Surface seawater medium-term of physical, chemical and biological samples (0.5 m deep) using a Nansen bottle with reverse thermometer outside and in the bottom (water/sediment interface) by scuba diving using a 2 liters polyethylene bottle (three samples); salinity, DO and nutrients were determined ashore as described in [37] ; to chlorophyll a was applied the method described in [38] ; for temperature was used an inversion thermometer fixed to the outside of a Nansen bottle. The physicochemical parameters are then: Sea Surface Temperature (SST), salinity, DO, nitrogen as ammonium cation (), nitrite (NO₂) and nitrate (NO₃), and ortho-phosphate (PO₄). The biological variables are composed by chlorophyll a (mg/m³) measurements as estimation of microalgal biomass but probably it also contains all free living autotrophic bacteria of water column both influenced qualitative and quantitatively by nutrient entrances that on the other hand, supply it self as feeding material for meroplankton larvae which are expressed in numbers of organisms per cubic meter of water and were collected by means drag plankton net of 100 mesh and fixed in 10% formalin and then counted under microscope. These data were collected with weekly frequency from July 21, 1999 to June 28, 2007 in Anjos Bay.

The five grid points (Figure 1) obtained from reanalysis data are located on the ocean (four points) and coastal (one point) regions, bounded at 21˚S, 25˚S and 39˚W towards the southeastern Brazilian coast (i.e. 6-hourly (UTC) atmospheric pressure (P_A and P_B), zonal (u) and meridional (v) wind components (1, 2 and 3 points) at 10 m above the ground; hourly tide gauge records and forecasting from the Arraial do Cabo station near Anjos Bay, and 6-hourly (UTC) atmospheric pressure from São Pedro d’Aldeia (SPA) meteorological station).

The QuikSCAT vector wind product of Remote Sensing Systems (RSS), available daily on a 0.25˚ grid, was obtained from the National Aeronautics and Space Administration (NASA)-Jet Propulsion Laboratory (JPL) and Physical Oceanography Distributed Active Archive Center (PO.DAAC). It was used Level 3 (L3) 25 km grid products from two passages per day (08 and 20 UTC) near the point of interest. High-resolution QuikSCAT vector wind fields suitable for coastal applications and studying of smaller oceanic processes have been produced by combining scatterometer measurements with a regional mesoscale model [39] or by use of “slices” [40] . The meteo-oceanographic dataset was selected from the same period.

Surface wind stress provides the most important forcing of the ocean circulation, while the fluxes of heat, moisture and momentum across the air-sea boundary are important factors in the formation, movement, and modification of water masses and the intensification of storms near coasts and over the open oceans [41] . Therefore, the zonal (zws) and meridional wind stress (mws) for each grid point were calculated using the follows equations:

where:

(air density);

W = intensity of the wind (m∙s⁻¹) calculated from zonal (u) and meridional (v) wind components; C_d = 1.1 + 0.053 (drag coefficient for the southeast Brazil coast, [42] . The unit used for wind stress is N∙m⁻², where 1 hPa is equal to 10² N∙m⁻².

Direction and wind speed as well as the meteorological residual were calculated from the wind dataset and the tide gauge records, respectively.

The basic statistical analysis was applied in the water samples and the meteo-oceanographic time series for the period from July 1999 to June 2007. Some outliers were identified and substituted by the average values between the previous and the following weekly data (Table 2).

The dynamic of aquatic systems depends of the environmental changes that affect species in many different ways, altering their productivity and interactions with other species [43] . Therefore, the seasonal patterns of extreme values of meteorological forcing and water samples were investigated to verify the relationship between nutrient concentrations, larvae and chlorophyll a.

Seasonality of extreme values and maximum occurrence of northeasterly wind distributions were carried out to verify the wind stress patterns in the periods of upwelling as well as the association between variables and minimum of sea surface temperature. Therefore, a correlation matrix was then applied to the extreme dataset and then extracted the relationships between these variables for the marine environment related to the studied region.

The multivariate method was applied because it treats several dimensions simultaneously and can take into account the variability between the different locations (spatial scale) and the variability between successive samples (temporal scale). Cluster analysis, for example, is a multivariate technique used to group objects into classes (clusters) on the basis of similarities within a class and dissimilarities between different classes. In hierarchical cluster analysis (HCA), a dendogram is drawn with samples plotted in clusters on the y axis and linkage distances plotted on the x axis. The linkage distances between the clusters illustrate relative similarities in the characteristics of the samples. In this work, the Ward’s method was used to form the clusters [44] as well as Pearson-r distances as the similarity measure. HCA is also used to discover groups of similar patterns when extreme values occur. Numerical analysis procedures, such as PCA, have been developed to interpret large space/ time dataset, which can decompose total variance into spatial and temporal variances. The principal axis method was used to extract the components, and this was followed by an orthogonal rotation [45] . PCA is a variable reduction procedure [46] and establishes a set of orthogonal factors based on a correlation matrix, providing information about similarities and redundancies of the samples, using Varimax normalized rotation. This method was then applied to describe more clearly the structure and composition of the study area ecosystem based on the relationship between physical, biological and chemical variables.

Large scale meteo-oceanography patterns are in accordance with the OA interactions through the SASA, a predominant air mass above the central region of the South Atlantic Ocean basin, centered near 30 degrees latitude that induces the currents of upper ocean due to the wind-driven forces [47] . Wind events induce different disturbances of the water mass structure depending on the season.

The seasonality of NE extreme wind stress in all points follows the seasonal pattern expected for the study area with peaks during spring (Figure 2). The average for spring season presents a value of 0.11 Nm⁻² on the oceanic region (ws_2, ws_3 and Q) and 0.07 Nm⁻² in coastal area (SPA and ws_1) with 74% in northeasterly. The ws_1 grid point, over the continent, has the lowest peaks in relation to other reanalysis points but the max-

imum values are observed in spring, with the highest value in November (0.19 Nm⁻²). The ws_2 point presents high peaks in spring and summer; August (0.33 Nm⁻²), September (0.29 Nm⁻²) and January (0.25 Nm⁻²). The ws_3 point presents the highest value in June (0.35 Nm⁻²), corresponding to the southwesterly, the other peaks are verified in April (0.28 Nm⁻²), August (0.32 Nm⁻²), spring and summer (Figure 2(a)). The ws_SPA point presents a peak in June (0.26 Nm⁻²), also corresponding to the southwesterly wind and in spring. The QuickSCAT (Q) point presents maximum values in March (0.42 Nm⁻²), June (0.37 Nm⁻²), August (0.41 Nm⁻²) and

September (0.44 Nm⁻²), all corresponding to the northeasterly winds (Figure 2(b)). The results obtained in the coastal area suggest the influence of the sea-breeze during the spring-summer as [34] . Sometimes when strong NE winds persist for several days, strong upwelling can develop with surface temperatures dropping near the coast, close to Cabo Frio [33] . The transport of BC follows the curve of annual variation of wind shear over the subtropical basin with a maximum during the summer and minimum during the winter [48] .

The curves for the pressure grid points show peaks between July and September (Figure 2(c)). It can be observed the relationship between the pressure and wind stress for the same months, referring to the pattern of the SASA that is a semi-stationary meteorological system, presenting a well emphasized seasonal movement. During the summer this system, over the continent, becomes weaker than winter, moving southerly and on its western side the winds blow northeasterly towards the southeastern coast of South America and they are more intense in the southeastern coast of Brazil [49] . This seasonal variability is one of the most important factors related to the occurrence of upwelling in the region of Arraial do Cabo, where the winds blow along the coastline from north to south push the surface waters offshore on spring-summer periods [11] .

The small number of occurrence of extreme southwesterly winds (Figure 3) verified in all points is associated with the number of frontal systems that reach the region, unlike of extreme northeasterly winds that presents higher percentage around 53% and 71% for ws_3 and SPA, respectively (Figure 3(a)). The predominance of NE winds indicates the seasonal variability of the SASA that leads upwelling in this region.

The seasonality of the extreme SST presents a well-defined pattern with maximum peaks from February to July, reaching the maxima value (24.5˚C) in April. The minimum peaks occur from September to January, reaching the minima value (15.9˚C) in November. It can be related to the presence of the phenomenon known as thermal inversion seasonal, which is typical on the region, when the TW and SACWS alternate, producing layers below the thermocline with higher temperatures in winter and lower ones in summer [50] . Another important aspect in seasonality of SST is the interaction of currents, such as wind-induced flows, near the continental shelf [51] related to the prevailing of the SASA with E-N winds blowing offshore the coastline and consequently leading to

occurrence of coastal upwelling. According to [52] the SST values are quite low from September to February and they start to increase gradually, reaching the maximum values in April.

Extreme seasonal variations were compared between SST, Chlorophyll a, Total Nitrogen (NO₂, NO₃, and NH₄), DO, PO₄, and peaks of larvae. DO presents maximum peak in February and June and two minimum peaks in April and July. Comparing the curves of DO with SST is verified values almost constant during spring and summer (Figure 4(a)). Total Nitrogen (Figure 4(b)) presents a small peak in February and another large one in September while phosphate presents a very sharp peak in September (Figure 4(c)). Chlorophyll a (Figure 4(d)) presents higher seasonal peak in February, being in accordance with DO. In this month occurs an increase of incidence of solar radiation with greater amount of nutrients and consequently increase of phytoplankton [53] . Smaller peaks are also observed in July, September and November. Studies commented by [54] affirm the existence of a negative correlation between Chlorophyll a and wind, in other words, when the wind is strong, primary production decreases. It can be explained due to the whitecapping breaking in the shallow water bodies lead to the production of short period waves under strong wind conditions that can prevent the transmission of light for phytoplankton [55] .

The maximum peaks of larvae are verified in November (Bryozoa, Isognomon, Mytilidae and Ostreidae), in December (Cypris) and in March (Bivalvia, Ascidiacea and Decapoda). These larvae as the other variables show a relationship with SST in the period of occurrence of upwelling with predominance of SACW in the region. However, Cirripedia and Polychaeta present peaks in July, they are found in warm and shallow waters (Figure 4(e)) In this month verifies the presence of the TW (warm) which at the western boundary is transported southward by the BC [56] . Cirripedia are sessile invertebrate organisms (barnacles) that cover substantial area of the substratum on intertidal and subtidal zones. They are important in determining and monitoring environmental impacts in coastal areas [57] . Although there are no further reports of economic damage, it is known that the hulls of ships, oil platforms, pipelines and other plant artificial substrates available in the marine environment may be completely covered by the barnacles causing corrosion of metals and an increase in maintenance costs [58] . The Polychaeta is a bristle-worm with species that live in the coldest ocean temperatures and other which tolerate the extreme high temperatures [59] .

Measurement of joint range of variables is performed numerically by means of correlation coefficients that represent the degree of association between two continuous variables. The correlation is simply the tendency of the variables present their joint variation. Thus, the measure of correlation does not necessarily indicate that there is evidence of causal relationships between two variables. Evidence of causal relations must be obtained from the knowledge of the involved processes. The correlation coefficients calculated between extreme values of the physical and biological variables are presented in Table 3.

The results show small variability in spatial distributions of presented correlations. Correlations of the physical variables with nutrients and larvae present a relatively similar pattern for all points. These results suggest a

relationship with the extremes that occur in the upwelling process due to the positioning of SASA, sea level lowering and drop in SST. This leads to a linear relationship between them, showing that these variables have an interaction with the variations of sea level and temperature as well as the pressure and wind stress related to the meteorological forcing. The nutrients (PO₄, NO₂, NO₃ and NH₄), Chlorophyl a, DO and some larvae present positive (nutrients) and negative correlations (Table 3) with the wind stress points around 0.50. The maximumcorrelation is verified between NH₄ and ws_SPA (0.71).

The sea level lowering is in accordance with the increasing of nutrients and meroplankton larvae in the study area. The minimal meteorological residual presents a high correlation with NH₄ (−0.71), Bryosoa (−0.57) and NO₃ (−0.52), showing the influence by the SASA over the southeast coast of Brazil near Arraial do Cabo. These nutrients stimulates phytoplankton production, mainly larger species such as mesozooplankton and fisheries, being verified in several upwelling zones around the world such as in the SE Pacific Ocean, NE Pacific Ocean, NE Atlantic Ocean and Indian Ocean [10] .

These results are important on recovery as well as in monitoring environmental impacts in coastal regions and these larvae can be used as indicator organisms in monitoring programs due to their local distribution, lifestyle and tolerance to a large range of environmental conditions.

In PCA and factors analysis, the most commonly criteria used for solving the number of components is the eigenvalue criterion, for which only the components with eigenvalues greater than 1 are retained [60] .

In the present study, twenty-five variables were analyzed and six components were extracted with this criterion. Then the components that contain a greater variance than the original standardized variables are kept. The first component can be expected to have large amount of the total variance. Each succeeding component will account for progressively smaller amounts of variance. Therefore, only the first six components extracted have eigenvalues greater than 1 and account for 84.5% of the total variance in the dataset (Table 4). To maximize the variance of the first six principal axes, the Varimax normalized rotation was applied. The first three components are the most important and explain 25.1%, 18.8% and 12.9% of the variance, respectively (Table 4). Components 4 - 6 are not as important, and each of these three components explains a small amount of variance, between 10.2% and 8.1%. The first component explains the greatest amount of the variance, and is characterized by highly negative loadings in PA, PB and PSPA (Table 5). This statistical strategy gives greater importance to the pressure in the ordering and feature extraction of the spatial distribution. It suggests that the SASA plays a vital role in the Atlantic’s mean climatology. Variations in the position of the SASA are linked to seasonal variability of wind stress in the southeast coast of Brazil, leading to occurrence of upwelling. Because of this association, component 1 can be defined as the “pressure component” in reference to the main feature of the study area, associated with variations in the SASA pattern.

The second component is characterized by highly positively loadings in Total Nitrogen, Bryosoa and ws_1 (Table 5), and by highly negatively loadings in SST, being then mainly attributed to temperature, wind and nutrients related to local variables. These abiotic variables characterize a local marine ecosystem and are related to the presence of different water masses, being this component denominated “water masses component”, explaining the presence of SACW that is very important in upwelling periods.

The third component is characterized by positively loadings in Decapoda and Ascidiacea and by negatively loading in Cirripedia (Table 5). This component gives greater importance to the larvae and can be denominated “biotic component”, corresponding to the larvae economically important to the region as crustaceans. Cirripedia and Ascidiacea are a classes of zoobenthos and are found in intertidal zones. Ascidiaceas, for example, are larvae bioindicators that react to environmental changes with high filtration, playing a significant role in water purification. They influence the amount of nutrients and pollutants in suspension [61] [62] . But they grow rapidly and have a long reproduction period, becoming invasive potential, contributing to incrustation in the port regions [63] . Each of the last three components explains about 10.2%, 9.4% and 8.1% of variance, respectively and are clearly characterized by loadings in Salinity and DO (component 4), Bivalvia and Cypris (component 5) and Polychaeta (component 6) related to the marine conditions, indicating a physical and biological behavior for both components (Table 4).

The main result of the HCA performed on the 25 variables is the dendrogram (Figure 5). For this study, the Pearson-r distance was chosen as similarity measurement, between sampling sites. The sampling sites with the larger similarity are first grouped and next are joined with a linkage rule. The steps are repeated until all observations have been classified. Ward’s method was more successful to form clusters with the samples and are more or less homogenous, compared to other methods. Ward’s method is distinct from other linkage rules because it uses an analysis of variance approach to evaluate the distances between clusters [60] .

The classification of the samples into clusters is based on a visual observation of the dendrogram. The dendrogram shows two big clusters, one of most physical variables in one side and biological ones at the other, corresponding to the macrostructure of ecosystem (Figure 5). The degree of refinement of similarities can be expressed through a line that is drawn across the dendrogram at a linkage distance [64] . In this study, it was used a linkage distance of about 1.5 (Figure 5). Thus, samples with a linkage distance lower than 1.5 are grouped into several clusters. This line allows a division of the dendrogram into six clusters. Fewer or greater number of clusters could be defined by moving the position of the line up or down on the dendrogram. This subjective evaluation made HCA a semi-objective method.

Observation of the dendrogram reveals some indications of the level of similarity between the six clusters (Figure 5). Samples from cluster 1 (SST, residual, Cirripedia and Polychaeta), 2 (wind and nutrients) and 3

(pressure) are linked to the other clusters at an elevated distance, indicating that these samples are distinct from the ones of the other three clusters, i.e. cluster 4 (Salinity and DO), cluster 5 (Chlorophyll a, Decapoda, Ascidiacea, Bivalvia and Cypris) and cluster 6 (Mytilidae, Ostreidae, Bryozoa and Isognomon). It can thus be expected that the biological samples of cluster 5 would have similarities with the ones of cluster 6 because they are linked at a low distance.

The dendrogram is close to the PCA components, presenting clusters and loadings corresponding to six groups and factors, respectively, showing physical variables in one side and biological ones at the other.