Vegetation Affects Timing and Location of Wetland Methane Emissions

Common assumptions about how vegetation affects wetland methane (CH4) flux include acting as conduits for CH4 release, providing carbon substrates for growth and activity of methanogenic organisms, and supplying oxygen to support CH4 oxidation. However, these effects may change through time, especially in seasonal wetlands that experience drying and rewetting, or change across space, dependent on proximity to vegetation. In a mesocosm study, we assessed the impacts of Typha on CH4 flux using clear flux chamber measurements directly over Typha plants (“whole‐plant”), adjacent to Typha plants (where roots were present but no stems; “plant‐adjacent”), and plant‐free soils (“control”). During the establishment phase of the study (first 30 days), the whole‐plant treatment had ~5 times higher CH4 flux rates (51.78 ± 8.16 mg‐C m−2day−1) than plant‐adjacent or control treatments, which was primarily due to plant‐mediated transport, with little contribution from diffusive‐only flux. However, porewater CH4 concentrations were relatively low directly below whole‐plant and in neighboring plant‐adjacent treatments, while controls accumulated a highly concentrated reservoir of porewater CH4. When the water table was drawn down to simulate seasonal drying, reserve porewater CH4 from control soil was released as a pulse, equaling the earlier higher CH4 emissions from whole‐plants. Plant‐adjacent treatments, which had neither plant‐mediated CH4 transport nor a concentrated reservoir of porewater CH4, had low CH4 flux throughout the study. Our findings indicate that in seasonal wetlands, vegetation affects the timing and location of CH4 emissions. These results have important mechanistic and methodological implications for understanding the role of vegetation on wetland CH4 flux.


Introduction
Wetlands take up carbon dioxide (CO 2 ) via primary productivity and emit methane (CH 4 ) via anaerobic decomposition (Segers, 1998). Relative to other ecosystems, wetlands cover only 5-8% of the terrestrial landscape (Mitsch & Gosselink, 2007) but play a significant role in climate regulation (Bastviken et al., 2011;Dean et al., 2018). Estimates of wetland greenhouse gas flux and understanding of carbon storage are limited by high uncertainty. This uncertainty is, in part, due to variation attributable to vegetation-effects on CH 4 production, consumption, and transport, which ultimately influence cumulative wetland CH 4 emissions (Carmichael et al., 2014).

The Role of Vegetation on CH 4
Production and transport of CH 4 from wetlands are controlled by a series of interacting biophysical mechanisms. Production of CH 4 is dependent on carbon availability, anoxic conditions, soil pH, reduction-oxidation (redox) potential, and temperature (Bridgham et al., 2013;Neubauer et al., 2005), while CH 4 emissions are controlled by diffusion, convection, ebullition, and ventilation (Bridgham et al., 2013;Kayranli et al., 2010;Yavitt & Knapp, 1998). Wetland vegetation influences many of these mechanisms. In terms of CH 4 production and consumption, vegetation provides carbon substrates to fuel microbial processes (Christensen et al., 2003). These carbon substrates are derived from decomposition of dead plant materials or more directly through root exudates (Carmichael et al., 2014;Tittel et al., 2019). However, the lacunar air ventilation system of many wetland species that supplies oxygen (O 2 ) to the rhizosphere can also reoxidize alternative electron acceptors such as sulfate, which inhibits CH 4 production due to microbial competition for carbon substrates (Dalcin Martins et al., 2017;Neubauer et al., 2005;Sutton-Grier & Megonigal, 2011). In addition, plant transport of O 2 stimulates aerobic oxidation of CH 4 by methanotrophic bacteria (Conrad, 2009;Faußer et al., 2012;Laanbroek, 2010). It has been reported that 30-90% of CH 4 produced in the anaerobic environment is oxidized before reaching the atmosphere (Holzapfel-Pschorn et al., 1986). In terms of CH 4 transport, the same air ventilation system for O 2 also allows porewater CH 4 to diffuse directly from the rhizosphere to the atmosphere, bypassing the hydrologic diffusive barrier along the sedimentwater-atmosphere continuum (Bendix et al., 1994;Colmer, 2003;Knoblauch et al., 2015;Laanbroek, 2010). Thus, a large portion of CH 4 released from wetlands has been attributed to plant-mediated transport Carmichael et al., 2014;Shannon et al., 1996) at rates faster than diffusion through the water column. Clearly, there are several interactions and tradeoffs among mechanisms that influence the ultimate role of vegetation in wetland CH 4 emissions. Given these mechanistic links between vegetation and CH 4 , it is not surprising that wetland CH 4 flux has been linked to phytomass and net primary productivity (Bhullar, Iravani, et al., 2013;Cheng et al., 2007;Christensen et al., 2003;Turetsky et al., 2014).

Spatial and Temporal Considerations on the Role of Vegetation on CH 4
While many studies have examined individual plant-mediated mechanisms of CH 4 production, consumption, and transport, relatively few studies address how these mechanisms interact through space and time to influence cumulative CH 4 emissions (Neubauer et al., 2005). For example, transport through macrophyte stems can occur after emergence, during peak growth, and following senescence, albeit the degree of transport is dependent on phenology (Kim et al., 1999;Yavitt & Knapp, 1998). In contrast, nonvegetated areas of wetlands emit CH 4 through transport processes independent of plant phenology, such as diffusion and ebullition (Chanton et al., 1989), which may occur across different time frames during the growing and nongrowing seasons. Plant-mediated transport can deplete porewater CH 4 directly below plants (Shannon et al., 1996), which may create a spatial porewater CH 4 gradient causing lateral CH 4 transport toward plants and away from nonvegetated locations adjacent to plants. This loss of porewater CH 4 adjacent to plants not only affects CH 4 emissions in nonvegetated locations but also has important implications for the placement of chamber-based measurements of CH 4 flux. A chamber over plant stems captures the combined effects of plant-mediated transport and carbon substrate supply (Hu et al., 2016;Jeffrey, Maher, Johnston, Kelaher, et al., 2019;Kankaala et al., 2005;Martin & Moseman-Valtierra, 2015;Milberg et al., 2017), while a chamber adjacent to plants only captures the effects of carbon substrate supply (Lawrence et al., 2017;Picek et al., 2007).
Finally, hydrological dynamics interact with wetland vegetation and CH 4 production, consumption, and transport (Bansal et al., 2018;Kim et al., 2012). The role of plant-mediated transport may diminish when the diffusion barrier is lost during drying events (Bhullar, Iravani, et al., 2013), which can occur daily in coastal wetlands or seasonally in ephemerally ponded wetlands. Similarly, the presence of vegetation may enhance recovery of CH 4 following rewetting through supply of carbon substrates or delay recovery through supply of O 2 (Kim et al., 2012). Linking these spatiotemporal considerations has important implications for interpreting and assessing the overall effects of vegetation on wetland CH 4 emissions and will strengthen our understanding and modeling of wetland carbon cycling.
The role of vegetation in CH 4 dynamics is increasingly relevant as wetland plant communities shift, often toward dominant invasive macrophytes such as Typha × glauca, Phragmites australis, and Phalaris arundinacea (Bansal et al., 2019;Lawrence et al., 2017;Rey-Sanchez et al., 2018;Zedler & Kercher, 2004), and as practitioners try to balance multiple ecosystems services during management (Badiou et al., 2011;Eviner et al., 2012). The objective of this study was to improve our mechanistic understanding of how emergent vegetation affects CH 4 flux over space, time, and variable hydrological conditions. Specifically, we sought to quantify how CH 4 flux rates change from directly over vegetation compared to adjacent or in plant-free soils and explore temporal changes in CH 4 flux rates through stages of vegetation and hydrologic phenology in a simulated seasonal wetland mesocosm.

Study Site and Species
The Prairie Pothole Region (PPR) is the largest wetland ecosystem in North America covering over 820,000 km 2 and consists of millions of glacially formed, depressional wetlands (Dahl, 2014). Hydroperiods of PPR wetlands range from ephemeral (~2 weeks) to permanent (year-round) with a majority (87%) of these basins classified as seasonal, emergent wetlands (Dahl, 2014). The seasonality of PPR wetlands is not unique; many wetlands worldwide experience hydrologic drawdown during parts of the year (Galatowitsch, 2012;Kayranli et al., 2010;Kim et al., 2012). The PPR region is particularly important for migratory bird and waterfowl habitat, but PPR wetlands are also important to national, continental, and global carbon budgets (Euliss et al., 2006) and have been identified as CH 4 hot spots (Bansal et al., 2016;Tangen et al., 2015). Typha is a dominant genus in the region, with a recent expansion of nonnative T. angustifolia and hybrid T. × glauca since the 1960s (Ralston et al., 2007;Stewart & Kantrud, 1971). Typha can influence soil organic carbon content via litter accumulation (Vaccaro et al., 2009) and can increase soil CH 4 emissions (Lawrence et al., 2017). We chose to use Typha in this study because of not only its prevalence in the PPR but also its ubiquitous distribution in freshwater wetlands worldwide (Bansal et al., 2019).

Mesocosms, Treatments, and Design
We used a mesocosm approach to isolate mechanisms of vegetation effects on CH 4 emissions and avoid confounded sources of variation that occur under field conditions, such as episodic weather events, soil porosity heterogeneity, and seasonal temperature variability. Mesocosms were established in 40-L glass aquaria (30 × 50 × 25 cm). Soils for the mesocosms were collected from a wetland at Northern Prairie Wildlife Research Center in Jamestown, North Dakota, United States (46°52′N, 98°38′W). The wetland has a dominant ring of emergent macrophytes (mostly Typha) around the edge and open water in the center. The mineral soils of PPR wetlands are classified within the Mollisol order, Aquoll suborder (Soil Survey Staff, Natural Resources Conservation Service, 1999). Soils for this study consist of black and very dark gray silty clay loam of the Parnell series (Soil Survey Staff, Natural Resources Conservation Service, 2020). In autumn 2017, following Typha senescence, soils were collected from the top 20 cm of sediment where water depth was~50 cm and stored in buckets under saturated anoxic conditions. Saturated soils were passed through a 6-mm sieve to remove coarse vegetation and debris and homogenized. We assumed that any low-molecular-weight, labile carbon substrates from previous plant exudates were consumed before soil was distributed in mesocosms; however, nonlabile C likely remained (Waldo et al., 2019). Each mesocosm was filled with uniform soil to a depth of 10 cm.
Our experiment consisted of three treatments: "whole-plant", "plant-adjacent," and "control". In total there were 18 replicates (n = 6 per treatment). The whole-plant and plant-adjacent treatments were located within the same mesocosm separated by an acrylic barrier that extended 2 cm into the sediment but not completely to the mesocosm bottom. The control treatments were located in separate mesocosms with no plants (Figure 1). Two control treatments were placed in a single mesocosm with a complete physical barrier between each side. Mesocosms were rearranged in the laboratory periodically to avoid confounding effects of environmental heterogeneity in the laboratory.
Mesocosms were kept in laboratory conditions, with an air temperature~23°C, and under full spectrum LED lights (KingLED, Shenzhen, China; VYPRx PLUS LED, Fluence Bioengineering, TX, United States). Lights were kept on a 12-to 16-hr photoperiod. An opaque polyvinyl chloride (PVC) collection chamber base (diameter = 20 cm) was permanently placed in the center of each replicate for gas flux measurements. Supports were used to keep each base raised above the bottom of the aquarium to allow lateral movement of water and root growth. PVC porewater sampling pipes of 2.5-cm diameter with 1-mm slits around the bottom 2 cm (Geoprobe, Salinas, KS, United States) were installed in all treatments for dissolved gas porewater measurements. Porewater sampling pipes were capped between sampling events. PVC pipes of 5-cm diameter with 1-mm slits around bottom 2 cm were inserted into a subset of two mesocosms per treatment for dissolved oxygen (DO) probes.
The whole-plant treatments were planted with T. latifolia (Roundstone Seeds, KY, United States) seedlings that were germinated in potting soil and reached an approximate plant height of 7 cm. As plants grew, their roots and rhizomes were free to grow into the plant-adjacent treatment of the mesocosm. Shoots that emerged in the plant-adjacent treatment were clipped weekly to maintain a roots/rhizomes-only condition. The control mesocosms were maintained plant free by clipping any germinates from the remnant seed bank. Excessive surface algae were removed manually from all mesocosms when present.

Timeline
The experiment took place from autumn 2017 to spring 2019, for a total of 400 days. There were five phases related to vegetation phenology or hydrological conditions:~1 month of establishment (days 1-28), 8 months of growth (days 29-274),~1 week of surface water drawdown (days 275-286),~1 week of dry soil (days 287-295), and~3 months after rewetting (days 296-400) ( Table 1). Each phase had periodic CH 4 flux measurements: establishment (three rounds of flux measurements), growth (four rounds), drawdown (two rounds), dry soil (one round), and rewetting (three rounds). Uptake of CO 2 rapidly increased in the whole-plant treatment during the establishment phase as plants grew in height; the growth phase was distinguished from the establishment phase as the point when CO 2 uptake by plants leveled off (see Bansal et al., 2020 for CO 2 and N 2 0 flux data). Dissolved CH 4 was measured in surface water during the establishment, growth, and rewetting phases and in porewater during the growth and rewetting phases. Dissolved O 2 was measured in porewater during the growth, drawdown, dry, and rewetting phases. Porewater CH 4 and DO were not sampled during the establishment phase due to limited sampling time and supplies. Water levels were maintained at~4 cm depth throughout the study by adding deionized water using a watering container, except during drawdown and dry soil phases. During the drawdown phase, soils were allowed to dry until the water table was below the sediment surface, with 40-50% volumetric water content (10-cm Hydrosense II, Campbell Scientific, Logan, Utah). We characterized the dry soil phase by cracks in the soil surface and <40% volumetric water content. Soils were rewetted to~4-cm standing water for the rewetting phase.
At the conclusion of the experiment, soil samples (~10 g) were collected and analyzed for percent organic matter; soils were sieved, dried at 105°C for 72 hr then 500°C for 5 hr. Another set of sieved soil samples (~10 g) were analyzed of % C (combustion method, North Dakota State University Soil Testing Lab). Belowground biomass was assessed in three of the mesocosms with plants to confirm lateral root and rhizome growth from the whole-plant into the plant-adjacent treatments. All soil was rinsed from plant structures, roots and rhizomes were separated, and all material was dried at 60°C for 72 hr. While we did not measure root and rhizome growth throughout the study, the first clipping of shoots in the plant-adjacent treatment began within the first 30 days of the study, indicating roots and rhizomes had grown into the plant-adjacent treatment during the establishment phase.

Methane Flux Measurement, Surface Water and Porewater Gas Concentrations
CH 4 gas flux measurements were conducted periodically throughout the experiment (see section 2.3). Gas flux was measured using a closed static chamber system with a high-frequency infrared gas analyzer (Gasmet DX4040, Gasmet Technologies Oy, Helsinki, Finland), which has an accuracy of 61.6 ppb for CH 4 . Measured gas flux over plants thus integrates flux due to plant-mediated transport and flux due to diffusion through the water column. Care was taken to minimize disturbance to soil and plants to diminish artificially induced ebullition. Due to our mesocosm and chamber design, determining experimental from natural ebullition was not possible, and measurements indicating ebullition (i.e., concentration increases >10 ppm over 1 min) were remeasured. A clear acrylic chamber (diameter = 20 cm, height = 20 to 100 cm, varied with height of plant, Figure 1) was placed onto the PVC base for 10-30 min. The taller chambers placed over the plant treatments included two internal fans (688 mm 2 ; placed 30 cm from bottom of chamber) to circulate air within the chamber, while there were no fans in the shorter control and plant-adjacent sampling chambers to avoid any turbulence at water surface. Air and soil temperature (Fluke 54 II B thermometer, Fluke Co, WA, United States) were recorded during each flux sampling. Gas concentrations were corrected based on chamber volume, pressure, and air temperature using the ideal gas law, and flux rates were calculated as the change in gas concentration over time and checked for appropriate fit using the HMR package in R 3.5.1 (Pedersen et al., 2010). The HMR package uses mean square errors to recommend a linear, nonlinear (Hutchinson & Mosier, 1981), or no flux fits based on the value of an exponential parameter. We visually inspected graphs of each flux measurement to manually select the recommended fit or a more appropriate fit (i.e., if p values and confidence intervals for nonlinear and linear fit were similar, we picked the more conservative linear fit). Out of 312 total flux measurements, 6% were nonlinear fits and 81% were linear fits. Thirteen percent of our measurements had no flux. Surface water CH 4 concentrations were used to calculate diffusive flux rates using a k 600 value of 0.01 (Bansal et al., 2020). We recognize there is error associated with parameters of the diffusive flux calculation (i.e., if we had used a k 600 value of 0.1 or 0.001). We worked to minimize this error by choosing an appropriate k 600 value of 0.01 m hr −1 by using k 600 :area relationships calculated from previous research on PPR wetlands (Dalcin Martins et al., 2017) and literature values from research on similar small ponds (Holgerson & Raymond, 2016).
Surface water and porewater dissolved gas concentrations were measured using the headspace equilibration method (Hope et al., 1995;Jahangir et al., 2012). Duplicate surface water samples were collected 2 cm below the surface, and duplicate porewater samples were collected from the bottom of the 2.5-cm diameter PVC sampling pipes after evacuation of standing water. Nitrogen gas (35 ml) was added to the surface water or porewater samples (25 ml) and was mixed via vigorous shaking for at least 3 min. Headspace gas was analyzed on a gas chromatograph (SRI Instruments, CA, United States), equipped with a flame ionization detector and electron capture detector; eight standards were run every 11 samples to generate calibration curves for both detectors. Optical sensors (accuracy of ±0.3 mg-O 2 L −1 , which is~3.7% DO at 25°C; PME minidot, Vista, California) measured continuous DO at 1-hr intervals. Surface water electrical conductivity and pH (accuracy of ±2%FS and ±0.01 pH; ExStick EC500, Extech Instruments, Nashua, NH, United States) were recorded in February 2018 and 2019 and October 2018, with average values of and 2,835 ± 35 μS m −1 and 8.7 ± 0.07 pH, respectively.

Statistical Analyses
To test for treatment differences in soil % C, soil organic matter, and belowground biomass harvested at the end of the experiment, we used analysis of variance (ANOVA). We used linear mixed effects ANOVA to test the effects of treatment, phase, and their interaction on CH 4 flux rates and dissolved CH 4 concentrations. Mean values of gas flux rates or dissolved gas concentrations across rounds within each phase for each Replicate was considered a random effect to account for repeated measurements over the course of the study. Least significant difference tests were used to compare means among treatments within phases. Data were log transformed to meet ANOVA assumptions of homoscedasticity of error variance and normality. Analyses were conducted using the lmertest package in R 3.5.1 (Kuznetsova et al., 2017), with Kenward-Roger estimates for degrees of freedom, and assessed significance at α = 0.05. All mean values are reported as mean ± standard error throughout results.
To model CH 4 flux rates over time (to help separate signal from noise), data were fitted to a mixed general additive model, with day of experiment, treatment, and their interaction as fixed effects and replicate as a random effect, using the mgcv package in R 3.5.1 (Wood, 2017) using a Gaussian family object. The modeled CH 4 flux rates were used to calculate cumulative CH 4 emissions over time for each treatment. It is important to note that day of experiment indirectly accounts for our experimental manipulation of water level (intended to mimic seasonal draw down of temporary wetlands).
To investigate relationships between porewater DO and CH 4 , we used Spearman rank correlation with the CH 4 flux or porewater concentration and daily mean of porewater DO. To investigate diurnal patterns of DO, we calculated the amplitude of each day's measurement from each chamber as the difference between the daily maximum and minimum concentrations. Then we calculated phase mean value of amplitude for each chamber and tested for treatment differences using linear mixed effects ANOVA to test the effects of treatment, phase, and their interaction on DO amplitude.

Methane Flux, Surface Water and Porewater Gas Concentrations
There were significant effects of treatment, phase, and treatment by phase interaction on CH 4 flux rates (Table 2 and Figure 2a). During the establishment phase of the study, the whole-plant treatment had CH 4 flux rates 4-6 times higher than the plant-adjacent and the control treatments (Figures 2a). Measured CH 4 flux rates from the whole-plant treatment were 96% greater than calculated diffusive flux rates (calculated using surface water CH 4 concentrations), indicating plant-mediated CH 4 transport as the dominant process contributing to measured flux during establishment (Figure 3). The average chamber-measured CH 4 flux rate for the control treatment during the establishment phase was 13.18 ± 9.35 mg-C m −2 day −1 , and the average calculated diffusive flux rate was 6.18 ± 1.37 mg-C m −2 day −1 ; thus, chamber-based measurements had a higher magnitude and also greater variability (Figure 3). The discrepancy is likely due to sampling differences; for measured fluxes, the "foot print" of flux was the entire surface area of chamber area, giving a more integrated measurement that may have also included microbubbles of CH 4 in addition to diffusive flux, as opposed to the two single syringes of water collected from surface water for calculated diffusive flux. Likewise, there is a known time frame over which chamber flux was measured, while the calculated diffusive rate is based on an estimated piston velocity, which is related to tank size and assumed laboratory conditions. The control treatment CH 4 flux rates increased almost threefold from the

10.1029/2020JG005777
Journal of Geophysical Research: Biogeosciences establishment phase up to 30.88 ± 10.31 mg-C m −2 day −1 in the growth phase. During the growth phase, all three treatments had similar flux rates (Figure 2a). From the growth phase to the water table drawdown phase, the whole-plant treatment CH 4 flux rates declined, the plant-adjacent flux rates remained consistent, while the control treatment CH 4 flux increased to a rate (432.50 ± 9.35 mg-C m −2 day −1 ) that was 14 times greater than during the growth phase (Figure 2a). During drawdown, the control treatment flux rates were significantly higher than the whole-plant treatment flux rates. Finally, during the dry and rewet phases, CH 4 flux rates were low and similar among treatments (Figure 2a).
The modeled temporal patterns of CH 4 flux rates over the course of the study for the whole-plant treatment were relatively high in the beginning and then steadily declined; for the plant-adjacent treatment, flux rates were consistently low; and for the control treatment, flux rates were low in the beginning, peaked in middle, and then low at the end (Figure 4a). This temporal pattern of CH 4 flux rates resulted in relatively high cumulative CH 4 emissions early in the study from the whole-plant treatment, but overall similar cumulative CH 4 emissions between the whole-plant and control treatment by the end of the study (Figure 4b). There were much lower cumulative CH 4 emissions from the plant-adjacent treatment over the course of the study compared to the whole-plant or control treatments (Figure 4b).
Porewater dissolved CH 4 concentrations were significantly affected by treatment, phase, and treatment by phase interaction (Table 2 and Figure 2b). During the growth phase, porewater CH 4 concentration was lowest in the whole-plant treatment, highest for the control treatment, and intermediate for the plant-adjacent treatment ( Figure 2b). All treatments had low and similar porewater CH 4 concentrations during the rewetting phase. There were no treatment differences (F 2,3 = 1.12, p = 0.43) in porewater DO during the study. There were differences (F 3,8 = 84.27, p < 0.001) in porewater DO concentrations among phases, which were 0.7 ± 0.001% during the growth phase, increased to 18 ± 0.85% during the drawdown phase, up to 90 ± 0.38% during dry phase, and then returned to 0.6 ± 0.001% after rewetting. There was no relationship between CH 4 flux and DO (r s = 0.22, p = 0.11; Figure S1 in the supporting information) or between porewater CH 4 and DO (r s = 0.11, p = 0.61), and no differences in diurnal amplitude in DO among treatments (F 2,3 = 0.04, p = 0.96; data not shown).

Temporal and Spatial Effects of Vegetation on CH 4 Flux
Our study demonstrates how CH 4 emissions are affected by the linkage between plant-mediated transport, porewater CH 4 concentrations, and the timing of water inundation. The elevated flux rates that we measured during Typha establishment align with the current consensus that vegetation acts as a conduit for CH 4 transport from the rhizosphere to the atmosphere at rates much higher than diffusive-only flux (Bendix et al., 1994;Colmer, 2003;Laanbroek, 2010). However, high rates of plant-mediated transport appeared to coincide with depleted porewater CH 4 reservoir below and around vegetation, which corresponded with lower flux rates from the whole-plant and plant-adjacent treatments later in the experiment. In contrast, nonvegetated controls had a highly concentrated CH 4 porewater reservoir that likely contributed to higher flux rates during simulated draw down of the water table that typically occurs later in the growing season in seasonal wetland. Ultimately, there were comparable total CH 4 emissions between the  Table 2). Phases were establishment ("est"), growth, drawdown ("down"), dry, and rewet. Note the log scale on the y axis of CH 4 flux rates. SE = standard error.
whole-plant and the control treatments. This finding challenges the common assumption that actively growing emergent vegetation increases CH 4 emissions from wetlands Carmichael et al., 2014), as the generalization may not apply in seasonally inundated wetlands (Altor & Mitsch, 2006. However, it should be noted that the soils used in our mesocosms were collected from locations that historically had vegetation; thus, there may have been residual, older plant-C inputs to fuel CH 4 production in our plant-free control treatment. Through combined measurements of dissolved porewater CH 4 and CH 4 flux rates, our study revealed potentially significant mechanisms in plant-mediated CH 4 flux. Low porewater CH 4 concentrations below plants was somewhat unexpected, as vegetation is known to supply fresh carbon substrates (i.e., root exudates) to fuel methanogenic CH 4 production (Christensen et al., 2003;King & Reeburgh, 2002;Knorr et al., 2008;Waldo et al., 2019). On the other hand, root oxygenation can increase CH 4 oxidation (Conrad, 2009;Faußer et al., 2012;Jeffrey, Maher, Johnston, Maguire, et al., 2019;Laanbroek, 2010) or alter availability of redox-active substrates (i.e., Fe (OH) 3 ) to favor nonmethanogenic microbial respiration (i.e., reduction of Fe (OH) 3 to Fe +2 ; Roden & Wetzel, 1996;Neubauer et al., 2007), both of which would result in lower porewater CH 4 below plants. However, DO concentrations were <1% in all treatments when soils were saturated, there were no treatment differences in diurnal DO patterns, and there was no relationship between DO and porewater CH 4 , all suggesting that there was not enough porewater DO to suppress CH 4 production or to enhance CH 4 oxidation. Thus, by observing high plant-mediated flux rates in combination with subsequent depleted porewater CH 4 , we suggest that CH 4 was being transported to the atmosphere relatively quickly after it was produced in the sediment below plants. Similar patterns of decreased sediment CH 4 in vegetated compared to nonvegetated sites have been observed in brackish tidal systems (Chanton et al., 1989;Gross et al., 1993) and in northern peatlands (Shannon et al., 1996). Plant-mediated transport allows CH 4 to quickly bypass the primary zone of oxidation that occurs near the water table surface (Bendix et al., 1994;Colmer, 2003;Knoblauch et al., 2015;Laanbroek, 2010). Therefore, our data suggest that Typha may reduce CH 4 oxidation temporally (through short residence time of porewater CH 4 that is quickly fluxed out via stems) and spatially (by allowing CH 4 to avoid the zone of oxidation). Even so, measurements of porewater CH 4 residence times, isotopic tracers, redox-active elements, or finer resolution oxygen probes would be necessary to confirm this finding with certainty.
Porewater CH 4 concetrations were moderately lower in the whole-plant treatment than in the plantadjacent treatment, which had similar root biomass as the neighboring whole-plant soil. This finding suggests that higher CH 4 flux via plant-mediated transport is, in part, enhanced by a supply of CH 4 that is produced and laterally transported via roots from adjacent soils within wetlands. This effect may have been magnified in our experiment due to the shallow depth of our aquarium soils. Root growth was bound by mesocosm volume and soil depth; therefore, the distance required for porewater CH 4 to diffuse via roots is likely shorter than in a natural wetland. For example, the flux rates we observed from the plant-adjacent and soil control treatments during establishment and growth phases (10-20 mg CH 4 -C m −2 day −1 ) were lower than flux rates observed across the PPR regions from 2003 to 2016 (21-270 mg CH 4 -C m −2 day −1 ; Tangen & Bansal, 2019a, 2019b, suggesting that there may be higher rates of CH 4 flux from plant-adjacent soils in a natural wetland than in our mesocosm study. The flux rates we observed from the whole-plant treatment during establishment and growth phases (20-80 mg CH 4 -C m −2 day −1 ) were similar in magnitude and less variable than Gleason et al. (2009) observed in a field assessment of vegetated wetland plots (50-250 mg CH 4 -C m −2 day −1 ). Thus, our measurements are in a similar range as natural

10.1029/2020JG005777
Journal of Geophysical Research: Biogeosciences wetlands in the PPR region, with smaller variation than natural wetlands, which allows us to make mechanistic inferences. To our knowledge, this is the first study to assess direct comparisons of CH 4 flux between plant stems and plant-adjacent soils in the PPR region. To better assess the complete impact of vegetation on wetland CH 4 budgets, future field studies should consider a "zone of influence" surrounding aerenchymous wetland plants, which accounts for root and stem transport of porewater CH 4 .
During a drying event, the water table drops, hydrostatic pressure is lowered, and the hydrologic barrier to free gas flow is removed, thereby allowing a rapid, pulsed release of porewater CH 4 to the atmosphere (Roslev & King, 1996). The rapid release of CH 4 during drying not only is dependent on water table dynamics but also requires a sufficiently large and concentrated porewater CH 4 reservoir to fuel the CH 4 pulse. Our study suggests that a drying-induced pulse of CH 4 may be considerably dampened in the presence of vegetation due to depletion of porewater CH 4 , possibly from plant-mediated transport, oxidation, or changes in redox conditions. In contrast, a drying-induced CH 4 pulse may be a dominant factor controlling annual CH 4 budgets for unvegetated wetlands (or unvegetated location within wetlands) if there is a sufficiently large reservoir pool of porewater CH 4 .
Water-logged soils facilitate anoxic conditions that promote CH 4 production. However, as exemplified in the present study and other recent research, the timing of inundation affects the magnitude of CH 4 emissions and how vegetation affects CH 4 flux. Altor and Mitsch (2006) found greater CH 4 emissions from permanently inundated areas than from intermittently flooded areas of experimental riparian marshes and found no difference in CH 4 flux between vegetated and nonvegetated plots in intermittently flooded areas. In Typha-dominated constructed wetlands, plants enhanced CH 4 emissions in wetlands with 40-60% soil moisture content but had no effect in a wetland with <40% soil moisture (McInerney & Helton, 2016). Still, more field studies are needed from seasonal wetlands to further elucidate how plant-mediated fluxes change across variable hydrology in natural conditions (Beringer et al., 2013;D Kim et al., 2012).
Recovery of CH 4 production and emissions following rewetting after drying is also potentially affected by the presence of wetland vegetation. During drying, there is consumption of labile carbon substrates by aerobic respiration, regeneration of alternate electron acceptors, reduction in size of methanogenic communities, and increased CH 4 oxidation, all of which delay recovery of CH 4 emissions by days to months following rewetting (Boon et Tian et al., 2012). Vegetation can speed up the rate of recovery by priming microbial activity via fresh carbon substrates (Ström et al., 2005;Waldo et al., 2019) but can also delay recovery by supplying O 2 for methanotrophic communities (Faußer et al., 2012) and by extending dry conditions through transpiration. In our study, O 2 levels were minimal in all treatments following rewetting, and none of the treatments exhibited high CH 4 flux rates or accumulated a porewater CH 4 reserve after 3 months. Although not significant, we observed a slight trend of lower flux rates and lower porewater CH 4 in the whole-plant treatments during the rewetting phase. The lack of CH 4 recovery after rewetting may be a result of limitations on carbon substrate from sources such as groundwater and decomposing plant material not being replenished in mesocosms as they would in a natural wetland, or because the fresh carbon inputs did not have enough time to accumulate for microbial consumption (Neubauer et al., 2005;Sutton-Grier & Megonigal, 2011;Updegraff et al., 1995).

Modeling and Management Implications
Budgets of CH 4 at local to global scales are often estimated using process-based models such as DNDC (Li, 2000), Ecosys (Grant & Roulet, 2002), and CLM4Me (Riley et al., 2011). These models couple biological and physical processes and are highly sensitive to water table depth, vegetation, and various CH 4 transport pathways. A recent review on CH 4 models identified a need to increase understanding of individual CH 4 processes over vertical and horizontal space, as well as hot moments and hot spots, as crucial for improving model predictions . In this paper, we outline several relevant mechanisms to help improve the underlying assumptions that drive these models. In particular, we demonstrate how vegetation can influence the timing of CH 4 emissions and residence time of porewater CH 4 , lateral transport of porewater CH 4 , and effects of water table drawdown. Our findings also have implications for empirically derived, data-driven models that predict flux at annual time steps over entire wetland systems. We demonstrate the dynamic temporal and spatial interaction between wetland vegetation and hydrology on seasonal CH 4 fluxes (Riley et al., 2011). Thus, process-based and empirical CH 4 models need to consider intra-annual hydrologic dynamics (e.g., drying and rewetting events) and spatial wetland heterogeneity (e.g., vegetated and nonvegetative cover, aerenchymous and nonaerenchymous vegetation types) for accurate predictions of CH 4 budgets.
Wetland drainage and restoration also have important impacts on CH 4 budgets. While wetland drainage generally decreases CH 4 emissions (Tangen et al., 2015), we show how drainage can produce a temporary, large pulse of CH 4 , offsetting the subsequent decrease in CH 4 flux. While the flooding that is required to restore wetlands may increase CH 4 emissions (Audet et al., 2013;Tuittila et al., 2000), our results suggest that rewetting wetlands later in the season could result in lower CH 4 emissions relative to early-season flooding. A better understanding of the mechanisms affecting CH 4 recovery could aid in reducing CH 4 emissions with wetland restoration (Jerman et al., 2009;Runkle et al., 2019). More research is needed, ideally using high-frequency data such as from eddy covariance flux towers (e.g., FLUXNET-CH 4 database, Knox et al., 2019) to understand short-term effects of drying and rewetting on CH 4 flux.
Wetland management actions to control problematic invasive macrophytes include crushing, cutting above/below water, harvesting, and herbicide (Bansal et al., 2019;Carson et al., 2018;Hazelton et al., 2014;Keyport et al., 2019). These manipulations of emergent vegetation could affect plant-mediated CH 4 production, consumption, and transport (Zhu et al., 2007). In terms of ecosystem services, the role of vegetation on CH 4 flux and wetland carbon budgets (and consequently carbon storage potential) should be considered when deciding management actions (Eviner et al., 2012). Our study highlights how the timing of management may affect CH 4 . For example, cutting and drowning Typha stems during their establishment may prevent the large CH 4 flux rates we observed early on in our study, but this action may only be effective in permanently inundated wetlands. If a wetland dries later in the season, there may be a CH 4 release that eliminates any benefit from early season cutting. Including the effect of vegetation on the timing of CH 4 fluxes, drying-induced pulses of CH 4 , and rewetting recovery time frames will improve wetland CH 4 budgets and management decisions.

Data Availability Statement
The data are in the ScienceBase repository and are available at Bansal et al. (2020).