velocity in the
first 50 m depth (double that at 100 m).
[42] Figure 7b shows that the general situation in
November 2007 is close to the one observed in July 2007.
The SST also shows a marked Catalan front in the north
associated with the NC cooling. In the south, the BC is
further away from the coast (about 10 km). A current loop,
which extends at 60 km from the Mallorca coast, is clearly
observed in both SST and SAGC velocity. This loop joins
the western branch of an eddy based north of the BC.
Unfortunately, the length of the glider cruise does not allow
us to capture the vertical structure of this eddy. When we
look at the glider transect it can be seen that the BC surface
intensity is approximately double that of July 2007, moreover the current has a deeper expression extending to ∼100 m.
The computed velocities from remote sensing, indicate that
the BC flows at distance higher than 10 km to the Mallorca
coast whereas it was close along the coast in July 2007, which
is also confirmed by Vabs .
[43] In April 2008, the general circulation characteristics
are quite different (compare Figure 7c). The SST image
shows a marked Balearic front whereas the Catalan front is
much weaker. Moreover, the glider cruise intercepts three
major dynamical structures: the BC in the south, the NC in
the north, and an eddy in between. The across‐track
velocities associated with these features appear clearly both
at surface and along the whole 180 m water column in the
glider CTD and GPS measurements. The velocities associ-
9 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Figure 7
10 of 17
C10029
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Table 4. Mean and Standard Deviation of Glider GPS (Vabs ) and Conductivity‐Temperature‐Depth (CTD, Vgz180) Currents for Both Go
and Return Cruisesa
Jul 2007 Dt = 6.8 d
Nov 2007 Dt = 2.6 d
Go
Return
Diff.b
Go
Return
Mean
STD
4.4
4.7
4.0
8.0
0.4
3.5
11.5
7.8
−9.8
10.4
Mean
STD
12.2
2.5
17.7
4.5
−5.5
2.1
19.6
7.1
20.2
9.5
Apr 2008 Dt = 14.4 d
Diff.b
CTD
21.3
9.0
GPS
−0.7
3.0
Jun 2008 Dt = 2.4 d
Go
Return
Diff.b
Go
Return
Diff.b
1.2
7.1
2.6
7.8
−1.4
3.7
13.8
24.4
11.8
11.3
2.0
28.0
4.7
13.1
7.5
13.4
−2.8
8.8
16.5
2.0
24.5
1.1
−8
0.9
a
In cm/s.
Diff. column corresponds to the difference between the “go” and “return” cruises. Given their short length in relation to the spatial filtering used (20 km),
only the means are significant for the November 2007 and June 2008 missions, both of which are affected by the windows edge.
b
ated with the eddy are quite strong, being of the same order
of magnitude as the BC (Vg0 < z180<100 > 5 cm/s). In the north,
within the 20 km Iberian coastal band, the NC signature is
also well identified showing a velocity with an opposite
direction and less amplitude than in the BC.
[44] In June 2008, remote sensing shows (Figure 7d) that
the situation is close to the one observed in July and
November 2007 with a marked Catalan front at north.
Between BC and NC, eddies are observed both east and
west of the glider track. Even though the glider cruise is
relatively short (42 km), the BC is captured. The transect
shows a progressive decrease of BC intensity in the coast–
open ocean direction, which is in agreement with the altimetric observations. It follows from altimetry that this
decrease is probably due to the geometrical orientation of
currents. Indeed, far away from the coast, the currents are
again almost parallel with the glider position, thus the
across‐track component cannot be well captured.
[45] Those comparisons show good qualitative agreement
and potential complementarily information arising from
remote sensing and glider measurements. However, even if
Vgz180 and the altimetric SAGC seem to be of same order of
magnitude, we believe that they are both here underestimated
(see stronger and almost colocalized Vabs in Figure 7). In
addition to the synopticity issues, this specific point will be
discussed in the next section through direct comparisons
between Vgz180 at surface (from CTD glider), altimetric
SAGC (from (M)SLA interpolate at the glider location) and
Vabs (from GPS glider positioning).
4.2. Variability at Glider Mission Temporal Scale
and Synopticity Issues
[46] Whereas along‐track altimetry allows an instantaneous measurement (every 35 days for Envisat), glider
cruises take several days (see Table 2) to complete a “go”
transect (northward: Mallorca–Iberian coast direction) and
“return” transect (southward: Iberian coast–Mallorca direc-
tion). This raises the crucial issue of temporal lags between
the glider and altimetric measurement. In this respect, direct
comparisons between go and return measurements of glider
are used here to provide a first assessment of the time delay
impact. Table 4 synthesizes the statistical characteristics,
differences of go and return velocities and the associated
mean time delay for the four glider missions.
[47] Table 4 shows that, the mean differences of Vgz = 0180
(glider CTD current) between the go and return trips range
from 0.4 to 23 cm/s. Apart from the mission of November
2007, the mean differences are less than 2 cm/s. The mission
of November 2007 is short (44 km), so a strong and quick
modification of BC at surface or in its baroclinic structure
would largely affect averages and explains the difference
between go and return (∼21.3 cm/s). For the mission of July
2007, the STD difference of Vgz = 0180 is 3.5 cm/s which is
equivalent to the one obtained in April 2008. For the whole
missions the mean and STD differences of Vabs (glider GPS
current), between go and return transects, are relatively
small compared to the initial values.
[48] Figure 8 is an illustration complementing Table 4 and
showing the spatial variations between go and return
velocities from both altimetric (M)SLA (SAGC), GPS
(Vabs ), and CTD measurements (Vgz = 0180) in July 2007 and
April 2008.
[49] Figure 8 indicates that even if the same kind of
dynamical patterns are observed, there are some major differences. In July 2007, Vgz = 0180 (Figure 8a, left) shows
equivalent magnitudes at go and return times whereas Vabs
(Figure 8b, left) decrease at return time over the whole
transect. In April 2008, even if the BC and eddy signatures
are observed both at go and return time, significant differences appear (Figures 8a and 8b, right). This seems to be
due to the eddy displacement and also to modifications in
the intensity, depth, and position of the BC. It is therefore
important to keep in mind that subsampled processes occurring at the glider mission temporal scale (<3 weeks) could
Figure 7. Surface circulation from remote sensing and the associated vertical structures from glider “go” (northward) missions of (a) July 2007, (b) November 2007, (c) April 2008, and (d) June 2008. Top images in Figures 7a–7d show SAGC in
cm/s (shaded arrow) derived from map of sea level anomalies ((M)SLA) overlapped by snapshot sea surface temperature
(SST, °C) and satellite altimetric tracks (black, Jason‐1; white, Envisat; the Envisat track 773 is highlighted in red). Glider
displacement (black curve) with corresponding GPS raw current (Vabs , black arrows). Bottom images in Figures 7a–7d show
projected across‐track glider CTD velocity (Vgz180, cm/s) as a function of latitude (°N) and depth (in m). For both top and
bottom images in Figures 7a–7d the areas inside the white squares indicate the Balearic Current (BC), pink squares indicate
anticyclonic eddies, and brown squares the Northern Current.
11 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Figure 8. LP‐filtered 20 km across‐track currents (cm/s) interpolated at the Envisat track location in
(left) July 2007 and (right) April 2008 as a function of latitude (°N) for both for “go” (thin curves)
and “return” (southward; thick curves) missions: (a) surface glider CTD current (Vgz = 0180, blue curves),
(b) glider GPS current (Vabs , black curves), (c) altimetric across‐track SAGC derived from (M)SLA
(brown curve), with time interpolated at glider passage.
12 of 17
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
potentially affect the comparison between glider and along‐
track altimetry in section 5. They can also be the consequence
of inexact colocalization between the two passages. In addition to high‐frequency 3D processes, the differences could
also have several causes that are difficult to distinguish such
as error in glider measurements, especially concerning positioning at surface during data transmission or from attitude
sensors providing glider heading, pitch and roll (see
Merckelbach et al. [2008] for technical details). For the altimetric SAGC derived from (M)SLA (Figure 8c) no significant differences are observed between go and return missions,
certainly because of the temporal smoothing effect due to the
optimal interpolation (OI; 7 day windows) procedure.
[50] The altimetric SAGC from (M)SLA show a satisfactory agreement with Vgz = 0180 and Vabs for the July 2007
and April 2008 missions except in the 40 km coastal band
where the altimetric SAGC shows weaker gradients and
amplitudes. Moreover, the magnitude of Vabs appear to be
significantly stronger (>30%) compared to both Vgz = 0180
and altimetric SAGC. There are several hypotheses that can
explain these discrepancies; they will be discussed in the
next section through the assessment of our new glider and
altimetry SAGC processing. In addition to physical content
nonhomogeneity and errors related to Vabs , differences
observed between Vgz = 0180 and Vabs could result from
a reference level issue (issue specifically addressed in
section 5.1.1). Concerning, the observed underestimation of
altimetry SAGC toward Vabs , it could be partially due to
spatial smoothing. Indeed, (M)SLA has been smoothed
through an OI procedure which uses a correlation length
scale of 100 km. Moreover, before OI, the altimetric along‐
track data (refer to the track positions in Figure 7) has been
low‐pass filtered with a cutoff of 42 km. This is why, in the
next sections, we will compute altimetric SAGC directly
from along‐track data following the methodologies described
in section 3.1. Impact of the MDT, spatial filtering, sampling, and editing will be evaluated by direct comparisons
with glider SAGC.
5. Cross‐Validation of Along‐Track Altimetry
and Glider Velocities
5.1. Assessment of the Glider Velocity Processing
5.1.1. Impact of the Reference Level Correction
[51] Figure 9 shows comparisons between altimetric (built
from 20 and 1 Hz edited along‐track data) and temporal
interpolated glider SAGC (using go and return transects) for
the missions of July 2007 and April 2008 with and without
the application of RLC (impact of HFGC is very small here
and thus will be discussed independently in section 5.1.2).
[52] From Figures 9a and 9b it follows that there is a very
good coherency between glider and altimetric velocities,
coherency that is better than the observed with (M)SLA in
the previous section. The signals are very well phased both
with and without the application of RLC.
[53] In July 2007, a progressive decrease of velocity is
seen in the Mallorca coast–open ocean direction. Between
40.0° and 40.2°N (i.e., inside the BC current), the glider and
altimetry curves show a very similar slope. After 40.2°N,
the velocity increases slightly in altimetry data whereas this
is not the case with glider data. In April 2008, the BC is well
identified. This current is associated with a relatively strong
C10029
velocity (>15 cm/s) which progressively decreases moving
offshore from the Mallorca coast. The anticyclonic eddy is
observed in the two data sets; this dynamical structure,
which crosses the Envisat track, is centered at about 40.5°N
in latitude and its associated velocity is between −20 and
20 cm/s (see Figures 9a and 9b).
[54] Even if glider and altimetric velocity show a good
agreement, significant differences are observed if RLC is
not applied. In July 2007, the altimetric SAGC is stronger
than Vgz = 0180 (by 7.5 cm/s on average, see Figure 9a);
however, the application of the RLC increases the mean
glider velocity and therefore improves the agreement with
altimetry (mean difference of 3.5 cm/s, see Figure 9b).
Moreover, the percentage of STD explained by glider in
altimetry becomes 75% with RLC application against 68%
without RLC. In April 2008, the use of the RLC also improves significantly the agreement between glider and
altimetry by increasing the glider velocity magnitudes (STD
of 6.9 cm/s without RLC against 14.9 cm/s with). The
percentage of STD explained by glider SAGC in altimetric
SAGC is thus again improved (47% with RLC against 6%
without).
[55] For missions of November 2007 and June 2008 the
length of missions are too short for statistical comparisons
of STD (see section 3.1.3). However, we can look at the
mean SAGC inside the BC. For June 2008, the RLC also
allows significant increases in the agreement between glider
and altimetric SAGC with a mean difference of 17.4 cm/s
without RLC against 0.7 cm/s with RLC (for a mean current
of ∼30 cm/s). In November 2007, Vgz = 0180 shows a mean
velocity of 5 cm/s. If the RLC is added, the glider SAGC
becomes 25.5 cm/s, which is 13.9 cm/s more than for
altimetry. This difference could be due to high‐frequency
dynamics not captured at the time of altimetric satellite
passage. Indeed the mean difference between the go and
return glider cruise is large for this mission (see Table 4)
which support this suggestion.
[56] To sum up, the application of RLC enables the general improvement of the agreement between the glider and
altimetry currents. However, it is important to note that for
the whole mission the glider SAGC magnitudes are stronger
than the altimetric ones. In addition to the combined effects
of different errors (such as residual coastal altimetry contamination, inaccuracy of the MDT, and compass error for
gliders), the underestimation observed in altimetry could be
partially due to its inexact colocalization with the glider that
can be influenced and partially advected by the strong currents (see differences between lineal and total glider cruise
distances in Table 2). In April 2008, such differences are
particularly observed at the eddy location. There, the altimetric SAGC underestimation could also come from the
cyclostrophic acceleration, neglected in the geostrophic
framework computation. This component, in the case of an
anticyclonic eddy with the characteristics of the one studied
here could be around 10%–20% of the geostrophic part of
the velocity [Gomis et al., 2001].
5.1.2. Impact of the High‐Frequency Geophysical
Correction
[57] The HFGC consists of the Ekman and MOG2D
velocities which have been interpolated spatially and temporally along the glider position measurements. Table 5
summarizes their statistical characteristics for the missions
13 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Figure 9
14 of 17
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Table 5. Ekman and Barotrophic Current Statistical Characteristics
Interpolated (in Space and Time) at Go and Return Glider Positionsa
Jul 2007
Go
Mean
STD
0.3
0.1
Mean
STD
0.1
0.1
Apr 2008
Return
Ekman
0.1
0.0
MOG2D
0.1
0.1
Go
Return
0.5
0.2
0.2
0.2
0.1
0.2
0.2
0.0
a
In cm/s.
of July 2007 and April 2008 and highlights that both Ekman
and barotropic high‐frequency velocities are very small
(subcentimetric) for the considered period.
[58] The impact of HFGC is therefore marginal, especially
compared to the RLC (small for Ekman and negligible for
MOG2D). Moreover, the Ekman and MOG2D velocities are
less than the amplitude of errors associated with both glider
and altimetric velocity computations (problem of spatial
synopticity, noise, etc.). In this respect, even if the use of
HFGC is theoretically justified, the present comparison
between glider and altimetry does not allow us to evaluate
quantitatively its impact. However, the small reduction of the
glider and altimetric SAGC mean differences for all missions
(between 0.3 and 0.6 cm/s) gives few positive insights.
5.2. Assessment of the Coastal Altimetric Velocity
Processing
[59] In the coastal zone, altimetric velocities are particularly affected by noise measurements (see section 3.1.3)
while this is less the case for glider velocities. In this regard,
the glider velocities are now used in order to evaluate the
impact of high‐frequency sampling combined with our editing strategy in order to capture small scale coastal signatures.
We will also briefly discuss the influence of MDT on glider‐
altimetry comparisons.
5.2.1. Impact of the MDT
[60] Table 6 summarizes statistical comparisons of the
impact of the MDT in the glider and altimetric SAGC
comparisons and shows that when the velocity derived from
the MDT is added to the altimetric surface geostrophic velocity
anomaly, agreement with the glider SAGC is improved. Indeed
the mean difference between the two velocities is strongly
reduced for all missions. This reduction is of ∼10 cm/s for
missions of July 2007, November 2007, and June 2008 while
it is ∼2 cm/s for the mission of April 2008. Concerning the
STD, the use of MDT also enables a decrease in the STD
difference for the missions of July 2007 and April 2008.
[61] The improvements in terms of mean difference show
that MDT performance is satisfactory for wavelengths
C10029
>40 km (approximately the length of the November 2007
and June 2008 missions). The smaller spatial scale variations cannot be correctly reproduced and thus well evaluated
in term of STD given the spatial smoothing and OI procedures used to build the MDT [cf. Rio et al., 2007].
5.2.2. Impact of Sampling and Editing Strategy
[62] In section 3.1.4, we highlighted that both editing
strategy and sampling play key roles in the altimetric velocity
calculation, especially in the coastal zone. From Figures 9b
and 9c, it follows that altimetric SAGC computed from edited 20 Hz altimetric data shows better agreement with the
glider SAGC than when using 1 Hz data. This is especially the
case in the 50 km coastal zone where a relative strong BC,
observed by glider in July 2007, is also captured with 20 Hz
data but not seen with 1 Hz data, while out of this area it is
interesting to note that the 1 and 20 Hz velocities are consistent. In April 2008, 20 Hz along‐track data show SAGC
close to that observed by the gliders, the 1 Hz velocity also
reproduces the BC in the coastal zone; however, the current
exhibits less amplitude than with the glider and 20 Hz data
(10–20 cm/s less). The 1 Hz velocities seem also to underestimate the eddy intensity between 40.1°N and 40.4°N,
whereas the 20 Hz data shows a better agreement with the
glider SAGC. When (M)SLA or standard 42 km smoothed
1 Hz along‐track data are used for comparison with the gliders (not shown) the results are worse due to too much spatial
smoothing.
[63] From these previous examples, it appears that the
20 Hz altimetric data associated with an adapted editing
strategy can recover more data close to the coast and also
produce better constraints on coastal altimetric SAGC
computations.
6. Conclusions and Perspectives
[64] In the present study, we analyzed the compatibility
between glider and altimetry measurements in view of
developing an integrated coastal observing system. Altimetric data have been specifically processed to assess a new
methodology dedicated to glider SAGC computation. This
approach also provided interesting insights into the present
limitations and potential improvements of coastal altimetry.
The analysis was two‐way, using altimetry in support to the
glider velocity processing and glider measurements in support of coastal altimetric techniques. We believe that such an
approach is necessary before undertaking multisensor data
fusions or applications aiming at a quantitative understanding
of coastal physical processes.
[65] As a first conclusion, we have shown that simple new
processing strategies allow us to generate glider and altimetric SAGC with very good agreement (percentage of STD
explained >40% over spatial scales of 20 km). For the glider
data set, the importance of the reference level and synopti-
Figure 9. Glider versus altimetric across‐track 20 km LP‐filtered currents (cm/s) for (left) July 2008 and (right) April 2008
missions. From top to bottom: (a) altimetric SAGC derived from edited 20 Hz data (red curve) vs. surface glider CTD current
(Vgz = 0180, thin blue curve), (b) SAGC derived from edited 20 Hz data (red curve) vs. glider SAGC with application of reference
level correction (RLC) and high‐frequency geophysical correction (HFGC; blue curve), (c) SAGC derived from 1 Hz data
(thin pink curve) vs. glider SAGC with application of RLC and HFGC (blue curve).
15 of 17
C10029
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Table 6. Statistical Comparisons Between Glider (With Application of RLC and HFGC) and Altimetric (20 Hz) SAGC by Using or Not
Using the Mean Velocity Derived From the MDT
Jul 2007
No MDT
MDT
Apr 2008
Mean Diff.
STD Diff.
Nov 2007 Mean Diff.
Mean Diff.
STD Diff.
Jun 2008 Mean Diff.
13.6
3.5
2.9
1.4
23.2
13.9
4.8
2.8
8.6
7.8
9.5
0.7
city issues, due to baroclinic short timescale signals, have
been confirmed, whereas Ekman and the high‐frequency
barotropic component show marginal impacts in our case
studies. For altimetry, the use of MDT improve the spatial
consistency between the altimetric and glider SAGC despite
a lack of spatial resolution that could entail an underestimation of the mean circulation. The use of edited 20 Hz
rather than standard 1 Hz and (M)SLA altimetry allows us to
go closer to the coast, to increase altimetry‐glider agreement
and to keep physical signals otherwise eliminated. The main
benefit of using 20 Hz data lies more in the greater quality
control procedure possibilities with more data, than in spatial scale resolving, given that a 20 km LP filtering is
applied.
[66] The combined use of altimetry and glider is promising, allowing observation of coherent general characteristics
of the circulation in the Balearic Sea both at the surface and
for the whole 180 m water column. High‐resolution
hydrographic fields from glider and altimetry have revealed
the presence of permanent BC and NC signals and nonpermanent signals, such as the relatively strong anticyclonic
eddy intercepted in April 2008. The synoptic view from
remote sensing during the glider missions also suggested a
more detailed picture of the mesoscale structures that may
play a key role in the exchanges and transport of heat and
other biogeochemical properties across the basin. The anticyclonic eddies observed during these missions confirm past
studies such as the work of Pascual et al. [2002]. This raises
the need to quantify the recurrence of such structures that do
not appear in the standard circulation scheme but that have
impacts on the general circulation in the Mediterranean.
Thus, the use of altimetry in combination with gliders could
play a major role in revising the general pattern and better
characterizing the associated forcing of the Balearic Sea
circulation.
[67] As a general conclusion, this study highlights the
need for high‐resolution coastal measurements made possible thought the development of synergic approaches and
the combined use of observing systems at several spatial/
temporal scales. Future works could evaluate and compare
velocities derived from new coastal altimetric products
(PISTACH, COASTALT, etc.) with glider velocities following the approach described in this paper. On a longer term,
the Surface Water and Ocean Topography (SWOT) mission
can provide very valuable information, complementary to
in‐situ measurement, for monitoring mesoscale and submesoscale coastal processes. In a more general framework,
we believe that multisensor approaches, combining complementary in‐situ and remote sensing, will contribute to a
better understanding of physical and multidisciplinary
processes within the coastal domain. In combination with
modeling, multisensor data can significantly help to quantify
changes in coastal systems, to comprehend the mechanisms
that regulate them, and to forecast their evolution.
Notation
DHz180 dynamic height derived from the glider conductivity‐
temperature‐depth (CTD) profile at depth = z meters
(with a reference level at 180 m); also refer to Figure 6.
Vgabs surface absolute geostrophic current (SAGC).
Vgz180 glider CTD current: cross‐track baroclinic component of geostrophic velocity derived from the glider
CTD profile at depth = z meters (with a reference
level at 180 m); also refer to Figure 6.
Vabs glider GPS current:180 m depth average absolute
velocity derived from glider GPS positions; also refer
to Figure 6.
[68] Acknowledgments. We deeply acknowledge the valuable contribution of M. Martínez‐Ledesma and B. Casas during the data acquisition
phase of the glider missions. Special thanks are also due to B. Garau for
processing glider data, to E. Heslop for valuable editing comments, and
to R. Escudier and L. Carrère for providing the Ekman and barotropic currents,
respectively. The altimeter data were produced by Ssalto/Duacs and distributed
by Aviso with support from CNES. Sea surface temperature images were
acquired and processed by EUMETSAT. This work has been partially funded
by the CICYT‐funded projects COOL (CTM2006‐12072/MAR) and
ECOOP (CTM 2007‐31006‐E) and by EU‐funded projects FP6 ECOOP
and FP7 MyOcean.
References
Anzenhofer, M., C. K. Shum, and M. Rentsh (1999), Coastal altimetry and
applications, Tech. Rep., 464, Geod. Sci. and Survey., Ohio State University, Columbus.
Bethoux, J. P. (1980), Mean water fluxes across sections in the Mediterranean Sea, evaluated on the basis of water and salt budgets and of
observed salinities, Oceanol. Acta, 3, 79–88.
Birol, F., M. Cancet, and C. Estournel (2010), Aspects of the seasonal variability of the Northern Current (NW Mediterranean Sea) observed by
altimetry, J. Mar. Syst., 81, 297–311, doi:10.1016/j.jmarsys.2010.01.005.
Bouffard, J., S. Vignudelli, P. Cipollini, and Y. Menard (2008), Exploiting
the potential of an improved multimission altimetric data set over the
coastal ocean, Geophys. Res. Lett., 35, L10601, doi:10.1029/
2008GL033488.
Bouffard, J., L. Roblou, F. Birol, A. Pascual, L. Fenoglio‐Marc, M. Cancet,
R. Morrow, and Y. Ménard (2010), Introduction and assessment of
improved coastal altimetry strategies: Case study over the northwestern
Mediterranean Sea, in Coastal Altimetry, edited by S. Vignudelli et al.,
chap. 12, Springer, New York, in press.
Carrère, L., and F. Lyard (2003), Modeling the barotropic response of the
global ocean to atmospheric wind and pressure forcing: Comparisons
with observations, Geophys. Res. Lett., 30(6), 1275, doi:10.1029/
2002GL016473.
Cipollini, P., et al. (2010), The role of altimetry in coastal observing systems, in Proceedings of OceanObs’09: Sustained Ocean Observations
and Information for Society, vol. 2, edited by J. Hall, D. E. Harrison,
and D. Stammer, Eur. Space Agency Publ., WPP‐306, in press.
Cleveland, W. S., and S. Devlin (1988), Locally weighted regression: An
approach to regression analysis by local fitting, J. Am. Stat. Assoc., 83,
596–610.
Durand, F., D. Shankar, F. Birol, and S. S. C. Shenoi (2008), Estimating
boundary currents from satellite altimetry: A case study for the east coast
of India, J. Oceanogr., 64, 831–845, doi:10.1007/s10872-008-0069-2.
16 of 17
C10029
BOUFFARD ET AL.: SYNERGY BETWEEN ALTIMETRY AND GLIDER
Durand, F., D. Shankar, F. Birol, and S. S. C. Shenoi (2010), Spatio‐temporal
structure of the East India Coastal Current from satellite altimetry, J. Geophys. Res., 114, C02013, doi:10.1029/2008JC004807.
Ekman, V. W. (1905), On the influence of the Earth’s rotation on ocean‐
currents, Ark. Mat. Astron. Fys., 2(11), 1–52.
Escudier, R. (2009), Estimating surface ocean currents using a multisensor
approach, in Internship Report of ENSTA Engineering School, edited by
A. Pascual and J. Bouffard, Institut Mediterrani d’Estudis Avançats,
Mallorca, Spain.
Font, J. (1990), A comparison of seasonal winds with currents on the continental slope of the Catalan Sea (northwestern Mediterranean), J. Geophys. Res., 95(C2), 1537–1546.
Font, J., J. Salat, and J. Tintoré (1988), Permanent features of the circulation in the Catalan Sea, Oceanol. Acta, 9, 51–57.
Gomis, D., S. Ruiz, and M. A. Pedder (2001), Diagnostic analysis of the 3D
ageostrophic circulation from a multivariate spatial interpolation of CTD
and ADCP data, Deep Sea Res., Part I, 48, 269–295, doi:10.1016/S09670637(00)00060-1.
Hodges, B. A., and D. M. Fratantoni (2009), A thin layer of phytoplankton observed in the Philippine Sea with a synthetic moored array of
autonomous gliders, J. Geophys. Res., 114, C10020, doi:10.1029/
2009JC005317.
Jordi, A., G. Basterretxea, and S. Anglès (2009), Influence of ocean circulation on phytoplankton biomass distribution in the Balearic Sea: Study
based on sea‐viewing wide field‐of‐view sensor and altimetry satellite
data, J. Geophys. Res., 114, C11005, doi:10.1029/2009JC005301.
La Violette, P. E., J. Tintoré, and J. Font (1990), The surface circulation of
the Balearic Sea, J. Geophys. Res., 95, 1559–1568.
Le Traon, P.‐Y., Y. Faugère, F. Hernandez, J. Dorandeu, F. Mertz, and
M. Ablain (2003), Can we merge Geosat follow‐on with TOPEX/Poseidon
and ERS‐2 for an improved description of the ocean circulation?, J. Atmos.
Oceanic Technol., 20, 889–895.
Le Traon, P.‐Y., and F. Ogor (1998), ERS‐1/2 orbit improvement using
TOPEX/POSEIDON: The 2 cm challenge, J. Geophys. Res., 103(C4),
8045–8057, doi:10.1029/97JC01917.
Lueck, R. G., and J. J. Picklo (1990), Thermal inertia of conductivity
cells: Observations with a Sea‐Bird cell, J. Atmos. Oceanic Technol.,
7, 756–768.
Merckelbach, L. M., R. D. Briggs, D. A. Smeed, and G. Griffiths (2008),
Current measurements from autonomous underwater gliders, paper presented at Current Measurement Technology, 2008, 9th Working Conference on Current Measurement Technology, Inst. of Electr. and Electr.
Eng., New York.
Morison, J., R. Andersen, N. Larson, E. D’Asaro, and T. Boyd (1994), The
correction for thermal‐lag effects in Sea‐Bird CTD data, J. Atmos. Oceanic
Technol., 11, 1151–1164.
Onken, R., A. Alvarez, V. Fernandez, G. Vizoso, J. Tintore, P. Haley, and
E. Nacini (2008), A forecast experiment in the Balearic Sea, J. Mar.
Syst., 71, 79–98, doi:10.1016/j.jmarsys.2007.05.008.
Pascual, A., B. Buongiorno Nardelli, G. Larnicol, M. Emelianov, and
D. Gomis (2002), A case of an intense anticyclonic eddy in the Balearic
Sea (western Mediterranean), J. Geophys. Res., 107(C11), 3183,
doi:10.1029/2001JC000913.
Pascual, A., D. Gomis, R. L. Haney, and S. Ruiz (2004), A quasigeostrophic analysis of a meander in the Palamós canyon: Vertical velocity,
geopotential tendency, and a relocation technique, J. Physical Oceanogr.,
34, 2274–2287.
Pascual, A., C. Boone, G. Larnicol, and P. Y. Le Traon (2009), On the quality
of real‐time altimeter gridded fields: Comparison with in situ data, J. Atmos.
Oceanic Technol., 26, 556–569, doi:10.1175/2008JTECHO556.1.
Pinot, J. M., J. Tintore, and D. Gomis (1994), Quasi‐synoptic mesoscale
variability in the Balearic sea, Deep Sea Res., Part, 1, 41, 897–914,
doi:10.1016/0967-0637(94)90082-5.
C10029
Pinot, J. M., J. L. López‐Jurado, and M. Riera (2002), The CANALES
experiment (1996‐1998). Interannual, seasonal, and mesoscale variability
of the circulation in the Balearic Channels, Prog. Oceanogr., 55, 335–370,
doi:10.1016/S0079-6611(02)00139-8.
Poulain, P. M., R. Gerin, E. Mauri, and R. Pennel (2009), Wind effects on
drogued and undrogued drifters in the eastern Mediterranean, J. Atmos.
Oceanic Technol., 26, 1144–1156.
Powell, B. S., and R. R. Leben (2004), An optimal filter for geostrophic
mesoscale currents from along‐track satellite altimetry, J. Atmos. Oceanic
Technol., 21, 1633–1642.
Pujol, M. I., and G. Larnicol (2005), Mediterranean Sea eddy kinetic
energy variability from 11 years of altimetric data, J. Mar. Syst., 58,
121–142, doi:10.1016/j.jmarsys.2005.07.005.
Ray, R. D. (1998), Spectral analysis of highly aliased sea‐level signals,
J. Geophys. Res., 103(C11), 24,991–25,003, doi:10.1029/98JC02545.
Rio, M.‐H., and F. Hernandez (2003), High‐frequency response of wind‐
driven currents measured by drifting buoys and altimetry over the world
ocean, J. Geophys. Res., 108(C8), 3283, doi:10.1029/2002JC001655.
Rio, M. H., P. M. Poulain, A. Pascual, E. Mauri, G. Larnicol, and R. Santoleri
(2007), A mean dynamic topography of the Mediterranean Sea computed
from altimetric data, in situ measurements and a general circulation model,
J. Mar. Syst., 65, 484–508, doi:10.1016/j.jmarsys.2005.02.006.
Robinson, A., W. Leslie, A. Theocharis, and A. Lascaratos (2001),
Encyclopedia of Ocean Sciences, in Mediterranean Sea Circulation,
pp. 1689–1706, Academic, London.
Rubio, A., B. Barnier, G. Jordà, M. Espino, and P. Marsaleix (2009), Origin
and dynamics of mesoscale eddies in the Catalan Sea (NW Mediterranean):
Insight from a numerical model study, J. Geophys. Res., 114, C06009,
doi:10.1029/2007JC004245.
Ruiz, S., A. Pascual, B. Garau, I. Pujol, and J. Tintore (2009a), Vertical
motion in the upper ocean from glider and altimetry data, Geophys.
Res. Lett., 36, L14607, doi:10.1029/2009GL038569.
Ruiz, S., A. Pascual, B. Garau, Y. Faugere, A. Alvarez, and J. Tintoré
(2009b), Mesoscale dynamics of the Balearic Front, integrating glider,
ship and satellite data, J. Mar. Syst., 78, S3‐S16, doi:10.1016/j.
jmarsys.2009.01.007.
Ssalto/Duacs (2006), Ssalto/Duacs user handbook: (M)SLA and (M)ADT
near‐real‐time and delayed‐time products, Rep. SALP‐MU‐P‐EA‐
21065‐CLS, Aviso, Ramonville‐Saint‐Agne, France.
Tintoré, J., D. P. Wang, and P. E. La Violette (1990), Eddies and thermohaline intrusions of the shelf‐slope front off Northeast Spain, J. Geophys.
Res., 95(C2), 1627–1633.
Ursella, L., P.‐M. Poulain, and R. Signell (2006), Circulation in the northern
and middle Adriatic Sea using surface drifters: Mesoscale to seasonal variabilities, J. Geophys. Res., 111, C03S04, doi:10.1029/2005JC003177.
Vignudelli, S., P. Cipollini, L. Roblou, F. Lyard, G. P. Gasparini, G. Manzella,
and M. Astraldi (2005), Improved satellite altimetry in coastal systems:
Case study of the Corsica Channel (Mediterranean Sea), Geophys. Res.
Lett., 32, L07608, doi:10.1029/2005GL022602.
Volkov, D., G. Larnicol, and J. Dorandeu (2007), Improving the quality of
satellite altimetry data over continental shelves, J. Geophys. Res., 112,
C06020, doi:10.1029/2006JC003765.
J. Bouffard, A. Pascual, S. Ruiz, and J. Tintoré, Marine Technologies,
Operational Oceanography, and Sustainability Department, Mediterranean
Institute for Advanced Studies, 21 Miquel Marquès, E‐07190 Mallorca,
Spain. (jerome.bouffard@uib.es)
Y. Faugère, CLS Space Oceanography Division, 8‐10 Hermès, F‐31526,
Ramonville CEDEX, France.
17 of 17