Circulation over the continental shelf of the western and southwestern Gulf of Mexico

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2011jc007007, 2011 Circulation over the continental shelf of the western and southwestern Gulf of Mexico Jean Dubranna, 1 Paula Pérez Brunius, 1 Manuel López, 1 and Julio Candela 1 Received 28 January 2011; revised 15 April 2011; accepted 21 April 2011; published 5 August [1] The circulation over the continental shelf break of the western and southwestern Gulf of Mexico is inferred from the analysis of drifter trajectories and months of continuous current measurements at seven different locations. The interpretation of the data is backed up by satellite altimetry, coastal sea level from tide gauges and wind model outputs. In accordance with previous numerical results, subinertial surface currents are driven by the wind along the shelves of the states of Tamaulipas and Veracruz. Northern wind regimes would force southward currents, whereas southern wind regimes would force northward currents at the surface but southward near the bottom, through a process involving Ekman drift and geostrophic balance. Our results show, however, that alongshore current variations are not correlated with the wind over the Western Campeche Bank. In addition, we identify other sources of current forcing. The transient eddies that collapse along the continental shelf can force strong alongshore currents and overwhelm the influence of established wind regimes. Their erratic occurrence is likely to be a major factor of interannual variability of the alongshore currents. Also, we point out the existence of coastally trapped waves generated by the wind in the northern shelf of Tamaulipas and propagating down to the Western Campeche Bank. The period of these waves ranges between 6 and 10 days, with phase speeds in the 4 m/s range. Citation: Dubranna, J., P. Pérez Brunius, M. López, and J. Candela (2011), Circulation over the continental shelf of the western and southwestern Gulf of Mexico, J. Geophys. Res., 116,, doi: /2011jc Introduction [2] Despite the continental shelf circulation has been well documented in the American part of the Gulf of Mexico (GoM), namely, north of 26 N, information about its Mexican counterpart is still underrepresented in the literature. Most of the studies in the Mexican waters of the GoM are actually dedicated to the characteristics of the offshore circulation. The northern continental shelf of the Mexican GoM is approximately 100 km wide, thinning southward to about 30 km at the latitude of the southern Bay of Campeche (BoC). From there, it widens again to reach about 200 km at the level of the Western Campeche Bank (WCB, western part of the Yucatan Peninsula). [3] Very few measurements have been collected in the continental shelves of Tamaulipas and Veracruz (hereafter referred as TAVE shelves, between 18.2 N and 26 N), and of the WCB. The ones that exist [Gutierrez de Velasco et al., 1992, 1993] suggest a seasonal reversal of the shelf currents 30 km off Tuxpan (21.6 N, 97.1 W), with down coast (upcoast) flows in fall winter (spring summer). Here, downcoast (up coast) flow refers to alongshore displacement with 1 Centro de Investigación Científica y de Educación Superior de Ensenada, Departamento de Oceanografía Física, Ensenada, México. Copyright 2011 by the American Geophysical Union /11/2011JC the coast on the right (left). Boicourt et al. [1998] first suggested the possibility for the down coast current to extend from the Texas shelf down to Tuxpan (20.6 N, 97.2 W) during fall winter. Following the tracks of drifters released in the Louisiana Texas shelf, Walker [2005] found that about half of them entered the Louisiana Texas down coast coastal current that extends to depths around 10 to 30 m during fall and winter, and ended up in Mexican waters. [4] Zavala Hidalgo et al. [2003] and Morey et al. [2005] report the most comprehensive results about the circulation and water properties over the continental shelf of the western and southwestern GoM. They explore the salinity, temperature, sea level and current fluctuations using a 10 year simulation of the GoM by a very high resolution model and hydrographic data. The model results corroborate the existence of a strong wind driven seasonal signal in the circulation of the surface waters over the shelf. During spring and summer, the average wind flows toward the west and northwest, driving up coast currents from the southern BoC shelf up to the northern TAVE shelf. In fall and winter, the direction shifts and the wind blows south to southwestward forcing a down coast current from the northwestern GoM down to the southern BoC. Most intense northerly winds are associated with the coming of atmospheric cold fronts traveling from the northwest continental United States. This reversal comes with the rise of the coastal water level which reaches an annual peak in September and October from about 1of17

2 Galveston (29.28 N W) down to the north of the Yucatan Peninsula. On the WCB, the model results show an up coast surface current year round, due to the relative orientation of the coast with the dominant winds. As a result, the down coast flow of the TAVE shelf converges with the up coast flow of the WCB between 93 W and 94 W during fall and winter causing offshore currents in the area. Once the waters have crossed the shelf break, they are likely to be transported by the mesoscale eddy field of the BoC. [5] Along with the wind stress, the circulation over the shelf is locally affected by mesoscale eddies [Boicourt et al., 1998], such as the permanent as well as time dependent cyclonic circulation in the Bay of Campeche (BoC), south of 22 N, mostly documented by Vazquez de la Cerda et al. [2005] and DiMarco et al. [2005]. From the study by Gutiérrez de Velasco and Winant [1996], this cyclone is thought to be forced by the local positive wind stress curl constrained by the presence of the high coastal Sierra Madre mountain range that channels the wind toward the Isthmus of Tehuantepec. Vazquez de la Cerda et al. [2005] confirm that the seasonality of the cyclonic circulation consists in an intensification in winter and a weakening in summer, in accordance with the seasonality of the wind stress curl. They also point out that the nonseasonal variations are related with synoptic events such as smaller scale eddies passing by. The energetic Loop Current Eddies (LCE) coming from the eastern GoM are anticyclonic and can also influence the shelf circulation, mostly north of 22 N. Walker [2005] studied the effect of wind and eddies on the shelf and slope circulation in the northwestern GoM and she mentions the seaward entrainment of drifters located in the Mexican continental shelf between 25 N and 26 N by cyclone/anticyclone pairs that influence the shelf circulation in the area. Specifically, 60% of the drifters entering the Mexican slope were exported seaward. Finally, the buoyancy gradients due to numerous fresh water inputs by the river mouths along the Mexican coast have a low impact on the shelf circulation [Zavala Hidalgo et al., 2003]. [6] This paper compiles on site measurements in the water column collected from November 2007 to July 2009, satellite altimetry data and the trajectories of surface drifters to study the circulation over the shelf of the western and southern GoM, south of 26 N. After a description of the data, we focus on the monthly circulation of the surface waters in relation with the wind forcing in order to compare our measurements with the model climatology presented by Zavala Hidalgo et al. [2003] and Morey et al. [2005]. Then, we describe the different circulation regimes observed during the measurement period, in relation to various forcing factors. A variable alongshore wind component blowing over a sloping and stratified continental shelf is a wellknown generation factor of coastally trapped waves [Gill and Clarke, 1974; Clarke, 1977]. Considering the variable wind conditions as well as the bathymetric profile of the continental shelf along the TAVE coast, such processes could be forced in the area. Besides, Zavala Hidalgo et al. [2003] first mentioned the possibility of the variability of the current along the WCB being produced by remote forcing due to coastal waves associated with the circulation variations on the TAVE shelf. We therefore examine the existence, propagation and generation of coastally trapped waves in the investigated area in the fifth section of this article. We then discuss and compare with previous studies the different processes found in our measurements, followed by general conclusions. 2. Data 2.1. Moorings [7] A total of seven moorings were deployed between November 2007 and July 2009 at seven different sites along the 130 m isobath of the western and southwestern GoM, corresponding approximately to the continental shelf break (Figure 1). Two of them have collected current profiles from the end of November 2007 to mid July 2009 (about 19 months) at stations LNK and CTZ, and the remaining 5 were deployed from about the end of July 2008 to mid July 2009 (about 12 months). [8] Each mooring was equipped with an Acoustic Doppler Current Profiler (ADCP) looking upward (Workhorse 300 khz from RDI) with measurements averaged over 8 m bins, and another looking downward (Workhorse 600 khz from RDI) with measurements averaged over 0.5 m bins, focusing on the bottom boundary layer. Both ADCPs were positioned at 14 m above the seafloor. [9] Current values correspond to 1 h average of 45 evenly spaced measurements over the sampling interval. Measurement bins located close to the boundaries (bottom and surface) or to other instruments near the 600 khz ADCPs were discarded due to evidence of signal contamination. The shallowest measurement bins consequently range between 12 and 20 m, and the deepest between 125 and 135 m. After removing the contaminated bins, the U and V components of the currents have been low pass filtered with a cutoff frequency of 48 h, in order to damp the tidal and inertial motions. [10] To our knowledge, this is the most comprehensive current data set collected in this area of the GoM. Several authors have actually raised the lack of onsite measurements in the Mexican waters of the GoM [Boicourt et al., 1998; Zavala Hidalgo et al., 2003, 2006] Wind [11] The wind speed vectors at 10 m elevation were collected from the North American Regional Reanalysis model (NARR) outputs, provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their Web site ( gov/psd/). The data cover the period between 1 November 2007 and 31 July 2009, with a 3 h time step and an approximate resolution of degrees. Each component was low pass filtered with a cutoff frequency of 48 h to match the filtering undergone by the current measurements. [12] The wind components along the Mexican coastline of the GoM were linearly interpolated from the 4 closest outputs of the NARR grid. The alongshore wind stresses were then calculated using the formula by Large and Pond [1981], and monthly averaged over the measurement period Drifters [13] A set of trajectories of surface drifters flowing in the continental shelf region during the period analyzed was included in the analysis of the surface currents. The data set forms part of an ongoing surface drifter program designed to monitor the surface currents over the deep waters of the Bay of Campeche. Monthly airborne deployments of three 2of17

3 Figure 1. Position of the moorings along the 130 m isobath. Shown are the 500, 1000, 1500, 2000 and 3500 m isobaths. Moorings marked with a circle were deployed from November 2007 to July Moorings marked with a square were deployed from July 2008 to July to five surface drifters have taken place south of 21 N over bottom depths greater than 1000 m since late September Of 154 drifters successfully launched by August 2010, 36 flowed into waters less than 200 m deep. Their trajectories in shallow water (shallower than 200 m) were selected for this study. [14] The drifters consist of a cylindrical shaped buoy hull (96.5 cm long and 12.4 cm in diameter) with a 45 m nylon tether line attached to a 1.2 m diameter paradrogue, that serves both to protect the buoy when water landing as well as a drogue to reduce slippage of the buoy in the water (Far Horizon Drifter, Horizon Marine Inc.). Hourly positions are recorded by a GPS receiver, and transmitted via Argos [Anderson and Sharma, 2008]. [15] Data gaps less than 6 h long were interpolated and a quality control process eliminated data of drifters on land, stuck on the bottom or onboard vessels, speeds exceeding 3 m/s or showing conspicuous peaks, and erroneous positions that result from bad fixes from the GPS receiver. Velocities were estimated from the hourly data using a central difference scheme. The hourly data from the drifters were separated in two groups depending on the time of the year they were collected: spring summer (March to August) and fall winter (September to February), and for each group another selection divided the data set into up coast or downcoast flowing periods. 3. Monthly Dynamics 3.1. Wind Forcing [16] The monthly mean along coast wind stress observed in the two years analyzed can be divided into 3 regional regimes (Figure 2). [17] Along the State of Tamaulipas (22.5 N to 26 N, moorings PER and LMP), the wind blew up coast from February to August, and during November and December, with greatest intensity from June to August. It was weakly down coast in January and experienced a dramatic inversion from up coast to down coast during September and October. [18] For the State of Veracruz (18 N to 22.5 N, moorings ARN, LNK, IT1 and CTZ), the monthly mean alongshore wind stress was mostly weak from February to May, although 2 northerly wind events could be spotted monthly on average. From June to August, it increased toward the north (moorings ARN and LNK) taking a steady up coast direction, and remained weak near the BoC (moorings IT1 and CTZ). The down coast reversal that occurred in September and October was intense, taking greatest values in September between LNK and IT1 due to favorable orientation of the coastline relative to the wind. These stronger values correspond to the most intense monthly alongshore wind stress experienced during the period analyzed over the TAVE and WCB shelves. In November and December, the intensity dropped back to weak values. [19] In the area of the WCB (mooring IT2), the monthly mean winds blew up coast year round with moderate intensities Surface Currents [20] As stated before, the shallowest measurements correspond to the current averaged over a bin 8 m thick, centered between 12 and 20 m depth depending on the mooring. Those measurements will be referred as surface currents. Due to the planning of the deployment and recovery operations, the current velocity statistics for July, August and September are only partial for some moorings. PER and LMP have either no or less than 5 days of measurement 3of17

4 Figure 2. Alongshore wind stress component averaged monthly over the measurement period. Positive (Negative) values indicate up coast (down coast) direction. during these months. ARN, LNK, IT1, CTZ and IT2 had between 14 and 28 days of measurement during July, IT2 had 20 days in August. [21] The correlation coefficient between the monthly alongshore wind stress and the averaged alongshore surface current is significant at the 95% level with 12 degrees of freedom at ARN (0.77) and LNK (0.78) along the TAVE shelf. These values are lower than the ones obtained with modeled currents by Zavala Hidalgo et al. [2003], which were in the 0.90 range in the same area. Specifically, no significant correlation was obtained at the northern locations of the TAVE shelf, namely PER and LMP. In the southern BoC, CTZ and IT2 have significant correlations of 0.61 and 0.67 respectively. At a location close to IT2 however, Zavala Hidalgo et al. [2003] did not find a significant correlation between the wind stress and the surface currents. Also, no significant correlation was found at IT1 (0.47), despite the fact that this station is located very close to CTZ, with similar wind regime. [22] The monthly averaged surface currents and corresponding standard deviation ellipses are presented in Figure 3. As in most coastal environments, the mean of the currents and the principal axis of the standard deviation ellipses are mostly oriented in the alongshore direction for all months, meaning that the water motions are essentially alongshore. [23] The moorings situated in the outer continental shelf of the BoC, namely CTZ, IT1 and LNK, mostly experience low surface currents on average, with a high variability. The mean velocities are typically in the 5 to 15 cm/s range from November to May, with standard deviations about twice as large. The alongshore current at LNK follows the wind stress trends at the same position, increasing in January and changing direction several times between November and May. According to the correlation coefficient values mentioned earlier, the relation between the wind and the surface currents is highest at this mooring. The currents at CTZ and IT1 show no preferential direction or particular intensification, possibly due to low alongshore wind stresses, but have a greater standard deviation in November. Current speeds at CTZ, IT1 and LNK increase slightly during summer to about 15 to 20 cm/s and are clearly pointing up coast, in agreement with the increase of the up coast wind stress observed from June to August. The corresponding standard deviations decrease to an average of 10 cm/s. [24] The currents between ARN and LMP are always in phase regarding their alongshore direction, except in November when they are down coast at LMP and up coast at ARN. They are flowing up coast from about December to August with intensities and standard deviations greater than the ones observed in the continental shelf of the BoC. In December and April at LMP, or in February and March at ARN, the mean current velocities increase to 30 to 40 cm/s, although the wind stresses remain low or at most moderate during the same months. An obvious discrepancy between the wind stress and the current occurs in January, when the current is heading north at about 26 cm/s on average at LMP while the wind stress is down coast. The up coast current velocities stabilize around 30 to 40 cm/s with lower standard deviations, during late spring and summer, as the summer wind regime develops. [25] The mean currents at PER are lower than 20 cm/s and mostly lower than 10 cm/s from November to May, flowing up coast from November to January and down coast from February to May. The standard deviations are around 10 cm/s to 15 cm/s. Although the alongshore wind stress is among the most intense at this mooring, the correlation coefficient between the wind and the alongshore current is not significant. The variations of the wind and the alongshore current are therefore disparate, and they are flowing in opposite directions several times between November and May (January, March, April, May). However, in June and 4of17

5 Figure 3. Monthly averaged surface currents and corresponding standard deviation ellipses. Shown are the 500, 1000, 1500, 2000 and 3500 m isobaths. 5of17

6 July, the current turns up coast and accelerates to around 40 cm/s, in conjunction with the intensification of the southerly wind stress. [26] The currents in the southwestern Campeche Bank were recorded by IT2. They are oriented up coast during the entire year except in September and October. The mean current varies between 4 cm/s and 13 cm/s between December and June, with standard deviations between 8 cm/s and 20 cm/s. The low velocities do not reflect the intense alongshore wind forcing experienced at this mooring. The current accelerates to about 20 cm/s in summer while its standard deviation drops to 5 cm/s on average. A possible convergence can be seen in November between IT2 (up coast, 18.9 cm/s) and CTZ (down coast, 13.3 cm/s) Drifters [27] The trajectories of the surface drifters on the continental shelf are shown in Figure 4. During spring summer, the drifters headed up coast 70% of the time. North of 21 N, all nine drifters flowed up coast except for one, which for a short period headed in the other direction. Speeds increased from south to north, with values above 80 cm/s recorded by four drifters off the Tamaulipas shelf. Most of the up coast trajectories occurred between April and June. South of 21 N, drifters moved both in the up coast and down coast directions, their speeds lower compared to their northern counterparts, with the exception of a down coast flowing drifter that recorded speeds over 80 cm/s for a short time. Most of the down coast data were collected between April and May. [28] During fall and winter, a majority of drifters headed down coast (74% of the drifter data), five of them recording speeds over 80 cm/s at various locations all along the shelf, reaching sustained speeds over 80 cm/s between 19 N and 22 N. The down coast data were nearly evenly distributed among the months corresponding to this period. By contrast, up coast flows occurred only 26% of the time, located exclusively north of 22 N, with the majority of the data collected between September and November. 4. Description of Specific Circulation Regimes 4.1. Fall Event [29] On Figures 2 and 3 we can observe a dramatic reversal of the alongshore wind stress and currents along the TAVE and WCB shelves in September and October. This fall directional shift is one of the most significant seasonal features of the circulation over the continental shelf of the GoM. It distinguishes the spring summer conditions, with both wind and currents flowing up coast, from the fall winter conditions with wind and currents flowing down coast. During this time period (fall and winter), the shelf circulation is under the influence of strong northerly wind events forced by cold atmospheric fronts traveling southward, inducing important variability in the current and the atmospheric forcing. [30] The down coast component of the wind stress varied little or slightly increased between PER and LNK, but it increased notably farther south to reach a maximum between LNK and IT1 due to the coastline orientating more parallel to the wind. Then, it decreased slightly but remained high in the region of IT1 and CTZ. The intensification of the alongshore wind in the southern TAVE shelf is consistent with the highest alongshore current velocities recorded at LNK in September, and IT1 and CTZ in October with values around 40 to 60 cm/s. These are the strongest monthly averaged speeds reached at these locations over the measurement period by at least a factor of 2. The current reversal is also visible at IT2 with values between 13 and 20 cm/s on average, although the wind kept an up coast orientation year round. [31] Figure 5 describes the variations of the most significant physical parameters in the fall of 2008, during the wind and current reversal. Although not all the parameters have values during this period, the same timeline has been kept for all of them to facilitate the tracking of the significant events, spatially and temporally. [32] Displayed as bold solid lines is the daily alongshore wind stress at the position of each mooring, with positive (negative) values being up coast (down coast) (Figures 5a 5g). The wind was blowing up coast at all the moorings until the end of August, and shifted to a steady down coast direction in mid September, except at IT2. It kept this direction during the second half of September and most of October, although some up coast wind events could be found during this last month especially at the northernmost moorings. Some episodic intensifications of the down coast wind stress occurred on 18 and 27 September (W1 and W2 respectively) and around 17 and 29 October (W5 and W6 respectively). At the beginning of November, the wind shifted again to a mostly up coast direction at all moorings. [33] The variations of the water level anomaly at the coast taken from the tide gauge of the City of Veracruz, as seen from Figure 5h (solid line), show a very good correlation with the low passed alongshore wind stress of IT1, also presented in Figure 5h (dotted line). Specifically, the shift in wind direction in September coincided within 5 days with the water level anomaly shifting from negative to positive. Then, the persistence of the down coast wind speed in September led to a gradual increase of the water level at the coast, caused by the accumulation of water by onshore Ekman transport. It decreased by the end of September and the beginning of October as the wind stress decreased, before rising again in the second half of October following the wind stress reinforcement. [34] Figures 5a 5g show the alongshore current velocity profiles as measured by the moorings. Red stands for downcoast currents and blue stands for up coast currents; white patches stand for contour speeds below 10 cm/s. The general pattern of the alongshore current velocity is consistent with the wind driven, low frequency variations of the coastal water level. The current shifted from up coast to down coast at the beginning of September, during which the overall velocities were greatest. They decreased slightly between the end of September and the first half of October, before increasing again during the second half of October. The current turned up coast again as well as the wind by the time the water level anomaly became very low at the beginning of November. This current pattern was observed at least along the southern TAVE shelf (moorings ARN to CTZ), possibly extending farther north considering that the alongshore wind trends were similar along the western coast. However, the absence of measurements during this period at PER and LMP prevents us from having a clear idea of the current behavior in the northern TAVE shelf during the wind reversal. 6of17

7 Figure 4. Drifter trajectories as functions of time of the year and along coast direction, color coded by speed. Solid black circles show the last position of the drifter (end of trajectory) next to the corresponding drifter number. Data collected during (left) March through August and (right) September through February. Histograms show the proportion of data heading (top) up coast or (bottom) down coast for the corresponding time of the year. Thin black lines correspond to portions of trajectories either on waters deeper than 200 m or heading in the opposite along coast direction. [35] Superimposed with the underlying variations of the water level, alongshore wind stress and current velocity, higher frequency variations of these parameters can also be seen. The major ones are related with intense wind events lasting about two or three days, traveling southward and highlighted by the W marks on Figures 5a 5g. Those northern winds blew down coast over the TAVE shelf but up coast with globally lower stress in the western Campeche Bank (IT2), due to different orientation of the coastline. They systematically came with an increase of the water level anomaly at the coast by about 50 to 150 mm recorded by the tide gauge of the City of Veracruz (arrows on Figure 5h) and stronger down coast currents. [36] The first short event occurred the first time the wind significantly shifted from up coast to down coast all along the TAVE shelf around 7 September (W1). The down coast current acceleration was coherent through the entire water column. It was first detected at ARN, and about 72 h later at IT2 which corresponds to an average down coast propagation speed of about 2.9 m/s between the two moorings. The greatest velocities (in the 60 cm/s range) were reached at the level of LNK, then weakening toward IT2. The sea level rise at the City of Veracruz and the current acceleration measured at IT1 were simultaneous. [37] Another example of such propagation and coherence between the wind, sea level height and alongshore currents can be observed at the end of October, with the current signal propagating from LMP (possibly PER) to IT2 at about 5.4 m/s on average (W6). At the northern moorings, the current acceleration was sharp, with maximum speeds increasing between LMP and LNK and steady down coast wind stress. Farther down coast, the current acceleration was more scattered in time, with decreasing maximum speed while the wind stress faded. At the level of IT2, most of the signal was already dissipated. Similar description can be made for W5, although the maximum current speed was reached at LNK and the intense down coast wind stress held all along the TAVE shelf. 7of17

8 Figure 5. Color maps of the alongshore current velocities measured by the moorings in cm/s: (a) PER, (b) LMP, (c) ARN, (d) LNK, (e) IT1, (f) CTZ, and (g) IT2. Superimposed are the alongshore wind stresses at the position of the moorings. Positive (negative) values are up coast (down coast). The beginning of events of interest are marked by a vertical solid line marked with a W. Arrows show the maximum mean current speed following each W event. (h) The low passed (cutoff period is 30 days, dotted line) and band passed (periods between 2 and 30 days, solid line) sea level anomaly (SLA) at the City of Veracruz which is located between LNK and IT1. SLA corresponding to the W events are marked by an arrow. Low passed (cutoff period is 30 days) alongshore wind stress at the City of Veracruz is represented by a dashed line. Arrows show the positive SLA associated with each W event. [38] Although the current was flowing down coast during most of the time in fall, some periods of up coast currents and baroclinic circulation can also be seen. Specifically, during the weak W4 event, when a down coast current acceleration propagated at least from ARN down to CTZ and came with an elevation of the sea level at the coast, the signal was interrupted by an up coast current at LNK. However, this upcoast current did not prevent the down coast signal from 8of17

9 propagating farther south and was very likely forced by a local anticyclonic eddy present in the area of LNK at this time. [39] Only one drifter entered the area shallower than 200 m during the current reversal in early September It arrived in the vicinity of the 500 m isobath on 12 September, at the position of LMP. There, its speed increased from about 0.2 m/s to as much as 0.8 m/s and its direction changed dramatically from northwest to exclusively down coast. The drifter entered the continental shelf (shoreward of the 200 m isobath) at the latitude of ARN. There, its velocity varied between about 0.2 and 1.05 m/s in a succession of accelerations and decelerations. When passing by ARE on 16 September, the speed of the drifter was very similar to the current velocity measured by the mooring at 50 m depth. Both were about 0.7 m/s. The greatest speed was reached when the drifter approached LNK on 20 September. Again, very good consistency was obtained between the velocity of the drifter and of the mooring, recording respectively 1.05 and about 0.8 m/s at 50 m depth and up to 1 m/s at the surface. The drifter then spent 4 days traveling between LNK and IT1, with speeds in the 0.3 m/s range, and experienced an intense acceleration in the vicinity of IT1 and CTZ, consistently with the measurements at the two moorings. After entering the bay in front of the City of Coatzacoalcos (19.15 N W), the drifter s speed decreased to less than 0.4 m/s The Spring/Summer Regime [40] From a climatological point of view, upwelling favorable winds on the TAVE shelf mostly blow during summer conditions (about April through August) and have a southerly to southeasterly orientation. The current velocity profiles of the moorings LMP, ARN and LNK are shown in Figures 6a, 6c and 6e, with the daily alongshore wind stress superimposed, from mid March to July The color code is the same as for Figure 5. Also plotted are the water temperature variations at 14 m above the seafloor measured at the same moorings. [41] According to our data, southerly (up coast) winds prevailed from mid April onward, occasionally interrupted by intense down coast events. Most of the time at ARN and LNK the currents flowed up coast near the surface and reversed down coast near the bottom. This configuration became more and more obvious southward, and as summer approached, with the nodal point of the current reversal gradually moving closer to the surface. Similar vertical structure was also seen at LMP from about the end of May onward. The most intense up coast currents were reached at LMP with values up to about 90 cm/s, decreasing southward to around 25 cm/s at LNK. Although the up coast currents at ARN and LMP were very consistent with the wind forcings, they were likely to be enhanced by the presence of the anticyclonic LCE Cameron, which collapsed along the continental shelf at the latitude of ARN in February From March to May, this eddy gradually traveled north, impacting the circulation in the vicinity of LMP. [42] Sharp periods (about 2 days) of down coast currents through the entire water column came with the intense down coast wind events. Maximum speeds were often reached at middepth or below and varied between about 40 and 60 cm/s, being highest to the south of the area. The water temperature systematically warmed up 2 to 5 degrees during those northerly winds as revealed by the Figures 6b, 6d and 6f. The most significant warming events were observed at ARN at the end of March and the beginning of April, when the near bottom temperature at 130 m reached about 24 C, which is similar to the mean summer temperature at 40 m depth according to the numerical results of Zavala Hidalgo et al. [2003]. For the other warming events, the temperatures increased from less than 20 C, to values between 21 and 23 C. On the contrary, the up coast winds tended to maintain water temperatures in the 17 to 20 C range, which is about 1 to 4 degrees less than the climatology of Zavala Hidalgo et al. [2003] Influence of Small and Medium Scale Eddies [43] From February to May 2009, the Loop Current Eddy (LCE) Cameron collapsed against the continental shelf of the GoM at the latitude of ARN and gradually traveled up coast while decaying. The mean Sea Surface Height Anomaly (SSHA) in February and May 2009 is shown in Figure 7 and illustrates the existence of the anticyclone paired with a cyclonic eddy located farther north, at the level of PER. Considering the position of the two eddies, they are expected to have influenced the circulation at PER, ARN and LMP. During the same period, the monthly wind stress headed up coast at LMP and was very low at ARN as seen in Figure 2. [44] In February and March, the alongshore surface currents at ARN and LMP were qualitatively consistent with the alongshore wind stress, running up coast. However, from a quantitative point of view, the current at ARN was faster than at LMP by at least a factor of 2, when the LCE was located in its vicinity. The influence of Cameron on the intensity of the up coast flow is strongly confirmed by the trajectory and speed of the drifter that entered the continental slope and shelf in February (Figure 7). According to the altimetry, the westward displacement of the drifter was likely driven by Cameron s circulation as it approached the northern BoC. Once entering the shelf, the drifter experienced an up coast acceleration from about 0.5 m/s when drifting shoreward, to 1.07 m/s upon arrival to the vicinity of ARN. Speeds near 1 m/s were maintained until it reached the convergence area between the anticyclone and the cyclone. There, the velocity dropped to 0.4 m/s in conformity with the lower current recorded by LMP at the same time. [45] In April and May, the decaying LCE traveled north, closer to LMP. The monthly surface currents at this mooring increased to about 0.5 m/s under the influence of the anticyclone. Again, the motion of the drifter that entered the continental slope and shelf in May readily illustrates the influence of the local eddies on the slope and shelf circulation. It flowed exclusively up coast with increasing speed from LNK up to the convergence area between the two eddies, located south of PER. The highest speeds (in the 1 m/s range, up to 1.18 m/s) of the drifter were reached while traveling along the western edge of the decaying LCE. The speed dropped upon arrival to the convergence area, as it was advected seaward by the circulation associated with the anticyclone and the adjacent cyclone. There, the motion of the drifter was dominated by the eddy current as it circled cyclonically and took a down coast/shoreward direction along the western edge of the cyclone, in accordance with 9of17

10 Figure 6. (a, c, and e) Color maps of the alongshore current velocities measured by the moorings LMP, ARN and LNK, respectively, in cm/s; the color scale is the same as for Figure 5. Superimposed are the alongshore wind stresses at the position of the moorings. Positive (negative) values are up coast (down coast). (b, d, and f) Temperature at 14 m above the seafloor at the position of the moorings. the direction of the currents measured at PER. The drifter was then expelled a second time from the outer shelf after entering the convergence zone. [46] From February to May, the cyclonic eddy paired with the LCE remained in the vicinity of PER. The currents measured by the mooring kept flowing down coast during that period, in accordance with the cyclonic circulation forced by the eddy, but against the high to moderate alongshore wind stress. 5. Vertical Structure of the Currents 5.1. Empirical Orthogonal Function Decomposition [47] Empirical orthogonal function (EOF) analysis in the time domain has been performed on the alongshore current profiles following the method described in Emery and Thomson [1998]. The results of the decomposition are presented in Figure 8, and Table 1 gives the value and lag times of the maximum correlation coefficient between every pair of first principal components. [48] The first EOFs account for 85 to 91% of the total variance of the alongshore velocity profiles. The spatial modes outline is very similar from one mooring to another, showing a quasi barotropic profile with little variations within the upper 60 to 80 m of the water column, then a gradual decay bottomward. [49] The first principal components represent the variations with time of the first EOFs. Positive values account for up coast currents, and negative values account for downcoast currents over the entire water column. Most of the general features already reported for the surface currents are also observed in the variations of the first principal component, some of which are (1) the down coast currents observed almost all along the coast in September October, with the principal components taking important negative values, (2) the effect of the LCE collapsing on the continental break in late February 2009, driving up coast currents at ARN, and (3) the increase of the up coast currents during spring summer in the TAVE shelf. 10 of 17

11 Figure 7. Average sea surface height anomalies in (left) February and (right) May Solid (dashed) lines are positive (negative) anomalies; thick line is the isoanomaly 0. Superimposed is the path of the two drifters that entered the continental shelf in the same period. They entered the continental shelf to the south of the TAVE shelf and traveled up coast. The color scale stands for the velocity of the drifters, in m/s. [50] Very little variability of the principal component was experienced to the north of the TAVE shelf, in the area of PER, as can be seen from the variance preserving spectra of Figure 8, confirming the trend already observed at the surface. Also, almost no significant correlation has been found between this mooring and the others located down coast as can be seen on Table 1. [51] Low but significant correlation is found between LMP and the moorings farther south. There the variability rises, most of which is being captured by periods greater Figure 8. (left) First empirical orthogonal function at moorings PER to IT2. (middle) First principal components and percent of variance explained by mode 1. (right) Variance preserving spectra of first principal components. 11 of 17

12 Table 1. Maximum Correlation Coefficients (r max ), Lags and Corresponding Propagation Speeds Between the First Principal Components a PER LMP ARN LNK IT1 CTZ IT2 Mooring r max Lag r max Lag r max Lag r max lag r max Lag r max Lag r max Lag PER LMP ARN LNK IT CTZ IT C a Lags are in hours. Correlation coefficients and lags below the 95% significant level were discarded. Positive lag means that the mooring under the column heading is leading the mooring under the row heading. C values correspond to the experimental propagation speed (m/s), computed through linear regression between the alongshore distance separating 2 moorings, and the lag of maximum correlation. Only the 6 last moorings had a sufficient number of experimental points for a linear regression to make sense. than 5 days. Specifically, the 5 to 8 day and 15 to 20 day bands seem to take most of the variance from LMP down to IT2, with greatest values at ARN, decreasing down coast. The correlations within the pool of the 6 southernmost moorings are systematically significant with largest values encountered between the moorings of the Tamaulipas and BoC shelves. They take values between 0.36 and 0.92, with 0.60 on average (0.69 when only the moorings of Tamaulipas and BoC shelves are considered). The associated lags are consistently positive in the down coast direction and negative in the up coast direction, characterizing a downcoast propagation rate in the variations of the alongshore current. Also, for the six down coast moorings, the lag of maximum correlation depends linearly on the alongshore distance, with an average slope of 3.59 m/s Coherence of the Barotropic Signal Along the Coast [52] Considering the low correlation of all the moorings with PER, the latter has been discarded from the present analysis. From the 6 time series of principal components determined independently at the 6 remaining moorings, we can look at the cross coherency between the principal component time series at LMP and any other mooring (Figures 9a 9e). [53] The cross coherency of the alongshore current at LMP with all the other moorings is significant at the 95% level for periods ranging from 6.1 to 10.7 days. A second band, around the 3 day period, also shows some significant coherency except at LNK. This band however seems of secondary importance compared to the 6.1 to 10.7 day band which shows the greatest and most consistent values from one mooring to another. The coherence in this band remains high all along the TAVE shelf, with highest values at LNK. For further analysis, the periods marked by the solid circles in Figures 9a 9e were investigated. They are 8.5, 7.1 and 6.1 days and were picked off because they have the highest squared coherency on average, over the entire set of moorings. [54] Figures 9f 9h show the lag observed between two time series of principal components at a specific frequency (or period), plotted against the alongshore distance separating the moorings corresponding to those time series. The lags were determined from the phase difference between the 2 moorings considered, at the appropriate frequency. With 6 series to analyze and accepting all of the positive phase differences, there are 15 different pairs of time series. Some degree of redundancy is introduced in the process as, for example, propagation from ARN to LNK and from LNK to IT1 requires propagation from ARN to IT1 provided that the coherences are large. However, including all the displacement/phase points allows to form an unbiased estimate of the alongshore propagation rate. Phase differences corresponding to squared coherency values below the 95% confidence level were discarded. [55] At a period of 6.1 days, all the displacement/lag pairs are tightly aligned along a propagation rate of about 4.35 m/s, calculated from the orientation of the principal axis of variance of the point cloud. Its slope implies a downcoast propagation of the up coast signal. At 7.1 and 8.5 day periods, the scatter of points around the fitted line is slightly larger than at 6.1 days, and the phase propagation speeds decrease significantly to 3.88 and 3.82 m/s respectively with a down coast direction. A striking result is the low scattering of the experimental points reported on the dispersion diagram, Figure 9i. The three wave numbers were computed for each frequency as k i = w/c i,c i being the phase speed estimated and reported on Figure 9f 9h. The phase speed deduced from this diagram, 4.00 m/s, is the harmonic mean of the C i. [56] The squared coherency between the alongshore wind stress and the first principal components is shown in Figure 10. It is significant in the 6.1 to 8.5 day band whether the local wind (dotted lines) or the wind at LMP (solid line) is considered. Only at IT2 is the squared coherency between the local wind stress and the principal component not significant at 6.1 days. At all moorings, the wind stress at LMP appears to have better coherence with the alongshore currents than the local wind. The coherence is particularly high at ARN, LNK, IT1 and CTZ in the 6.1 to 7.1 day band when considering the wind at LMP. From the phases of the coherence analysis, the wind signal at LMP would lead the current signal by about 12 h at the same mooring. Then, the lags with the moorings farther down coast increase with the distance from LMP at a mean rate of about 4.2, 4.3 and 4.6 m/s for 6.1, 7.1 and 8.5 day period, respectively. [57] The squared coherency spectra between the alongshore wind stress at LMP and the sea surface height anomaly at Port Isabel (located at the U.S. Mexico border, at about 26 N) and the City of Veracruz are plotted on Figure 11a. When compared with the sea level in Veracruz 12 of 17

13 However, no significant coherence is found between the sea level height anomalies at both locations in the 6 to 10 day band. Figure 9. (a e) Cross coherencies of the first principal components at LMP with the other moorings. The dotted lines stand for the 95% confidence level for 16 degrees of freedom. The solid circles identify the frequencies used in Figures 9f 9i. (f h) Phase difference of first principal components plotted against alongshore displacement. Every possible pairs of displacement/phase between the moorings south of LMP with coherence over the 95% confidence level are reported for each frequency. The solid line stands for the phase speed computed from the major axis of the variance ellipse. (i) Experimental dispersion diagram, with bold line having a slope equal to the harmonic mean of the individual phase speed estimates, listed on Figures 9f 9h. (solid line), the coherence values are significant in the 20 day, the 6 to 10 day and about 3 to 4 day bands. The 6 to 10 day band is barely significant, but it takes values much higher than when the analysis is carried out between the same wind and the sea level at Port Isabel (dotted line), for which the coherence is not significant in that band. At both locations the highest coherences are reached in the 20 day and 3 to 4 day bands, taking similar values at the City of Veracruz and Port Isabel. The cross correlation function between the sea level height at Veracruz and Port Isabel shows two maxima taking values of 0.76 and 0.74 with Port Isabel leading Veracruz by 72 and 11 h respectively. Most of the correlation value is captured by coherent variations of the sea level heights in periods longer than about 20 days or in the 3 to 4 day band as can be seen in Figure 11b. 6. Discussion 6.1. Comparison With Climatological Currents [58] Our observations generally agree with the seasonal trends of up coast and down coast flows shown by the numerical results of Zavala Hidalgo et al. [2003], particularly along the Tamaulipas and Veracruz shelves. In their case, the surface currents were driven by the climatological wind stress, which changes from blowing up coast to down coast in September, returning to the up coast conditions in May. Drifter trajectories suggest that the flow is continuous all along the shelf in both seasons. In addition, the measured sea level changes along the TAVE shelf were in phase with the low frequency variations of the observed down coast currents during fall winter. Both results are consistent with the idea of shelf flow driven by cross shelf pressure gradients. Considering the high correlation we observed between the wind stress and the sea level variations in fall, we suggest that the along coast wind stress would cause the variations of the coastal sea level through Ekman transport. The cross shelf pressure gradient would, in turn, force the low frequency variations of the alongshore current. [59] The period covered by the observations also shows significant differences with the modeled climatological currents. Along the Mexican/U.S. border (mooring PER), for example, surface current variations show weak dependence with the local winds, and were mostly driven by the presence of eddies interacting with the shelf during most of the measurement period. The next two paragraphs address further differences. [60] The fall winter wind reversal was 4 months shorter than the climatological one, with winds blowing down coast only in September and October. During that period, the surface currents flowed down coast all along the shelf from the Mexican/U.S. border to the WCB. The strongest downcoast winds and currents were observed along the shelf of Veracruz, specifically in the western and southern BoC area. This contrasts with the weak mean currents observed in the southern shelf during the rest of the year and showing no preferred direction of flow. Both the wind and current reversals were in phase within about 5 days according to Figure 5h. Similar swift reversal along the Louisiana Texas shelf in fall were reported by Nowlin et al. [1998], suggesting a continuity of this phenomenon all along the shelves of the northern and western GoM. [61] The currents measured in the WCB flowed down coast between September and October, while the local alongshore wind was blowing in the opposite direction. This contrasts with the modeling results of Zavala Hidalgo et al. [2003], which show both currents and winds flowing up coast throughout the climatological year. Hence, our measurements do not suggest an obvious convergence of the shelf flow caused by opposing currents in the southern Gulf of Mexico during the fall winter. Instead, we observe convergence of the mean flow in the southern shelf during the strong wind reversals of September and October, caused by differences in strength rather than in direction of the mean currents. 13 of 17