Evaluating the Feedback of the Reservoir Methane Cycle to Climate Warming under Hydrological Uncertainty

: Freshwater reservoirs are widely recognized as methane (CH 4 ) emission hotspots. Existing research has shown that temperature and hydrological conditions signiﬁcantly affect wetland CH 4 cycling processes. However, the feedback of the CH 4 cycle to climate warming remains unclear for deep reservoirs where seasonal water thermal stratiﬁcation exists. This study combined a reservoir CH 4 cycling model and a Statistical DownScaling Model (SDSM) to evaluate reservoir CH 4 cycling feedbacks under multiple climate change scenarios while accounting for hydrological uncertainty. Daily air temperatures in 2100 were predicted by the combination of the CanESM5 model and a SDSM. To address hydrological uncertainty, we selected three representative hydrological years (i.e., wet, normal, and dry) to create hydrological scenarios. Results showed that annual sediment CH 4 production increased with warming, ranging 323.1–413.7 × 10 3 t C year − 1 among multiple scenarios. Meanwhile, the CH 4 oxidation percentage decreased with warming, which meant warming promoted sediment CH 4 release non-linearly; 67.8–84.6% of sediment ebullient ﬂux was ultimately emitted to the atmosphere (51.3–137.7 × 10 3 t C year − 1 ), which showed ebullition was the dominant emission pathway. Higher air temperatures and drier conditions generally promote reservoir emissions. This study is helpful for predicting reservoir emissions while directing decision-making for reservoir sustainability.


Introduction
Methane (CH 4 ) is a typical greenhouse gas (GHG), contributing to approximately 21% of the increase in radiative forcing since pre-industrial times [1]. Compared to carbon dioxide (CO 2 ), the climate warming potential of CH 4 is greater (i.e., by a factor of 28). Moreover, CH 4 concentrations in the near-surface atmosphere have increased by a factor greater than 1.5 compared to pre-industrial levels [2]. Freshwater systems contribute greatly to the global CH 4 budget, being responsible for approximately 42% of all global emissions from natural and anthropogenic sources [3]. Among these sources, reservoirs that are constructed by means of river damming have been recognized as CH 4 emission hotspots [4,5]. A recent modeling assessment estimated that global reservoirs emitted 1076 Tg of CO 2 -equivalent in 2020, of which CH 4 contributed 70% [6]. Moreover, under the ongoing expansion of hydropower engineering projects, reservoirs will likely have an increasing effect on global warming over the coming decades, while their CH 4 emissions are expected to increasingly dominate reservoir-induced effects [7]. Therefore, the quantitative prediction of reservoir emissions under conditions of environmental change is critical to accurately assessing reservoir-based climatic effects while also estimating global budgets.
Damming changes upstream river environments from lotic to lentic, leading to significant organic carbon (C) sedimentation and accumulation in reservoir sediment [8]. This makes reservoir sediment CH 4 hotspots [9,10]. Many studies have investigated reservoir emission budgets and CH 4 cycling processes. Generally, organic C is degraded and mineralized to generate CH 4 in anoxic sediment, except for a small fraction that is permanently buried [11]. Then, CH 4 is partially oxidized in oxic sediment [12], while the remaining fraction is released from sediment via diffusion and ebullition [13]. Extensive research has been conducted to measure and monitor CH 4 emissions in rivers and reservoirs [13][14][15]. Meanwhile, modeling approaches have been developed and applied to simulate CH 4 cycling processes and calculate their respective budgets [16,17]. It has been widely found that reservoir CH 4 emissions are subject to considerable variability and are regulated by multiple influencing factors [4,18,19].
Climatic factors (i.e., air temperature, precipitation, atmospheric pressure, etc.) have frequently been shown to affect CH 4 cycling processes and emissions in aquatic ecosystems [3,20,21]. Temperature affects the rate of microbiological processes, thereby influencing both CH 4 production and consumption. Wetland experiments have shown that climate warming leads to a disproportionate increase in CH 4 production over oxidation, which is indicative of a positive feedback loop (i.e., where warming can cause increases in wetland emissions) [3]. Moreover, temperature affects CH 4 solubility and thereby impacts CH 4 release pathways (i.e., through diffusion or ebullition) [22]. On the other hand, precipitation affects hydrological conditions, which strongly impact CH 4 cycling processes [20]. It has recently been shown that both water depth variation and submerged conditions affect C emissions in reservoirs and wetlands [5,23]. The effects of temperature increases on wetland emissions vary with wetland water tables [24].
The effect of climate warming on CH 4 cycling is complex in deep reservoirs due to the existence of seasonal thermal stratification [25]. The water level has a significant effect on water and sediment temperature distribution [26], thereby affecting reservoir emission feedback as air temperatures increase. Global warming is increasingly impacting the world's environments, where hydrological conditions are subject to significant variability and uncertainty. CH 4 emission feedback to climate change has been explored for wetlands with comparatively low water depths [2,24,27]. However, deep reservoirs differ from wetlands in multiple ways, such as water depth, sediment condition, and biological communities. Feedbacks of sediment CH 4 processes and emissions on climate warming have rarely been investigated in deep reservoirs, especially under hydrological uncertainty. It is important to understand reservoir feedbacks to project future global reservoir emissions under climate change and to direct anthropogenic emission reduction planning to achieve sustainable development goals.
This study combines a reservoir CH 4 cycling model and a Statistical DownScaling Model (SDSM) to evaluate reservoir CH 4 cycling feedback to climate warming under hydrological uncertainty. We selected the Danjiangkou Reservoir, China's second-largest reservoir, as a case study. The projections of air temperature in 2100 under the shared socioeconomic pathways (SSPs) simulated by the CanESM5 model were downscaled on the SDSM platform. To address hydrological uncertainty, three representative hydrological years (i.e., wet, normal, and dry) were selected through means of historical reservoir inflow analysis. After combining temperature and hydrological scenarios, the reservoir CH 4 cycling model simulates how climate change influences spatiotemporal CH 4 cycling through hydrodynamic and biogeochemical processes in the Danjiangkou Reservoir. This study can help predict reservoir CH 4 emission dynamics and direct decision-making in the pursuit of reservoir sustainability.

Study Area
The Danjiangkou Reservoir (Figure 1), located in the Yangtze River basin, is China's second-largest reservoir. It plays a vital role in providing water to the South-North Water Transfer Project. At its normal pool level (i.e., 170 m), its storage capacity is 29.05 × 10 9 m 3 , with a total area of 1023 km 2 . The main functions of the reservoir are to provide water, generate hydropower, control floods, and discharge environmental flow. For the reasons explained below, the Danjiangkou Reservoir is a good case study candidate. Firstly, along with agricultural development, the land adjacent to its upstream reaches is high in soil organic carbon content and subject to water degradation and soil erosion [28]. This eroded soil is then washed into the Han River, where it is trapped by the Danjiangkou Reservoir. This increases organic C content in reservoir sediment, which provides substrates for methanogens to produce CH 4 [9]. Secondly, hypolimnetic hypoxia in this deep reservoir is induced by seasonal water stratification and algal growth [25]. Incremental air temperature lengthens the period of hypolimnetic hypoxia [29] and stimulates algal growth [30], which may promote CH 4 production in reservoir sediment [31,32]. Finally, precipitation variation affects runoff in upstream river reaches and water storage levels in the reservoir [33], which could influence reservoir sediment CH 4 emission pathways [34,35].

Study Area
The Danjiangkou Reservoir (Figure 1), located in the Yangtze River basin, is China's second-largest reservoir. It plays a vital role in providing water to the South-North Water Transfer Project. At its normal pool level (i.e., 170 m), its storage capacity is 29.05 × 10 9 m 3 , with a total area of 1023 km 2 . The main functions of the reservoir are to provide water, generate hydropower, control floods, and discharge environmental flow. For the reasons explained below, the Danjiangkou Reservoir is a good case study candidate. Firstly, along with agricultural development, the land adjacent to its upstream reaches is high in soil organic carbon content and subject to water degradation and soil erosion [28]. This eroded soil is then washed into the Han River, where it is trapped by the Danjiangkou Reservoir. This increases organic C content in reservoir sediment, which provides substrates for methanogens to produce CH4 [9]. Secondly, hypolimnetic hypoxia in this deep reservoir is induced by seasonal water stratification and algal growth [25]. Incremental air temperature lengthens the period of hypolimnetic hypoxia [29] and stimulates algal growth [30], which may promote CH4 production in reservoir sediment [31,32]. Finally, precipitation variation affects runoff in upstream river reaches and water storage levels in the reservoir [33], which could influence reservoir sediment CH4 emission pathways [34,35].

Reservoir Methane Cycling Simulations
The Environmental Fluid Dynamics Code (EFDC) is a surface water modeling system originally developed by Hamrick [36]. The EFDC has been used to build models for different waterbody types, including rivers [37], reservoirs [38], lakes [30], estuaries [39], etc. For this study, we used a quasi-three-dimensional reservoir model developed on the EFDC platform for the Danjiangkou Reservoir, which has previously been calibrated and validated [40]. The reservoir model couples hydrodynamic, water quality, and sediment

Reservoir Methane Cycling Simulations
The Environmental Fluid Dynamics Code (EFDC) is a surface water modeling system originally developed by Hamrick [36]. The EFDC has been used to build models for different waterbody types, including rivers [37], reservoirs [38], lakes [30], estuaries [39], etc. For this study, we used a quasi-three-dimensional reservoir model developed on the EFDC platform for the Danjiangkou Reservoir, which has previously been calibrated and validated [40]. The reservoir model couples hydrodynamic, water quality, and sediment diagenesis modules on the EFDC to simulate interactions of multiple variables in the water body and sediment. The hydrodynamic module simulates water flow and material transportation in the horizontal direction with orthogonal curvilinear coordinates and in the vertical direction with sigma coordinates. The water quality module simulates C, nitrogen (N), and phosphorus (P) transformations, dissolved oxygen (DO) variation, and algal dynamics. Meanwhile, these organic matter forms are composed of labile particulate, refractory particulate, and dissolved organic matter. The sediment diagenesis module simulates organic and inorganic matter flux at the sediment-water interface and biogeochemical processes in sediment. More details on EFDC theory can be found in the existing document [41].
To simulate CH 4 production and consumption in reservoir sediment as well as CH 4 emissions from sediment to the overlying water and subsequently to the atmosphere, the specific processes are introduced as follows. Particulate organic carbon (POC), derived from terrestrial organic matter, wastewater, and/or autochthonous algae, deposits in sediment. Next, labile and refractory POC are degraded, and CH 4 is produced within anoxic sediment, while permanently buried POC is only negligibly degraded. Then, the produced CH 4 moves up into the upper oxic sediment layer, which is partly oxidized by oxygen diffusing from overlying water. The remaining CH 4 is released into the water column via diffusion or ebullition. Additionally, the rising bubbles (ebullition) lose a fraction of CH 4 in the water column, and the remaining CH 4 is subsequently released into the atmosphere. Following the empirical model proposed by Ostrovsky [42], we calculated the CH 4 ebullient flux to the atmosphere.
During the process of building the reservoir model, we first input the boundary of Danjiangkou Reservoir into the EFDC platform, which was drawn from the map using the Universal Transverse Mercator (UTM) Coordinate System. Before this, the coordinate system of the map needed to be converted from the World Geodetic System (WGS) to the UTM Coordinate System due to standard cartography ( Figure 1) using WGS. Subsequently, bottom elevation, hydrological, meteorological, and water quality data were input to the EFDC software (version 10.3) step by step, and the water body of the reservoir was divided into six layers averagely in order to simulate the change of variables in the vertical direction [40]. Additionally, calibration and verification had been done previously [40,43]. To improve its running efficiency, the model was run by setting the dynamic time-step to 2 s as the initial value.

Air Temperature Prediction and Hydrological Uncertainty
This section introduces the prediction method that we used for the air temperature above the surface of the Danjiangkou Reservoir in 2100. In this study, we chose the modeling results of air temperature simulated by the newest version of the Canadian Earth System Model (CanESM5) because its early versions had great performance in simulating climate change in China [44][45][46]. Additionally, according to historical runoff, we selected representative hydrological years to address hydrological uncertainty. Through runoff data from different hydrological years, discharges from the Danjiangkou Reservoirs were simulated by a reservoir operation model [25]. The corresponding air temperature, runoff, and discharge data for all scenarios were input into the Danjiangkou Reservoir model developed on the EFDC. The model was then run in all scenarios. Following this, we analyzed the spatiotemporal feedback of reservoir CH 4 cycling to climate warming during all three hydrological years.

Near-Surface Air Temperature Prediction
The Statistical DownScaling Model (SDSM) [47], a hybrid model between a stochastic weather generator and multiple regression equations, enhances the efficiency of generating multiple, low-cost, and single-site scenarios of surface weather variables under present and future climate forcing. To generate the scenarios, the first step was to obtain observed near-surface air temperature and predictor variable data in the study area. The daily air temperature monitored by the Laohekou Meteorological Station between 1979 and 2014, which this study regarded as the air temperature above the surface of the Danjiangkou Reservoir, was obtained from the National Centers for Environmental Information (NCEI), the United States of America (https://www.ncei.noaa.gov/maps/daily, accessed on 26 April 2023). For this study, we selected one grid box covering the Laohekou Meteorological Station from all the grid boxes on the Canadian Climate Impacts and Scenarios (CCIS) project website (https://climate-scenarios.canada.ca/?page=pred-cmip6, accessed on 26 April 2023). Next, in this grid box, we downloaded modeling results from the National Centers for Environmental Prediction-Department of Energy (NCEP-DOE) Reanalysis 2 model between 1979 and 2014, which contains data sets of 26 predictor variables. In the same grid box, we simultaneously downloaded other predictor variable data sets, namely, modeling results under the SSPs (i.e., SSP1-2.6, SSP2-4.5, and SSP5-8.5) simulated by the CanESM5 model between 2015 and 2100. The second step was to develop a statistical model. To do so, we screened 26 predictor variables simulated by the NCEP-DOE Reanalysis 2 model on the SDSM platform to find predictor variables whose optimal statistical relationships corresponded to observational data. Two predictor variables were subsequently selected: air temperature at 2 m (temp) and mean sea level pressure (mslp). The observational and the selected predictor variable data were, respectively, sectioned into two: (1) that which ranged between 1979 and 1996 and (2) that which ranged between 1997 and 2014. The first observational and predictor variable data were used to develop the statistical model and calibrate it on the SDSM platform. The second predictor variable data were input into the statistical model to obtain predictand, which was used to calculate deviations from the observed data. Lastly, after verification, we input temp and mslp under the SSPs simulated by the CanESM5 model into the statistical model to obtain the air temperature above the surface of the Danjiangkou Reservoir in 2100.

Hydrological Uncertainty Analysis
The main source of uncertainty between Global Climate Models (GCMs) and within the GCM was runoff projections under climate change, which were obtained by the combination of a downscaling method and a hydrological model [48]. Additionally, hydrological uncertainty distinctly increased over time under future climate scenarios and was larger under high than low emission scenarios [49]. Given the difficulty in predicting precipitation, we used historical runoff data from the Han River as inflow to the Danjiangkou Reservoir when predicting the impact of climate warming on reservoir sediment CH 4 cycling processes. To comprehensively account for future variation, we selected three hydrological years (i.e., wet, normal, and dry) from a total of fifty years of runoff data, following the method proposed by Cai and Rosegrant [50]. Based on the SSPs under the different hydrological years, a total of nine scenarios were selected for this study (Table 1), which were subsequently used to analyze climate warming impacts on reservoir sediment C cycling processes under different hydrological conditions.

Air Temperature Prediction
The predictor variables (i.e., temp and mslp) simulated by the NCEP-DOE Reanalysis 2 model and observed daily near-surface air temperature data between 1979 and 1996 were Sustainability 2023, 15, 9197 6 of 14 used to calibrate the statistical model for downscaling on the SDSM platform. Then, the predictor variables and air temperature data between 1997 and 2014 were used to verify the statistical model. As Figure 2 shows, the statistical model could perform well during the downscaling process. The determination coefficient (R 2 ) and curve slope of the daily mean air temperature are 0.95 and 0.99, respectively (Figure 2a). The variance of the monthly mean air temperature ranged from 0.34 to 0.88 (Figure 2b). nomic pathway (SSP) 1-2.6, SSP2-4.5, and SSP5-8.5 were 17.7 °C, 19.7 °C, and 22.6 °C, respectively. c Percentage of methane (CH4) oxidation, CH4 diffusive flux, or CH4 ebullient flux in CH4 production.

Air Temperature Prediction
The predictor variables (i.e., temp and mslp) simulated by the NCEP-DOE Reanalysis 2 model and observed daily near-surface air temperature data between 1979 and 1996 were used to calibrate the statistical model for downscaling on the SDSM platform. Then, the predictor variables and air temperature data between 1997 and 2014 were used to verify the statistical model. As Figure 2 shows, the statistical model could perform well during the downscaling process. The determination coefficient (R 2 ) and curve slope of the daily mean air temperature are 0.95 and 0.99, respectively (Figure 2a). The variance of the monthly mean air temperature ranged from 0.34 to 0.88 (Figure 2b).  Near-surface air temperatures in 2100 under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 simulated by the CanESM5 model were downscaled by using the statistical model ( Figure 3). Annual mean air temperatures under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 were 17.7 • C, 19.7 • C, and 22.6 • C, respectively. Although the annual mean air temperature under the SSP2-4.5 was significantly higher than that under the SSP1-2.6 (2.0 • C), the mean difference between them was negligible from mid-June to August (0.6 • C). Generally, the daily air temperature under the SSP5-8.5 was distinctly the highest among the three SSPs. SSP2-4.5 was significantly higher than that under the SSP1-2.6 (2.0 °C), the mean difference between them was negligible from mid-June to August (0.6 °C). Generally, the daily air temperature under the SSP5-8.5 was distinctly the highest among the three SSPs.

Dynamic Variations of Sediment Methane Budgets
Results show that annual sediment CH4 cycle flux increased under the warmer SSPs ( Figure 4 and Table 1). As the annual mean air temperature increased from 17.7 °C (SSP1-2.6) to 19.7 °C (SSP2-4.5), annual CH4 production and ebullition increased by 10.4 × 10 3 t C year −1 per °C and 4.3 × 10 3 t C year −1 per °C, respectively. Likewise, as annual mean air

Dynamic Variations of Sediment Methane Budgets
Results show that annual sediment CH 4 cycle flux increased under the warmer SSPs ( Figure 4 and Table 1). As the annual mean air temperature increased from 17.7 • C (SSP1-2.6) to 19.7 • C (SSP2-4.5), annual CH 4 production and ebullition increased by 10.4 × 10 3 t C year −1 per • C and 4.3 × 10 3 t C year −1 per • C, respectively. Likewise, as annual mean air temperature increased from 19.7 • C to 22.6 • C (SSP5-8.5), annual CH 4 production and ebullition increased by 17.5 × 10 3 t C year −1 per • C and 9.2 × 10 3 t C year −1 per • C, respectively. These modeling results indicated a significant nonlinear relationship between annual CH 4 production (or ebullition) and annual mean near-surface air temperature (Figure 4). Additionally, the CH 4 oxidation growth rate was slower compared to CH 4 production among all SSPs (Figure 4), resulting in a decrease in the CH 4 oxidation percentage of CH 4 production (Table 1). Meanwhile, the CH 4 ebullient flux percentage in CH 4 production increased under the warmer SSPs, while the CH 4 diffusive flux percentage in CH 4 production decreased.  Furthermore, hydrological conditions influence the response of sediment CH4 cycling processes to climate warming. Under the same SSP, even though the differences in annual CH4 production (or oxidation) among the hydrological years were insignificant (less than 5%), the CH4 oxidation percentage in CH4 production decreased as annual runoff decreased (Table 1). In summary, when the annual mean air temperature increased and runoff decreased, the CH4 oxidation percentage in CH4 production decreased. Likewise, the variations in hydrological conditions significantly affected CH4 diffusive (8.1-12.6%) and ebullient (23.6-28.9%) fluxes. Accompanied by a decrement in runoff, the CH4 ebullient flux percentage in CH4 production increased rapidly, while the CH4 diffusive flux percentage in CH4 production decreased slowly (Table 1). This means that diffusion was not always the dominant reservoir sediment CH4 emission pathway, while annual CH4 ebullient flux surpassed diffusive flux and dominates CH4 emissions during the dry year.

Spatiotemporal Feedback of Sediment Methane Production and Oxidation
Generally, daily reservoir sediment CH4 production increased as air temperatures  Furthermore, hydrological conditions influence the response of sediment CH 4 cycling processes to climate warming. Under the same SSP, even though the differences in annual CH 4 production (or oxidation) among the hydrological years were insignificant (less than 5%), the CH 4 oxidation percentage in CH 4 production decreased as annual runoff decreased (Table 1). In summary, when the annual mean air temperature increased and runoff decreased, the CH 4 oxidation percentage in CH 4 production decreased. Likewise, the variations in hydrological conditions significantly affected CH 4 diffusive (8.1-12.6%) and ebullient (23.6-28.9%) fluxes. Accompanied by a decrement in runoff, the CH 4 ebullient flux percentage in CH 4 production increased rapidly, while the CH 4 diffusive flux percentage in CH 4 production decreased slowly (Table 1). This means that diffusion was not always the dominant reservoir sediment CH 4 emission pathway, while annual CH 4 ebullient flux surpassed diffusive flux and dominates CH 4 emissions during the dry year.

Spatiotemporal Feedback of Sediment Methane Production and Oxidation
Generally, daily reservoir sediment CH 4 production increased as air temperatures increased from January to July and decreased as air temperatures decreased from August to December under the three SSPs (Figures 3 and 5a). This indicated that the methanogenic rate varies with air temperature. Accounting for air temperature differences among the SSPs, daily CH 4 production under the SSP2-4.5 was generally higher than under the SSP1-2.6 from January to early June (Figures 3 and 5a). Between mid-June and August, the mean air temperature under the SSP2-4.5 was slightly higher compared to the SSP1-2.6 (0.6 • C). This means that there was no significant difference in CH 4 production (29.1 t C day −1 ). Meanwhile, daily CH 4 production was generally highest under the SSP5-8.5 before September, while mean CH 4 production was 444.8 t C day −1 (or 415.8 t C day −1 ) higher compared to the SSP1-2.6 (or SSP2-4.5) from mid-June to August. After August, differences in daily CH 4 production among the three SSPs were not significant. Two reasons could explain this phenomenon. Firstly, under the warmer SSPs, given that more POC was consumed in reservoir sediment from January to August, especially in the shallow-water zone and the drawdown area ( Figure 6), there was less POC for CH 4 production after August. Secondly, less CH 4 was produced under the cooler SSPs. Because CH 4 was mainly produced in sediment from mid-June to August, differences in CH 4 production among the SSPs during this period mainly resulted in a nonlinear relationship between annual CH 4 production and annual mean air temperature. Secondly, less CH4 was produced under the cooler SSPs. Because CH4 was mainly produced in sediment from mid-June to August, differences in CH4 production among the SSPs during this period mainly resulted in a nonlinear relationship between annual CH4 production and annual mean air temperature.   Temporal and spatial CH 4 production showed significant differences under the same SSP during all three representative hydrological years (Figures 5b and 6). Under dryer conditions, the sediment temperature response rate to air temperature variation increased in both the reservoir's shallow-and deep-water zones, causing a time-lag decrease in CH 4 production variation. Moreover, compared with CH 4 production in shallow-and deep-water zones, CH 4 production in the drawdown area is highest during the period when temperatures increased and lowest during the period when temperatures decreased. These led to larger intra-year variations of CH 4 production under drier conditions (Figures 5b and 6). Additionally, among the three hydrological years, CH 4 production differences during the former period were nearly cancelled out by differences during the latter period, leading to non-significant differences in annual CH 4 production under the same SSP.
In general, the feedback trend in daily sediment CH 4 oxidation to warming under all three hydrological conditions was similar to that of CH 4 production ( Figure 5). Nevertheless, daily CH 4 oxidation under the SSP5-8.5 was slightly higher than the other SSPs from September to October (Figure 5c). This is because methanotrophic rates increased under the warmer SSPs, which were subject to similar daily CH 4 production among all SSPs during this period. Besides, the CH 4 oxidation percentage in CH 4 production in the reservoir's drawdown area is inversely proportional to air temperature. That is, as the air temperature increased, the percentage of CH 4 oxidation decreased. Under the warmer SSPs, this phenomenon became more apparent.

Feedback of Sediment Methane Emissions
The feedback trend in daily sediment CH 4 diffusive flux to warming was similar to that of CH 4 production (Figures 5a and 7a). Additionally, mean diffusive flux between mid-June and August under the SSP5-8.5 was significantly larger, being 85.5 t C day −1 and 80.1 t C day −1 larger than that under the SSP1-2.6 and SSP2-4.5, respectively. In other words, the difference in mean diffusive flux during that period between SSP1-2.6 and SSP2-4.5 was insignificant (Figure 7a). Besides, the impacts of hydrological conditions on diffusive flux differed from CH 4 production under the same SSP (Figures 5b and 7b). Due to different runoff during three hydrological years, the differences in water level became apparent. At the same time, as air temperature increased, the differences in daily CH 4 diffusive flux became gradually significant ( Figure 7b). As the air temperature decreased, the differences became insignificant (Figure 7b). The maximum differences occurred in mid-September. That is, daily CH 4 diffusive flux in mid-September during the wet year is 129.9 t C day −1 and 283.1 t C day −1 larger than that during the normal and dry years, respectively. Consequently, different hydrological conditions could enlarge the differences in daily CH 4 diffusive flux under the SSPs. diffusive flux became gradually significant (Figure 7b). As the air temperature decreased, the differences became insignificant (Figure 7b). The maximum differences occurred in mid-September. That is, daily CH4 diffusive flux in mid-September during the wet year is 129.9 t C day −1 and 283.1 t C day −1 larger than that during the normal and dry years, respectively. Consequently, different hydrological conditions could enlarge the differences in daily CH4 diffusive flux under the SSPs. Compared with the feedback trend in daily sediment CH4 diffusive flux to warming, daily sediment CH4 ebullient flux increased or decreased more rapidly as air temperature ( Figure 7c). Additionally, mean ebullient flux between mid-June and August under the SSP5-8.5 was significantly larger, being 231.6 t C day −1 and 227.0 t C day −1 larger than that under the SSP1-2.6 and SSP2-4.5, respectively. During this period, the differences in CH4 ebullient flux between SSP5-8.5 and SSP1-2.6 (or SSP2-4.5) were more significant than those in diffusive flux. Compared Figure 7b and Figure 7d, in general, ebullition was the main sediment CH4 emission pathway in the summer. Moreover, the variations in hydrological conditions affected the time duration of ebullition as the main emission pathway, lasting nearly four months, less than three months, and several days during the dry, normal, and wet years, respectively. Additionally, the increment in time duration resulted in reservoir CH4 released into the atmosphere increasing sharply.

Reservoir Methane Emission to the Atmosphere
When CH4 was transported from sediment to the water column and then to the atmosphere via diffusion, most was oxidized in the water column while only 0.1-0.2 × 10 3 t C year −1 was released to the atmosphere under the SSPs during all three hydrological years (less than 0.2%). Nevertheless, CH4 in bubbles is seldom released into the water column. In other words, most CH4 in bubbles is released into the atmosphere under the SSPs during all three hydrological years (51.3-137.7 × 10 3 t C year −1 , 67.8-84.6%). Consequently, the CH4 ebullient flux feedback to climate warming is more important than that of diffusive flux, the former of which will mainly affect future climate change. CH4 ebullient flux to the atmosphere was higher under the warmer SSPs (i.e., 76.8, 82.7, and 103.3 × 10 3 t C year −1 under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 during the normal year, respectively). Meanwhile, CH4 ebullient flux to the atmosphere increased as runoff decreased (i.e., 56.8, Compared with the feedback trend in daily sediment CH 4 diffusive flux to warming, daily sediment CH 4 ebullient flux increased or decreased more rapidly as air temperature ( Figure 7c). Additionally, mean ebullient flux between mid-June and August under the SSP5-8.5 was significantly larger, being 231.6 t C day −1 and 227.0 t C day −1 larger than that under the SSP1-2.6 and SSP2-4.5, respectively. During this period, the differences in CH 4 ebullient flux between SSP5-8.5 and SSP1-2.6 (or SSP2-4.5) were more significant than those in diffusive flux. Compared Figure 7b and Figure 7d, in general, ebullition was the main sediment CH 4 emission pathway in the summer. Moreover, the variations in hydrological conditions affected the time duration of ebullition as the main emission pathway, lasting nearly four months, less than three months, and several days during the dry, normal, and wet years, respectively. Additionally, the increment in time duration resulted in reservoir CH 4 released into the atmosphere increasing sharply.

Reservoir Methane Emission to the Atmosphere
When CH 4 was transported from sediment to the water column and then to the atmosphere via diffusion, most was oxidized in the water column while only 0.1-0.2 × 10 3 t C year −1 was released to the atmosphere under the SSPs during all three hydrological years (less than 0.2%). Nevertheless, CH 4 in bubbles is seldom released into the water column. In other words, most CH 4 in bubbles is released into the atmosphere under the SSPs during all three hydrological years (51.3-137.7 × 10 3 t C year −1 , 67.8-84.6%). Consequently, the CH 4 ebullient flux feedback to climate warming is more important than that of diffusive flux, the former of which will mainly affect future climate change. CH 4 ebullient flux to the atmosphere was higher under the warmer SSPs (i.e., 76.8, 82.7, and 103.3 × 10 3 t C year −1 under the SSP1-2.6, SSP2-4.5, and SSP5-8.5 during the normal year, respectively). Meanwhile, CH 4 ebullient flux to the atmosphere increased as runoff decreased (i.e., 56.8, 82.7, and 110.6 × 10 3 t C year −1 under the SSP2-4.5 during the wet, normal, and dry year, respectively). Thus, annual CH 4 ebullient flux (137.7 × 10 3 t C year −1 ) was highest under the SSP5-8.5 during the dry year, accounting for 33% of CH 4 production. Likewise, annual CH 4 ebullient flux (51.3 × 10 3 t C year −1 ) was the lowest under the SSP1-2.6 during the wet year, accounting for 16% of CH 4 production.

Conclusions
Climatic and hydrological conditions influence both reservoir methane (CH 4 ) cycling processes and emissions. This study combined a reservoir CH 4 cycle model and a Statistical DownScaling Model (SDSM) to evaluate reservoir CH 4 cycling feedbacks under multiple climate change scenarios while accounting for hydrological uncertainty. China's Danjiangkou Reservoir was used as a case study to evaluate the impact of climate change on spatiotemporal reservoir CH 4 cycling processes. Due to differences in daily summer air temperatures under the three SSPs, annual CH 4 production and ebullition showed a significant and increasing nonlinear relationship with the annual mean near-surface air temperature. The annual CH 4 oxidation percentage in annual CH 4 production decreased when the annual mean air temperature increased and runoff decreased. At the same time, the phenomenon that the CH 4 oxidation percentage in CH 4 production in the reservoir's drawdown area was inversely proportional to air temperature became more apparent under the warmer SSPs. Besides, although diffusion was the main CH 4 emission pathway in sediment during the wet and normal years, CH 4 ebullient flux surpassed diffusive flux and dominated CH 4 emissions during the dry year. Likewise, less than 0.2% of CH 4 released from sediment via diffusion was released to the atmosphere (0.1-0.2 × 10 3 t C year −1 ), while 67.8-84.6% of CH 4 released from sediment via ebullition was released to the atmosphere (51.3-137.7 × 10 3 t C year −1 ). Thus, the CH 4 ebullient flux feedback to climate warming was more important than that of diffusive flux. CH 4 ebullient flux increased as air temperatures increased, and runoff decreased. That is, annual CH 4 ebullient flux was highest (137.7 × 10 3 t C year −1 ) under the SSP5-8.5 during the dry year, while it was lowest (51.3 × 10 3 t C year −1 ) under the SSP1-2.6 during the wet year. This means that high air temperatures and low runoff aggravated CH 4 emissions as well as climate warming, while low air temperatures and high runoff were beneficial in reducing reservoir CH 4 emissions. This study could help researchers predict reservoir CH 4 emission dynamics while directing decision-making policies for reservoir sustainability.
This study did not account for the impacts of future precipitation changes on reservoir CH 4 cycling processes due to the difficulty of accurate precipitation prediction. Meanwhile, POC loading varied in reservoir sediment under different hydrological conditions. This could affect the modeling results of reservoir CH 4 cycling flux. After the improvement in GCMs and decrease in uncertainty, accounting for the impacts of multiple factors (air temperature, precipitation, and POC loading affected by runoff) on reservoir CH 4 cycling will be a prospective research topic. Besides, reservoir CH 4 emissions could be effectively controlled by reservoir operation policy adjustments. A reservoir operation policy that considers both economic benefit improvements and CH 4 emission reductions is urgently needed to mitigate the positive feedback between reservoir CH 4 emissions and climate warming.  Data Availability Statement: Partial data can be obtained from the websites and references mentioned in the text. Partial data on the case study are available from the corresponding author upon reasonable request.