The role of hydrodynamic and biogeochemistry on CO 2 flux and p CO 2 at the Amazon River mouth

Recent estimates indicate that the lower Amazon River outgasses significant amounts of carbon dioxide (CO2) that was not previously accounted for the global inland water carbon budget. Detailed evaluation of seasonal variability and controlling mechanisms behind the CO2 fluxes in this large and complex area remains incomplete. Previous observations 10 throughout the Amazon basin showed that higher CO2 fluxes (FCO2) and partial pressure of CO2 (pCO2) occur during high water and higher wind intensity seasons. The influence of wind and water speed, depth of water column, as well as respiration of allochthonous and autochthonous organic matter, are frequently assigned as the main control variables. Here, we assess the influence of a set of biogeochemical and hydrodynamic parameters on the seasonal variation of FCO2 and pCO2 near the Amazon River mouth. FCO2, pCO2 and biogeochemical and hydrologic analyses were carried out from 2010 15 to 2016 during four different hydrological periods per year (N=25) in the North Channel of the Amazon River mouth. FCO2 and pCO2 were used as dependent variables and analyzed against 33 biogeochemical, hydrodynamic and meteorological parameters along the hydrological seasons. The highest FCO2 and pCO2 was obtained at high discharge season (11.28 ± 7.82 μmol m2 s and (4575 ± 429 μatm, respectively) when most of these parameters tend to be higher. Among the 33 parameters analyzed, the significant correlations with FCO2 and pCO2 (p<0.05) observed were for water and air temperatures, dissolved 20 oxygen, dissolved organic carbon, nitrate, dissolved inorganic nitrogen and pH. These variables could be considered suitable predictors for estimating pCO2 and FCO2 in the Amazon River mouth area. For a better estimation and understanding of carbon budgets in tropical rivers it is still required to verify and to quantify more deeply the relationship among CO2 evasion and others hydrodynamic, meteorological and biogeochemical variables. Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-465 Manuscript under review for journal Biogeosciences Discussion started: 10 December 2018 c © Author(s) 2018. CC BY 4.0 License.


Introduction
Rivers play an active role in the global carbon budget, connecting terrestrial ecosystems to the atmosphere and ocean.The Amazon River is responsible for 20 % (120000 m 3 s -1 ) of the water discharge to Western Tropical North Atlantic Ocean, being the largest source of freshwater to the ocean (Molinier et al., 1996;Goes et al., 2014).
Inland running waters transported roughly 5.7 Pg C annually.This value represents approximately 75 % of this amount returned to the atmosphere as CO2 (Le Quéré et al., 2015), with the terrestrial plants being the major source of carbon processed in the river (Richey et al., 2002;Cole et al., 2007;Aufdenkampe et al., 2011).
The most recent estimates shows that Amazon running waters outgasses 1.39 Pg C year −1 , increasing the global emissions to 2.9 Pg C year −1 from inland waters.This difference represents an increase of about 43 % (Sawakuchi et al., 2017).This value is also approximately an equivalent amount of CO2 as the rainforest sequesters on an annual basis (Brienen et al., 2015).
Therefore, understanding the drivers controlling pCO2 and CO2 outgassing is important to improve regional and global scale carbon budget estimates.
Climatic and geomorphological features play an important role on water turbulence and chemistry, which in turn directly affects gas fluxes to the atmosphere.In large rivers, wind speed is one of the major controls on gas transfer velocity (K600) (Borges et al., 2004;Alin et al., 2011;Rasera et al., 2013;Sawakuchi et al., 2017).Likewise, other physical variables such as depth and width of the channel, and the fetch length, vertical mixing and water velocity are also related to the evasive flux (Raymond and Cole, 2001;Rasera et al., 2008;Liu et al., 2017).In streams and small tributaries, seasonality, water velocity and depth are the most important physical parameters, while for mainstem and estuaries wind speed is the main driver related to the gas flow at the air-water interface (Alin et al., 2011).
However, most studies in the Amazon River basin are concentrated in the central region, which represents ≈ 30% of the total basin area (Hedges et al., 1986;Hedges et al., 2000;Moreira-Turcq et al., 2003).On the other hand, studies in the lower Amazon River are still relatively scarce, with low temporal resolution due to limited logistical access.Besides that, usually, these studies do not consider the influence of a large range of variables (biogeochemical, hydrodynamic and meteorological) and local factors such as macro and meso tides, strong trade winds and large air-water surface area that promote favorable conditions for CO2 outgassing (Sawakuchi et al., 2017).The present study aimed to evaluate and identify the most important hydrodynamic, biogeochemical and meteorological factors related to the pCO2 and FCO2 at the Amazon River mouth.

Study area and sampling procedures
The measurements were carried out from 2010 to 2016 in the deepest point of the North Channel in the Amazon River mouth (00°02'872" S 051°04'616" W) near the city of Macapá (Fig. 1) (Table 1 and 2, S1).The North channel represents ~30 % of the Amazon River discharge (Ward et al., 2015).This area is influenced by meso tides that promotes a ~3m variation in river depth amplitude, reversing the river flow with no salinity intrusion (Sawakuchi et al., 2017).In 150 km downstream from Macapá the North Channel remains to widen and channelize between large islands until being completely disconnected from land and the riparian zone/floodplains (Ward et al., 2015;Sawakuchi et al., 2017).Water entering the ocean continues fresh, at the surface, as much as 60 km offshore from this point (Molinas et al., 2014).
Located in the equatorial region, the climate in the study area is characterized by a rainy tropical monsoon climate, type "Am" (Köppen, 1948), warm and humid, with little temperature variation.The annual average air temperature is 26.0 ± 0.4 °C, with maximum and minimum values between 31.5 ± 0.7 °C and 22.0 ± 0.3 °C, respectively (Costa et al., 2013).
Field sampling was carried out based on the hydrographic periods developed by Richey et al., (1990) and following the protocols described in the CAMREX project (Carbon in the Amazon River Experiment), over the last 30 years (Richey et al., 1986(Richey et al., , 1990)).
Sampling campaigns of 27 biogeochemical, 5 kinetic and 2 thermodynamic variables (Table 1) were done between 2010 and 2016 (N = 25).Surface in situ measurements and samples were collected for the biogeochemical analysis.The resulted data set of each field campaign was combined according to four hydrological seasons: rising water season (from Jan to Mar), high water (from Apr to Jun), falling water (from Jul to Sep) and low water (from Oct to Dec) (Valerio et al., 2018).The data analysis was carried out considering the average values of each of the 25 campaigns for the identification of the most influencing parameters on pCO2 and FCO2 and for the understanding of their seasonal variability.

pCO2, CO2 flux and estimation gas transfer velocity (K600)
pCO2 was measured using a headspace equilibration chamber filled with glass beads to increase the exchange surface and reduce the volume of air.The chamber was coupled to an Infrared Gas Analyzer (LICOR Instruments, LI-820) as described by Frankignoulle et al., (2001).Surface water was collected at approximately 30 cm depth by a submersible pump (Rule 360 Bilge) passing through the chamber from top to bottom at a flow rate of 1.5 L min -1 .A diaphragm mini-pump (AS-200; Spectrex) promotes a closed air circuit (150 mL min -1 ) circulating through the equilibrating chamber through a moisture trap (Drierite) connected to an Infrared Gas Analyzer (Rasera et al., 2008, Sawakuchi et al., 2017).
CO2 fluxes were measured using a floating chamber (volume 10.6 L and 0.1251 m 2 area) (Rasera et al., 2008).The floating chamber was coupled to a portable Infrared Gas Analyzer using a diaphragm minipump (AS-200; Spectrex) and a desiccating water trap (Drierite).Flux measurements were started after the concentrations were stable with the atmosphere concentration was registered.Chamber measurements were done for approximately 5 minutes and repeated for three times.
The FCO2 in the water-air interface and K600 were calculated as described by Sawakuchi et al (2017), based on the following main equation: where d(pCO2)/dt is the slope of the CO2 accumulation in the chamber (µatm h −1 ), V is the chamber volume (m 3 ), TK is the air temperature (in Kelvin, K), A is the surface area of the chamber at the water surface (m 2 ), and R is the gas constant (L atm K −1 mol −1 ) (Frankignoulle, 1988).
Triplicate samples were filtered through a pre-combusted GF/F filters (Whatman) into 30 mL pre-combusted glass vials with Teflon septa and preserved in the field with 30 µL of 50 % Hydrochloric acid (HCl) for determination of dissolved organic carbon (DOC), total dissolved nitrogen (TDN), dissolved organic nitrogen (DON) and dissolved inorganic nitrogen (DIN) with a Shimadzu TOC-VCPH.
Coarse and fine suspended sediment concentrations (CSS and FSS, respectively) were determined by filtering a known volume of water, using a churn splitter to homogenize the samples into a 0.63 µm sieve to obtain the coarse fraction and then through pre-weighted 0.45 μm cellulose acetate membrane filters to obtain the fine fraction similarly to Leite et al., (2011).
The consumption of oxygen over a 24 h period was used for the calculation of respiration rates.Three initial and final replicate samples were incubated in 60 mL acid-washed Biological Oxygen Demand bottles in the dark in river water, held at ambient temperature.After these steps, oxygen concentrations were measured by Winkler Titrations method (Wetzel and Likens 1991;Ellis et al., 2012) using a Hach titrator.

Hydrodynamic and meteorological variables
Discharge, water velocity, and river depth were measured across the North Channel transect (Fig. 1) during all cruises using a Sontek River Surveyor M9 Portable nine-beam 3.0 MHz/1.0MHz/ 0.5 MHz Acoustic Doppler Current Profiler (ADCP).
The precise calculation of the mean discharge, besides velocity and depth, during a complete semidiurnal tide cycle, was completed in ~ 10 to 13 hours.Normally, the net discharge measurement process requires an effort of 8 to 11 trips between the left and right banks of the Amazon River (Ward et al., 2015;Sawakuchi et al., 2017).

Statistical analyses
To verify the normality of the data distribution the Shapiro-Wilk test was performed (α <0.05), in which null hypothesis of normality was rejected.Nonparametric correlation (Spearman) among FCO2 and thermodynamic, kinetic and biogeochemical parameters were performed with a 95 % confidence interval and level of significance α <0.05.The Kruskal-Wallis analysis (p <0.05) was performed to verify the significant differences of FCO2, pCO2 and net discharge for each hydrological season (high, falling, lower and rising waters).All the statistical analyzes were performed with R-project software, version3.3.2 (Develoment Core Team, 2018).

Biogeochemical and hydrodynamic seasonal variability
The regular seasons oscillation, mainly between 2014 and 2016, is shown by the annual discharge hydrograph at North Channel (Fig. 2).Discharge varied from a low of nearly 33,000 m³ s -¹ at the end of low water season in december of 2010, to a high of about 141,000 m³ s -¹ in may of 2014, at the middle of high water season.At rising and high water, discharge reached 82,389 ± 34,715 m³ s -1 and 104,637 ± 17,505 m³ s -1 (Fig. 5), respectively.The discharge reduced roughly 50 % at low water, showing a significant variation between the seasons (Kruskal-Wallis, p <0.05).
The highest CO2 fluxes were observed during high discharge season (11.28 ± 7.82 μmol CO2 m -2 s -1 , Fig. 3).FCO2 reached 24.3 μmol CO2 m -2 s -1 in months with high discharge and was as low as 2.20 μmol CO2 m -2 s -1 in months with low discharge, showing an expressive reduction (68 %) of the outgassing potential, with a significant seasonal variability between months with high and low discharge (Kruskal-Wallis, p <0.05).
pCO2 was higher at high water season (4575 ± 429 μatm, Fig. 3) followed by falling water (3080 ± 957 μatm, Fig. 3).During maximum discharge, the pCO2 (4885 µatm) was about 11 times that of the concentration in the atmosphere (390 µatm, Richey et al., 2002).Even at low discharge season, when the availability of organic matter related to DOC, CSS and FSS concentrations was significantly reduced (Fig. 4), the surface water pCO2 was at least 3 times higher than the atmosphere (Fig. 3, Table 2).
The highest pH and DO values (7.08 ± 0.38 mg L -1 and 6.69 ± 0.33 mg L -1 ) were measured at low water, when pCO2 and FCO2 were lowest (Table 2).

Relationship between FCO2, pCO2, biogeochemical and hydrodynamic variables
Discharge presented a positive correlation with pCO2 (Spearman R = 0.75, p <0.001, Fig. 5).But we could not find a significant correlation with FCO2 (p > 0.05, Table 3).The seasonal pattern of wind speed and K600 observed (Fig. 6) corroborated the assumptions of the expected influence of the wind on the CO2 transfer rate in estuarine environments (Broecker and Siems, 1984;Wanninkhof 1992, Wanninkhof andMcGillis 1999).However, we could not find a significant correlation between FCO2 and wind speed (p > 0.05) (Table 3).
A negative correlation was registered for both pCO2 and FCO2 with dissolved oxygen (Table 3, Fig. 8) and pH (Table 3).
Considering that a positive correlation between DOC and pCO2 was observed, these results highlighted the influence of the biological processes, especially organic matter degradation on the CO2 outgassing (Varol and Li, 2017).On the other hand, a significant correlation between pCO2 and respiration was not observed.This is different from what we hypothesized based in the assumption that in the river mainstem heterotrophic respiration is often considered to be a primary mechanism for aquatic CO2 supersaturation (Mayorga et al., 2005;Ward et al., 2013).

Discussion
The highest CO2 evasive flux and pCO2 were observed in high discharge season.These results are consistent with previous studies in large tropical rivers (Richey et al., 2002;Li et al., 2012;Sawakuchi et al., 2017).Our observed correlations between pCO2 and FCO2 with DOC, air and water temperatures, pH and DO have also been shown in other tropical systems (Rasera et al., 2013;Li et al., 2012;Liu et al., 2017).
The Amazon River pCO2 is characterize by a dynamic balance between inputs from respiration of organic matter in the mainstem and floodplains and outputs from outgassing due to in situ primary production (Devol et al. 1987;Richey et al., 2002;Mayorga et al. 2005;Engle et al., 2008;Melack et al., 2009;Silva et al., 2009;Abril et al., 2014;Gagne-Maynard et al., 2017;Sawakuchi et al., 2017;Ward et al., 2018).This balance is directly influenced by the hydrological cycle leading to a seasonal variability of FCO2 and pCO2, as we observed.Rainfall is the main factor that controls the inputs of allochthonous organic matter during rising and high waters to the mainstem, while floodplains have a major role during falling water season (Richey et al., 2002;Abril et al., 2014).
The higher CO2 outgassing observed at falling water (Fig. 3) could be attributed to the influence of organic matter retention in floodplains, were intense autotrophic production occurs and when the water level start to decrease the transfer of organic matter to rivers increased (Mortillaro et al., 2012).

Biogeochemical Controls
The breakdown of molecules derived from terrestrial plants within the fluvial heterotrophic network is a potentially dominant source of CO2 for the Amazon river mainstem (Duarte and Prairie, 2005;Battin et al., 2009;Abril et Ward et al., 2013Ward et al., , 2016)).As such, a positive correlation between pCO2 and consequently FCO2 with respiration was expected, however, this was not observed.This can be related to the bacterial growth rates, which were higher during highrising and high-falling water than during low-rising water.Patterns of community respiration were the opposite, with respiration rates highest during low-rising water.The combination of high rates of bacterial production and low rates of respiration during high water suggests that bacterial growth efficiencies were maximal during high water (Benner et al., 1995).It is important to consider, however, possible methodology inconsistences.Ward et al., (2018) showed that the titration method may lead to an underestimation of the respiration rates in the lower Amazon River, which could be the cause of the lack of correlation between these parameters.For example, it was observed that respiration rates were tightly linked to how rapidly they were mixed in carefully controlled incubation chambers and the authors further concluded that river flow rates may control microbial respiration rates by controlling the suspension of sediments and particle-bound microbes (Ward et al., 2018).
The contribution of allochthonous organic matter to the river, indicated by lignin phenol compositions, has been shown to be highest during rising and high water (Ward et al., 2015), and, likewise, we registered highest concentration during high, rising and falling waters (5.02 ± 0.72 mg L -1 , 4.64 ± 0.6 mg L -1 and 4.46 ± 0.62 mg L -1 respectively, Table 2).At low water, the DOC concentrations decreased 27 %, to 3.66 ± 0.45 mg L -1 (Fig. 4).These elevated DOC levels, and previously measured lignin phenols, indicate more abundant substrates for microbial decomposition, which is coincides with high pCO2 and FCO2.
These results indicate that most of the organic matter in the river may be carried from land to the water by lateral transport (Ward et al., 2017) and can have a strong influence on organic matter concentration in the water and consequently, how terrestrial environments and floodplains may be major drivers controlling pCO2 and FCO2 (Mayorga et al., 2005;Ward et al., 2013Ward et al., , 2016Ward et al., , 2017;;Abril et al., 2014).
The positive relationship between DOC and pCO2 can be related to the organic carbon limited bacterial growth and respiration in whitewater rivers, Benner et al. (1995) observed that the quality of the bulk DOC is the main factor limiting bacterioplankton growth in white Amazonian rivers.The greater residence time of the water influenced by tidal cycles and the influence of the flood pulse contribute to the higher concentrations of DOC at the Amazon mouth (Thomaz et al., 1998;Farjalla et al., 2002).The annual peak in high water, which typically is in June, results in the inundation of large areas of the floodplains and an accelerated exchange of materials between river and floodplains, besides the daily exchanges influenced by the tidal regime that causes a ~3 m variation in river amplitude (Benner et al., 1995;Sawakuchi et al., 2017).
Although others studies in the Amazon Basin, mainly in lakes, did not find significant correlation between DOC and pCO2 (Marotta et al., 2010;Pinho et al., 2015;Valerio et al., 2018 ) the hydrodynamic characteristics (i.e.tidal cycle and low water velocity) and the greater bioavailability of organic matter due increased concentrations of DOC along the river continuum (Cunha et al., 2012;Seidel et al., 2015;Ward et al., 2015) create specific conditions that contribute to the heterotrophic metabolism and high concentration and evasion of CO2 at the mouth of the Amazon River.A negative correlation was found between FCO2, pCO2 and pH (Table 3).In situ respiration and photosynthesis can be the major drivers controlling pH at non anthropized river systems, and thus regulating the chemistry equilibrium of the dissolved inorganic carbon (DIC) system (Cole and Caraco, 2001;Rasera et al., 2013, Varol andLi, 2017).Respiration of allochthonous and autochthonous organic matter produce CO2 that decreases pH and increase pCO2 in the river, aquatic photosynthesis increases pH and reduce dissolved CO2 level contributing to the precipitation of carbonates (Li et al., 2012).
The negative correlation between DO and CO2 flux can be related to the oxidation of the organic matter carried out to the river (Varol and Li, 2017).In rivers, the organic matter is available in dissolved and particulate fractions, where the dissolved fraction is more easily assimilated by decomposers, while, the particulate organic matter is slowly broken down into the dissolved, through exoenzymes released by bacteria in the aquatic environment (Pusch et al., 1998).The availability of DOC would control the aquatic microbial heterotrophic production, which would consequently consume O2 and release CO2.
FCO2 was negatively related to conductivity, DIN and NO3 -(Table 3).The concentrations of carbonate species (HCO3 -and CO3 2-) and some ions (Na+, Ca+, Mg+, K+ and SO4 2 ) determines the conductivity in the aquatic environment, which were observed with highest concentrations during the rising and low water, which could be related to the influence of groundwater and lower dilution capacity due discharge reduction (Neiff, 1990;Thomaz et al., 2007;Silva et al., 2013;Cunha and Sternberg, 2018).NO3 -concentrations showed the same pattern, considering that increasing nitrogen could elevate aquatic photosynthesis and reduce pCO2, limitation of bacterial growth by nitrogen and organic carbon was also observed in some freshwater ecosystems (Wang et al., 1992;Benner et al., 1995;Elser et al., 1995;Li et al., 2012).During the low water the CSS and FSS concentrations decreased 55 % and 33 %, respectively, which could be related to the limited floodplain connectivity, promoting the higher light input in the water column (Gagne-Maynard et al., 2017).

Hydrodynamic and Meteorological Controls
As observed in our study, pCO2 had a strong and positive correlation with discharge (R = 0.75, p <0.01, Fig. 5) and a negatively correlation with water temperature (R = -0.68,p <0.01).The pCO2 can be influenced by rainfall intensity, flood pulse and river discharge that controls the input of organic matter in river system (Finlay, 2003).In addition, the combination of hydrodynamic and meteorological parameters, such as water velocity, water temperature and wind speed determine the air-water turbulence which controls the CO2 outgassing (Barth and Veizer, 1999;Alin et al., 2011).
The local precipitation in the basin added to the upstream hydrological characteristics (geomorphology, size, relief, slope of the channel, land and soil cover use, etc) regulate the net discharge in a certain stretch of the river.In turn, depending on the characteristics of the water body (river, estuary, lake), there may be interactions between the wind and the surface of the water, generating more turbulence and promoting synergistic effects that tend to increase the mass transfer coefficient in the air-water interface, consequently, these interactions increase the CO2 outgassing (Alin et al., 2011;Varol and Li, 2017).
However, this relation between water velocity and FCO2 was not observed in the studied site near the Amazon River mouth due the lower water velocity (roughly 0.7 m s -¹, Table 1) in comparison to upstream sections of the Amazon, where no tidal effect is present, i.e.Óbidos with 1.92 m s -¹ (Sawakuchi et al., 2017).
Our results showed that water and air temperature presented a strong influence on FCO2 (Table 3) and K600 (Fig. 8).As observed in other tropical rivers (Li et al., 2012, Liu et al., 2017) the mean water temperature is considered relatively high at the Amazon River (Moreira-Turcq, 2013).Thus, water temperature and turbulence, both factors can reduce the gases solubility in the water leading to a higher exchange ratio (for instance, CO2 and DO) and in addition accelerating the decomposition of organic matter (Alin et al., 2011, Li et al., 2012, Ward et al., 2018).
In estuaries wind speed is often the major source of turbulence on the water surface (Alin et al., 2011).A high rate of gaseous transfer, at any wind speed, can be favored when both wind and water currents have opposing directions.These effects result in the sum of both velocities and, consequently, tend to intensify the level of turbulence at the interface airwater (Zappa et al., 2007;Beaulieu et al., 2012;Sawakuchi et al., 2017).However, the absence of correlation with U10 and FCO2 in our investigation may be explained by the use of wind velocity data, obtained from weather stations located on land relatively far from sampling points, leading to the use of possible non-representative data.

Conclusion
Higher pCO2 and CO2 outgassing were observed during the high waters season highlight the strong influence of the hydrological cycle.This coincides with the peaks of dissolved organic matter concentration, water and air temperature contributing to the increase in pCO2.
These assumptions are confirmed by the positive correlation between FCO2 and pCO2 with DOC, and negative correlation with water and air temperatures.Consequently, these variables could be considered suitable predictors for estimating pCO2 and FCO2 in the Amazon River mouth area.For a better estimation and understanding of carbon budgets in tropical rivers it is still required to verify and to quantify more deeply the relationship among CO2 evasion and others hydrodynamic, meteorological and biogeochemical variables.Our results can support the development, adjustment and parameterization of regional models of CO2 emissions in the study area, providing a significant contribution for the understanding of carbon cycle of the Amazon River mouth, reducing the FCO2 estimates errors for different temporal scales in large tropical rivers.

Figure 2 :
Figure 2: Discharge at North Channel with open cycles showing the data of time series samplings.

Figure
Figure 3: Seasonal variability of FCO2 and pCO2.

Figure 4 :
Figure 4: Seasonal variability of DOC, FSS and CSS.