Introduction

Rivers and estuaries play an important role in the transport and transformation of carbon from the continent to the adjacent coastal zone and typically act as sources of CO2 to the atmosphere1,2,3,4. Estuaries are often strongly influenced by human activities. Attention has been focused on estuaries as potentially important sites of organic matter cycling in the coastal zones in recent decades. Estuaries are regarded as sources of CO2 to the atmosphere because of their elevated loads of detrital organic matter that drives intense respiration5.

The exchange of CO2 between inner estuaries and the atmosphere has been recently evaluated using data from 106 systems6. The estimated global emissions of 0.26 Pg C year−1 from these systems is lower than previous estimates, which have been in the range of 0.27 to 0.60 Pg C year−11,2,7,8,9. However, all recent studies have reported higher values than those first reported (0.1 Pg C year−1)10.

Very little is known about the seasonal and interannual variability of the CO2 flux at the air–water interface in rivers and estuaries of Brazil and particularly in the northeastern region of the country. Recently, two studies in tropical estuaries in northeastern Brazil reported values of 30 and 35.6 mmol m−2 d−111,12.

The climatic diversity found in the basins of the Atlantic coast of Brazil is a function of their position in relation to atmospheric circulation systems such as the South Atlantic tropical cyclone and the polar anticyclone. The basins of the Atlantic coast present a range of climates from the warm tropical climate of the south to humid and semi-humid climates as well as semi-arid regions in the north13.

To quantify CO2 emissions from the Brazilian estuaries, water samples were collected monthly from 6 estuaries (12 months - 72 observations). Additionally, we estimated the amount of carbon emissions from rivers and estuaries using measurements of salinity, temperature, alkalinity and pH at low tide from 22 systems (78 observations) along the northern and northeastern Brazilian coast during the wet and dry periods (Fig. 1).

Figure 1
figure 1

The Köppen climate classification29 of the Brazilian estuaries considered in this study: Am (tropical wet climate), Af (tropical rainforest climate), Aw (tropical wet and dry or savanna climate) and BSh (semi-arid climate).

Map generated with Surfer 9.0® software.

Full size image

The objectives of this study are to characterize and quantify the variability of CO2 emissions and the variables associated with the carbonate system in the tropical systems located in northern and northeastern Brazil. The characterization of these emissions is associated with population density, climate type and estuarine gradient. Finally, we assess the contribution of excess CO2 input by rivers to the total CO2 emission to the atmosphere from the estuaries through correlation with others species of the carbonate system, as well as oxygen and nutrients.

Results

Climatology and fluvial discharge

The mean annual water discharge from the Brazilian estuaries studied here varies from 1.0 to 5,946 m3 s−1, contributing with 8,740 m3 s−1 (276 km3 year−1) to the Atlantic Ocean (Table 1). The observed pluviometric precipitation and evaporation were not significantly different from their historical averages (1982–2012) (t-test; α = 0.05). The relationship between river discharge and pCO2 was explored for each system (18 systems; Table 2). The river discharge was positively correlated with pCO2 values in 44.4% of the systems. However, the systems that showed a positive correlation in Table 2 showed no association with population density or a specific climate type.

Table 1 Drainage area, population density (inhabitants/km2) and mean values of river discharge (annual average), salinity, temperature, total alkalinity, pCO2 and CO2 fluxes in northern and northeastern estuaries in Brazil
Full size table
Table 2 Mean values of river discharge (Q), sample number (n), salinity, equations of fit of the pCO2 as a function of Q, r2 (p > F) and pCO2 values. The best goodness of fit is shown in bold (r2)
Full size table

Distribution of physical and chemical parameters

The salinity varied widely along the coast (0.2–46 units), indicating a strong interaction with precipitation and evaporation as well as soil characteristics of the drainage basins (Table 1). Low salinities were found in the upper estuaries and in Guajará Bay (in the northern region). High salinities (38.8–46 units) were observed in the latitudinal band between −4.8° and −5.1° (the Apodi, Conchas, Cavalos and Assu Estuaries). The water balance was negative in these estuaries because evaporation was 8.5 fold higher than precipitation. The Spearman correlation coefficient values indicated that salinity was positively correlated (α = 0.05) with temperature, TA, DO% and pH and negatively correlated with pCO2 and CO2 fluxes (FCO2).

Water temperatures along the coast ranged from 24.6 to 29.5°C. The estuaries of the northern region (Guajará Bay, Marapanim and São Marcos Bay) had temperatures that were 3.6°C higher (averages: 28.9–25.3°C) than the estuaries located at the other extreme of the latitudinal band (Cururupe, Acuípe and Caravelas).

The pH values in the monitored estuaries varied widely (from 6.60 to 8.20) (Table S1; Supplementary Material). Low values (<7.0) were observed in the northern (Guajará Bay) and southern (the Real, Cururupe and Acuípe Estuaries) regions, whereas high values (>7.8) were observed in the northeastern estuaries (Catu, Jaguaribe, Apodi, Conchas and Cavalos) located between latitude −3.9° and −5.1°. The correlation between pH values and DO% values was low (r2 = 0.49; p < 0.0001) (Fig. 2a).

Figure 2
figure 2

The relationship between DO (%) and pH (a) and pCO2 (b).

Full size image

The estuarine systems had a mean Dissolved Oxygen (DO) saturation of 67 ± 18% and a mean DO concentration of 146 ± 60 μmol L−1 (Table S1; Supplementary Material). Values lower than 156 μmol L−1 are below the limit recommended in the Brazilian legislation14. In this study, 50% of the samples tested had DO concentrations below this limit (156 μmol L−1). A regression analysis demonstrated a strong negative exponential correlation between DO% and pCO2 (r2 = 0.92; p < 0.0001) (Fig. 2b).

TA and DIC in the Brazilian monitored estuaries (Table 1 and Supplementary Material S1) varied widely between 131 and 2,800 μM for TA and between 241 and 2,527 μM for DIC. High DIC concentrations (>2,000 μM) were observed in the northeastern estuaries (Apodi, Conchas, Cavalos, Assu, Potengi and Caravelas), whereas lower values (<400 μM) were observed in the northern and southern estuaries (Guajará Bay and Cururupe). TA and DIC were strongly correlated with salinity and pH (Spearman's correlation; r = 0.7; α = 0.05; p < 0.0001).

pCO2 and CO2 emissions (FCO2)

The pCO2 values (Table 1) varied widely between 162 and 8,638 μatm in northern and northeastern estuaries. The pCO2 values of estuaries that receive low water discharges (<50 m3s−1) were not significantly different from those of estuaries that receive high water discharges (>50 m3s−1) (α = 0.05; p < 0.7). A Spearman correlation analysis showed a strong negative correlation (α = 0.05) between pCO2 and salinity, DO% and pH. The pCO2 in the water was significantly different in the dry and rainy seasons (α = 0.05; p < 0.01). Monthly mean pCO2 in the rainy season was 1.5-fold higher than in the dry season. According to the results of Kruskal–Wallis tests (α = 0.05; p < 0.001), the estuaries laying in different climate types (Af, Am, Aw and BSh) had significantly different pCO2 values. pCO2 values were 2 to 3.3-fold higher in estuarine systems located in the Af, Am and Aw climate zones than those of the system located in the BSh climate zone (Fig. 3). The Kruskal–Wallis test also revealed significant differences (α = 0.05; p < 0.01) in the pCO2 values of the different estuary types (tidal fresh/upper, inner, middle and lower estuaries). The pCO2 values in the lower systems were 1.5–2.6-fold lower than those in the tidal fresh/upper, inner and middle estuaries (Fig. 3).

Figure 3
figure 3

pCO2 and FCO2 (averages values) according to climatic type and estuary type for the northeastern Brazilian estuaries.

n indicates the number of samples.

Full size image

The FCO2 of the monitored Brazilian estuaries ranged between 0.9 and 175.2 mmol m−2 d−1 (Table 1; Fig. 4) with a mean rate of 55 ± 45 mmol m−2 d−1. CO2 emissions significantly differed (α = 0.05; p < 0.01) seasonally, with monthly rates 1.8-fold higher during the rainy season when compared with the dry season (Fig. S1; Supplementary Material). The Kruskal–Wallis tests (α = 0.05; p < 0.01) for the climate types (Af, Am, Aw and BSh) demonstrated significant differences in the CO2 fluxes. CO2 emissions in the estuaries characterized by Af, Am and Aw climates were 2.2–2.8-fold higher than those in the system within the BSh climate (Fig. 3). The Kruskal–Wallis test also demonstrated significant differences in CO2 emissions (α = 0.05; p < 0.001) according to estuarine type (inner, middle or lower), where the lower systems showed 2–3.6-fold lower rates (Fig. 3).

Figure 4
figure 4

FCO2 levels of the northeastern Brazilian estuaries.

The numbers represent the estuaries shown in Fig. 1 (figure generated with Surfer 9.0 ® software).

Full size image

CO2 emissions from 4,222 km2 of estuarine area in the northern Brazilian were estimated to be 1.01 Tg C y−1.

Discussion

The processes that partition dissolved CO2, pCO2 and DIC in estuarine environments are complex and are influenced by a combination of factors and natural processes, such as the types of rock and soil in the hydrographic basin, the water–atmosphere interactions and the oxidation/reduction reactions of anthropogenic inputs4. However, most of these processes are controlled primarily by the pH. The results of this study indicate that the typical tropical coastal systems in the northern and northeastern regions of Brazil reflect significant heterotrophy in terms of the CO2 fluxes (average: 55 ± 45 mmol m−2 d−1) and the high levels of pCO2 observed (average: 1,564 μatm). Analyses were conducted to determine whether the rock and soil types influence the water CO2 content. No correlation between fluvial discharge and pCO2 was observed, suggesting that the transport processes via fluvial runoff are not the primary factors responsible for the high concentrations of dissolved CO2 found in the Brazilian estuaries monitored in this study. Other studies have shown that the lack of correlation between these two variables is characteristic of basins that are located in populous industrialized areas15. For example, in the Zenne River (an urban river in Brussels, Belgium), periods of high fluvial discharge have not been found to be correlated with high pCO2 values16. PCA analysis results (Fig. 5) showed a strong positive correlation between rainfall and pCO2 in Guajará Bay (in the northern region). Guajará Bay had the highest river discharge of the systems studied (5,946 m3 s−1) as a result of the high runoff of the Pará River. In the northern region (systems 1 to 3; Fig. 1 and Fig. 3), tropical climate (Af and Am types) predominates precipitation in this region is 2.3-fold higher than in the northeastern region (systems 4 to 28; Fig. 4). However, the evaporation in the northeastern region is 2-fold higher than that in the northern region. The BSh climate (4 systems) strongly influenced this result. The climate in this region is semi-arid (BSh type) and the estuaries are surrounded by shrimp farms, salt ponds and agricultural areas. The mean annual freshwater discharge in the region was 4.3 m3s−1 and the rate of evaporation was 8.5-fold higher than the precipitation measured during this study. The difference between the mean temperatures of the northern and northeastern regions was 1.3°C. The reported mean temperatures of the water masses adjacent to the coasts of the northern and northeastern regions of Brazil were 27.6°C and 27.2°C17,18, respectively. We obtained similar results for the estuaries of the northern and northeastern regions (28.5°C–27.6°C; Table 1).

Figure 5
figure 5

Graphic representations of the results of the principal component analysis (PCA).

The four groups correspond to the four associations: rainfall and pCO2 (within the square area); fluvial discharge (Q) and estuarine area (within the smaller oval); evaporation, DO%, temperature (T°C), pH, TA and salinity (within the larger oval); and population density and FCO2 (within the central area of the graph). The parameters are indicated by triangles (Δ) and the estuaries are indicated by filled circles ().

Full size image

A strong negative correlation between DO% and pCO2 (Fig. 2b) was detected, suggesting that intense organic matter decomposition in the estuaries, either in the water column or the sediment, results in a decrease in pH (Fig. 2a) and enhanced pCO2 values. Other processes such as direct groundwater discharge, biogeochemical reactivity at seepage faces or benthic organic matter processing could also influence the observed surface water composition in these estuaries19,20,21. Additionally, the AOU (Table S1; Supplementary Material) was positively correlated with the excess CO2 (r2 = 0.51; p < 0.0001), which indicates that the organic load that enters the systems has a significant impact on the microbial processes that are associated with the carbon and nitrogen cycles. The DIP measurements were strongly correlated with pCO2 values in systems where the population density was >100 inhabitants/km2 (r2 = 0.51; p < 0.002). This nutrient enrichment can lead to the excessive production of algal biomass and consequently to eutrophication. Under high CO2(aq) values, which result from bacterial respiration, carbonic acid is produced and the water pH decreases15. If the characteristics of a water body are such that the CO2 concentration in the aqueous media is close to equilibrium with the atmospheric CO2 levels, strong correlation among TA, pCO2 and pH can be expected. However, the complex set of processes that lead to the production or consumption of CO2(aq) and the production or removal of carbonic acid (HCO3−) prevent the establishment of a correlation among TA, pH and pCO24,15. TA exhibited a strong correlation with pH but not with pCO2. High TA values (average: 1,298 μmol L−1) may therefore be associated with the high values of HCO3− that are found mainly in the northeastern systems, which represent 90% of the total DIC; CO32− represents 6% of the total DIC and the remaining 4% is associated with CO2(aq). These values are consistent with those recorded for highly polluted systems, including tropical rivers, such as the Capibaribe and Tietê Rivers in Brazil4,22 and temperate systems, namely, the Scheldt estuary and its tributaries Zenne and Dijle (Belgium and the Netherlands)16.

DIC concentrations in the Brazilian estuaries (average: 1,229 μmol L−1) were higher than those reported for tropical rivers (average: 695 μmol L−1)23 and lower than those reported for Indian estuaries (average: 1,777 μmol L−1)24.

In the inner and upper estuaries, the impoverishment in oxygen due to organic matter remineralization is accompanied by an increase in DIC, TA and pCO2 and a decrease in pH. TA concentrations may have also been affected by the incorporation of fulvic and humic acids originated from the mangrove areas or urban sewage11, contributing to the high values of TA as noted above.

The average pCO2 in the Brazilian estuaries varied widely, reaching levels up to 8,638 μatm. The pCO2 levels observed in the Brazilian estuaries are within the reported range for other estuaries around the world. A recent study estimated that on average, upper estuaries around the world have a pCO2 of 3,033 ± 1,078 μatm6. This value is similar to that observed in the present study (average: 2,674 μatm) for the tidal fresh and upper estuaries (Fig. 3). Our pCO2 values for the middle and lower estuaries (average: 1,783 and 747 μatm, respectively) were also similar to those reported for other estuaries around the world (2,277 ± 626 and 692 ± 178 μatm, respectively)6 (Fig. 3). Similar values to those presented here were reported by recent studies for tropical estuaries in the northeastern region of Brazil (4,700 ± 2,420 and 2,200 ± 360 μatm, respectively)11,12. The pCO2 supersaturation found in the upper estuaries is not unusual, as estuaries are generally highly heterotrophic ecosystems where the organic carbon carried by rivers is partially remineralized. The spatial pCO2 patterns of the Brazilian estuaries exhibited trends similar to those observed for other estuaries around the world.

The average FCO2 from the Brazilian estuaries is 54.4 ± 45 mmol CO2 m−2 d−1. This value is 2-fold higher than that reported for Indian estuaries (27 mmol CO2 m−2 d−1)24 and 0.73-fold higher than that presented for tidal systems and small deltas (74 ± 55 mmol CO2 m−2 d−1)2. A recent study using global-scale segmentation found that the average CO2 emission rate from small deltas and tidal systems was 16.5 mol m−2 y−1 (45 mmol CO2 m−2 d−1)25. This value is very close to that observed in the present study. Tidal systems and small deltas represent over 90% of the estuarine systems of the northern and northeastern regions of Brazil.

Another recent study indicated that upper estuaries are strong sources of CO2 (106.8 mmol CO2 m−2 d−1), middle estuaries are moderate sources of CO2 (47.9 mmol CO2 m−2 d−1) and lower estuaries are weak sources of CO2 (23 mmol CO2 m−2 d−1). These values are very similar to those obtained in our study (85, 67.2 and 23 mmol CO2 m−2 d−1, respectively)26. According to the Kruskal-Wallis and Dunn tests, the estuarine groups in this study showed significant differences (p < 0.05) among the upper, middle and lower types. Tidal fresh and inner estuaries emit CO2 to the atmosphere via heterotrophic ecosystem metabolism9,27. However, our data (pCO2 values and CO2 fluxes) showed high standard deviation values, which complicates comparisons.

We refer to 6 reports indicating that tropical estuaries (<23.5° latitude) exhibit an average emission rate of 23.9 ± 28.5 mol CO2 m−2 d−1 (65.4 mmol CO2 m−2 d−1), which is very similar to the values obtained in the present study for the climatic types Af (69.3 mmol CO2 m−2 d−1), Am (60 mmol CO2 m−2 d−1) and Aw (53.7 mmol CO2 m−2 d−1). According to the Kruskal-Wallis and Dunn tests, the climatic groups showed significant differences (p < 0.05), primarily between the Af and BSh types. The spatial FCO2 patterns observed in the northern and northeastern Brazilian estuaries were similar to those observed in tropical regions in other studies, indicating similar hydrological and biological processes, with the exception of the BSh climatic type. As noted above, the semi-arid estuaries in this region are surrounded by shrimp farms, salt ponds and agricultural areas. The rates of FCO2 determined for these systems were similar to those determined for the lower estuaries. As can be noted in Fig. 5, the estuaries in this region exhibited positive correlations among salinity, TA, pH and evaporation.

Finally, Brazilian estuaries contribute 1.01 Tg C y−1 (52%) less CO2 to the atmosphere than that reported for Indian estuaries (1.92 Tg C y−1)24. However, the estuarine area of the Brazilian systems presented here is 6.4 times lower than that of Indian estuaries. Small deltas and tidal systems worldwide are estimated to contribute 120 Tg C y−1 as CO2 to the atmosphere from an area of 360 × 103 km22.

Our study suggests that the estuaries in northern and northeastern Brazil act as sources of CO2 to the atmospheric. The range of pCO2 levels observed was similar to that reported for other inner estuaries, except in the case of a few estuaries of the BSh climatic type for which lowest levels of pCO2 were determined.

The FCO2 observations made in this study contribute to the characterization of humid tropical estuarine systems that previously represented an important geographical information gap.

Methods

Geographic coverage and types of systems studied

The Brazilian coastal zone spans from latitude 4° N to latitude 34° S. Due to the length of the coast, the interface zone from the continent to the ocean includes a wide range of climatic and geographic regions that vary in their hydrologic, geologic, geomorphologic and oceanographic characteristics28. The geographic area covered in this study included small deltas, tidal systems and large rivers, which constituted 28 systems along a 17.2° latitudinal band (−0.6°; −47.6° to −17.75°; −39.2°), representing 42% (3,875 km) of the total length of the Brazilian coast (9,198 km). In the northern region, three systems (covering a total area of 2,870 km2) were studied, whereas in the northeastern region, 25 systems (covering a total area of 1,351.5 km2) were studied. The population density in the basins adjacent to these systems were <100 inhabitants/km2 for 12 systems, 100 ≤ inhabitants/km2 < 1,000 for 13 systems and >1,000 inhabitants/km2 for 3 systems (Table 1).

According to the Köppen climate classification system29, 7 of the systems are classified as climate type Am (tropical wet climate), 6 systems as type Af (tropical rainforest climate), 11 systems as Aw (tropical wet and dry or savanna climate) and 4 systems as BSh (semi-arid climate). According to this classification system, tropical climates and dry (arid and semi-arid) climates are present in our study area (Fig. 1).

According to an estuarine gradient classification system30, 8 of the systems are characterized as tidal fresh and upper estuary type, 11 belongs to the inner estuary type, 3 to the middle estuary type and 6 to the lower estuary type (Table 1).

River discharge data were obtained from the database of the Brazilian National Water Agency31 and demographic data were obtained from a database of the Brazilian Institute of Geography and Statistics32.

Sampling, analysis and calculation methods

In situ measurements were conducted from a boat during 12 continuous months (July 2012–June 2013) in estuarine portions (0.5 ≤ salinity < 35) of the 6 systems of the coast of the state of Pernambuco (−7.55°; −34.8° to −8.85°; −35.1°). Temperature and salinity (conductivity) were measured using a CTD (Sea Bird Electronics SBE911plus® - Washington, USA). Measured salinity was verified via comparison with the chlorinity, which was determined using AgNO3 titration33. pH was measured on the NBS scale on board and after sample collection using a pH/ion analyzer 350 and a Ross combination electrode (Orion ®). The precision and the accuracy of the pH measurements were ±0.005 units and 0.1%, respectively. In the laboratory, total alkalinity (TA) was determined using analytical procedures described by the US Geological Survey (USGS) for water samples (‘ANC–alkalinity and acid neutralizing capacity’) using the inflection point titration method34. The precision and accuracy of TA determinations were ±20 μmol kg−1 and 5%, respectively. Dissolved inorganic nutrients (phosphate (DIP) and silicate) were analyzed according to Reference 33, after filtration of the samples using 0.47-mm Whatman® GF/C glass fiber filters. The precision was 0.01 μmol for DIP and 0.03 μmol for silicate. The accuracy was ±2% for DIP and ±6% for silicate.

Additionally, we estimated the amount of carbon emissions from 22 estuaries on the Brazilian coast (several months between 2001 and 2012) using measurements of salinity, temperature, phosphate, silicate, dissolved oxygen (DO), alkalinity and pH together with data stored at a monthly resolution from monitoring records of sources linked to states and universities in Brazil. The relative oxygen saturation (DO%) in the water was calculated from the DO concentrations and oxygen solubility in the water at the field temperature and salinity. The apparent oxygen utilization (AOU) was computed using oxygen solubility equations35.

pCO2, FCO2 and DIC were computed using the salinity, temperature, nutrients (phosphate and silicate), pH and TA measurements with dissociation constants given for a salinity range of 0 to 5036 in the CO2calc® program37. Daily and monthly wind speed and barometric pressure data were obtained from the Brazilian Meteorological Institute (INMET) for stations that were close to the respective estuaries. The molar fraction of atmospheric CO2 (XCO2) was obtained from NOAA (http://esrl.noaa.gov) (ppm) and pH2O = water vapor pressure (μatm) was obtained from Reference 38. The gas transfer rate was calculated using the equation given in Reference 39.

Statistical analyses (regression analyses, t-tests, Spearman's correlation, principal component analysis or PCA and descriptive statistics) were performed using the XLSTAT® 2010 software.