A model study of the Angola Benguela Frontal Zone: Sensitivity to atmospheric forcing



[1] A regional ocean model is applied to the tropical South East Atlantic Ocean in order to investigate the sensitivity of the intensity and position of the Angola Benguela Frontal Zone. The results suggest that the position of the Angola Benguela Frontal Zone is mainly determined by opposing northward and southward flow within 2 degrees off the coast. The windstress curl plays a major role in altering the position of the Angola Benguela Frontal Zone, as it controls the southeastward flow of the South Equatorial Counter Current. The opposing northward flow is mainly the result of geostrophic adjustment to coastal upwelling driven by the meridional windstress. The strength of the frontal zone is found to be related to the strength of the southerly windstress that controls the coastal upwelling.

1. Introduction

[2] The Angola Benguela Frontal Zone (ABFZ) is an important frontal feature situated between 15–17° S off the west coast of Africa [e.g., Shannon et al., 1987; Meeuwis and Lutjeharms, 1990; Kostianoy and Lutjeharms, 1999; Lass et al., 2000; Veitch et al., 2006]. It separates the warm saline equatorial current system from the cold and highly productive Benguela upwelling system and hence is a vital system boundary, whose fluctuation in position or intensity seems to affect local fisheries and rainfall variability [e.g., Gammelsrod et al., 1998; Binet et al., 2001; Rouault et al., 2003].

[3] The ABFZ reveals a typical annual cycle characterized by meridional frontal movements accompanied by changes in the cross-thermal gradient. The front is most clearly defined in austral summer, when it reaches its southernmost position, while it is less intense and positioned farthest north in winter [Meeuwis and Lutjeharms, 1990]. Superimposed onto the annual cycle is variability ranging from intra-annual to interannual time scales.

[4] Kostianoy and Lutjeharms [1999] found short term frontal variations being correlated to the pressure gradient driven by the South Atlantic Anticyclone (SAA). Major southward frontal shifts have been observed during warm events in the tropical South East Atlantic Ocean, such as Benguela Niños [Shannon et al., 1986; Veitch et al., 2006]. On time scales up to millennia, [Kim et al., 2003] found evidence of thermal gradient changes of the front due to changes in the intensity of the trade wind system.

[5] Despite its inherent variability, one of the remarkable characteristics of the frontal zone is that it is confined to such a relatively narrow band of latitudes. However, to date surprisingly little is known about mechanisms responsible for keeping the ABFZ in its position. Shannon et al. [1987] suggested that the frontal position is likely to be maintained by a number of factors (e.g. coastline orientation, bathymetry, stratification and windstress). The coast changes direction at Cape Frio (18.5° S, 12° E) and Tambua (16° S, 12° E) and a dramatic change in the shelf width and the steepness of the continental slope is apparent between 15–20° S. The southeasterly trade winds drive coastal upwelling to the south of the ABFZ, while to the north of the front they turn clockwise inducing negative windstress curl. All of these features may possibly act to maintain or influence the position of the ABFZ. However, Meeuwis and Lutjeharms [1990] concluded that the position of the front was solely due to the opposing flows of the Angola and Benguela Currents.

[6] In an attempt to better understand the underlying processes responsible for the formation and maintenance of the ABFZ, a regional ocean model has been applied to the tropical South East Atlantic region. As far as the authors are aware, this is the first attempt to model the ABFZ. The sensitivity of the position and intensity of the front to changes in windstress is investigated. Our approach may thus contribute to an improved understanding of the impact of remote climate signals, such as ENSO, on the ABFZ and the ecosystem.

2. Model and Experimental Design

[7] The regional ocean modeling system (ROMS) is used in this study. A detailed description of ROMS physics and its parameterizations are given by Penven et al. [2005]. The model domain extends zonally from 5° W to 20° E and meridionally from 30° S to 7° N and thus includes equatorial dynamics. The horizontal grid spacing is 1/3°, which is smaller than the baroclinic Rossby radius of deformation in that area (R ≈ 40–60 km) and is also fine enough to sufficiently resolve the main ABFZ features. In the vertical, 32 levels were used. Bottom topography in the model is derived from a 2' resolution database [Smith and Sandwell, 1997]. The model is nested within an ocean general circulation model (ORCA2 Colberg et al., 2004), which provided temperature and salinity for initialisation and monthly climatologies of temperature, salinity and velocities, as well as fresh water fluxes for the open boundaries and upper ocean forcing. Heat fluxes are taken from the NCEP/NCAR reanalysis [Kalnay et al., 1996].

[8] In the reference run, monthly mean windstress is computed from QuikSCAT satellite scatterometer data [Liu et al., 1998], covering the period from October 1999 till March 2003.

[9] Three sets of experiments have been conducted. Set 0 is the reference experiment forced with the above described configuration, which has been integrated for 32 years. The model solution reaches a statistical equilibrium after 2 years integration, similar to other ROMS configurations [e.g., Penven et al., 2005]. Note that Colberg [2006] showed that the model ABFZ and northern Benguela region compare favourably with available observations in terms of the annual cycle of upper ocean temperature and currents. Set 1 consists of two experiments in which the magnitude of the windstress has been modified so as to simulate a stronger and weaker trade wind system, respectively. To do so, an ellipsoid-like shaped windstress anomaly has been superimposed on the monthly climatological windstress. To avoid strong gradients, which may lead to unwanted divergence, the transition zone between the anomaly and the original windstress has been smoothed. The resulting values of the windstress and superimposed anomaly are 50% stronger (weaker) for experiment 1 (2). Set 2 also consists of two experiments, which aim to simulate an anomalous northward or southward position of the South Atlantic Anticyclone (SAA). To do this, the windstress has been respectively shifted meridionally by 1.5° northwards and southwards of its mean position. All Set 1 and Set 2 experiments were integrated for 12 years.

3. Reference Experiment (Set 0)

[10] Figure 1a shows the annual mean meridional temperature gradient in the frontal area at a depth of 20 m for the reference experiment. We use maxima in the thermal gradients to determine the fronts in the region. Multiple fronts are evident between 12° S and 19° S. The strongest front (the middle front) is situated between 16° S and 17° S, in good agreement with observations of the ABFZ from ship measurements and satellite imagery [e.g., Shannon et al., 1987; Kostianoy and Lutjeharms, 1999; Lass et al., 2000; Veitch et al., 2006]. This front is clearly defined up to 10° E, but can be traced farther westwards to at least 6° E. Its orientation is east-west, consistent with satellite maps of SST for the period 1982 to 1985 [Meeuwis and Lutjeharms, 1990]. In the following discussion, we focus mainly on the evolution of the middle front which, as shown later, may be determined by the dynamics of opposing geostrophic flows. The apparent northern and southern fronts in Figure 1a are the northern rims of local upwelling cells. Kostianoy [1997] termed the southern front the Benguela Upwelling Front (BUF) as it is formed by the northern edge of the Cape Frio upwelling cell.

Figure 1.

Annual mean of the model (a) meridional temperature gradient (dT/dy) and (c) windstress curl with geostrophic velocities (arrows) at 20 m superimposed, respectively. The zero contour line is dashed - dotted for Figure 1a. The contour interval is 1° C/100 km for Figure 1a. A grey-scale/colour-scale in the HTML and on-line PDF version in 10−7 Nm−3 is given to the right of Figure 1c. Hovmöller plot for the annual cycle of (b) dT/dy and (d) meridional geostrophic velocity averaged between 11–14° E. The zero contour line is omitted for Figur es 1b and 1d. The contour interval is 1° C/100 km for Figure 1b and 0.04 ms−1 for Figure 1d.

[11] A Hovmöller diagram of the meridional temperature gradient (Figure 1b) indicates that the middle front is weakest and farthest north in July consistent with observations [Shannon et al., 1987; Meeuwis and Lutjeharms, 1990]. The southernmost position appears to be in early spring (October) and in January, which only partly agrees with observations, as most authors suggest that the front is furthest south during summer [Shannon et al., 1987; Meeuwis and Lutjeharms, 1990]. However, the difference between the southernmost position of the model front in October and January is small. Strongest frontal gradients occur in late summer/early autumn (March to April), which marks the beginning of the upwelling season off Namibia and is in good agreement with observations [e.g., Shannon et al., 1987; Meeuwis and Lutjeharms, 1990; Kostianoy and Lutjeharms, 1999]. Maximum values of the middle front suggest a temperature difference of more than 4° C per 1° of latitude similar to what is observed [Shannon et al., 1987; Meeuwis and Lutjeharms, 1990].

[12] The annual mean geostrophic velocity associated with the sea surface elevation (Figure 1c) indicates a northward directed jet-like coastal current that encounters a somewhat stronger southward current between 16° S and 17° S (the approximate position of the ABFZ). The northward directed jet is likely caused by coastal upwelling [see, e.g., Philander and Yoon, 1982], which reduces the sea surface elevation there, leading to an onshore pressure gradient, which in turn drives a geostrophically balanced coastal jet. This picture is in agreement with the observations of Shannon et al. [1987]. The upwelling mechanism does not necessarily depend on the curl of the windstress but rather on the alongshore winds apparent between 17–18° S (not shown). These winds drive an offshore Ekman flow (not shown), which leads to divergence near the coast and requires upwelling due to mass conservation.

[13] The southward flow may be related to cyclonic windstress curl (Figure 1c), which is prevalent north of 15° S. However, closer to the coast, strongest southward flow exists just north of the maximum negative windstress curl (between 15–17° S), which implies a lowered sea surface height due to upwelling, and hence may act to steer the prevailing currents.

[14] A Hovmöller plot of the meridional geostrophic velocity within 2° off the coast, encompassing the equatorward coastal jet, is shown in Figure 1d. Strongest northward flow within the middle front area (17–19° S) occurs during April, which coincides with maximum alongshore windstress (not shown) and strong windstress curl (Figure 1c). Maximum southward flow (15–16° S) is also evident during April and again between August and November, similar to that observed for the maximum windstress curl (not shown), hence suggesting a possible connection as suggested above. The meridional movement of the opposing flow that characterises the annual cycle is in close agreement with the annual migration of the ABFZ.

4. Sensitivity to Windstress Forcing

[15] Strengthening of the mean trade winds by 50% results in a firmly established main front (Figure 2a), which is positioned at around the same latitude as the front simulated by the reference experiment. However, the gradient is slightly stronger during the first half of the year, in agreement with Kim et al. [2003], who suggested that an enhanced trade wind system may lead to an intensification of the ABFZ due to the combined effect of an enhanced heat transport into the tropical South Atlantic via the South Equatorial Current (SEC) and stronger upwelling. However in the model, the gradient increases only marginally, because the idealised wind system does not affect the SEC due to the model setup. Figure 2a indicates a reduced meridional movement of the front throughout the year, with a smaller northward shift in June, but a slightly enhanced southward shift of the front in October. The position of the southern front does not change a great deal, but is slightly weaker than that of the reference experiment.

Figure 2.

Hovmöller plot for the annual cycle of dT/dy, averaged between 11–14° E. (a) Experiment 1 of Set 1 (enhanced trade winds), (b) Experiment 2 of Set 1 (reduced trade winds), (c) Experiment 1 of Set 2 (northward shifted SAA), and (d) Experiment 2 of Set 2 (southward shifted SAA). Positive (negative) contour lines are solid (dashed). The zero contour line is omitted. The contour interval is 1° C/100 km.

[16] A reduction of the trade winds by 50% leads to a much weaker single front between 16–18° S (Figure 2b), as a result of reduced upwelling. The weak front is most pronounced from late summer through to late spring with associated temperature gradients about three times smaller than in the reference experiment.

[17] The main front in Experiment 1 of Set 2 (mean winds shifted north by 1.5°, Figure 2c) is positioned between 15° S and 16° S and is thus shifted northwards by slightly more than 1° compared to the reference experiment. Hence, it has responded almost linearly to the meridional shift in windstress but is not as wide as the front in the reference experiment and is stronger throughout the year. Thus, northward shifted wind forcing leads to a more distinct front, whose magnitude is more intense and uniform throughout the year. The southern front (18–19° S) is evident throughout the year, but has not changed its position in response to the anomalous windstress forcing. Maximum gradients in this front are apparent in late summer/early autumn and show a meridional migration similar to the middle front. A weak northern front is also apparent throughout the year except in late winter/early spring.

[18] The southward shifted SAA (Experiment 2 of Set 2, Figure 2d) results in a relatively weak southern front (18–20° S), which is just the mean front shifted southwards by about 1.5°.

[19] In order to assess the mechanisms potentially associated with the modeled fronts, the geostrophic velocities and windstress curl in each case are shown in Figure 3. The flow pattern in the northern part of the region for each experiment resembles the one evident in the reference experiment. Thus, the eastward flowing South Equatorial Counter Current (SECC) is apparent north of 15° S, turning southwards between 8° E and 11° E. The southward flow splits into an eastern and a western branch just to the north of maximum windstress curl. For each experiment, the western branch turns southeastwards towards the coast, approximately at the location of the ABFZ. Furthermore, the geostrophic flow towards the coast follows the maximum negative windstress curl (Figures 3a to 3d) in each case. Hence, it is suggested that the windstress curl may be responsible for the observed poleward flow in the ABFZ area. The confluence between the south and northward flows then determines the position of the ABFZ.

Figure 3.

Annual mean of windstress curl with geostrophic velocities (arrows) superimposed. (a) Experiment 1 of Set 1 (enhanced trade winds), (b) Experiment 2 of Set 1 (reduced trade winds), (c) Experiment 1 of Set 2 (northward shifted SAA), and (d) Experiment 2 of Set 2 (southward shifted SAA). A grey-scale/colour-scale in the HTML and on-line PDF version in 10−7 Nm−3 is given to the right of each plot. Note the different scales in the printed journal version for each plot.

[20] Immediately south of the ABFZ area, northward geostrophic flow is apparent at the coast for each experiment except for Experiment 2 of Set 1 (reduced trade winds, Figure 2b). This flow pattern is what is expected for alongshore winds that induce coastal upwelling and thus a geostrophic justified coastal jet. However, in Experiment 2 of Set 1 (Figure 2b), the flow is poleward due to the dramatic reduction of upwelling favourable winds.

5. Summary and Discussion

[21] In order to investigate the response of the ABFZ towards changes in windstress, a regional ocean model (ROMS) has been applied to the tropical South East Atlantic. The model successfully simulates the occurrence of a sharp thermal front situated at approximately 16.5° S, which is identified as the ABFZ.

[22] Several experiments have been conducted in order to assess the sensitivity and intensity of the ABFZ in the presence of changing windstress. In summary, the model results suggest that the position of the front is closely tied to opposing north and southward geostrophic flow within 1–2 degrees off the coast. The opposing flows appear to respond to the overlying atmospheric circulation. Strong anticyclonic windstress curl in the frontal area acts to steer the South Equatorial Counter Current (SECC) southeastwards towards the Angolan coast, while maximum alongshore windstress further to the south enables coastal upwelling, which triggers the equatorward coastal jet. This mechanism is similar to what has been observed during an upwelling event in the ABFZ [Mohrholz et al., 2001]. It is evident that changes in the meridional position of the windstress curl lead to changes in the position of the ABFZ.

[23] The intensity of the ABFZ is found to be mainly related to the generation of upwelled water towards the south of the ABFZ, which is determined by the strength of the meridional wind field to the south of the front. As a result, the ABFZ is sensitive to changes in both the intensity and location of the trade winds. The model results agree partially with Kim et al. [2003], who suggested that changes in trade winds alter the intensity of the ABFZ, due to the combined effect of enhanced upwelling and changes in heat transport into the tropical South East Atlantic. However, in our study it is found that the response of the ABFZ towards increased or reduced winds is not linear. Enhanced winds only marginally increase the temperature gradient, while reduced winds lead to a greatly weakened front. Differences between the model results and Kim et al. [2003] are likely due to fact that the model setup does not account for an adjustment of the equatorial currents towards changes in the trade wind system.

[24] As far as the authors are aware, this study is the first model investigation of the ABFZ. The results of this study expand our understanding of the mechanisms leading to the formation of the ABFZ and its sensitivity to variability in the forcing.

[25] Furthermore, our results imply a connection between a localized phenomenon, such as the ABFZ, and the large scale atmospheric circulation. Hence, South Atlantic climate variability [Venegas et al., 1997; Sterl and Hazeleger, 2003; Colberg and Reason, 2006]; ENSO related variability [e.g., Reason et al., 2000; Colberg et al., 2004]; and long term climate scenarios that involve changes of the intensity and position of the South Atlantic Anticyclone [e.g., Kim et al., 2003] are likely to affect the position and strength of the ABFZ.

[26] Low oxygen events are known to have devastating effects on the Benguela Current large marine ecosystem (BCLME), whereas the Angola Gyre is known to be a major source for oxygen depleted water in the region [e.g., Shannon et al., 1987]. Our study suggests that a southward shift of the ABFZ may be the result of a southward shifted South Atlantic Anticyclone and, furthermore, may be associated with an anomalous southward protrusion of the Angola Current. Hence, more oxygen depleted water may be advected into the area. Thus, with respect to the large biodiversity of the BCLME, further investigations should focus on an improved understanding of the relation between the ecosystem and variability in the ABFZ.


[27] We would like to thank Pierrick Penven for initial help with the ROMS model. Financial support from the Benguela Current Large Marine Ecosystem (BCLME) is gratefully acknowledged.