Effect of small water retention structures on diffusive CO₂ and CH₄ emissions along a highly impounded river

ABSTRACT

The impoundment of running waters through the construction of large dams is recognised as one of the most important factors determining the transport, transformation, and outgassing of carbon (C) in fluvial networks. However, the effects of small and very small water retention structures (SWRS) on the magnitude and spatiotemporal patterns of C emissions are still unknown, even though SWRS are the most common type of water retention structure causing river fragmentation worldwide. Here we evaluated and compared diffusive carbon dioxide (CO₂) and methane (CH₄) emissions from river sections impounded by SWRS and from their adjacent free-flowing sections along a highly impounded river. Emissions from impounded river sections (mean [SE] = 17.7 [2.8] and 0.67 [0.14] mmol m⁻² d⁻¹, for CO₂ and CH₄, respectively) never exceeded those from their adjacent free-flowing river sections (230.6 [49.7] and 2.14 [0.54] mmol m⁻² d⁻¹). We attribute this finding to the reduced turbulence in impounded river sections induced by SWRS compared to free-flowing river sections (i.e., physical driver). Likewise, the presence of SWRS favoured an increase of the concentration of CH₄ in impounded waters, but this increase was not sufficient to cause a significant influence in the CH₄ efflux from the downstream free-flowing river sections. By contrast, this influenced the larger-scale longitudinal patterns of dissolved CH₄, which exhibited a clear shifting pattern along the study stretch, modulated by variables associated with the presence of SWRS, such as higher water residence times, higher sedimentation rates, and higher temperatures. Overall, our results show that the presence of SWRS can modify the concentrations of C gases in highly impounded rivers but exerts a minor influence on diffusive C emissions.

KEYWORDS:

• carbon dioxide
• dam
• discontinuum
• fluvial network
• fragmentation
• greenhouse gas emissions
• impoundment
• methane
• regulation

Introduction

Inland waters are active components of the global carbon (C) cycle that transform, store, and outgas more than half of the C they receive from terrestrial ecosystems (Cole et al. 2007, Battin et al. 2009a, Tranvik et al. 2009, Aufdenkampe et al. 2011). According to recent global estimates, fluvial networks outgas 2.1 PgC y⁻¹ in the form of carbon dioxide (CO₂), of which 1.8 PgC y⁻¹ is emitted from streams and rivers (Raymond et al. 2013) and 0.32 PgC y⁻¹ from lakes and reservoirs (Raymond et al. 2013). Similarly, fluvial networks outgas 0.09 PgC y⁻¹ in the form of methane (CH₄), of which 0.02 PgC y⁻¹ is emitted from streams and rivers (Stanley et al. 2016) and 0.07 PgC y⁻¹ from lakes and reservoirs (Bastviken et al. 2011, Deemer et al. 2016). Despite the increasing availability of data covering both spatial and temporal variability, fundamental uncertainties persist regarding the magnitude and spatiotemporal patterns of C emissions along highly impounded rivers where lotic and lentic sections interact in series.

Because of the high human demand for energy and water, many fluvial networks worldwide have been regulated with a variety of hydraulic structures ranging from large dams to smaller reservoirs, impoundments, and weirs (Nilsson et al. 2005, Döll et al. 2009). Thus, the natural condition of many fluvial networks as uninterrupted water conduits draining water from the catchments to the ocean (Vannote et al. 1980, Frissell et al. 1986) has been replaced by an alternating series of lotic and lentic reaches (Ward and Stanford 1983, Stanford and Ward 2001). Disruption of the water flow by dams changes the hydrological dynamics of rivers (Kondolf and Batalla 2005, Grill et al. 2015), thereby altering the water physicochemistry (Ward and Stanford 1983) and the transport of suspended particles (Syvitski et al. 2005, Houser et al. 2010), ultimately leading to multiple consequences for the structure and function of aquatic biological communities (Haxton and Findlay 2008, Ponsatí et al. 2015, von Schiller et al. 2015, Proia et al. 2016).

The increase in water residence time caused by river impoundment favours the mineralization of organic matter (OM) and CO₂ supersaturation by increasing the interaction between organic substrates and biological actors (Battin et al. 2009b, Acuña and Tockner 2010, Crawford et al. 2016). Moreover, sediments, OM, and nutrients tend to accumulate in the lentic waterbodies upstream from water retention structures because of the high deposition rates typically occurring in these types of aquatic ecosystems (Syvitski et al. 2005, Maeck et al. 2013). The trapping of OM often leads to the development of anaerobic sedimentary environments where active methanogenesis takes place. This enchancing process leads to subsequent CH₄ supersaturation of the water column and release of CH₄ to the atmosphere (Keller and Stallard 1994, Giles 2006, Delsontro et al. 2010).

Most quantitative estimates of C emissions from impounded waters have been obtained for very large (>10⁴ km²), large (10⁴–10² km²), and medium-sized (100–1 km²) reservoirs (St. Louis et al. 2000, Barros et al. 2011, Deemer et al. 2016). These estimates indicate that the depositional zones close to the river inflow, where fresh OM enters the reservoir (i.e., the riverine–lacustrine transition zone; Wetzel 2001), are active compartments in terms of CO₂ and CH₄ production and emission, despite their relatively small areal coverage (Beaulieu et al. 2016). In smaller impoundments (<1 km²), the riverine–lacustrine transition zone can cover a larger fraction of the reservoir’s surface area and still host intense CO₂ and CH₄ production, making these systems potential “hot spots” for C emissions to the atmosphere (Maeck et al. 2013).

Small water retention structures (SWRS) include small to very small impoundments with an impounded area <0.1 km² and a volume <0.2 hm³ (Lehner et al. 2011). These systems, despite their small global areal extent (3.8% of the global reservoir surface area; Downing et al. 2006, Lehner et al. 2011), represent one of the most common features in freshwater landscapes (99.5% of the total number of reservoirs worldwide; Downing et al. 2006, Lehner et al. 2011). They remain overlooked, however, or even ignored in most biogeochemical studies in fluvial networks (Downing et al. 2006). SWRS are also of increasing concern because their cumulative effects may be considerable (Downing et al. 2006). For example, Harrison et al. (2009) showed that SWRS play an important role in the regional and global removal of nitrogen from surface water. Casas-Ruiz et al. (2015) showed that the presence of SWRS generates changes in the composition and degradation of dissolved OM in rivers. Similarly, Proia et al. (2016) found that the presence of SWRS modifies the natural dynamics of C processing along the river network by altering the structure and activity of microbial communities. These findings lead to the intriguing question of whether SWRS can affect the magnitude and patterns of diffusive CO₂ and CH₄ emissions as larger impoundments do.

Here we investigated the effect of SWRS on the concentrations and diffusive emissions of C gases in an intensively impounded stretch of a Mediterranean river. To this end, we measured the dissolved concentrations and diffusive emissions of CO₂ and CH₄ in impounded and free-flowing riverine sections on a seasonal basis. We hypothesised that the presence of SWRS would lead to higher water residence times and higher OM accumulation, thus enhancing aerobic heterotrophic respiration and anaerobic methanogenesis in both the water column and the sediments of impounded waters. We therefore predicted high concentrations and emission rates of CO₂ and CH₄ in impounded river sections associated with SWRS compared to their upstream free-flowing sections. We also hypothesised that the CO₂ and CH₄ accumulated in impounded river sections would enhance emissions from the free-flowing sections located downstream from the dams. Thus, we expected higher CO₂ and CH₄ emissions from free-flowing riverine sections located downstream than from those located upstream of SWRS.

Methods

Study site and sampling design

The Fluvià River (northeastern Iberian Peninsula) is 97 km long with a mean annual flow of 3.6 m³ s⁻¹ at the catchment outlet (data from 2004 to 2014, Catalan Water Agency; http://aca.gencat.cat/ca/inici). The Fluvià River catchment (990 km²) is mostly calcareous, with some areas (<15%) of siliceous materials (Cartographic and Geological Institute of Catalonia, 2006, http://www.icc.cat) and covered with mixed forests (78%), agricultural (19%), and urban (3%) areas (Land Cover Map of Catalonia, Centre of Ecology and Forestry Research of Catalonia, 2009, http://www.creaf.cat). The climate is typically Mediterranean. The mean monthly air temperature ranges from 6 °C in January to 26 °C in July, and the mean annual precipitation is 660 mm, with rainfall primarily occurring in autumn and spring and occasional storms in summer (data from 2004 to 2014, Catalan Water Agency). The water flow of the Fluvià River has been deeply modified due to the high human demand for energy and water (García-Ruiz et al. 2011). Its fluvial network presents up to 61 SWRS that alter the flow dynamics from its headwaters to the river mouth (Pavón 2010).

This study focused on a highly impounded 36 km stretch situated in the upper part of the Fluvià river main stem (Fig. 1a). To cover a wide spectrum of hydromorphological and trophic conditions (Table 1), we selected 11 SWRS to sample in spring (28–30 Apr), summer (2–4 Sep), and winter (9–11 Dec) 2014. At each site, we sampled the stagnant water impounded in the SWRS and the free-flowing riverine sections located upstream and downstream (Fig. 1b). The distance between the SWRS and the downstream sampled free-flowing river section was between 20 and 380  m.

Figure 1. Location of (a) the Fluvià River catchment (northeastern Iberian Peninsula), with the position of the studied small water retention structures (SWRS; black circles, n = 11). See Table 1 for a detailed hydromorphological and physicochemical description of the waters impounded in the studied SWRS. (b) Scheme of a SWRS sampling unit (i.e., upstream free-flowing river, impounded water, and downstream free-flowing river) sampled at each study site.

Table 1. Hydromorphological and physicochemical characteristics of the 11 studied impoundments. DOC = dissolved organic carbon, TDN = total dissolved nitrogen.

	Hydromorphological characteristics				Physicochemical characteristics
Impoundment	Surface area (ha)^a	Volume (hm^3)^a	Mean depth (m)^a	Water residence time (h)^b	Temperature (°C)^b	Conductivity (µS cm^-1)^b	DOC (mg L^-1)^b	TDN (mg L^-1)^b	Chlorophyll a (µg L^-1)^b
1	0.8	0.05	1.5	10.0 (6.9–13.2)	13.4 (7.8–18.4)	546 (480–623)	2.9 (2.1–4.0)	3.4 (1.4–4.8)	1.5 (1.0–2.1)
2	0.2	0.01	1.1	0.6 (0.4–0.9)	9.9 (5.8–19.3)	550 (477–622)	2.8 (1.7–3.3)	3.4 (1.5–5.1)	1.9 (0.3–3.4)
3	0.1	0.03	1.1	0.3 (0.2–0.4)	13.7 (6.7–20.3)	543 (472- 627)	2.2 (1.4–2.8)	3.6 (1.9–4.9)	2.1 (0.2–3.5)
4	0.2	0.03	1.3	1.3 (1.0–1.4)	14.9 (7.3–22.2)	751 (653–854)	2.2 (1.3–2.7)	3.8 (1.9–5.1)	3.3 (0.1–5.5)
5	0.3	0.02	1.3	1.5 (0.9–2.0)	14.5 (7.6–21.6)	725 (663–762)	2.4 (1.5–2.9)	3.6 (1.9–4.7)	2.1 (0.3–4.5)
6	0.2	0.02	0.8	0.6 (0.5–0.7)	15.9 (7.9–23.9)	813 (784–833)	2.1 (1.6–2.8)	3.0 (1.8–4.2)	1.8 (0.2–3.4)
7	0.2	0.03	0.7	0.8 (0.3–1.6)	17.2 (9.4–24.9)	882 (821–946)	2.1 (1.7–2.7)	2.9 (1.8–3.9)	2.3 (1.6–3.0)
8	1.9	0.04	1.8	10.0 (8.2–13.6)	21.2 (10.3–25)	933 (891–975)	2.0 (1.6–2.5)	2.8 (1.7–4.0)	4.0 (2.3–6.2)
9	1.6	0.02	1.3	4.3 (2.6–5.2)	16.7 (9.3–23.6)	1038 (915–1170)	2.5 (2.0–3.1)	1.6 (1.4–1.8)	21.9 (1.2–48.1)
10	2.7	0.01	0.6	3.2 (2.8–3.4)	12.9 (7.9–17.9)	682 (678–685)	2.0 (1.6–2.6)	2.3 (1.5–3.4)	6.1 (0.9–12.2)
11	9.6	0.12	0.7	40.8 (37.0–45.5)	13.2 (6.4–20.0)	575 (516–634)	2.4 (2.0–2.7)	2.1 (1.1–3.2)	11.4 (0.6–27.6)

^aMean of the 3 sampling dates. The temporal variation was negligible for these parameters

^bMean and range (in brackets) of the 3 sampling dates

Determination of diffusive CO₂ and CH₄ fluxes

In the impounded river sections, we determined the CO₂ flux across the water–air interface by the enclosed chamber method (Frankignoulle 1988). Briefly, we monitored the CO₂ gas concentration in an opaque floating chamber every 4.8 s with an infrared gas analyser (EGM-4, PP-Systems, USA). In all cases, flux measurements lasted until a change in CO₂ of at least 10 µatm was reached, with a maximum duration of 600 s and a minimum of 300 s. We calculated the CO₂ flux from the rate of change of CO₂ inside the chamber as follows:(1) where is the change in CO₂ concentration in the chamber along time in µatm s⁻¹, V and S are the volume and surface area of the chamber (27.1 dm³ and 19.4 dm², respectively), T is the air temperature in Kelvin, and R is the ideal gas constant (L atm K⁻¹ mol⁻¹). Positive CO₂ flux values represent efflux of gas to the atmosphere while negative CO₂ flux indicate influx of gas from the atmosphere. We performed 2 measurements in the central part of the impounded water after flushing the chamber with ambient air between consecutive measurements.

At each SWRS, we determined the partial pressure of CO₂ and CH₄ in the water (, ) and in air (, ) at the same location as we took the flux measurements. We measured the with an infrared gas analyser (EGM-4, PP-Systems, USA) coupled to a membrane contactor (MiniModule, Liqui-Cel, USA). The water was circulated by gravity through the contactor at 300 mL min⁻¹, and the equilibrated gas was continuously recirculated into the infrared gas analyser for instantaneous measurements (Teodoru et al. 2010). Measurement accuracy of the infrared gas analyser is estimated to be within 1% over the calibrated range. We determined the by the headspace equilibrium technique and gas chromatography described by Striegl et al. (2012). Briefly, we collected 40 mL of water with a 60 mL polypropylene syringe creating a headspace with ambient air of 3:2 ratio (water sample to ambient air). To facilitate the kinetics of equilibration between the liquid and the gas phase, we vigorously shook the syringe for 1  min and then submerged it for 30  min at constant water temperature. We then transferred the 20 mL of equilibrated gas to a preevacuated gas-tight glass tube (Exetainers, Labco Limited, UK). The CH₄ samples were analysed in the laboratory with a gas chromatograph coupled to a Flame Ionization Detector (Trace GC Ultra, Thermo Fisher Scientific, USA). Measurement accuracy of the gas chromatograph is estimated to be within 4% over the calibrated range. For and , the air samples were taken ∼1  m above the water surface layer and directly analysed with the infrared gas analyser and the gas chromatographer, respectively.

We used the CO₂ flux measured with the chamber to derive the gas transfer velocity of CO₂ () from Fick’s law of gas diffusion:(2) where is the specific gas transfer velocity for CO₂ (m d⁻¹), CO₂ flux is the chamber-measured CO₂ flux between the surface water and the atmosphere (mmol m⁻² d⁻¹), and is the Henry’s constant (mmol µatm⁻¹ m⁻³) adjusted for salinity and temperature (Weiss 1974, Millero 1995). Because the gas transfer velocity is temperature and gas dependent, we standardised to a Schmidt number of 600 (; m d⁻¹), which corresponds to CO₂ at 20 °C in freshwater:(3) where Sc is the Schmidt number of a given gas at a given water temperature (Jähne and Münnich 1987, Wanninkhof 1992, Bade 2009).

At each SWRS, we calculated the diffusive CH₄ flux across the water–air interface (mmol m⁻² d⁻¹) following equation 2(2) and combining , , and the chamber-derived obtained by applying equation 3(3) for CH₄.

In the free-flowing riverine sections upstream and downstream the SWRS, we determined the diffusive CO₂ and CH₄ flux across the water–air interface (mmol m⁻² d⁻¹) using equation 2(2) . At each river section, we determined , , , and following the same technique described above, and we estimated the river section gas transfer velocity from the section slope (s; m m⁻¹) and the mean section water velocity (v; m s⁻¹) with equation 4(4) in Raymond et al. (2012):(4) The was transformed to and following equation 3(3) . Note that the applicability of indirect and in the free-flowing riverine sections based on stream morphometric and hydrologic variables (Raymond et al. 2012) has previously been tested and validated in the same river network by Gómez-Gener et al. (2016).

Statistical analyses

We investigated the overall effect of the SWRS on the CO₂ and CH₄ fluxes, concentrations, and gas transfer velocities in each season by using a one-way repeated measures analysis of variance (ANOVA). We performed subsequent post hoc comparisons (Tukey’s honest significant differences test) to evaluate the specific effect of SWRS on (1) the impounded water CO₂ and CH₄ fluxes, concentrations, and gas transfer velocities by comparing impounded waters and free-flowing upstream reaches, and (2) the downstream C fluxes, concentrations, and gas transfer velocities by comparing upstream and downstream free-flowing river sections adjacent to the SWRS.

We assessed the relative influence of and on the CO₂ efflux using simple linear regression models. We used the same statistical approach to assess the relative influence of and on the CH₄ efflux.

We built 2 partial-least squares regression models (PLS) to identify the potential drivers of impounded water and . PLS is a regression extension of a principal component analysis (PCA) that allows the exploration of relationships between multiple and collinear independent (X; potential explanatory variables listed and described in Supplemental Table S1) and dependent (Y; and ) data matrices. This linear, multivariate model produces latent variables (i.e.; PLS loadings) representing the combination of X-variables that best describe the distribution of observations in a Y space (Wold et al. 2001). For each Y-variable, the best PLS model was selected by iteratively removing X-variables to maximise the goodness of fit (R₂Y) and the predictive ability (Q₂Y) of the model. In each case, Q₂Y was determined by comparing modelled and actual Y observations through an iterative, cross-validation process. We identified the influence of each X-variable by using variable influence on the projection (VIP) scores, calculated as the sum of square of the PLS weights across all components. Variables with high influence on impounded water and were identified as VIP > 1, variables with moderate influence as VIP between 1 and 8, and low influence as VIP < 0.8 (Wold et al. 2001).

All statistical analyses were conducted in the R statistical environment (R Core Team 2013) using the vegan package (Oksanen et al. 2015), except for PLS, which was analysed with the software XLSAT (XLSTAT 2015.2.01, Addinsoft SRAL, Germany). When necessary, we log-transformed data to meet parametric assumptions. In all cases, differences were considered statistically significant when p < 0.05.

Results

Diffusive CO₂ and CH₄ emissions

The surface waters impounded in the 11 SWRS were net emitters of CO₂ (mean [SE] = 17.7 [2.8] mmol m⁻² d⁻¹, range = −2.1 to 86.2 mmol m⁻² d⁻¹; Fig. 2a) and CH₄ (0.67 [0.14] mmol m⁻² d⁻¹, range = 0.01–3.29 mmol m⁻² d⁻¹; Fig. 2b) to the atmosphere in the 3 studied seasons. The only exception to this pattern occurred at site N5, which acted as a sink of CO₂ in winter (−2.1 mmol m⁻² d⁻¹). Nonetheless, the CO₂ and CH₄ effluxes from the impounded river sections never exceeded those from their adjacent free-flowing river sections, which were also always net emitters of both C gases to the atmosphere (Fig. 2a–b). More precisely, the efflux from impounded waters was significantly lower than their upstream and downstream free-flowing river sections in the case of CO₂ (ANOVA, F = 10.26, p < 0.001, n = 33; Fig. 2a) but comparable in the case of CH₄ (ANOVA, F = 4.82, p = 0.23, n = 33; Fig. 2b).

Figure 2. Mean (a) CO₂ efflux, (b) CH₄ efflux and CH₄ efflux expressed as CO₂-equivalents (CO₂-equivalents = CO_2e; 1 g CH₄ = 28 g CO_2e; IPCC 2013), (c) partial pressure of CO₂ in water (), (d) partial pressure of CH₄ in water (), (e) gas transfer velocity of CO₂ (), and (f) gas transfer velocity of CH₄ () of the 11 SWRS grouped by sampling units (i.e., upstream river, impoundment water, and downstream river) during the 3 sampled seasons (i.e., spring, summer, and winter). Error bars represent standard error (SE). Significant differences of reported parameters between SWRS units (p < 0.05, Tukey’s post hoc test after repeated measures ANOVA) are marked with different letters above the bars.

No significant differences among SWRS sampling units (i.e., upstream free-flowing river section, impounded river section, and downstream free-flowing river section) in terms of (ANOVA, F = 0.08, p = 0.92, n = 33; Fig. 2c) were detected. By contrast, we detected significant differences among SWRS sampling units in terms of (ANOVA, F = 7.47, p < 0.001, n = 33; Fig. 2d). Specifically, we detected a significant increase in in the impounded waters in relation to their upstream free-flowing river sections (ANOVA, post hoc, p < 0.001, n = 33; Fig. 2d). In spring, this effect was enhanced and translated into an increase of in the flowing sections downstream the SWRS (ANOVA, post hoc, p = 0.01, n = 33; Fig. 2d). This impounded-waters increase was, however, not sufficient to significantly influence the CH₄ efflux of the downstream riverine sections (ANOVA, post hoc, p = 0.18, n = 33; Fig. 2b).

The surface water turbulence (measured as and ) was significantly lower in the impounded river sections than in free-flowing river sections for both CO₂ (ANOVA, F = 18.36, p < 0.001, n = 33; Fig. 2e) and CH₄ (ANOVA, F = 18.32, p < 0.001, n = 33; Fig. 2f). These 2 parameters govern the diffusive CO₂ and CH₄ emissions (see equation 2(2) ), and the surface water turbulence (measured as and ) showed a significant positive relationship with both the CO₂ efflux (CO₂ efflux = 31.1 + 46.2; r² = 0.61, p < 0.001, n = 33; Supplemental Fig. S1a) and the CH₄ efflux (CH₄ efflux = 31.1  + 46.2; r² = 0.61, p < 0.001, n = 33; Supplemental Fig. S1b).

Longitudinal patterns of and

Distinct longitudinal patterns along the studied 36 km stretch were identified for and (Fig. 3). The showed a relatively smooth downstream pattern during the 3 investigated seasons (Fig. 3a–c). By contrast, the presence of SWRS not only influenced the dynamics of at the local scale (discussed earlier) but also led to abrupt fluctuations in the longitudinal pattern (Fig. 3a–c). Although this longitudinal pattern was consistent in the 3 studied seasons, a more marked fluctuating pattern was observed during winter (Fig. 3f).

Figure 3. Downstream longitudinal patterns of the partial pressure of CO₂ in water () in (a) spring, (b) summer, and (c) winter and of the partial pressure of CH₄ in water () in (d) spring, (e) summer, and (f) winter across the 11 small water retention structure (SWRS) units (i.e., upstream river, impoundment water, and downstream river). The continuous solid lines represent the mean impoundment water or . Water flow direction goes from left to right.

Drivers of impounded water and

The PLS models identified distinct combinations of factors as predictors of impounded water and (Fig. 4a, Supplemental Table S2). The PLS model for extracted 2 components from the data matrix that explained 68% of the variance (Fig. 4a, Supplemental Table S2). The most influential factors (VIP > 1) on were surface area, water residence time, , electrical conductivity, and alkalinity, while the moderately influential factors (VIP between 0.8 and 1) on were surface water oxygen saturation, dissolved organic carbon (DOC) concentration, and pH (see the direction and the relative strength of the effect of each factor on the in Supplemental Table S2).

Figure 4. Loadings plot of the partial least squares (PLS) regression analysis for (a) partial pressure of CO₂ in water () and (b) partial pressure of CH₄ in water (). The graph shows how the Y-variable (squares) correlates with X-variables (circles) and the correlation structure. X-variables are classified according to their variable influence on projection value (VIP): highly influential (black circles), moderately influential (grey circles), and less influential (white circles). The X-variables situated near Y-variables are positively correlated to them and those situated on the opposite side are negatively correlated (see Supplemental Tables S1 and S2 for explanation of abbreviations and summary of PLS models, respectively).

The PLS model for extracted 2 components from the data matrix that explained 69% of the variance (Fig. 4b, Supplemental Table S2). The most influential factors (VIP > 1) on were surface area, water residence time, total dissolved nitrogen concentration, DOC concentration, surface water temperature, and concentration of Chl-a in suspension (see the direction and the relative strength of the effect of each factor on the in the Supplemental Table S2); no moderately influential factors were obtained from the PLS model for .

Discussion

Low diffusive CO₂ and CH₄ emissions from impounded waters

Disruption of water flow by retention structures changes the hydrological dynamics of rivers, leading to higher water residence times and higher OM deposition rates (Syvitski et al. 2005, Acuña and Tockner 2010, Maeck et al. 2013). The intense accumulation of OM in lacustrine sediments typically boosts aerobic and anaerobic heterotrophic processes (Maeck et al. 2013), consequently favouring the enrichment of surface waters in CO₂ and CH₄ and ultimately leading to high C emissions to the atmosphere (Cole et al. 2000, St. Louis et al. 2000, Duarte and Prairie 2005, Giles 2006). Contrary to our expectations, however, the diffusive CO₂ and CH₄ emissions from impounded waters in our study never exceeded those from adjacent upstream free-flowing riverine waters (treated as reference systems without SWRS influence). We attribute this finding to the reduced turbulence in impounded river sections induced by SWRS compared to free-flowing river sections. Our results agree with previous findings that emphasised the importance of the gas transfer velocity as a major driver of diffusive CO₂ and CH₄ emissions from fluvial networks encompassing lakes and reservoirs (Guérin et al. 2007, Lundin et al. 2013, Crawford et al. 2014, Gómez-Gener et al. 2015).

Note that the total C emitted to the atmosphere and associated atmospheric effect from the studied sections is mainly dominated by the CO₂ form (Fig. 2a–b, principal y-axis). The ∼30-fold higher global warming potential of CH₄ compared to that of CO₂ over a 100-year time horizon (IPCC 2013) may, however, substantially balance the atmospheric contribution of both C forms (Fig. 2b, secondary y-axis). Therefore, the results from this study also emphasise the importance of including CH₄ fluxes when aiming to understand the effects of the global warming on aquatic–atmospheric interactions in highly impounded rivers.

For CH₄, we must consider that our estimates in the studied impounded waters did not account for the ebullitive efflux (i.e., CH₄ bubbling). The ebullitive CH₄ efflux has been described as a major pathway for CH₄ efflux in very large, large, and medium-sized reservoirs (>1 km²; Fearnside and Pueyo 2012, Sobek et al. 2012, Beaulieu et al. 2016, Deemer et al. 2016), small impounded systems (<1 km²; Maeck et al. 2013), and very small systems previously examined in the same river (<0.1 km²; Gómez-Gener et al. 2015). Specifically, Gómez-Gener et al. (2015) showed that the ebullitive CH₄ efflux from 5 impounded systems in the same river network substantially contributed to the total CH₄ efflux to the atmosphere. Therefore, based on their results, we may expect that the inclusion of ebullition estimates (e.g., using inverted funnel-style bubble traps) in our sampling design would probably have led to higher total CH₄ emission rates in the studied impounded waters compared to the free-flowing river sections.

Low effect of SWRS on downstream diffusive CO₂ and CH₄ emissions

Flow regulation by dams not only affects C emissions from the surface waters of reservoirs but may also influence the C emissions in the riverine sections downstream from the reservoir (Guérin et al. 2006, Kemenes et al. 2007). Emissions downstream of reservoirs have generally been poorly studied and are usually not taken into account in reservoir C balances (Barros et al. 2011). The few works reporting C emissions downstream of reservoirs, mainly conducted in large reservoirs situated in the tropic region, showed that downstream C emissions can account for 7–25% of total CO₂ emissions from reservoirs (Abril 2005, Guérin et al. 2006) and 50–90% of total CH₄ emissions (Abril 2005, Guérin et al. 2006, Kemenes et al. 2007).

In the present study, we found no significant differences in CO₂ or CH₄ effluxes between free-flowing river sections located upstream and downstream of SWRS, even though significant enrichment in CH₄ in the impounded surface water was evident. We propose that this finding may be at least partially due to the loss of CO₂ and CH₄ along the distance between the SWRS and the downstream sampling stations. We acknowledge that although the distance between the SWRS and the downstream sampling stations was relatively short (never exceeding 380  m), a nonquantifiable part of CO₂ and CH₄ could have been emitted along this distance (e.g., in the waterfall associated with the SWRS; Galy-Lacaux et al. 1997, Wehrli 2013, Deshmukh et al. 2015). Therefore, this result suggests the need to reduce the distance with the downstream sampling station and to include CO₂ and CH₄ emissions from the waterfall and turbine in this type of quantification.

Routing vs. local controls on the longitudinal patterns of dissolved CO₂ and CH₄ concentrations

Conceptual models of longitudinal patterns in streams and rivers are framed by the interplay of 2 contrasting perspectives. Some frameworks emphasize flow as an integrator over space (Vannote et al. 1980, Montgomery 1999; i.e., routing control), but other frameworks highlight patchiness and abrupt spatial changes (Frissell et al. 1986, Thorp et al. 2006, Winemiller et al. 2010; i.e., local control). These models are based on gradient analyses in which rivers are ultimately viewed as uninterrupted continua; however, few fluvial networks remain free from flow regulation over the entire course but, contrarily, typically result in an alternating series of lentic and lotic reaches (Ward and Stanford 1983, Nilsson et al. 2005, Döll et al. 2009).

Here we found that the downstream longitudinal patterns of dissolved C gases in water differed considerably depending on the gas species. CO₂ concentrations varied moderately and homogeneously both at the local scale (i.e., across SWRS sampling units) and at the river network scale (i.e., along the 36 km study stretch). This pattern suggests that although local controls may influence CO₂ concentrations in river water, the longitudinal dynamics of this gas have a routing control (Montgomery 1999) where the CO₂ lost in riverine sections is constantly replenished by CO₂ produced internally or imported from the catchment (Hotchkiss et al. 2015, Gómez Gener et al. 2016). By contrast, CH₄ concentrations showed a more variable pattern both at local and river network scales, and this shifting longitudinal pattern suggests that the concentration of CH₄ is highly dominated by local controls such as the presence of SWRS (Stanley et al. 2016). Likewise, the difference in between impounded and riverine sections also highlights the high reactivity and dynamism of this gas, which within a few meters can be generated and rapidly emitted or consumed via oxidation (Guérin et al. 2006, Kemenes et al. 2007).

Regulation of dissolved CO₂ and CH₄ concentrations from impounded waters

Surface water supersaturation of dissolved CO₂ and CH₄ in impounded waters was persistent along the studied stretch and across seasons. Nonetheless, the results from the PLS models suggest that the drivers of gas supersaturation in SWRS differ between CO₂ and CH₄.

CO₂ supersaturation in impounded waters was explained by a combination of factors related to internal metabolic processes (i.e., oxygen saturation level, dissolved organic carbon, water residence time, and system surface area) and factors that may be related to hydrological inputs of CO₂ from soil respiration and mineral weathering within the catchment (i.e., surface water alkalinity, pH, and electrical conductivity). These results support the idea that direct terrestrial or geological origin of river (Stets et al. 2009, McDonald et al. 2013, Hotchkiss et al. 2015, Marcé et al. 2015) does not contradict the widespread notion of net heterotrophy leading to CO₂ supersaturation in different aquatic ecosystems (Duarte and Prairie 2005, Lapierre et al. 2013). In the case of , our results suggest that the presence of SWRS, despite their relatively small water retention capacity, can still produce effects on impounded water similar to those previously found in larger impoundments (Delsontro et al. 2010, Maeck et al. 2013, Beaulieu et al. 2016), for example by enhancing anaerobic processes in the sediments (Maeck et al. 2013, Crawford and Stanley 2016).

Conclusions and implications

Because of the attenuated turbulence induced by SWRS, the diffusive CO₂ emission from impounded river sections was significantly lower than that from free-flowing river sections. Contrarily, no reduction in diffusive CH₄ emissions associated with the presence of SWRS was detected, likely because the significant enrichment in CH₄ at the impounded river sections compared to the free-flowing river sections compensated the physical effect on diffusive CH₄ emissions. The significant enrichment in CH₄ detected in the impounded waters was not sufficient, however, to significantly increase the diffusive CH₄ efflux from the downstream free-flowing riverine sections. Nonetheless, it influenced the overall longitudinal patterns of CH₄ concentration along the 36 km study stretch. These longitudinal patterns of CH₄ concentration along the study stretch, which were driven by local factors associated with the presence of SWRS, emphasise that viewing a river as a continuum is insufficient to describe the true spatial complexity of C gas fluxes in fluvial networks.

Acknowledgements

This research was funded by the Olot City Council through an Oriol de Bolòs research grant to BO and was supported by the Spanish Ministry of Economy and Competitiveness (FUNSTREAM CGL2014-58760-C3-1R). LG was supported by an FPI predoctoral grants (BES-2012-059743). We thank Silvia de Castro for field and laboratory assistance.

ORCID

Lluís Gómez-Gener http://orcid.org/0000-0003-3279-3589