Methane emissions due to reservoir flushing: a significant emission pathway?

Reservoirs represent a globally significant source of the greenhouse gas methane (CH₄), which is emitted via different emission pathways. In some reservoirs, reservoir flushing is employed as a sediment management strategy to counteract growing sediment deposits that threaten reservoir capacity. Reservoir flushing utilizes the eroding force of water currents during water level drawdown to mobilize and transport sediment deposits through the dam outlet into the downstream river. During this process, CH₄ that is stored in the sediment can be released into the water and degas to the atmosphere, resulting in CH₄ emissions. Here, we assess the significance of this CH₄ emission pathway and compare it to other CH₄ emission pathways from reservoirs. We measured seasonal and spatial CH₄ concentrations in the sediment of Schwarzenbach Reservoir, providing one of the largest datasets on CH₄ pore water concentrations in freshwater systems. Based on this dataset we determined CH₄ fluxes from the sediment and estimated potential CH₄ emissions due to reservoir flushing. CH₄ emissions due to one flushing operation can constitute 7 %–14 % of the typical annual CH₄ emissions from Schwarzenbach Reservoir, whereby the amount of released CH₄ depends on the seasonal timing of the flushing operation and can differ by a factor of 2. Larger flushing events that mobilize deeper sediment layers lead to non-linear increases in CH₄ mobilization. This suggests that regular flushing of smaller sediment layers releases less CH₄ than removal of the same sediment volume in fewer flushing events of thicker sediment layers. However, additional indirect CH₄ emissions pathways contributing to the total CH₄ emissions may vary with the flushing operation. In other reservoirs with higher sediment loadings than Schwarzenbach Reservoir, reservoir flushing could cause substantial CH₄ emissions, especially when flushing operations are conducted frequently. Our study recognizes CH₄ emissions due to reservoir flushing as an important pathway, identifies potential management strategies to mitigate these CH₄ emissions and emphasizes the need for further research.

1 Introduction

Worldwide millions of reservoirs have been constructed (Lehner et al., 2011; Couto and Olden, 2018), and their number is projected to increase substantially in the near future (Zarfl et al., 2015) because of rising demand for hydropower. Besides being used for electricity generation and energy storage, reservoirs can serve multiple purposes such as water supply, flood control, irrigation and navigation (WCD (World Commission on Dams), 2000). In the past, hydropower was widely considered a greenhouse gas (GHG)-neutral form of energy (Hoffert et al., 1998). Today, we know that reservoirs represent a significant source of GHG emissions, especially of the potent GHG methane (CH₄), and researchers have estimated that reservoirs contribute around 10–52 Tg CH₄ yr⁻¹ to the global budget of atmospheric CH₄ (Lauerwald et al., 2023).

CH₄ can be emitted from reservoirs via different pathways such as ebullition, plant-mediated transport, diffusion across the water–atmosphere interface, and degassing during turbination and during spring or fall turnover as storage flux (Bastviken et al., 2011). CH₄ is typically produced in the anoxic part of the sediment (Le Mer and Roger, 2001). At oxic interfaces within the sediment, at the sediment surface or within the water column, CH₄ is at least partly oxidized by methane-oxidizing bacteria (Bastviken et al., 2002) before it reaches the water surface, from where it is emitted as diffusive flux to the atmosphere. CH₄ flux via ebullition and plant mediation, on the other hand, can bypass oxic interfaces, thus avoiding oxidation (Chanton and Whiting, 1995). In anoxic deep water of lakes and reservoirs, CH₄ typically accumulates and can reach large concentrations of stored CH₄ (Ragg et al., 2021). Intensive vertical mixing during spring and fall overturn causes rapid transport of the stored CH₄ to the water surface, from where it diffuses to the atmosphere (Bastviken et al., 2004). This storage flux can substantially exceed the annual diffusive emissions during stratified conditions (Encinas Fernández et al., 2014).

While most of these emission pathways usually exist in both lakes and reservoirs, the degassing of CH₄ during turbination occurs only in reservoirs where it can become the dominant source of CH₄ emissions (Kemenes et al., 2007; Harrison et al., 2021), especially if the turbinated water is drawn from anoxic deep water where large amounts of CH₄ are stored. Ebullition of CH₄ is often observed as the dominant emission pathway, especially in shallow reservoirs (DelSontro et al., 2010; Sobek et al., 2012). It is well known that changes in hydrostatic pressure can induce bubble formation and release from sediments (Maeck et al., 2014; Harrison et al., 2017; Encinas Fernández et al., 2020). Therefore, water level drawdowns in reservoirs can substantially increase the ebullition flux of CH₄ due to decreasing pressure, which can become particularly large during fall drawdown or occur regularly during diel pumped-storage operations, as well as hydropeaking and ship lock operations (Maeck et al., 2014; Harrison et al., 2017; Almeida et al., 2020; Encinas Fernández et al., 2020).

Growing sediment deposits in reservoirs, caused by particles introduced from the catchment and organic matter produced within the system, pose a challenge to maintaining reservoir capacity (e.g., by decreasing reservoir storage volume). Globally, reservoir storage capacity is decreasing and is estimated to be completely lost for most reservoirs within 200–300 years if sediment management strategies are not adopted (ICOLD, 2009). Sediment management strategies are applied in reservoirs to reduce sediment yield, route sediments or remove already-deposited sediment (Morris, 2020; Petkovšek et al., 2020). Deposited sediment can be removed by mechanical removal (e.g., dredging) or hydraulic flushing, which utilizes the eroding force of water currents. Drawdown flushing represents a typical form of hydraulic flushing, which consists of three steps: completely drawing down the water level, maintaining a free-flow state and recovering water levels (Kondolf et al., 2014). During the free-flow state, sediment is mobilized and flushed through a dam outlet to the downstream river section. Consequently, any CH₄ previously stored in the sediment pore water is also mobilized into the water stream and can eventually degas to the atmosphere at the water surface, during turbination or downstream of the reservoir, leading to CH₄ emissions. Sediment flushing operations have been conducted worldwide in reservoirs of all sizes, with (initial) storage capacities ranging widely from 0.8×10⁶ to 9640×10⁶ m³ (Sumi et al., 2017; Antoine et al., 2020). Among reservoirs with relatively regular flushing operations, flushing frequencies of between twice a year and once every 5 years have been reported (Brandt and Swenning, 1999; Chang et al., 2003; Kantoush and Sumi, 2010; Fruchard and Camenen, 2012; Grimardias et al., 2017; Sumi et al., 2017; Antoine et al., 2020), but there are also reservoirs that are flushed irregularly (Sumi et al., 2017). Reservoirs with higher sediment yield need to be flushed more frequently than others. For instance, the Tapu Reservoir in Taiwan is characterized by high sedimentation rates induced by heavy rainfall and steeply sloped mountains and was flushed 10 times between 1991 and 1996 (Chang et al., 2003). The total amount of flushed sediment depends on reservoir geometry, sediment characteristics and operation strategy (Morris, 2020; Petkovšek et al., 2020), and the conditions for a successful flushing operation are typically optimal in smaller reservoirs that are long and narrow (Kondolf et al., 2014).

In this study, we investigated CH₄ emissions due to reservoir flushing, a pathway that has not yet been included in estimates of CH₄ emissions from reservoirs. We determined the amount of CH₄ stored in the sediment pore water of a reservoir in different seasons and estimated potential CH₄ emissions resulting from reservoir flushing scenarios. Furthermore, we assessed the relative importance of reservoir flushing in comparison to other typical CH₄ emission pathways in reservoirs.

2 Material and methods

2.1 Study site

Field measurements were conducted in Schwarzenbach Reservoir, which is located in the northern part of the Black Forest in southwest Germany ($\mathrm{48}{}^{\circ }{\mathrm{39.334}}^{\prime }$ N, $\mathrm{8}{}^{\circ }{\mathrm{19.630}}^{\prime }$ E) at ∼660 m above sea level (m a.s.l.). The reservoir is operated as a pumped-storage hydroelectric energy system. At a maximum storage capacity of 14.4×10⁶ m³, the reservoir is 47 m deep and covers a surface area of ∼0.66 km². The reservoir's alkalinity and salinity are typically on the order of 0.3 mmol_eq L⁻¹ and 0.04 g kg⁻¹, respectively. The reservoir receives input from two natural creeks (Schwarzenbach and Seebach) located at the western end of the basin and from an artificial channel in the south (Raumünzachstollen) that collects water from the immediate catchment area (Fig. 1a). Due to a sediment trap located upstream of the artificial channel, the Seebach and Schwarzenbach are the only sources for sediment transport into the reservoir. For the pumped-storage operation, an inlet and outlet are based near the dam at ∼5 m above ground. The contributions of the individual inputs are described in Mouris et al. (2018). Since the dam was completed in 1926, the reservoir has been completely emptied on three occasions, in 1935, 1952 and 1997.

Figure 1 (a) Bathymetric map of Schwarzenbach Reservoir showing in- and outflows and the position of the sampling stations (c1–c10). The grey-shaded area symbolizes the estimated basin-wide erosion channel with a 60 m width. (b) Sediment altitude at the sampling stations and water level during the three field campaigns (April 2019, June 2020 and September 2020). The individual sampling stations are indicated above the panel.

2.2 Field measurements

Three field campaigns (April 2019, June 2020 and September 2020) were conducted at Schwarzenbach Reservoir, during which a total of 47 sediment cores were retrieved, i.e., 13 in April 2019 and 17 in each of the campaigns in June and September. The sediment cores were sampled along the transect close to the thalweg (Fig. 1a). At each sampling station, two sediment cores were retrieved, except for station c5 in April 2019 and June 2020 and station c3 in September 2020, when only one sediment core was retrieved each time. The transect length was ∼1.6 km, and the altitude of the sediment cores ranged between 626.4 and 652.7 m (Fig. 1b). In April 2019, the reservoir's water level was around 3 m higher than in June and September 2020. In April 2019, sediment cores were taken at stations c1, c2, c4, c5, c6, c7 and c9 and in June and September 2020 at stations c1, c2, c3, c5, c6, c7, c8, c9 and c10. Vertical temperature and dissolved oxygen (DO) profiles were measured in the water column at the respective stations with a multiparameter probe (conductivity–temperature–depth (CTD) probe, RBR Ltd., Ottawa, Canada, equipped with an RBRcoda³ T.ODO fast temperature and oxygen optode). The temperature of the water overlaying the sediment (∼0.5 m above the sediment surface) served as a proxy for the sediment temperature. Water samples for CH₄ analysis were taken using a 2 L water sampler (Limnos, Finland).

2.3 Measurements of CH₄ concentrations

Sediment cores were taken with a gravity corer equipped with a PVC liner of 600 mm length and an inner diameter of 58 mm. The liner was capped using a rubber stopper or a liner cap. Within 1–3 h of sampling, the sediment cores were processed at the nearest shore. Sediment subsamples were taken through 0.6 cm pre-drilled holes (1 cm vertical spacing) in the PVC liner using 1 mL cutoff syringes at sediment depths of 0.25, 1.25, 2.25, 3.25, 4.25, 5.25, 7.25, 10.25 and 15.25 cm. Each sediment subsample was immediately inserted into a 100 mL glass bottle (DWK Life Sciences GmbH, Germany). The glass bottles were filled completely with demineralized water and closed using a PTFE-coated silicone septum (DWK Life Sciences GmbH, Germany). The bottle was shaken vigorously, enabling the pore water to dissolve into the water. The procedure to measure the CH₄ concentration in the water was adapted from Hofmann et al. (2010). Briefly, 50 mL of liquid was sampled with a 50 mL syringe after sediment settling and injected into a 100 mL glass injection vial (DWK Life Sciences GmbH, Germany) containing 20–30 g of NaCl (≥99.5 %, p.a., ACS, ISO, Carl Roth GmbH, Germany) and 30–40 mL of demineralized water. After sample injection, the CH₄ degassed into the headspace (∼35 mL) due to the oversaturated NaCl solution. The bottles were stored upside down until further processing to prevent CH₄ loss over time through the septum. The CH₄ concentration in the equilibrated headspace of the injection vial was measured using a gas chromatograph equipped with a flame ionization detector (GC 6000, Carlo Erba Instruments, UK). The CH₄ concentration in the water sample was obtained by referring the measured CH₄ concentration in the headspace to the volume of the respective water sample. With the concentration of CH₄ in the water sample $C_{{\mathrm{CH}}_{\mathrm{4}},\text{ws}}$, the volume of the glass bottle V_gb, the volume of the sediment sample V_ss, the volume of the pore water V_pw and the porosity ϕ, the concentration of CH₄ per sediment volume $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ and the concentration of CH₄ in the pore water $C_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}$ were calculated as

$$\begin{array}{}\text{(1)}& {\displaystyle }{C}_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}={\displaystyle \frac{C_{{\mathrm{CH}}_{\mathrm{4}},\text{ws}}\cdot (V_{\text{gb}}-(V_{\text{ss}}-V_{\text{pw}}\left)\right)}{V_{\text{ss}}}}\left(\mathrm{mmol}{\mathrm{L}}^{-\mathrm{1}}\right),\text{(2)}& {\displaystyle }{C}_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}={\displaystyle \frac{C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}}{\mathit{\varphi }}}\left(\mathrm{mmol}{\mathrm{L}}^{-\mathrm{1}}\right).\end{array}$$

The porosity of the sediment describes the ratio of void volume that is occupied by the pore water to the total volume (Brimhall and Dietrich, 1987) and was calculated for each sediment subsample accordingly:

$$\begin{array}{}\text{(3)}& \mathit{\varphi }={\displaystyle \frac{V_{\text{pw}}}{V_{\text{ss}}}}\left(\text{–}\right).\end{array}$$

However, because the difference between V_ss and V_pw was small, the uncertainty in the volume ratios determined from weight measurements was rather large for individual sediment samples from specific depths and locations. Therefore, we determined the average vertical profile of porosity in the sediment by fitting a cubic function of depth to all porosity profiles from all sampling campaigns (Fig. S1 in the Supplement). The porosities of this average porosity profile were used to calculate the pore water concentrations $C_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}$. Vertical profiles of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}$ were obtained by measuring at sediment depths between 0.25 and 15.25 cm. Missing data were estimated by bilinear interpolation of the CH₄ distribution. Missing values at boundaries were estimated by extrapolation, assuming constant concentrations below the depth of the deepest available measurement. Interpolated and extrapolated values constituted ∼10 % of all data points.

2.4 Diffusive CH₄ flux from the sediment into the water column

Assuming molecular diffusion within the sediment and using Fick's first law of diffusion accounting for the porosity and the tortuosity of the sediment, the vertical flux of CH₄ in the sediment is given by Berner (1980):

$$\begin{array}{}\text{(4)}& F_{\text{sed}}=-\mathit{\varphi }\left(D_{{\mathrm{CH}}_{\mathrm{4}}}{\mathit{\theta }}^{-\mathrm{2}}\right){\displaystyle \frac{\partial C}{\partial z}}\left(\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}\right),\end{array}$$

where $D_{{\mathrm{CH}}_{\mathrm{4}}}$ is the molecular diffusivity of CH₄ in water (m² d⁻¹), ϕ is the porosity (–), θ is the tortuosity (–) and $\partial C/\partial z$ is the vertical gradient of the CH₄ pore water concentration (mmol m⁻³). The diffusive flux of CH₄ at the sediment–water interface, F_sed ($\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$), was assumed to correspond to the diffusive flux of CH₄ in the uppermost part of the sediment (see also Berner, 1980) and was determined using the gradient of the CH₄ pore water concentration obtained by linear regression of the $C_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}$ values from 0.25, 1.25 and 2.25 cm depth in the respective sediment core. $D_{{\mathrm{CH}}_{\mathrm{4}}}$ was calculated from the Schmidt number of CH₄ (Wanninkhof, 1992) and the viscosity of the water (Weast, 1988), taking temperature and salinity into account. We assumed the sediment's temperature and salinity were the same as in the water overlaying the sediment. As porosity, we used the median of the porosity measurements in the upper 2.25 cm of all sediment cores (ϕ=0.97). The tortuosity was determined using the porosity according to Boudreau (1997):

$$\begin{array}{}\text{(5)}& {\mathit{\theta }}^{\mathrm{2}}=\mathrm{1}-\ln \left({\mathit{\varphi }}^{\mathrm{2}}\right)\left(\text{–}\right).\end{array}$$

2.5 Stored CH₄ in the potentially eroded sediment volume due to reservoir flushing

For the estimation of the CH₄ storage in the sediment, missing values in the measured profiles of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ were estimated by interpolation and extrapolation using the same procedure as in the case of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{pw}}$. Additionally, profiles of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ were extended to 100 cm sediment depth assuming that $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ below 15 cm sediment depth is constant. Finally, the profiles of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ were linearly interpolated to obtain a regular 1 cm vertical resolution starting at 0.5 cm sediment depth. From these profiles, the total amount of CH₄ stored at different depths within the sediments of Schwarzenbach Reservoir was calculated by lateral and vertical integration.

Figure 2 Profiles of (a) temperature and (b) dissolved oxygen at station c9 during different seasons in Schwarzenbach Reservoir.

Assumptions about the flushing channel width (60 m) were based on reported values for Dashidaira Reservoir, a reservoir of similar size (storage capacity: 9×10⁶ m³; surface area: 0.35 km²) that has been extensively studied in terms of flushing operations (Kantoush et al., 2010; Esmaeili et al., 2015, 2017). Furthermore, the flushing channel was assumed to extend basin-wide along the thalweg of Schwarzenbach Reservoir (Fig. 1a). The total surface area covered by this channel was about 1.1×10⁵ m². We assumed that the CH₄ profile from a sampling station represents the CH₄ concentrations also in the sediment in the proximity of the sampling station. Respective sediment areas were calculated by extending laterally to the half distance of bordering stations. For the outermost stations, sediment areas extended to the respective end of the basin. With all stations representing complementary parts of the total sediment surface area, $C_{{\mathrm{CH}}_{\mathrm{4}},\text{sed}}$ profiles measured at the different stations and their corresponding sediment areas were used to calculate the amount of CH₄ stored per unit depth at different sediment depths across the entire channel, $S_{{\mathrm{CH}}_{\mathrm{4}}}$ (mmol m⁻¹). The average concentration and the total amount of CH₄ in the potentially eroded sediment volume due to reservoir flushing ($C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ (mol m⁻³) and $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ (mol), respectively) depend on the thickness of the flushed sediment layer. $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ was estimated from the stored methane $S_{{\mathrm{CH}}_{\mathrm{4}}}$ by integrating vertically down to the depth of sediment erosion during flushing (Z_e):

$$\begin{array}{}\text{(6)}& N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}\left(Z_{\mathrm{e}}\right)=\underset{\mathrm{0}}{\overset{Z_{\mathrm{e}}}{\int }}{S}_{{\mathrm{CH}}_{\mathrm{4}}}\left(z^{\prime }\right)\mathrm{d}{z}^{\prime }\left(\mathrm{mol}\right).\end{array}$$

$C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ was obtained by dividing $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ by the volume of mobilized sediment.

The relevance of $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ for overall CH₄ emissions from Schwarzenbach Reservoir was assessed by comparing the $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ stored in the potentially eroded sediment volume during a flushing event with other pathways of CH₄ emission from Schwarzenbach Reservoir (ebullition, diffusive CH₄ emissions from the reservoir surface and degassing during turbination) (Peeters et al., 2019; Encinas Fernández et al., 2020).

3 Results

3.1 Reservoir characterization

Schwarzenbach Reservoir is a pumped-storage system, and reservoir management can substantially change the water level. In April 2019, water levels were about 3 m higher than in June and September 2020 (Fig. 1b). The water column of Schwarzenbach Reservoir was stably stratified during all three campaigns (Fig. 2a). In June and September 2020, DO concentrations were oversaturated near the water surface and above 2 mg L⁻¹ throughout the entire water column (Fig. 2b). In April 2019 the DO sensor failed, but DO concentrations can be expected to be around 10 mg L⁻¹ as is indicated by a DO profile measured in April 2018 (Fig. 2b).

3.2 Spatiotemporal dynamics of CH₄ in sediment pore water

The CH₄ concentrations in the pore water of the sediment of the reservoir differed spatially and seasonally (Fig. 3). In April 2019 and June 2020, regions with particularly large CH₄ concentrations were located in the center of the transect at deep sediment layers, whereas at the same depth within the sediment, CH₄ concentrations near the dam and the western end of the basin were considerably lower (Fig. 3a and b). In September 2020, the largest CH₄ concentrations were still found in the middle of the transect but were, in general, more evenly distributed across the transect and larger at shallower depth than in June 2020 (Fig. 3c). For each sampling campaign, we determined a median CH₄ pore water profile by compiling the medians of all measurements from the same sediment depth into a vertical profile (Fig. 3d). With the progressing season, we observed an increase in median CH₄ pore water concentrations and stronger vertical CH₄ gradients in the uppermost 5 cm of the sediment. In April 2019, the median CH₄ concentration profile was characterized by relatively low concentrations and a weak vertical gradient in the upper sediment layers, followed by an almost linear increase in CH₄ concentrations. In contrast, the median CH₄ profiles in June and September 2020 both showed stronger gradients in the upper sediment layers with a more saturated curve in deeper layers. However, in September 2020, the overall CH₄ concentrations were much larger in all but the deepest sampled sediment layers. The mean and median CH₄ concentrations across all measurements were 0.27 and 0.16 mmol L⁻¹ in April 2019, 0.37 and 0.34 mmol L⁻¹ in June 2020, and 0.63 and 0.68 mmol L⁻¹ in September 2020, respectively, indicating an increase in overall CH₄ concentration in the pore water with the progressing season.

Figure 3 (a–c) Spatial distribution of CH₄ pore water concentration in the top 15 cm of the sediment along the transect in April 2019 (a), June 2020 (b) and September 2020 (c). The individual sampling stations are indicated above the panels. (d) Median CH₄ concentration of all pore water profiles within each campaign. Shaded areas display the 25th and 75th percentiles.

3.3 CH₄ storage estimation

To estimate the stored CH₄ that is potentially emitted due to sediment erosion during reservoir flushing of Schwarzenbach Reservoir, we calculated $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$, the amount of CH₄ stored between the top of the sediment and a hypothetical erosion depth within the 60 m wide erosion channel centered around the thalweg along the basin (Fig. 4; the channel is indicated in Fig. 1a). With the progressing season, $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ was consistently higher at all erosion depths. At an erosion depth of 15 cm, $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ was 6.5, 8.1 and 12.5 kmol CH₄ in April 2019, June 2020 and September 2020, respectively.

3.4 CH₄ fluxes at the sediment–water interface

Diffusive CH₄ fluxes at the sediment–water interface, F_sed, were determined for each campaign at each station along the transect (Fig. 5c). The overall variability in F_sed from all stations measured during a campaign increased with the progressing season from April to September (Fig. 5a). Compared to the range of F_sed values in April 2019, the range of F_sed values was ∼3.5 times larger in June 2020 and ∼4 times larger in September 2020. While the lowest measured F_sed values were very similar across the three campaigns (between 0.02 and 0.06 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$), the median and maximum CH₄ fluxes increased from April to September. The average CH₄ flux over all measurements was 1.02 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$. Sediment temperatures, approximated by water temperature ∼0.5 m above the sediment, increased with the season (Fig. 5b). In April 2019, the temperatures along the transect were rather uniform, ranging between 5.1 ^∘C (c9) and 5.9 ^∘C (c1). Temperatures were higher in June 2020, ranging from 7.7 ^∘C (c10) to 12.3 ^∘C (c1), and in September 2020, ranging between 10.1 ^∘C (c9) and 16.0 ^∘C (c1). In June and September 2020, temperatures were more elevated in the shallower part of the basin.

Figure 4 (a) Stored amount of CH₄ in the flushed sediment layer, $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$, as a function of layer thickness. The thickness of the flushed sediment layer corresponds to the erosion depth of the sediment in the 60 m wide erosion channel. (b) The average CH₄ concentration in the flushed sediment layer.

4 Discussion

4.1 Seasonal and spatial differences in pore water CH₄ concentrations

CH₄ pore water concentrations in Schwarzenbach Reservoir ranged between 0.002 and 1.530 mmol L⁻¹, which is comparable to what other studies have found in the sediment of freshwater systems (Schulz and Conrad, 1995; Murase and Sugimoto, 2001; Huttunen et al., 2006; Maeck et al., 2013; Norði et al., 2013; Donis et al., 2017). The CH₄ pore water concentrations typically increased with increasing sediment depth (Fig. 3), which is consistent with observations in other studies (Frenzel et al., 1990; Huttunen et al., 2006; Deutzmann et al., 2014). The CH₄ pore water concentrations showed seasonal and spatial differences (Fig. 3). The median CH₄ concentration in the pore water across all measured profiles in September 2020 (0.68 mmol L⁻¹) was almost 2 times larger than in June 2020 (0.37 mmol L⁻¹) and about 4 times larger than in April 2019 (0.16 mmol L⁻¹) (Fig. 3b). The increase in pore water concentration with the season may be explained by differences in CH₄ production within the sediment. Many studies have demonstrated that CH₄ production rates in sediments are enhanced at higher temperatures relative to CH₄ oxidation rates (Schulz et al., 1997; Lofton et al., 2014; Marotta et al., 2014; Sepulveda-Jauregui et al., 2018), although, in some systems, CH₄ oxidation rates may keep up with CH₄ production rates (Shelley et al., 2015). In Schwarzenbach Reservoir, sediment temperatures increased as the season progressed (Fig. 5b), supporting the hypothesis that higher production was responsible for larger pore water CH₄ concentrations later in the season. Enhanced CH₄ concentrations were present, especially in deeper sediment layers, which are more strongly affected by changes in production and less affected by CH₄ losses due to oxidation and vertical transport compared to the upper sediment layers. However, the spatial distributions of pore water CH₄ along the transect during the different campaigns cannot be explained by temperature alone. Sediment temperatures were elevated in the shallow water zone in June and September 2020, but CH₄ concentrations were generally largest near the center of the transect and lowest towards the western and eastern end of the basin (Fig. 3a). Availability of organic matter in the sediment is another factor that correlates with CH₄ production (Duc et al., 2010) or CH₄ concentrations (Murase and Sugimoto, 2001) in lake sediments and is considered a major limiting factor for CH₄ production (Segers, 1998). Settling of organic matter is known to correlate with the current velocity of the inflows (Kufel, 1991). Fewer organic particles might be deposited close to the inflows than in the open water because the water current velocity is likely to be smaller in the open water of the central basin than near the inflows.

4.2 Significance of the potentially released amount of CH₄ due to reservoir flushing

Assuming that the CH₄ stored in the sediment is completely released into the water during the sediment flushing process and degasses quickly to the atmosphere when the water is rapidly transported out of the reservoir, $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ provides a measure of the potential CH₄ emission from the reservoir due to flushing. The amount of CH₄ mobilized during a flushing operation depends on the depth of sediment erosion and on the season (Fig. 4a). Flushing operation eroding a 15 cm thick sediment layer in the assumed 60 m wide flushing channel along the thalweg of Schwarzenbach Reservoir implies a flushed sediment volume of 16.5×10³ m³, which agrees well with the study by Saam et al. (2019), who assessed the feasibility of flushing in Schwarzenbach Reservoir and obtained a flushed sediment volume of 13.6×10³ m³ in a simulation of a full-drawdown flushing scenario. At an erosion depth of 15 cm, the resulting $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ values are 6.5, 8.1 and 12.5 kmol CH₄ if the flushing operation is conducted in April, June and September, respectively. This suggests that conducting a reservoir flushing operation in spring rather than in late summer would reduce flushing-induced CH₄ emissions in Schwarzenbach Reservoir by a factor of 2.

Typical CH₄ emission pathways in Schwarzenbach Reservoir include CH₄ emissions due to ebullition (212 mol d⁻¹ during normal operation mode), diffusive CH₄ emissions (27 mol d⁻¹) and CH₄ emissions due to degassing during turbination (14 mol d⁻¹) (Peeters et al., 2019; Encinas Fernández et al., 2020). Together, CH₄ emissions by these pathways add up to 253 mol d⁻¹ or 92.4 kmol yr⁻¹. Hence, one flushing operation with 15 cm erosion depth potentially causes CH₄ emissions that would account for 7 % (flushing in April) to 14 % (flushing in September) of the typical annual CH₄ emission in Schwarzenbach Reservoir.

Figure 5 (a) Diffusive CH₄ flux at the sediment–water interface of each campaign. The middle lines of the boxes represent the medians; the boxes demarcate the 25th and 75th percentiles; and the whiskers extend to the most extreme data points, not considering outliers. (b) Water temperature 0.5 m above the sediment and (c) the mean diffusive CH₄ flux at each station of the transect with either one sediment core (indicated by black outline) or two sediment cores per station at each campaign (error bars represent the range of values). The individual sampling stations are indicated above the panels.

Not only the total amount of CH₄ but also the average CH₄ concentration in the flushed sediment increase with increasing erosion depth (Fig. 4b). This implies that overall less CH₄ is released if an equivalent sediment volume is eroded by several flushings that mobilize thin layers of sediment instead of a few flushings that mobilize thicker sediment layers. For instance, if flushing in April mobilizes the top 5, 10 or 100 cm of sediment, $C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ is 0.10, 0.27 and 0.74 mmol L⁻¹, respectively (Figs. 4b and S2 in the Supplement). Hence, the potential release of CH₄ during one flushing event of a 100 cm thick sediment layer is about 2.7 or 7.5 times larger than 10 flushing events of 10 cm and 20 flushing events of 5 cm thick sediment layers, respectively. Below ∼50 cm of sediment, $C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ is essentially constant (Fig. S2). Thus, $N_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ in flushing events that erode more than 50 cm of sediment increases essentially linearly with the volume of sediment eroded.

Note that Schwarzenbach Reservoir has been emptied only three times since its completion. However, many reservoirs worldwide are flushed regularly, such as Dashidaira Reservoir in Japan, a reservoir of similar size, where a flushing operation with full water level drawdown is conducted annually (Esmaeili et al., 2017; Sumi et al., 2017). In this reservoir, flushed sediment volumes of between 60×10³ and 590×10³ m³ with an average flushed sediment volume of 287×10³ m³ have been reported (Sumi et al., 2009; Esmaeili et al., 2017). Therefore, the average flushed sediment volume in Dashidaira Reservoir is 17–21 times larger than the estimated flushed sediment volumes in the study of Saam et al. (2019) and in this study.

Unfortunately, no data are available on CH₄ concentration in the sediment or CH₄ emissions of Dashidaira Reservoir. However, assuming that CH₄ concentrations in the sediment of Dashidaira Reservoir are comparable to those in Schwarzenbach Reservoir and that the erosion depth is 100 cm or more, the average concentration in the flushed sediment can be approximated by a $C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}$ of 0.73 mol m⁻³, which is the average CH₄ concentration (seasonal average) in a 100 cm thick flushed sediment layer in Schwarzenbach Reservoir (Fig. S2). Therefore, the average flushed sediment volume in Dashidaira Reservoir of 287×10³ m³ would contain 210 kmol CH₄. Because the reservoir is flushed each year, the potential emissions from this reservoir due to flushing would be 210 kmol yr⁻¹ or, with consideration of the reservoir's surface area, 0.6 $\mathrm{mol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{yr}}^{-\mathrm{1}}$, i.e., 2.3 or 4.3 times larger, respectively, than the typical annual emissions from Schwarzenbach Reservoir. Hence, CH₄ emissions due to reservoir flushing operations can contribute substantially to overall CH₄ emissions from reservoirs.

CH₄ emissions due to reservoir flushing can represent a significant contributing pathway to overall CH₄ emissions from reservoirs only if CH₄ production in the sediment can replenish the amount of CH₄ in the flushed sediment volumes between flushing operations. In Schwarzenbach Reservoir, the measured CH₄ flux at the sediment–water interface was 1.02 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$ (average over all measurements). Typically, 60 %–90 % of CH₄ produced in the sediment is oxidized (Le Mer and Roger, 2001). Compensating for the flux from the sediment and the loss due to oxidation of the produced CH₄ requires a CH₄ production rate per unit sediment surface of 2.6–10.2 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{d}}^{-\mathrm{1}}$. Considering a 100 cm thick layer of sediment, this implies a CH₄ production rate of 2.6–10.2 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{d}}^{-\mathrm{1}}$. This estimated production rate is compatible with the median production rate of 10 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{d}}^{-\mathrm{1}}$ obtained from direct measurements of production rates in sediments of lakes (the median of a data compilation of 93 samples of different lake sediments; Martinez-Cruz et al., 2018). At a CH₄ production rate of 2.6–10.2 $\mathrm{mmol}{\mathrm{m}}^{-\mathrm{3}}{\mathrm{d}}^{-\mathrm{1}}$, the time required to produce the average CH₄ concentration in the flushed sediment of $C_{{\mathrm{CH}}_{\mathrm{4}},\text{FSL}}=\mathrm{0.73}$ mol m⁻³ is 73–291 d. This estimation suggests that CH₄ production in the sediments provides sufficient CH₄ in the flushed sediment layers between annual flushing operations.

The CH₄ emissions associated with a flushing event may be smaller than the above-estimated potential CH₄ emissions because some of the mobilized CH₄ may be oxidized in the water before reaching the atmosphere. Most of the CH₄ is mobilized during the free-flow state of drawdown flushing, which is typically maintained for several hours, depending on the desired amount of flushed sediment (Kondolf et al., 2014). The residence time of the CH₄ after mobilization until spill-out at the dam may be in the range of a few hours; thus, the potential for oxidation of the mobilized CH₄ is rather limited.

4.3 Additional indirect CH₄ emissions due to reservoir flushing

In addition to CH₄ emissions driven by sediment erosion, the water level drawdown that is necessary during reservoir flushing operations may also lead to enhanced CH₄ emissions. During a drawdown flushing operation, the water is released through one or more outlets and the water level is decreased dramatically, resulting in additional CH₄ emissions, as the CH₄ stored in the released water can degas during turbination and downstream of the reservoir. Furthermore, as water level fluctuations in reservoirs can enhance CH₄ ebullition due to changing hydrostatic pressure (Harrison et al., 2017; Encinas Fernández et al., 2020), lowering the water level during the drawdown period would cause an additional CH₄ ebullition flux. If the ebullition originates from the sediment that is mobilized during flushing, the total CH₄ emissions during the entire operation do not differ substantially from the above estimates of CH₄ release during sediment erosion. However, CH₄ ebullition may originate from sediment layers deeper than the eroded sediment. As a result, total CH₄ emissions might be larger than the CH₄ emissions estimated by sediment erosion alone. The relative contribution of each pathway to the overall CH₄ emissions due to reservoir flushing may be an area for future studies. Lastly, throughout a flushing operation, large sediment areas may fall dry and become inundated again. Drying and rewetting of sediments are known to affect CH₄ emissions from these sediments and are an important area for more research (Yang et al., 2012; Kosten et al., 2018).

5 Conclusions

Assessing the significance of CH₄ emissions due to reservoir flushing in comparison to other pathways requires consideration of the flushing frequency, the flushed sediment volume, the erosion depth during the flushing event, and the vertical and lateral distribution of CH₄ concentrations in the sediment. In Schwarzenbach Reservoir, we estimated that one flushing operation could potentially cause CH₄ emissions that are comparable to 7 %–14 % of the reservoir's typical annual emissions. Seasonal differences in CH₄ concentrations in the sediment suggest that the timing of the flushing operation during the season affects the amount of CH₄ emitted. Because the CH₄ concentration increases with depth in the sediment, removal of the same sediment volume by regular flushing of shallow sediment layers might cause fewer CH₄ emissions than few flushings of thicker sediment layers. However, additional indirect CH₄ emissions may vary with the flushing operation and affect the total CH₄ emissions due to reservoir flushing. In Schwarzenbach Reservoir, flushing is not conducted regularly. Nevertheless, there are many reservoirs where flushing operations mobilize much larger volumes of sediment and are conducted more regularly than in Schwarzenbach Reservoir. In these reservoirs, CH₄ emission due to reservoir flushing is likely substantial, and thus their overall CH₄ emission will be underestimated if this pathway is ignored. Our findings highlight CH₄ emissions due to reservoir flushing as an important area for future research and emphasize the need for field studies during reservoir flushing operations to better understand underlying processes.