1. Introduction
The Brazilian Southeast continental shelf, in the vicinity of Cape Frio (Fig. 1), is characterized by strong upwelling, particularly during spring and summer (Castro and Miranda 1998). The large-scale atmospheric high pressure center located over the South Atlantic Ocean makes the prevailing surface winds blow from the northeast along the coast near Cape Frio (Stech and Lorenzzetti 1992). The northeasterly winds are upwelling favorable in the Southern Hemisphere and are highly correlated with surface cold water events near the cape (Allard 1955; Emílsson 1961; Ikeda et al. 1974). Winds blowing from the southwest (downwelling favorable) are frequent and intense during winter (although not predominant) because of an increase in the frequency of frontal passages. Upwelling events have been shown to strongly influence the local ecosystem, enriching the water column with nutrients and supporting the regional fisheries productivity (Matsuura 1996, 1998). The cold water plumes at the surface also modify surface heat fluxes, affecting the sea breeze circulation near Cape Frio (Franchito et al. 1998).
(top) Summertime and (bottom) wintertime (left) average SST (°C), (middle) average wind stress curl [(N m−2) (104 km)−1], and (right) wind stress curl standard deviation [(N m−2) (104 km)−1] derived from MODIS and QuikSCAT satellite observations. Vector-average wind stress for each season is overlain on left panels. Colorbar for bottom left panel is shown on the bottom. São Sebastião Island (SSI), Cape Frio (CF), Cape of São Tomé (CST), and Caravelas (CA) are shown.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
In addition to this traditional view of coastal upwelling, several other mechanisms may play an important role bringing the cold water to the surface near Cape Frio. Variations in vorticity resulting from curvature of the trajectory as the flow passes the cape (Arthur 1965) have been suggested to be one of the key factors controlling the magnitude of the upwelling at Cape Frio (Rodrigues and Lorenzzetti 2001). Other mechanisms include instabilities of the mean flow (Campos et al. 1995, 2000), changes in coastline orientation and alongshelf bottom topography modifying the alongshelf pressure gradient (Palma and Matano 2009), and wind stress curl–driven upwelling (Castelao and Barth 2006). Ekman pumping from wind stress curl, in particular, has been shown to be one of the dominant mechanisms driving vertical transport in other coastal systems (Pickett and Paduan 2003). Off Cape Frio, vertical transport driven by wind stress curl is only slightly smaller than vertical transport because of coastal upwelling during summer, effectively doubling the total vertical transport near the cape (Castelao and Barth 2006).
Regardless of the mechanism responsible for bringing the cold water to the surface near the coast, upwelling off Cape Frio is responsible for the development and maintenance of a strong sea surface temperature (SST) front separating cold waters inshore and warm waters influenced by the Brazil Current offshore. Several studies have shown that SST can have profound influence on wind stress variability throughout the World Ocean wherever there are strong SST fronts (e.g., Chelton et al. 2004; Xie 2004). In particular, Chelton and Xie (2010) summarize that surface wind increases over warm water in association with decreased stability through enhanced vertical mixing that deepens the atmospheric boundary layer and draws momentum from the upper boundary layer down to the sea surface. Over cold water, on the other hand, surface winds and wind stress decrease in association with increased stability that decouples the surface winds from the stronger winds aloft. Therefore, if winds blow along a SST front, high wind stress over warm water and low wind stress over cold water generate a curl that varies linearly with the crosswind component of the SST gradient. This effect has been well documented for open ocean conditions (O’Neill et al. 2003; Chelton et al. 2004; O’Neill et al. 2005, 2010; among others) and has recently been shown to be quite significant in the coastal ocean, including the vicinity of capes (Chelton et al. 2007). By this mechanism, an initially spatially uniform wind blowing from the northeast along the SST front off Cape Frio would be modified in a way that generates a negative anomaly in the wind stress curl (decrease in southward winds near the coast and increase offshore). In that scenario, the large-scale wind stress curl field would be modified so that anomalously negative intensifications in the wind stress curl would be a consequence of upwelling and the establishment of the SST front. A similar scenario can potentially occur in several other capes around the world (e.g., Cape Blanco off the U.S. West Coast; Cape Columbine off South Africa). Therefore, wind stress curl anomalies off Cape Frio can be either a cause or a consequence of upwelling.
The ocean–atmosphere interaction described above has already been shown to be important in another coastal region (Chelton et al. 2007). Here, the main goal is to quantify the relation between SST, SST gradients and wind stress curl variability near Cape Frio. Wind stress curl–driven upwelling has been shown to drive strong vertical transport near the cape (Castelao and Barth 2006). The relation between wind stress curl–driven upwelling and changes in SST is quantified, as well as the role of SST fronts and ocean–atmosphere interactions on wind stress curl variability.
2. Methods
Surface wind speed and direction are obtained from the SeaWinds scatterometer onboard the Quick Scatterometer (QuikSCAT) satellite. A detailed description of QuikSCAT is given by Chelton and Freilich (2005). The SeaWinds scatterometer is a scanning microwave radar that infers the surface winds stress from measurements of radar backscatter from the roughness of the sea surface at multiple antenna look angles (Naderi et al. 1991). Wind retrievals are accurate to better than 2 m s−1 in speed, and 20° in direction, which is essentially equivalent to the accuracy of in situ point measurements from buoys (Freilich and Dunbar 1999). Surface stresses are obtained from the equivalent neutral-stability 10-m winds using the modified Large-Pond drag coefficient for neutrally stable conditions (Large et al. 1994). The wind stress curl was computed within each measurement swath using centered differences. In the standard processing of QuikSCAT data used here, the spatial resolution is about 25 km and measurements closer than about 30 km to land are contaminated by radar backscatter from land in the antenna side lobes.
The region around Cape Frio is frequently cloudy, decreasing the opportunities for obtaining infrared sea surface temperature (SST) observations. Therefore, SST analyses from several sources are used here, in which in situ and remotely sensed data from a variety of satellites are blended together. Those SST analyses were produced by the Group for High-Resolution Sea Surface Temperature (GHRSST) (Donlon et al. 2007; http://www.ghrsst-pp.org), and are described in great detail by Reynolds and Chelton (2010). Two of the analyses are produced by National Oceanic and Atmospheric Administration (NOAA)’s National Climatic Data Center (NCDC), and combine in situ measurements and Advanced Very High Resolution Radiometer (AVHRR) data, or in situ, AVHRR and Advanced Microwave Scanning Radiometer (AMSR) data. Those are referred to as AVHRR-only and AMSR+AVHRR in Reynolds and Chelton (2010), respectively (the same notation is used here). A third analysis is produced by the Remote Sensing System (RSS) using AMSR, Tropical Rainfall Measuring Mission Microwave Imager (TMI), and Moderate Resolution Imaging Spectroradiometer (MODIS) data, and is referred to as RSS here. A fourth analysis, the Operational SST and Sea Ice Analyses (OSTIA), is produced by the Met Office using in situ, AVHRR, AMSR, TMI, Advanced Along Track Scanning Radiometer (AATSR), and geostationary Spinning Enhanced Visible and Infrared Imager (SEVIRI) data. The four SST analyses listed above are characterized by different degrees of smoothing associated with merging observations from different sources (Reynolds and Chelton 2010). For comparison, observations obtained directly from the infrared MODIS are also used.
Observations from April 2006 to November 2009 are used, when all SST analyses and QuikSCAT data are available simultaneously (with the exception of Fig. 11, right panel, where QuikSCAT observations from 1999–2009 are used). All observations were averaged onto a 0.25° latitude by 0.25° longitude grid and time-averaged as in Chelton et al. (2007). Specifically, SST measurements were first averaged in overlapping 3-day periods at daily intervals. The crosswind component of the SST gradient was computed within each QuikSCAT measurement from the instantaneous wind stress field and the 3-day-averaged SST field centered on the date of the QuikSCAT observation. Crosswind SST gradients are defined as the cross product
3. Results
The austral summer-averaged SST in the region is characterized by gradual cooling from north to south (Fig. 1). The dominant deviations from this pattern are the presence of cold water near Cape Frio, and the Brazil Current offshore, which transports warm water southwestward. Wind is predominantly from the northeast, which is upwelling favorable. Associated with this wind distribution is the presence of a strong intensification in the negative wind stress curl (i.e., favoring wind stress curl-driven upwelling in the Southern Hemisphere) in a band approximately 200 km wide from the coast and extending to the north and to the south of Cape Frio. The surface water is on average 1°–1.5°C cooler in the region of intensified negative wind stress curl (Fig. 2; note in Fig. 1 that the cooling is even stronger near to the coast, but no wind stress curl data are available because of the gap in QuikSCAT observations within 30 km from land). The wind stress curl far from the coast is positive and weak on average (Fig. 1). The wind stress curl near the coast is highly variable, with the standard deviation exceeding the averaged values.
Binned scatterplot of summertime wind stress curl vs average sea surface temperature (SST). The statistics were computed over the area shown in Fig. 6 (top panels). The points are the means within each bin computed from overlapping 29-day averages at 7-day intervals in the three December–March time periods during late 2006 to early 2009, color coded for the different SST analyses. The error bars represent the ±1 standard deviation over all OSTIA individual 29-day averages within each bin.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
During winter, SST is still lower near Cape Frio, although the coldest water is now found in the south associated with northward transport of shelf water originated farther south (Fig. 1; Piola et al. 2000). The wind stress curl remains negatively intensified near the coast, and standard deviations are even higher than during summer. The band of high standard deviation extends farther south during winter. For the most part, the average wind stress remains southwestward, which is upwelling favorable. The exception is the region north of about 16°–18°S, where the alongshore component of the average wind stress vector is weakly downwelling favorable. Even though winds are on average upwelling favorable over most of the region, previous studies have shown that the frequency and intensity of winds blowing from the south increase substantially during that time because of an increase in the frequency of synoptic weather systems passages (Castro and Miranda 1998).
The presence of cold water near Cape Frio is associated with a strong SST front. During summer, the front is mostly located near Cape Frio and is clearly associated with upwelling (Fig. 3). The outer edge of the front roughly follows the 200-m isobath, suggesting that flow topography interactions may play a role in the establishment and in the dynamics of the front (e.g., Rodrigues and Lorenzzetti 2001). Farther north, between 16° and 18°S, the average SST gradient is intensified near the 200-m isobath, again suggesting strong topographic control. During winter, on the other hand, the front extends southwestward along the shelf break (Fig. 4) and is associated both with upwelling near Cape Frio and with the northward transport of cold waters originated farther south (Piola et al. 2000; see also Fig. 1). The magnitude of the gradients farther north (16°–18°S) is much weaker compared to summer, possibly because upwelling favorable winds at that location are less frequent during winter (Fig. 1).
(top) Average and (bottom) standard deviation of summertime SST gradient magnitude [°C (100 km)−1] based on different SST analyses. The 200-m isobath contour is also shown.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
(top) Average and (bottom) standard deviation of wintertime SST gradient magnitude [°C (100 km)−1] based on different SST analyses. The 200-m isobath contour is also shown.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
There are substantial differences in SST gradients between the different SST analyses, however, for both summer and winter. The average background SST gradient is substantially higher in the MODIS data, especially in the region offshore of the 200-m isobath to the south of Cape Frio. AVHRR-only, AMSR+AVHRR, and OSTIA fields, in contrast, are characterized by low gradients everywhere, except near the coast and on the southeastern corner of the domain during winter, where gradients are intensified. RSS fields are somewhat between these two extremes. The same is true for the SST gradient standard deviation distribution. Values are higher and the field distribution is noisier for MODIS and RSS, and smoother for AVHRR-only, AMSR+AVHRR, and OSTIA. Analyses of individual fields suggest that a substantial fraction of this difference is due to cloud contamination on the MODIS data. Since the cloud mask is imperfect in MODIS, the SST gradient field is sometimes contaminated by erroneous gradients computed between water pixels and cloud pixels not tagged as clouds. Since the gradients in those cases can be quite large, this leads to a relatively large SST gradient background in the averaged field and to large standard deviation associated with time and spatial variability in cloud cover. The use of observations from microwave sensors (which can measure SST even during cloudy periods) and in situ data in some of the other SST analyses, in addition to smoothing (Reynolds and Chelton 2010), substantially decrease this problem. The smoothing leads to an underestimation of the magnitude of the SST gradients, however. During winter, for example, the magnitude of SST gradients near Cape Frio along the 200-m isobath is on average 28% smaller using AVHRR-only data compared to MODIS observations (Fig. 4). For the remainder of this manuscript, analyses focus on the higher-resolution observations from MODIS and on the smoother fields from AVHRR-only and OSTIA.
The coupling between crosswind SST gradients and wind stress curl can be clearly seen in two 29-day average fields during 2007 (Fig. 5). Regions with intensifications in the negative crosswind component of the SST gradient are in general also characterized by strong negative wind stress curl intensifications, both during summer and winter. During September 2007, positive crosswind SST gradients are also visually correlated with positive intensifications in the wind stress curl. It is interesting to note that during March 2007 all SST analyses are characterized by positive crosswind SST gradients near the coast off São Sebastião Island (see Fig. 1 for location), in a region where the wind stress curl is negative. Temperatures in that region increase toward the coast mainly because water offshore is cold because of southwestward advection of water upwelled near Cape Frio (see Fig. 1). Southwestward winds over the relatively warmer water near the coast are not intensified (which would generate positive wind stress curl), however, possibly because of the close proximity to the coast.
Wind stress curl [(N m−2) (104 km)−1] with contours of crosswind SST gradient overlain for two example 29-day averages based on different SST analyses. Negative crosswind SST gradient contours are shown in black, while positive crosswind SST gradient contours are shown in green. Contour interval in all cases is 0.5°C (100 km)−1, with the zero contour not plotted for clarity.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
There are substantial differences in the relation between crosswind SST gradients and wind stress curl between the different SST analyses. Summertime crosswind SST gradients are weaker in OSTIA observations compared to AVHRR-only data, suggesting a higher amount of smoothing in OSTIA fields. During winter, however, crosswind SST gradients from both SST analyses are roughly of the same magnitude. Crosswind SST gradients based on AMSR+AVHRR (not shown) and AVHRR-only data are generally very similar. Even though MODIS crosswind SST gradients are noisy (consistent with Figs. 3 and 4), they are still visually correlated with the wind stress curl distribution near the coast, where SST gradients are strong. Offshore, however, the weaker crosswind SST gradients are not strongly correlated with the wind stress curl distribution.
The intensity and region of maximum correlation between crosswind SST gradients and wind stress curl varies with season and SST analyses. For OSTIA and AVHRR-only data, correlations during summer are high near Cape Frio (Fig. 6). During winter, correlations are still relatively high near Cape Frio, but also over most of the shelf to the west of 44°W. Correlations based on MODIS observations are generally weaker, and the area of relatively high correlations is also smaller compared to OSTIA and AVHRR-only data. Spurious crosswind SST gradients associated with the imperfect cloud mask in MODIS will evidently disrupt the correlation with the wind stress curl field. Despite that, crosswind SST gradients and wind stress curl are approximately linearly related regardless of the SST analyses used (Fig. 7). Negative anomalies in the crosswind SST gradients, as often observed off Cape Frio, are associated with substantial negative wind stress curl anomalies. The slope of the regression between crosswind SST gradients and wind stress curl differs between the SST analyses, though. Slopes based on OSTIA and AVHRR-only data are generally similar, but slopes based on MODIS observations are substantially smaller.
Maps of the correlation between wind stress curl and crosswind SST gradients for (top) summer and (bottom) winter conditions based on different SST analyses. The boxes represent the regions over which the statistics in Figs. 2 and 7 were computed for the respective seasons. Statistics in Fig. 9 were computed over the common area defined by the boxes shown on summertime and wintertime panels.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
Binned scatterplot of (left) wintertime and (right) summertime (top) wind stress curl vs crosswind component of the SST gradient and (bottom) wind stress curl vs SST. Anomaly fields are used in all cases. Anomaly fields are defined to be the deviation of each wintertime or summertime 29-day average from the overall respective seasonal average. The statistics were computed over the area shown on Fig. 6 for each season. The points in each panel are the means within each bin computed from overlapping 29-day averages at 7-day intervals in the three time periods defined by December–March from late 2006 to early 2009 (summer) and by June–September during calendar years 2006–08 (winter), color coded for the different SST analyses. The error bars represent the ±1 standard deviation over all OSTIA individual 29-day averages within each bin. The slope s of the least squares fit line to the binned averages and the 95% confidence interval are labeled in each panel. The fit is restricted to (top) crosswind SST gradients between −1.5°C (100 km)−1 and 0.2°C (100 km)−1 and (bottom) wind stress curl between −2 N m−2 (104 km)−1 and 1 N m−2 (104 km)−1.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
The linear relationship between crosswind MODIS SST gradients and wind stress curl during winter seems to be somewhat weaker in the limit of strong gradients. The wind stress curl anomaly associated with strong positive crosswind SST gradients is smaller than expected based on the regression using MODIS observations, for example. Analysis of individual 29-day average fields show that crosswind SST gradients offshore are often large and noisy (e.g., Fig. 5), so it is possible that the deviation from the linear relationship is associated with problems in the MODIS cloud mask. It is important to point out, also, that few observations are available for large negative or positive gradients, so the average estimations are presumably less robust.
It is interesting to note that the wind stress curl is also highly correlated with SST itself near Cape Frio, São Sebastião Island and Caravelas during summer, and near Cape Frio and São Sebastião Island during winter (Fig. 8). Differences in the correlations between SST and wind stress curl between the different SST analyses are generally small. Slopes of the regression between wind stress curl and SST observations are different than zero (significant at the 95% confidence level), indicating that negative anomalies in the wind stress curl are generally accompanied by cold water anomalies (Fig. 7). The relationship is very close to linear for low values of wind stress curl, but it becomes very noisy in the limit of high wind stress curl. This is particularly true for high positive wind stress curl during summer. For all SST analyses used here, SST anomalies during that time are higher than expected based on the regression analysis. It is not clear if this departure is significant, however, because of the low number of observations with strong positive wind stress curl anomaly.
Maps of the correlation between wind stress curl and SST for (top) summer and (bottom) winter conditions based on different SST analyses. The boxes represent the regions over which the statistics in Figs. 2 and 7 were computed for the respective seasons. Statistics in Fig. 9 were computed over the common area defined by the boxes shown on summertime and wintertime panels.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
It is clear from these analyses that the wind stress curl is correlated both with the crosswind SST gradients and with SST itself. Time series of the spatial cross correlation between these variables reveal that the correlation between crosswind SST gradients and wind stress curl is stronger (Fig. 9). Cross correlations are statistically significant (at the 95% confidence level) most of the time, with coefficients generally larger than 0.5 and often reaching 0.75 during summer and winter. Weak correlations are generally observed during April–June and November. Consistently with Fig. 5, the correlation is generally weaker when crosswind SST gradients are computed using MODIS observations, since noise (presumably due to cloud contamination) is expected to disrupt the correlation. For all SST analyses, the degree of correlation between crosswind SST gradients and wind stress curl seems to be dependent on the wind stress directional steadiness (r = 0.58–0.63), defined to be the magnitude of the vector-average wind stress divided by the scalar average of the winds stress magnitude computed from the daily-averaged QuikSCAT data in each 29-day period. The wind stress directional steadiness generally peaks during summer, and reaches a minimum in April–May (see also Fig. 11, right panel).
Time series of statistics computed from QuikSCAT data and different SST analyses in the region defined by the common area between the boxes shown on summertime and wintertime panels in Fig. 6 from overlapping 29-day averages at 7-day intervals: (top to bottom) the wind stress directional steadiness, defined to be the magnitude of the vector-average wind stress divided by the scalar average of the wind stress magnitude computed from the daily-averaged QuikSCAT data in each 29-day period; the spatial cross correlations between wind stress curl and the crosswind component of the SST gradient; and the spatial cross correlation between wind stress curl and SST, color coded for the different SST analyses. Correlations significant at the 95% confidence level are plotted as solid lines on middle and bottom panels. Also shown on middle and bottom panels are the correlations between the correlation time series and the wind stress directional steadiness, color coded for the different SST analyses.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
Although cross correlations between wind stress curl and SST from the different analyses are comparatively weaker, they often reach 0.5, peaking at about 0.7 on a few occasions. Correlations are statistically significant for most of the summer. The importance of the wind stress directional steadiness on the correlation between wind stress curl and SST is small (r = 0.21–0.38). Once again, correlations based on MODIS observations are sometimes weaker than correlations based on OSTIA or AVHRR-only observations.
An empirical orthogonal function (EOF) decomposition of the wind stress curl field reveals the dominant spatial and temporal mode of variability in the system (Fig. 10). For all EOF analyses, the average field has been removed from the observations. The first wind stress curl EOF, which explains about 26% of the total variance, is characterized by large values near Cape Frio and Cape of São Tomé, and also to the south along the shelf break, from 23.5°S, 43°W to 27°S, 47°W. The amplitude time series shows large negative values during summer, indicating a peak in the negative wind stress curl near Cape Frio during that time (see also Fig. 1), and positive values during fall and winter, when the intensity of the negative wind stress curl near Cape Frio is smaller.
The dominant EOF of wind stress curl and crosswind component of the SST gradient based on different SST analyses from overlapping 29-day averages at 7-day intervals from early 2006 to late 2009. The fraction of the variance explained by the dominant EOF is shown in each panel. (bottom) The associated EOF amplitude time series are shown, color coded for each variable/SST analyses. Also shown in parenthesis are the correlation coefficients between the EOF amplitude time series of the crosswind component of the SST gradient and the wind stress curl, color coded for the different SST analyses. The first number is the correlation computed after the averages were removed, while the second is the correlation computed after the respective seasonal cycles were removed.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
EOF decompositions of the crosswind component of the SST gradient based on the different SST analyses were also performed (Fig. 10). The general patterns of the first EOF are similar among the different SST analyses. A strong intensification near Cape Frio and Cape of São Tomé is observed, which is similar to the pattern observed on the wind stress curl EOF. Intensification in the crosswind SST gradient EOF is also observed roughly along the 200-m isobath to the southwest of Cape Frio, especially for analysis based on OSTIA and AVHRR-only data. The spatial cross correlations between the wind stress curl EOF and the crosswind SST gradient EOF based on OSTIA and AVHRR-only are 0.60 and 0.53, respectively. For MODIS data, the spatial correlation is substantially smaller at 0.29.
The amplitude time series of the first EOF of the crosswind SST gradients based on the different analyses are also very similar to each other, and they are highly correlated with the amplitude time series of the wind stress curl EOF (r = 0.71–0.78). If the seasonal cycle is removed from the wind stress curl and the crosswind SST gradient observations before computing the EOFs, results remain similar, with the correlation between the EOF amplitude time series still high (r = 0.70–0.73).
4. Discussion and conclusions
Cold water is frequently found at the ocean surface near Cape Frio (Fig. 1). This is presumably due to several forcing mechanisms acting together, including coastal upwelling (e.g., Allard 1955; Emílsson 1961; Ikeda et al. 1974), interactions of the flow with topography (e.g., Rodrigues and Lorenzzetti 2001; Palma and Matano 2009), wind stress curl-driven upwelling (e.g., Castelao and Barth 2006), and instabilities of the mean flow (e.g., Campos et al. 1995, 2000). Regardless of the dominant mechanism, a strong sea surface temperature (SST) front is established associated with the upwelling region. The front is stronger during summer (Fig. 3), but its signature is still very robust during winter (Fig. 4), possibly due to the Brazil Current influence offshore.
Since prevailing winds are from the northeast, winds near Cape Frio often blow along the SST front. Analyses of satellite observations of winds from QuikSCAT and high-resolution SST observations from several sources reveal that intensifications in wind stress curl are strongly correlated with intensifications in crosswind SST gradients in the region (Fig. 5). This is consistent with previous studies that have shown that SST modification of atmospheric stability near the sea surface often leads to high wind stress over warm water and low wind stress over cold water, generating curl that varies linearly with the crosswind component of the SST gradient (O’Neill et al. 2003; Chelton et al. 2004; O’Neill et al. 2005; Chelton et al. 2007; among others). Unlike other coastal regions (e.g., California Current system; Chelton et al. 2007), however, the coupling off Cape Frio is not limited to summertime, being statistically significant during winter. As in other regions, the downwind component of SST gradients is highly correlated with the wind stress divergence (not shown). Empirical orthogonal function (EOF) decompositions of wind stress curl and crosswind SST gradients performed independently reveal that the patterns of variability in the fields are very similar (Fig. 10). The spatial cross correlations between the EOFs are about 0.6, while correlation coefficients between the amplitude time series of the wind stress curl and crosswind SST gradient EOFs are larger than 0.7.
The strength of the coupling as identified by the observations is dependent on the SST analyses used, however. The region off Cape Frio is characterized by substantial cloudy coverage, which limits the utility of high-resolution, infrared satellite SST observations. When SST is derived from MODIS, the correlation between crosswind SST gradients and wind stress curl is considerably weaker than when SST is derived from analyses that combine infrared and/or microwave satellite observations with in situ data (OSTIA and AVHRR-only, Figs. 6 and 9). This difference is at least partially due to noise in MODIS SST gradients (e.g., Fig. 5) associated with imperfections in the cloud mask for the region. The coupling coefficient quantified using different products is also dependent on the SST analyses used (Fig. 7).
The spatial cross correlation coefficients between crosswind SST gradients and wind stress curl off Cape Frio are roughly the same size as the coefficients for the California Current system (cf. Fig. 9, middle panel with Chelton et al.’s (2007) Fig. 9, bottom panel). However, it is interesting to note that, regardless of the SST product used, the coupling coefficient between crosswind SST gradients and wind stress curl anomalies off Cape Frio (Fig. 7) is substantially smaller than the coefficient estimated for the California Current system (Chelton et al. 2007). Chelton et al. (2007) computed crosswind SST gradients using AMSR-E observations only, which are of coarser resolution than the SST analyses used here. Therefore, one possible explanation to the observed difference in the coupling coefficients between the two regions is that it is a result of the degree of smoothing in the SST datasets used. Since the wind stress curl field remains the same regardless of the SST product used, weaker SST gradients resulting from smoother SST fields would result in larger coupling coefficients. Indeed, coefficients reported here based on the comparatively smoother OSTIA and AVHRR-only analyses are larger than coefficients computed based on higher-resolution MODIS SST fields (Fig. 7). If an even smoother dataset were used (AMSR-E observations only, for example), the coupling coefficient would presumably be even larger, bringing the results closer to Chelton et al. (2007)’s observations.
However, it is also possible that the difference in the coupling coefficients between the two regions is real, which would indicate that crosswind SST gradient anomalies off Cape Frio lead to smaller modifications in the wind stress curl field. Chelton et al. (2007) note that the coupling in the California Current system only occurs during summer. During that time, they show that the wind stress directional steadiness is very high, often greater than 0.9. Off Cape Frio, on the other hand, the variability in wind stress direction is much larger (Fig. 9). Since the correlation between the crosswind SST gradients and wind stress curl is strongly dependent on the wind stress directional steadiness (Fig. 9), it is possible that winds do not blow steadily along the SST front off Cape Frio long enough for the ocean–atmosphere interaction to fully develop, leading to a comparatively smaller coupling coefficient. To test this hypothesis, the coupling coefficient between OSTIA crosswind SST gradients and wind stress curl was computed as a function of the wind stress directional steadiness (Fig. 11, left panel). This evidently decreases substantially the number of available observations used in the regression analyses since only data collected during specific values of wind stress directional steadiness are used. To increase the number of available observations, and therefore the statistical significance of the fit, observations from different seasons were combined (note that coupling coefficients are very similar between winter and summer; Fig. 7, top panels). Analyses show that there is a strong relationship between wind stress directional steadiness and the coupling coefficient (Fig. 11). For wind stress directional steadiness smaller than 0.2, the coupling coefficient is small and not significantly different than zero (95% confidence level). This suggests that for winds that are highly variable, crosswind SST gradients do not affect the wind stress curl. As the wind steadiness increases, so does the coupling coefficient. For typical values of wind stress directional steadiness between 0.2 and 0.8 (Fig. 9), the coupling coefficient is about 1–1.2 (Fig. 11). This is consistent with the coefficients estimated using the entire dataset (Fig. 7). For high values of wind stress directional steadiness, however, the coupling coefficient increases substantially, approaching 1.7 (Fig. 11). This is closer to the values reported for the California Current system by Chelton et al. (2007). As mentioned before, the wind stress directional steadiness in the California Current system during summer is often larger than 0.9. The wind stress direction is much more variable off Cape Frio, where the wind stress directional steadiness rarely exceeds 0.8 (Fig. 9). This suggests that the wind stress do not blow long enough along SST fronts off Cape Frio for a complete adjustment to the SST influence to develop, except in a few occasions. In most instances, the coupling is not fully developed (Fig. 11), although it is still clearly significant (Figs. 7, 9, and 10).
(left) Binned scatterplot of coupling coefficient (10−2 N m−2 °C−1) between crosswind SST gradients and wind stress curl vs wind stress directional steadiness. The statistics were computed over the region defined by the common area between the boxes shown on summertime and wintertime panels in Fig. 6 from overlapping 29-day averages at 7-day intervals, with observations from different seasons combined. Black circles are the slope s of the least squares fit between crosswind SST gradients and wind stress curl using only observations collected for certain values of wind stress directional steadiness (0–0.2, 0.2–0.4, and so on). The error bars represent the 95% confidence interval. (right) Average wind stress directional steadiness for 1999–2009 from 29-day averages at 7-day intervals. Shaded gray represents the ±1 standard deviation.
Citation: Journal of Physical Oceanography 42, 11; 10.1175/JPO-D-11-0224.1
The wind stress directional steadiness was computed over 29-day periods at 7-day intervals (see methods for details). Because of the nonlinear nature of the computation, it is difficult to translate a wind stress directional steadiness of 0.8 (which appears to be a threshold for strong coupling between crosswind SST gradients and wind stress curl, Fig. 11) to an adjustment time scale. The large steadiness needed, however, suggests that at least a couple of weeks of steady winds along the SST front are needed for the coupling to fully develop. For winds varying on shorter time scales, the coupling coefficient is on average 30%–40% smaller (Fig. 11). Analysis of 10 years of wind stress directional steadiness (1999–2009) reveals that large values of wind steadiness (>0.8) occur relatively often (within one standard deviation from the average) during summer, from January to March (Fig. 11, right panel). During the other months, wind steadiness larger than 0.8 lies outside one standard deviation from the average. This suggests that, although coupling between crosswind SST gradients and wind stress curl occurs year-round off Cape Frio (Fig. 9), it is only fully developed during short periods during summer.
While the results show that much of the wind stress curl anomaly variability seems to be driven by variability in the crosswind SST gradients, analyses also suggest that wind stress curl leads to variability in SST itself, at least near Cape Frio (Fig. 8). The correlation between wind stress curl and SST itself is smaller than the correlation between crosswind SST gradients and wind stress curl (Fig. 9), but they can be quite large at times, often larger than 0.5. The coupling coefficient estimated thorough regression analyses (significant at the 95% confidence level) suggests that wind stress curl anomalies can lead to surface cooling of as much as 1°C (Fig. 7), which is a significant fraction of the temperature drop observed off Cape Frio (Fig. 1). Evidently, the variability around the averaged values used in the regression analyses is large, which increases the uncertainty in the role of wind stress curl anomalies on SST variability.
The statistically significant correlations between crosswind SST gradients and wind stress curl, and between wind stress curl and SST (Fig. 9), suggest the interesting possibility of a positive feedback between these variables. If cold water is found near the coast (initially due to coastal upwelling, flow topography interactions, Ekman pumping associated with a large scale wind pattern, or any other process) establishing a SST front, air–sea interactions will act to modify the wind field, increasing the magnitude of the negative wind stress curl (Fig. 7; assuming winds are from the northeast, as it is often the case). The negative wind stress curl can then drive a substantial vertical transport (Castelao and Barth 2006), which can potentially lead to further decreases in SST (Fig. 7). The influence on the subsurface hydrography and circulation of the region can be considerably larger than at the surface. A decrease in SST would lead to further increases in the negative wind stress curl anomaly, in a positive feedback. Disentangling these mechanisms with satellite observations alone is difficult, however, since they are strongly related. Better understanding these processes is important, though, because similar conditions favoring feedback between wind stress curl and SST variability are observed around several capes in upwelling regions (e.g., Cape Blanco off the U.S. West Coast, Cape Columbine off South Africa). Modeling efforts that properly represent the coupling between SST and wind variability (e.g., Jin et al. 2009; Perlin et al. 2011) could be used to quantify how much of the wind stress curl variability is due to crosswind SST gradients, and how much of the SST variability is due to wind stress curl variations. These models could also be used to better quantify the time scale involved in the response of the wind stress to SST fronts (Fig. 11).
In summary, wind stress curl can be either a cause or a consequence of upwelling near Cape Frio. If wind stress curl drives upwelling, one would expect it to be correlated with SST. If, on the other hand, wind stress curl develops as a consequence of upwelling (i.e., due to ocean–atmosphere interactions as the wind blows along a SST front), one would expect it to be correlated with crosswind SST gradients. Off Cape Frio, the wind stress curl is correlated with both SST and crosswind SST gradients. The correlation with crosswind SST gradients is considerably stronger, however, suggesting that the SST front often observed off Cape Frio modifies the wind stress distribution in the region. Wind stress curl anomalies are strongly correlated with crosswind SST gradients, both during summer and winter. The predominantly northeasterly winds are decreased near the coast over the cold upwelled waters, generating negative anomalies in the wind stress curl distribution. For most of the time, winds appear to be too variable for the ocean–atmosphere interaction to fully develop, leading to coupling coefficients between crosswind SST gradients and wind stress curl smaller than in other coastal regions (i.e., California Current system). During periods of high wind stress directional steadiness off Cape Frio, however, the coupling is substantially intensified, by up to 40%–75%. Since the wind stress curl variability is also correlated with SST itself, negative wind stress curl anomalies are associated with SST decreases, further enhancing upwelling. The strength of the coupling between these processes as identified in the observations is also dependent on which SST analysis is used, because of noise (possibly due to cloud contamination) in MODIS and smoothing in blended products (e.g., OSTIA, AVHRR-only).
Acknowledgments
This research was supported by NASA (Grant NNX10AE92G). The author gratefully acknowledges suggestions and comments from the anonymous reviewers that greatly improved the manuscript.
REFERENCES
Allard, P., 1955: Anomalies dans le temperature de l’eau de la mer observes au Cabo Frio au Bresil. Bull. Inf. Com. Oceanogr. Etude Cotes, 7, 58–63.
Arthur, R. S., 1965: On the calculation of vertical motion in eastern boundary currents from determinations of horizontal motion. J. Geophys. Res., 70, 2799–2803.
Campos, E., J. Gonçalves, and Y. Ikeda, 1995: Water mass characteristics and geostrophic circulation in the South Brazil Bight: Summer of 1991. J. Geophys. Res., 100 (C9), 18 537–18 550.
Campos, E., D. Velhote, and I. da Silveira, 2000: Shelf break upwelling driven by Brazil Current cyclonic meanders. Geophys. Res. Lett., 27, 751–754.
Castelao, R. M., and J. A. Barth, 2006: Upwelling around Cabo Frio, Brazil: The importance of wind stress curl. Geophys. Res. Lett., 33, L03602, doi:10.1029/2005GL025182.
Castro, B. M., and L. B. Miranda, 1998: Physical oceanography of the western Atlantic continental shelf located between 4°N and 34°S. The Sea, A. R. Robinson and K. H. Brink, Eds., Vol. 11, Wiley and Sons, 209–251.
Chelton, D. B., and M. H. Freilich, 2005: Scatterometer-based assessment of 10-m wind analyses from the operational ECMWF and NCEP numerical weather prediction models. Mon. Wea. Rev., 133, 409–429.
Chelton, D. B., and S.-P. Xie, 2010: Coupled ocean-atmosphere interactions at oceanic mesoscales. Oceanography (Washington, D.C.), 23, 52–69.
Chelton, D. B., M. G. Schlax, M. H. Freilich, and R. F. Milliff, 2004: Satellite measurements reveal persistent small-scale features in ocean winds. Science, 303, 978–983.
Chelton, D. B., M. G. Schlax, and R. Samelson, 2007: Summertime coupling between sea surface temperature and wind stress in the California current System. J. Phys. Oceanogr., 37, 495–517.
Donlon, C., and Coauthors, 2007: The global ocean data assimilation experiment high-resolution sea surface temperature pilot. Bull. Amer. Meteor. Soc., 88, 1197–1213.
Emílsson, I., 1961: The shelf and coastal waters off southern Brazil. Bull. Inst. Oceanogr., 11, 101–112.
Franchito, S., V. Rao, J. Stech, and J. Lorenzzetti, 1998: The effect of coastal upwelling on the sea-breeze circulation at Cabo Frio, Brazil: A numerical experiment. Ann. Geophys., 16, 866–881.
Freilich, M. H., and R. S. Dunbar, 1999: The accuracy of the NSCAT 1 vector winds: Comparisons with National Data Buoy Center buoys. J. Geophys. Res., 104 (C5), 11 231–11 246.
Ikeda, Y., L. B. Miranda, and I. C. Miniussi, 1974: Observations on stages of upwelling in the region of Cabo Frio (Brazil) as conducted by continuous surface temperature and salinity measurements. Bull. Inst. Oceanogr., 23, 33–46.
Jin, X., C. Dong, J. Kurian, J. McWilliams, D. Chelton, and Z. Li, 2009: SST–wind interaction in coastal upwelling: Oceanic simulation with empirical coupling. J. Phys. Oceanogr., 39, 2957–2970.
Large, W. G., J. C. McWilliams, and S. C. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys., 32, 363–403.
Matsuura, Y., 1996: A probable cause of recruitment failure of Brazilian Sardine (Sardinella aurita) population during the 1974/75 spawning season. S. Afr. J. Mar. Sci., 17, 29–35.
Matsuura, Y., 1998: Brazilian Sardine (Sardinella brasiliensis) spawning in the southeast bight over the period 1976/93, Rev. Brasil. Oceanography, 46, 33–43.
Naderi, F. M., M. H. Freilich, and D. G. Long, 1991: Spaceborne radar measurement of wind velocity over the ocean: An overview of the NSCAT scatterometer system. Proc. IEEE, 79, 850–866.
O’Neill, L. W., D. B. Chelton, and S. K. Esbensen, 2003: Observations of SST–induced perturbations of the wind stress field over the Southern Ocean on seasonal time scales. J. Climate, 16, 2340–2354.
O’Neill, L. W., D. B. Chelton, S. K. Esbensen, and F. J. Wentz, 2005: High-resolution satellite observations of SST modification of the marine atmospheric boundary layer over the Agulhas Return Current. J. Climate, 18, 2706–2723.
O’Neill, L. W., D. B. Chelton, and S. K. Esbensen, 2010: The effects of SST-induced horizontal surface wind speed and direction gradients on midlatitude vorticity and divergence. J. Climate, 18, 2706–2723.
Palma, E., and R. Matano, 2009: Disentangling the upwelling mechanisms of the South Brazil Bight. Cont. Shelf Res., 29, 1525–1534.
Perlin, N., E. Skyllingstad, and R. Samelson, 2011: Coastal atmospheric circulation around an idealized cape during wind-driven upwelling studied from a coupled ocean–atmosphere model. Mon. Wea. Rev., 139, 809–829.
Pickett, M. H., and J. D. Paduan, 2003: Ekman transport and pumping in the California Current based on the U.S. Navy’s high-resolution atmospheric model (COAMPS). J. Geophys. Res., 108, 3327, doi:10.1029/2003JC001902.
Piola, A., E. Campos, O. Möller Jr., M. Charo, and C. Martinez, 2000: Subtropical shelf front off eastern South America. J. Geophys. Res., 105, 6565–6578.
Reynolds, R., and D. Chelton, 2010: Comparisons of daily sea surface temperature analyses for 2007/08. J. Climate, 23, 3545–3562.
Rodrigues, R. R., and J. A. Lorenzzetti, 2001: A numerical study of the effects of bottom topography and coastline geometry on the Southeast Brazilian coastal upwelling. Cont. Shelf Res., 21, 371–394.
Stech, J., and J. Lorenzzetti, 1992: The response of the South Brazil Bight to the passage of wintertime cold fronts. J. Geophys. Res., 97 (C6), 9507–9520.
Xie, S.-P., 2004: Satellite observations of cool ocean–atmosphere interaction. Bull. Amer. Meteor. Soc., 85, 195–208.