The hydrographic basin of the Capibaribe River covers 7557.42 km² and is located in the coastal region of the State of Pernambuco, Brazil (Figure 1). Lithologically, the Capibaribe River basin is composed of 90% crystalline basement and 10% sedimentary basin. In the lower Capibaribe River, near the MRR, the Tertiary-aged Barreiras formation is represented by disperse sedimentary deposits that indistinctly cover the crystalline basement and the sedimentary basin. Natural vegetation covers 56.3% of the total drainage area of the Capibaribe River, with 38.8% represented by agricultural areas, 3.14% attributed to urban areas and 0.45% occupied by the continental water body [17]. Given its wide geographical coverage, the Capibaribe basin exhibits a complex environment that is characterized by climatic contrasts in the terrain, soils and vegetation cover and marked socioeconomic gradients. From its headwaters, located 240 km from the coastline, to its mouth in the MRR, the river crosses 42 municipalities, of which 15 are completely comprised within the basin, and 26 have their administrative center in this basin [17]. The total population of the basin is 1,450,000 inhabitants [18,19], for a mean population density in the basin of 190 inhabitants km⁻², with the highest concentration in the MRR (1000 inhabitants km⁻²).

Various activities occur in the basin, and the main activities include the production of food, non-metallic minerals, textiles, metallurgy, chemicals, pharmaceutical products, veterinary products, sugars/ethanol, leathers, plastic materials, drinks, transport materials and wood. The residual organic load that effectively reaches the body of water is estimated at 32.4 t biochemical oxygen demand (BOD) day⁻¹, of which 95.7% is of domestic origin and 4.3% is of industrial origin [19]. Furthermore, according to [17], the residual organic load can reach values of up to 28 t BOD day⁻¹ at collection station CB-72/CPRH (Figure 1).

The Capibaribe River basin exhibits a high spatial variability in precipitation, with mean annual values between 600 mm and 2400 mm (mean = 1133 mm from 1990 to 2010), with an increase in precipitation closer to the coast. This spatial variability reflects the influence of the two main atmospheric systems that affect the eastern edge of the northeast region of Brazil, the Intertropical Convergence Zone and the Atmospheric Easterly Waves [17,20]. Potential evapotranspiration is approximately 1700 mm, with a decrease near the coast, where the value falls to 1500 mm.

The mean annual temperature in the Capibaribe River basin varies between 20.4˚C and 26.1˚C, while the maximum temperature oscillates between 25.5˚C and 29.9˚C. Throughout the year, the temperatures in the region display a variability that can be represented by two seasons: a period with lower monthly means between April and September and the period from October to March, during which the mean temperature values increase. The winds vary between 2 m·s⁻¹ and 5 m·s⁻¹, with the months of August, September and December displaying the highest monthly wind speeds [21]. The mean fluvial discharge in the São Lourenço da Mata station (Figure 1) is 13.74 m³·s⁻¹, resulting in a specific flow

rate of 9.79 L·s⁻¹·km⁻² [17].

In the present study, the climatological and fluvial discharge variables were analyzed using a historical time series of 20 years (1990-2010), with a special focus on the period that included biogeochemical and river water quality data (2002-2010). These last data were sampled and analyzed bimonthly from 2002-2010 at Station CB-72 (Figure 1), which is under the responsibility of the State Agency of the Environment and Water Resources (Agência Estadual de Meio Ambiente e Recursos Hídricos—CPRH) [19]. As will be discussed later, the statistical significance between these two series of physical data (climatology and river flow) was assessed to detect trends and verify the climatological representativeness of the performed analyses for 2002 to 2010, for which river water quality data were available, making the simultaneous use of physical and biogeochemical data possible.

The fluvial discharge values were obtained from the database of the National Water Agency of Brazil (Agência Nacional de Águas—ANA) [22], while the data on air temperature and pluviometric precipitation were obtained from the Brazilian National Institute of Meteorology (Instituto Nacional de Meteorologia—INMET) [21].

The historical series of evaporation data was obtained from the Department of Water Resources (Secretaria de Recursos Hídricos—SRH) [17], and the evapotranspiration rate was determined as follows using the Hargreaves method [23] and data on the latitude, air temperature and relative humidity:

where EVTP (i) = Hargreaves’ potential evapotranspiration for month “i” (mm); FL = factor latitude (calculated as in [23]); TEMP = monthly mean temperature (˚C); and HU = mean monthly relative humidity (%).

The air-water CO₂ flux was determined based on bimonthly measurements (2002 to 2010) of pH, temperature, salinity and alkalinity that were obtained at station CB-72 of the CPRH in combination with climatological data (precipitation, wind intensity and atmospheric pressure) from the Meteorological Station of INMET (Figure 1).

The dissolution constants for carbonic acid that were used in the calculations were obtained from [24] for salinities of 0.1 - 50 and temperatures of 1˚C - 50˚C. The CO₂ solubility coefficient was calculated using the equations in [25], and the dissociation constants of sulfate [26] and borate [27] were calculated as follows:

where FCO₂ = CO₂ flux at the air-water interface (mmol·m⁻²·d⁻¹); k (CO₂) = gas-transfer rate (m·s⁻¹); KH = CO₂ solubility, calculated according to [25] (m·L⁻¹ atm⁻¹); and = difference between the partial pressure of CO₂ on the water surface (pCO_2(aq)) and in the atmosphere (pCO_2(air)) (μatm).

A positive value of indicates a liquid flux of water to the atmosphere. The partial pressure of atmospheric CO₂ was obtained through the following equation:

where P_atm = barometric pressure, obtained from the local meteorological data (atm); xCO₂ = mole fraction of atmospheric CO₂, obtained from NOAA (http://esrl.noaa. gov) (ppm); and pH₂O = water vapor pressure (μatm), obtained from [28].

The water vapor pressure was determined using the following equation:

where sst = the sea surface temperature of the body of water (˚K) and S = the salinity of the body of water.

The gas-transfer rate was calculated using the equation from [29] as follows:

where u₁₀ is the wind speed at 10 m from the surface of the body of water (m·s⁻¹) and Sc is the Schmidt number.

The results obtained from Equations (2)-(5), combined with the results from the analyses of the mineral nutrients [19], were used as input data for the CO₂ calc^Ò software [30]. This software uses the equations of the carbonate system to obtain pCO₂ in the water, , CO₂ aqueous, , dissolved inorganic carbon (DIC) and other parameters derived from the carbonate system.

The excess of CO₂ (CO_2(aq)) (μmol·kg⁻¹) is defined as the amount of DIC that is transferred as CO₂ to the atmosphere after attaining air-water equilibrium. The excess of CO₂ was calculated according to the equation from [31], as follows:

where is the concentration of total free CO₂ in

μmol·kg⁻¹ (i.e.,

in water)

and is the solubility coefficient of CO₂.

The apparent oxygen utilization (AOU, μmol·kg⁻¹) was determined based on the equation in [32]:

with the dissolved oxygen saturation (%O₂) calculated according to the following equation:

where [O₂]_eq = the dissolved oxygen (DO) concentration at equilibrium with the atmosphere, calculated based   on [33] and the local air pressure (μmol·kg⁻¹); [%O₂] = DO saturation (%); O₂ = in situ DO concentration (μmol·kg⁻¹); and = solubility of oxygen as a function of temperature and salinity (μmol·kg⁻¹).

The statistical analyses (t-test, trend, PCA and descriptive statistics) were performed using XLSTAT^® 2010 software. The non-parametric Mann-Kendall test was selected, which is widely utilized to detect monotonic trends in data series, without the need to specify whether these trends are linear.

The air temperature, pluviometric precipitation and evaporation did not show significant differences between the historical period (1990-2010) and the study period (2002-2010) (t-test; α = 0.05) (Figures 2(a)-(c)). The annual water balance was positive, with a seasonal variability characterized by 6 months of positive water balance (March to August) and 6 months of negative water balance (September to February). According to the Mann-Kendall trend test (α = 0.05), the observed data on precipitation, temperature and humidity did not show significant increasing or decreasing trends for these variables.

Fluvial discharges did not show significant differences between the historical period and the study period (t-test; α = 0.05) (Figure 2(e)). The mean flow rate for the study

period was 11.8 m³·s⁻¹, which was very similar to the value of 13.7 m³·s⁻¹ obtained for the historical period (1990-2010) [17]. Based on the Mann-Kendall test (α = 0.05), there was no significant trend (increasing or decreasing) in the flow rate data from 1990-2010.

The wind intensities show significant differences between the historical period and the study period (t-test; α = 0.05). The mean annual value recorded in the period from 2002-2010 of 2.2 m·s⁻¹ was lower than the historical mean value of 2.6 m·s⁻¹. All of the monthly wind intensities were lower than those recorded historically in the region (Figure 2(d)). Based on the Mann-Kendall test (α = 0.05; p < 0.0001), there was a positive (growing) trend among the historical wind data (1990-2010).

Water temperature and salinity did not show significant trends according to the Mann-Kendall test (α = 0.05), indicating the repetitive cyclical nature of these state variables in the study period. The coefficient of variation for both state variables was minimal (0.1 and 0.4, respecttively). According to the t-test (α = 0.05; p < 0.002), the mean seasonal water temperatures (March-August x September-February) showed significant differences between the wet and dry periods of the year. The amplitude of the means was 1.5˚C (wet = 27.1˚C and dry = 28.6˚C).

The mean salinity value was 0.3 ± 0.01 units, with maximum values in February (dry period) and minimum values in June (wet period), indicating that salinity follows the climatic seasonal cycle. The salinity in this section of the river was classified as freshwater (94% of the values were <0.5).

The mean BOD concentration was 4.4 ± 3 mg·l⁻¹ from 2002-2010. When this concentration was associated with fluvial discharge, a mean input of 53.1 t BOD·d⁻¹ in the Capibaribe River was obtained, which represents 63.8% more than the residual pollution load that was reported by the CPRH for 2010 (32.4 t BOD·d⁻¹). The data observed do not show significant differences (α = 0.05; p < 0.9) in the mean concentrations of organic load between the dry and wet periods. The data series for the BOD concentration of the period 2002-2010 showed a maximum value of 16.7 mg·l⁻¹ and a minimum of 0.5 mg·l⁻¹ (Figure 3(b)). Values above 5 mg·l⁻¹ are above the legal limit recommended by the Brazilian legislation for this type of water source [34]. According to the Mann-Kendall test, there is a positive trend in the BOD data (α = 0.05; p < 0.025), indicating a gradual increase in the levels of biodegradable organic material during the study period.

The dissolved oxygen saturation (DO%) showed a mean value of 44% ± 20%, with a mean DO concentration of (95 ± 70 μmol·kg⁻¹) (3.04 ± 2 mg·l⁻¹) during 2002- 2010. The highest monthly mean was recorded in October (during the dry period), while the lowest value was observed in April (during the wet period) (Figure 3(b)).

Values less than 5 mg·l⁻¹ are below the limit recommended by the Brazilian legislation [34]. During the study period, 78% of the samples exhibited concentrations below this limit (5 mg·l⁻¹).

According to the t-test (α = 0.05; p < 0.25), the data did not show significant differences between the dry and rainy periods. Based on the Mann-Kendall test, the DO% did not show significant alteration trends (α = 0.05) from 2002-1010.

The DIC concentration during 2002-2010 showed a smaller variation than the other parameters associated with the carbonate system (596 to 2,235 μmol·kg⁻¹; coefficient of variation = 0.2) (Figure 4(a)). The data did not show significant differences (α = 0.05; p < 0.2) between the dry and rainy periods. Based on the Mann-Kendall test (α = 0.05), the DIC did not show a positive trend during 2002-2010, suggesting a continuous input of carbon inorganic species throughout the years. The mean distribution of the compounds that comprise DIC was as follows: = 80.8%; = 0.2% and CO_2(aq) = 19% (Table 1). Of these compounds, and CO_2(aq) displayed a higher coefficient of variation and thus influence the variation of the DIC concentration in the Capibaribe River.

The TA showed a mean value of 851 ± 160 μmol·kg⁻¹ during 2002-2010 (Figure 4(a)). The estimated values showed significant differences between the dry and wet periods (α = 0.05; p < 0.004), with a higher mean during the dry period (920 μmol·kg⁻¹). The variance was 377 - 1350 μmol·kg⁻¹ (coefficient of variation = 0.2). Based on the Mann-Kendall test (α = 0.05), the TA did not show a positive trend during 2002-2010, indicating that the concentrations recorded in this period were constant. The highest monthly mean was observed in February (942 μmol·kg⁻¹), while the lowest monthly mean TA occurred in June (674 μmol·kg⁻¹).

The waters of the Capibaribe River show high concentrations of mineral nutrients that serve as important fertilizers in the adjacent estuarine and coastal waters [16]. The and concentrations measured during

Table 1. Variables associated with the carbonate system in the Capibaribe River, Brazil, during 2002-2010. Mean values in parentheses represent the proportion (%) of the total DIC.

2002-2010 were above the corresponding limits recommended by the legislation [34]. The period from October to February (the dry period) systematically showed the highest concentrations of dissolved nutrients, reaching 38 μmol·kg⁻¹ for and 580 μmol·kg⁻¹ for (Figure 4(a)).

The pH values measured in the Capibaribe River did not show significant differences (α = 0.05; p < 0.9) between the wet period and the dry period. The mean of the series was 7.0 ± 0.4; however, low values were observed in both climatic periods (Figure 4(b) and Table 1).

The concentrations showed significant differences (α = 0.05; p < 0.04) between the wet and the dry period (Figure 4(b) and Table 1).

The partial pressure of carbon dioxide in the Capibaribe River was highly variable throughout the study period (coefficient of variation = 0.8), ranging between 311 and 31,400 μatm, with a mean value of 8340 μatm (Figure 5(a) and Table 1). The pCO_2(aq) values did not show significant differences between the dry and rainy periods (α

= 0.05; p < 0.6). The highest monthly mean was recorded in April (17,440 μatm) (Table 1). Based on the MannKendall test (α = 0.05), the partial pressure of CO₂ did not show a significant positive or negative trend during 2001-2010, indicating that the concentrations were consistent throughout the analyzed temporal series.

The progression of the CO₂ flux (FCO₂) in the Capibaribe River was always positive during the study period, with a mean value of +225 mmol·m⁻²·d⁻¹, indicating a continuous release of CO₂ into the atmosphere (Figure 5(b)). The highest value calculated in the period was +995 mmol·m⁻²·d⁻¹ (April 2010), and the lowest flux was +4.9 mmol·m⁻²·d⁻¹ (August 2005), thus supporting the high coefficient of variation (Figure 5(b) and Table 1). Similarly to the water partial pressure, the FCO₂ values did not show significant differences between the dry and rainy periods (α = 0.05; p < 0.5). The highest monthly mean was also recorded in April (+426 mmol·m⁻²·d⁻¹), while the lowest monthly mean value was observed in August (+119 mmol·m⁻²·d⁻¹). Based on the Mann-Kendall test (α = 0.05), the CO₂ fluxes did not show trends (positive or negative) during 2001-2010, indicating an almost continuous release to the atmosphere over the period analyzed.