Radon‐traced pore‐water as a potential source of CO2 and CH4 to receding black and clear water environments in the Amazon Basin

Groundwater is a primary source of dissolved CO2 and CH4 in Amazonian headwaters, yet in higher order rivers, a groundwater/pore‐water source is difficult to constrain due to the high spatial and temporal heterogeneity of pore‐water exchange. Here, we report coupled, high resolution measurements of pCO2, CH4, and 222Rn (a natural pore‐water and groundwater tracer) during receding waters in the three major water types of the Central Amazon Basin: black (Negro River); clear (Tapajós River); white (Madeira River). Considerable spatial heterogeneity was observed in pCO2, CH4, and 222Rn concentrations ranging from 460 μatm to 8030 μatm, 7 nM to 281 nM, and 713 dpm m−3 to 8516 dpm m−3, respectively. The significant correlations between pCO2 and CH4 to 222Rn in the black and clear waters suggests that pore‐water further enhanced CO2 supersaturation by 18–47% and is a driver of CH4 dynamics in these waters.

Our understanding about the role of rivers in global greenhouse gas budgets is evolving. Recent estimates of CO 2 efflux from rivers vary in magnitude from 0.6 Pg CO 2 yr 21 to 3.9 Pg CO 2 yr 21 (Aufdenkampe et al. 2011;Raymond et al. 2013;Lauerwald et al. 2015;Drake et al. 2018) and estimated emission of CH 4 has recently been revised from 1.5 Tg C yr 21 to 26.8 Tg C yr 21 (Bastviken et al. 2011;Stanley et al. 2016; see also Kirschke et al. 2013). The tropics are the largest contributor of global CO 2 emissions from rivers, yet the region is underrepresented in global data sets and the source of large uncertainties Lauerwald et al. 2015). Constraining the relative contribution of the sources driving riverine CO 2 and CH 4 supersaturation and atmospheric exchange rates remains a challenge (Cole et al. 2007;Raymond et al. 2013;Borges et al. 2015Borges et al. , 2018Teodoru et al. 2015;Stanley et al. 2016). The Amazon river system is generally supersaturated in CO 2 and CH 4 , and is estimated to emit globally significant amounts of both gases (Richey et al. 2002;Melack et al. 2004;Rasera et al. 2013;Sawakuchi et al. 2014; Barbosa et al. 2016). The factors contributing to CO 2 supersaturation remain unclear and are likely spatially and temporally variable (Richey et al. 2009). Respiration of allochthonous (Mayorga et al. 2005) and autochthonous (Ellis et al. 2012) organic matter, carbonate weathering (Vihermaa et al. 2014), and contributions from wetlands and floodplains (Abril et al. 2014) have all been highlighted as sources of CO 2 to Amazonian rivers. Drivers of riverine CH 4 dynamics are more ambiguous, with recent studies highlighting hydrological drivers such as seasonal water stage and wetland-river connectivity (Sawakuchi et al. 2014;Borges et al. 2015;Barbosa et al. 2016). Despite these advances, large uncertainties remain in Amazonian CO 2 and CH 4 budgets (Richey et al. 2009;Melack 2016).
Pore-water and groundwater exchange have been shown to be the primary source of CO 2 and CH 4 in Amazonian headwaters (Johnson et al. 2008;Neu et al. 2011). In higher order Amazonian rivers, pore-water is hypothesized to contribute to CO 2 and CH 4 supersaturation, however, this exchange pathway is difficult to constrain beyond small streams due to high spatial and temporal heterogeneity (Cook et al. 2003). Radon-222 ( 222 Rn) is a natural tracer of any water that has been in contact with sediments (porewater and/or groundwater) and has been used to assess groundwater inputs into river and lakes (Cook et al. 2006;Burnett et al. 2010). More recently, 222 Rn has revealed how pore-water releases CO 2 and CH 4 to estuarine surface waters (Call et al. 2015;Maher et al. 2015;Sadat-Noori et al. 2016), but no similar investigations have been performed in the Amazon. Here, we define pore-water as the exchange of interstitial water into surface waters (i.e., a combination of meteoric and hyporheic exchange). We investigate whether radon-traced pore-water may be a source of CO 2 and CH 4 to major tributaries of the Amazon river system during receding waters spanning the three major water types (black, clear, white).

Methods
Longitudinal surveys were conducted in three major tributaries representing the three water types of the central Amazon Basin: black water (Rio Negro 150 km surveyed); clear water (Tapaj os 100 km surveyed); and white water (Madeira 100 km surveyed) (Fig. 1a-c). Each water type has unique chemical characteristics related to the geomorphological properties of their catchments (Sioli 1968;Junk et al. 2011). Briefly, black waters drain large areas of low-lying podzols and contain high levels of dissolved organic material. Clear waters drain Precambrian shields and are low in suspended sediments and organic material. White waters originate in the Andes Mountains and contain high sediment loads and nutrients. Extensive wetlands and floodplains exists in each basin (Junk et al. 2011;Hess et al. 2015), draining into the main river stems via a complex network of fluvial connections (Mertes et al. 1996). "Igarap es" are forest streams that drain straight to the river channel, or first to floodplain lakes.
Seasonal rainfall and Andes snow melt result in large oscillations in river water levels (Junk et al. 2011) causing the inundation of forests, wetlands, and floodplains across the basin (Hess et al. 2015). Surveys were conducted during receding waters during August 2015 and September 2015 (see Supporting Information for hydrographs for the black, clear, and white rivers). The main riverine channel was surveyed for each water type along with two lakes in black waters and one lake in clear waters. At the time of sampling, all lakes were connected to the main river channel. Black water lakes were surrounded by flooded forests (large trees, non-herbaceous) as was the clear water lake, however, the western flank was separated from the main channel by a sand bar with a single opening.
Water column pCO 2 , CH 4 , and 222 Rn were determined by continuously pumping water from a depth of 50 cm into two showerhead gas equilibration devices (GED) aboard a moving vessel that averaged 10.6 6 3.5 km h 21 and 5.4 6 2.9 km h 21 during river and lake surveys, respectively. Equilibrated headspace air was then pumped into an Off-Axis Integrated Cavity Output Spectrometer which measured CO 2 and CH 4 at 1 s intervals. A separate gas stream from the same GED was pumped to an automated 222 Rn-in-air analyser which logged data at 10 min intervals. Moving averages of 10 min and 30 min were applied to smooth pCO 2 and CH 4 concentrations based on experimentally determined gas equilibration times (Webb et al. 2016). A Hydrolab DS5 sonde logged temperature, every 5 min and a BBE Moldaenke Fluoroprobe logged fluorescence every 5 min. All average values reported in results are 6 95% confidence interval. All regression analyses having a p value of < 0.05 were deemed as being significant. No CH 4 data is available from white water due to instrument malfunction. Detailed descriptions of longitudinal surveys are provided as Supporting Information.

Partial pressure of CO 2
Across all water types, pCO 2 displayed considerable spatial variability (Fig. 1a-c). In the black waters, the range in pCO 2 spanned over 4000 latm, with highest pCO 2 observed where the Igarap e da Freguesia converges with the river at Novo Airão (NA) (Fig. 1a). localized areas of elevated riverine pCO 2 were also observed in the vicinity of the adjoining Igarap e Maraj a (IM, 7817 latm) and Igarap e Camar a (IC, 7435 latm), and at the confluence of the floodplain lake, Lago Acajatuba (LA, 7291 latm). Lowest black water pCO 2 was recorded in Lake 2, however, distinct areas of higher pCO 2 were observed in the southern perimeters of Lake 1 (max 7023 latm) and Lake 2 (max 6864 latm). Overall, average riverine pCO 2 was 26% higher than average lake pCO 2 (Table 1).
In the clear waters, lowest pCO 2 was observed in the river near Santarem and highest pCO 2 at the north-eastern end of Largo Verde (LV) (Fig. 1b). Upstream of LV, riverine pCO 2 gradually increased, peaking (1532 latm) nearby the adjoining igarap e. Other areas of elevated riverine pCO 2 were observed near a small lake at Pindobal (1013 latm) and in the vicinity of Cajutuba Beach (CB, 1174 latm). In contrast to black waters, average lake pCO 2 was 72% higher than average riverine pCO 2 in the clear water (Table 1). White river pCO 2 spanned only 470 latm ( Fig. 1c) with average pCO 2 over twofold higher than the average clear river pCO 2 but less than one third of average black river pCO 2 (Table 1). Significant inverse relationships (p < 0.05) were observed between pCO 2 and chlorophyll a (Chl a) in the black river, black lakes, and clear river (Fig. 2c,f).

CH 4 concentrations
Methane concentrations spanned a range of 274 nM, with lowest concentrations in downstream river locations and maximal concentrations in the lakes (Fig. 1d,e). Black river CH 4 concentrations were generally < 40 nM, however, concentrations up to 157 nM were observed in the vicinity of NA. Clear river CH 4 ranged from 15 nM to 41 nM with highest concentrations in the vicinity of CB and the igarap e. Overall, average CH 4 concentrations were higher in the lakes than in the rivers (Table 1).

Radon-222
A general trend of higher 222 Rn concentrations in upstream locations was evident in all water types (Fig. 1f-h). In black waters, localized areas of elevated concentrations were observed in the river at IC (3423 dpm m 23 ), IM (3389 dpm m 23 ), and NA (4647 dpm m 23 ), and maximum values in the lakes (L1: 6159 dpm m 23 , L2: 6506 dpm m 23 ). Clear water concentrations of 222 Rn were also highest in the lake, with distinct areas of higher riverine concentrations in the vicinity of CB (4778 dpm m 23 ) and the igarap e (7797 dpm m 23 ). Overall, average lake concentrations of 222 Rn were 40% and 60% higher than average riverine concentrations in the clear and black waters, respectively. The white river had the lowest range of 222 Rn (Fig. 1h) but the highest average riverine concentration of the three water types (Table 1). Significant positive relationships (p < 0.05) were observed between pCO 2 and CH 4 with 222 Rn in the rivers and lakes of the black and clear waters (Fig. 2a,b,d,e).

Discussion
Our study presents concurrent surface-water measurements of a natural pore-water tracer ( 222 Rn) with pCO 2 and CH 4 concentrations from waters of the central Amazonian basin. We build on an earlier study documenting 222 Rn concentrations in Amazonian rivers (Devol et al. 1987) by reporting high resolution measurements to map potential areas of increased pore-water to surface-water interactions. The significant positive correlations observed between pCO 2 and CH 4 with 222 Rn suggest pore-water may be a relevant source of pCO 2 and CH 4 during receding black and clear waters, providing a basis for designing future studies to quantify the influence of pore-water exchange in carbon budgets of Amazon waters.

CO 2 and CH 4 distribution
Wide ranges of pCO 2 have been reported from the diverse aquatic systems in the Amazon Basin (Richey et al. 2002;Rasera et al. 2013;Abril et al. 2014;Melack 2016). Seasonally, pCO 2 tracks the hydrograph (Richey et al. 2002(Richey et al. , 2009 and the results from this study are in general agreement with published observations in terms of season and water type. The average pCO 2 observed in the receding black and clear rivers (i.e., Negro and Tapaj os) is higher than those reported by Abril et al. (2014), likely due to our surveys extending further upstream where pCO 2 was considerably higher. Receding white river (Madeira) observations are higher than those reported at the mouth ( 1300 latm) by Abril et al. (2014), but lower than the 4100 latm reported further upstream by Almeida et al. (2017). Clear rivers had the lowest pCO 2 of the three water types which is consistent with other studies Rasera et al. 2013;Abril et al. 2014). The high-spatial resolution data from our study revealed higher pCO 2 upstream and close to igarap es confluence with the main channels, suggesting igarap es may be a source of CO 2 to the main channels. Data on Amazonian CH 4 concentrations are much sparser than pCO 2 , with large spatial and temporal variability of concentrations and associated fluxes (Melack et al. 2004;Sawakuchi et al. 2014;Borges et al. 2015;Barbosa et al. 2016). In contrast to the large spatially distributed measurements of previous studies such as Barbosa et al. (2016), which measured CH 4 at four locations along a 700 km transect of the Negro River and in 21 tributaries within the Negro basin, this article presents smaller scale CH 4 measurements. Methane distribution was characterized by a few localized areas of distinctly higher concentrations in lakes, which are known emitters of CH 4 (Crill et al. 1988;Devol et al. 1988), and where the igarap e joins the black river at NA.

Radon tracing of surface water CO 2 and CH 4 sources
Radon-222 is produced in sediments by the radioactive decay of radium-226 ( 226 Ra) and has a short half-life of 3.8 d. The noble gas is often highly enriched in groundwater/ pore-water and once discharged to surface waters the only losses are radioactive decay and atmospheric evasion (Cook et al. 2008). Radon-222 activities in surface waters integrate the various recent groundwater and pore-water exchange pathways, such as hyporheic exchange or the lateral flow from regional aquifers to the main channels. While it was beyond the scope of this initial study to differentiate between the different radon pathways, our observations imply pore-water connectivity in the river and lakes during receding waters.
Diffusion of 222 Rn from sediments can also be a source to surface waters. Based on the average 226 Ra content in sediments from Amazon floodplain lakes (2.09 6 1.55 dpm g 21 ; Sanders et al. 2017) and using the empirical equation to relate 226 Ra activity in sediments with 222 Rn diffusion (J diffusion 5 495. 226 Ra sed 1 18.2; see Burnett et al. 2003), we estimate a 222 Rn diffusion rate of 1053 dpm m 2 d 21 across the sediment interface. Water level data for the black, clear, and white rivers (no depth data for lakes) were estimated to be 16 m, 5.5 m, and 12 m, respectively (Supporting Information Fig. S2). Therefore, assuming homogeneous depth, the contribution of 222 Rn diffusion from sediments can sustain maximum river 222 Rn concentrations of 365 dpm m 23 , 1060 dpm m 23 , and 487 dpm m 23 , respectively. While the sediment 222 Ra content was determined from four sites within the Amazon basin, thus placing considerable uncertainty in our estimates, they indicate that the contribution of diffusion from sediments to surface water 222 Rn concentrations were 14%, 12%, and 30% in the black, white, and clear rivers, respectively. Therefore, most of the radon observed in the rivers seems to be sourced from advective pore-water or groundwater pathways.
To our knowledge, only one previous study has documented 222 Rn concentrations in the Amazon. Devol et al. (1987) reported a similar span in 222 Rn values (1400-9240 dpm m 23 ) from eight sites along a 1700 km transect of the Amazon River mainstream and at the mouths of seven tributaries during rising waters (February-March). Samples taken at the mouth of the Rio Negro (2050 dpm m 23 ) and Madeira (4450 dpm m 23 ) are within ranges observed in this study, however, the high-resolution measurements obtained here illustrate the high-spatial variability of 222 Rn that can occur at smaller scales, reflecting the heterogeneous nature of pore-water exchange with surface waters. The general trend of higher 222 Rn concentrations upstream from river mouths suggests greater pore-water influence on surface-water chemistry in these locations which is consistent with other studies in rivers and wetlands (Cook et al. 2003;Santos and Eyre 2011). While evasion could explain the reduced 222 Rn concentrations downstream, the observed trend cannot simply be explained by degassing and suggests 222 Rn inputs along the rivers sampled ( Fig. 1f-h; Supporting Information Fig.  S3). Adjoining lakes appeared to be subject to increased pore-water influences. Furthermore, river segments near igarap es had distinctly higher 222 Rn, suggesting these channels may drain surrounding soils. The hypothesized enrichment of 222 Rn in narrower and steeper-banked igarap es may be due to the larger sediment surface area relative to the overlying water and/or the expected increase in hydrostatic pressure with the surrounding water table during receding waters (i.e., increased pore-water discharge).
While pore-water inputs may be small relative to surfacewater processes, the significant positive relationship observed between pCO 2 and CH 4 with 222 Rn in the rivers and lakes of the black and clear waters suggests a common source (Fig. 2a,b,d,e). Other studies have used 222 Rn to suggest that groundwater is a significant source of CO 2 and CH 4 to surface waters (Atkins et al. 2017;Webb et al. 2017). Based on water flow through rates, Richey et al. (2002) estimated that CO 2 derived from soil respiration is exported to streams via the lateral flow of groundwater and could account for 25% of evasion from the waters of the central Amazonian basin. Using the y-intercept of the pCO 2 -222 Rn linear regression in the black river and lakes (Fig. 2a) and the average pCO 2 of each (Table 1), we find similar values. Average pCO 2 would be 21% lower than observed in the black river and 23% lower in the black lakes if there were no recent pore-water inputs ( 222 Rn approaching zero). In the clear river and lake, average pCO 2 may be 18% and 47% lower, respectively. This implies that while other sources contribute to CO 2 supersaturation, pore-water may be a relevant source of CO 2 in these receding waters. No significant relationships were observed in the white river which may be due to the limited spatial extent of the river studied. In addition to pore-water, primary production and respiration may also exert a strong control on pCO 2 in the Amazon. The significant inverse relationship observed between pCO 2 and Chl a in the black river, black lakes, and clear river (Fig. 2c,f), indicates primary production is an important controller of pCO 2 which is consistent with the recent findings of Amaral et al. (2018). This is particularly evident in black lakes where the concentrations of Chl a were considerably higher (relative to the black river, Table 1) and may explain the lower pCO 2 and weaker (albeit still significant) pCO 2 -222 Rn relationship (Fig. 2a).
Although wetlands are becoming increasingly recognized as an important source of CH 4 to adjoining rivers (Devol et al. 1990;Borges et al. 2015), Sawakuchi et al. (2014) suggested wetland-sourced CH 4 may not be as relevant during their study based on higher CH 4 fluxes during low waters vs. high waters in Amazonian Rivers. The observed CH 4 -222 Rn relationship (Fig. 2b,e) supports the hypothesis that porewater may be an important mechanism in driving riverine CH 4 dynamics (see review by Stanley et al. 2016) and may explain the decoupled wetland-river connectivity observed in Sawakuchi et al. 2014. Similarly to the tidal pump concept (see Stieglitz et al. 2013;Call et al. 2015), where surface water infiltrates sediments during incoming tides (rising waters) and then returns to surface waters during outgoing tides (receding waters), we hypothesize that such a process may be occurring at seasonal scales (as opposed to diurnal/ semi-diurnal tidal pumping) in the black and clear waters sampled during this study. Such a concept would result in a trend of increasing CH 4 concentrations as the water levels transitioned from high to low which was observed in white water rivers and floodplain lakes by Barbosa et al. (2016). Sawakuchi et al. (2014) and Barbosa et al. (2016) suggest that dilution and higher rates of CH 4 oxidation during high water may also explain higher concentrations during the low water period. Clearly, further seasonal studies on CH 4 concentrations in the Amazon basin are required to determine the main drivers of riverine CH 4 dynamics.

Conclusion
This study presents coupled, high resolution spatial measurements of pCO 2 , CH 4 , and 222 Rn of the major tributaries of the central Amazon basin on a scale of 100 km. Relationships suggest that pore-water may be a relevant source of CO 2 and CH 4 to the receding black and clear water tributaries of the central Amazon Basin. Igarap es appear to be sources of dissolved CO 2 and CH 4 to the main channels and we hypothesize that a portion of this CO 2 and CH 4 may be derived from draining surrounding soils. While this initial study cannot quantify pore-water exchange rates, it provides a basis for more extensive, quantitative studies on the role of pore-water in the Amazonian carbon cycle.