Hydrologic Lag Effects on Wetland Greenhouse Gas Fluxes

Hydrologic margins of wetlands are narrow, transient zones between inundated and dry areas. As water levels fluctuate, the dynamic hydrology at margins may impact wetland greenhouse gas (GHG) fluxes that are sensitive to soil saturation. The Prairie Pothole Region of North America consists of millions of seasonally-ponded wetlands that are ideal for studying hydrologic transition states. Using a long-term GHG database with biweekly flux measurements from 88 seasonal wetlands, we categorized each sample event into wet to wet (W→W), dry to wet (D→W), dry to dry (D→D), or wet to dry (W→D) hydrologic states based on the presence or absence of ponded water from the previous and current event. Fluxes of methane were 5-times lower in the D→W compared to W→W states, indicating a lag ‘ramp-up’ period following ponding. Nitrous oxide fluxes were highest in the W→D state and accounted for 20% of total emissions despite accounting for only 5.2% of wetland surface area during the growing season. Fluxes of carbon dioxide were unaffected by transitions, indicating a rapid acclimation to current conditions by respiring organisms. Results of this study highlight how seasonal drying and re-wetting impact GHGs and demonstrate the importance of hydrologic transitions on total wetland GHG balance.


Introduction
Concerns over climate change have stimulated considerable research focused on providing a better understanding of the key sources and sinks of biologically produced and consumed greenhouse gases (GHGs) such as carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O) [1][2][3]. Globally, wetland ecosystems have been recognized as prospective carbon storage sites where carbon uptake exceeds losses. Wetlands, however, have the potential to emit and consume significant amounts of potent GHGs such as CH 4 and N 2 O [3][4][5][6]. Therefore, the ability to accurately estimate the overall GHG balance of wetland ecosystems is necessary to determine their role as radiative sources or sinks.
Wetland systems often are distinguished by highly dynamic GHG fluxes that display substantial temporal and spatial variability. This variability has been associated with a myriad of factors including abiotic conditions (e.g., soil moisture and temperature), biotic communities (e.g., microbial, vegetation), emission pathway (diffusive flux, ebullition, and plant-mediated flux), wetland type (e.g., peatland, mineral-soil, tidal), and weather [4,[6][7][8][9][10][11]. Wetland hydrology, specifically inundation or ponding, is a primary factor that directly and indirectly influences the GHG and carbon balance of wetland systems [4,12]. Throughout the growing season, many wetlands are typified by fluctuations between ponded and dry conditions, which are often separated by a moving, hydrologic transition area, or moist-margin [13]. This intermediate area may function as a hot spot for GHG production and flux that differs from the adjacent ponded and dry areas [14]; thus, while the areal coverage of this transition area may be relatively small, it may have a disproportionate contribution to overall wetland GHG as a hot spot for GHG production and flux that differs from the adjacent ponded and dry areas [14]; thus, while the areal coverage of this transition area may be relatively small, it may have a disproportionate contribution to overall wetland GHG fluxes. Despite this, few studies have examined abiotic conditions or GHG fluxes of hydrologic transition areas.
The Prairie Pothole Region (PPR) of North America has been identified as an area of interest owing to its potential role in the continental GHG budget [6, 15,16]. The PPR, which encompasses approximately 800,000 km 2 of the United States and Canada, is characterized by millions of small, mineral-soil wetlands embedded within a landscape composed of cropland and grassland. Wetlands of the PPR can be highly productive, and consequently are capable of sequestering atmospheric carbon in soils and vegetation [15,16]. These wetlands can also be sources or sinks of GHGs, most notably CH4 and N2O. Studies have demonstrated considerable temporal variation in GHG flux of PPR wetlands, and have recognized relations between flux and factors such as soil moisture, temperature, water depth, and land use. Although we are unaware of any studies that specifically examine GHG flux from hydrologic transition areas of PPR wetlands, Bansal et al. [11] demonstrated differences in CH4 flux among saturation states (ponded, semi-saturated, and dry) of a seasonal PPR wetland, and Pennock et al. [8] associated N2O flux events with areas of PPR wetlands that were distinguished by rapid soil drainage.
Most wetlands in the PPR are classified as seasonal (i.e., ponds completely dry during most years) [17,18], with concentric zones consisting of distinct vegetation communities that vary based on hydrology and soil properties. PPR wetlands are fed through atmospheric inputs from snowmelt runoff and precipitation, and evapotranspiration exceeds precipitation in the region [13,19,20]. Therefore, a large proportion of wetlands draw down (i.e., water levels recede) or completely dry seasonally (Figure 1), depending on the magnitude of inputs (precipitation, runoff) and losses (evapotranspiration, recharge) [13]. As water levels fluctuate, the vegetation zones transition between wet, moist-soil, and dry conditions. Thus, PPR wetlands are ideal microcosms for studying the effects of hydrological transitions. The time series depicts water-level drawdown throughout the ice-free season, which is characteristic of many seasonal wetlands. In this example, water depths gradually declined from approximately 33 cm during April until the site dried during early July; the wetland remained dry throughout the remainder of the season. (b) Depiction of ponded (blue) and dry (white) portions of a depressional wetland basin along a seasonal hydrologic gradient from ponded to dry; the checkered area represents the transition area between the ponded and dry areas (i.e., wet to dry). PPR wetlands often begin the The time series depicts water-level drawdown throughout the ice-free season, which is characteristic of many seasonal wetlands. In this example, water depths gradually declined from approximately 33 cm during April until the site dried during early July; the wetland remained dry throughout the remainder of the season. (b) Depiction of ponded (blue) and dry (white) portions of a depressional wetland basin along a seasonal hydrologic gradient from ponded to dry; the checkered area represents the transition area between the ponded and dry areas (i.e., wet to dry). PPR wetlands often begin the ice-free season with high water levels from snowmelt runoff and spring precipitation. Water levels typically recede throughout the year because evapotranspiration exceeds precipitation. Previous research has demonstrated variation in GHG fluxes among vegetation zones and landscape elements of PPR wetlands [6, 16,[21][22][23]. Studies have also identified areas of partially-saturated soils as potential hot spots for GHG flux [8,11,14]. The goal of this study was to assess GHG fluxes from hydrologic and transition states (i.e., wet to wet, dry to wet, dry to dry, or wet to dry) of PPR wetlands. To accomplish this goal, we analyzed a published, long-term database consisting of GHG fluxes (CH 4 , N 2 O, CO 2 ) and associated hydrologic data from seasonal PPR wetlands [24].

GHG Data
We obtained data for this study from a published, comprehensive database of wetland GHG fluxes [24]. The studies that constitute this database have resulted in numerous peer-reviewed publications detailing GHG fluxes of PPR wetlands [6, 10,11,[25][26][27][28]. This database includes CH 4 , N 2 O, and CO 2 fluxes from nearly 200 seasonal and semipermanent (classification of [17]) PPR wetlands that were embedded in grasslands and croplands; data were collected between 2003 and 2016. Flux data were collected using the using the static-chamber approach and abiotic covariates, such as water depth, were provided at the individual sample location (i.e., gas-collection chamber; hereafter, chamber). In addition, we obtained topographic data describing the representative surface areas for each of the chambers (e.g., concentric chamber zones), which were based on mid-point elevations between adjacent chambers [26]. The topographic data were based on detailed surveys performed at each wetland catchment using a GPS survey system. Methodologies among the various studies that comprise these data remained consistent over time, and are detailed by Gleason et al. [25], Finocchiaro et al. [26], Tangen et al. [6], and Tangen and Bansal [24]. Below, we provide a general overview of the gas-collection methods.
Within each wetland catchment, samples were collected and abiotic data were measured along a transect spanning the wetland-upland gradient. Five sample locations were equally distributed between the approximate wetland center and edge, and another three were established at the toe-, mid-, and shoulder-slope positions in the adjacent uplands ( Figure 2). When water depths were less than approximately 5 cm, chamber bases were permanently installed at the beginning of each field season. Mid-day (9:00-16:00 h) flux measurements were made biweekly (every 2 weeks) throughout the ice-free season. For flux measurements, opaque polyvinyl chloride chambers (20-cm diameter, 20-cm height) were attached to bases, or when water depths exceeded 5 cm the chambers were attached to floats. When using floats, chambers were deployed and secured to previously-installed poles using a pully system to minimize disturbance to the underlying sediments. Chambers were deployed for approximately 30 min after which headspace gas samples were collected using a syringe through a rubber septum and transferred to pre-evacuated 10-mL glass bottles. Gas samples were analyzed within 3 weeks of collection using a gas chromatograph equipped with electron capture and flame ionization detectors (SRI 8610C, SRI Instruments, Torrance, CA).

Database Modification
The comprehensive database [24] was modified as follows. We included only seasonal wetlands owing to their regional abundance and hydrologic characteristics that result in distinctive water-level declines during most years. We retained only data for the five wetland-zone samples (chambers 1-5) collected from April through September, and for wetlands where data were available for all five sample locations (i.e., no missing values). For this subset of data, we determined the sample interval (days between repeated sample events) for corresponding samples, and removed samples separated by less than 12 or greater than 16 days (10th and 90th quantiles). The remaining data approximate the standard biweekly sampling scheme used for most of the associated studies [24]. We also removed data collected from drained wetlands because of the disrupted hydrology, which could confound analyses presented in this paper. The resulting database consisted of 9,840 samples (five samples/wetland/date) from 88 seasonal wetlands distributed across North and South Dakota, Minnesota, and Iowa. Of these wetlands, 17 were sampled for 1 year, 45 for 2 years, 20 for 4 years, and 6 for 8 years. Wetlands were distributed among various land uses including native prairie (n = 24), restored grasslands (n = 41), and cropland (n = 23). For the wetlands included here, average soil organic carbon, nitrogen, and bulk density in the upper 15 cm of the soil profile was 5.04% (se ± 0.23), 0.45% (se ± 0.02), and 0.95 g cm −3 (se ± 0.02), respectively [24]. Combining data from a range of studies facilitates the identification of robust trends across a range of climatic conditions, geographic locations, and soils that are representative of the PPR. The previously-mentioned topographic data were available for 84 of the 88 wetlands. On average, 74% of the 88 wetlands dried each year, although 0-100% dried during any given year. Moreover, 95% of the wetlands dried at least once during the various studies conducted from 2003-2016 (Table 1). A wetland was considered dry when all sample locations were dry (water depth = 0), although the soils may have been saturated or partially saturated.

Analyses
To facilitate analyses based on changing hydrologic conditions, we created a hydrologic state variable to categorize each sample based on water depths of the previous and current sample dates (~2-week interval) within each season. Sample locations that remained ponded or unponded between sampling dates were categorized as wet to wet (W→W) or dry to dry (D→D), respectively. Sample locations that transitioned between ponded and unponded conditions were categorized as wet to dry (W→D) or dry to wet (D→W).
To demonstrate seasonal hydrologic variability, we determined the areal extent of the four hydrologic states and calculated the average monthly percentages of total wetland surface area that were in each hydrologic state. Surface areas were determined using the aforementioned topographic survey data collected at each wetland. We also tested for differences in CH 4 , N 2 O, and CO 2 fluxes among the four hydrologic states using analysis of variance (ANOVA) methods. In the models, the repeated samples (several years and multiple dates within years) were treated as subsamples in time, and wetland and sample location (chamber) were considered random effects. Prior to analyses, CH 4 , N 2 O, and CO 2 fluxes were natural log transformed after adding a nominal value to make fluxes positive. Least square means (LSM) were back transformed for data presentation and statistical significance was inferred using α = 0.05. Furthermore, we calculated average monthly flux for each hydrologic state, as well as an overall flux for all states combined. Lastly, we estimated an average cumulative seasonal (April-September) flux (CH 4 , N 2 O, and CO 2 ) for each hydrologic state that incorporated the surface area of each state. To do this, we calculated average surface area (m 2 ) of Atmosphere 2019, 10, 269 5 of 13 each hydrologic state based on daily wetland totals; if a hydrologic state did not occur for a given wetland/date combination a value of zero was assigned for that state. We then scaled the average GHG flux (LSM; mg CH 4 /CO 2 or µg N 2 O m −2 hr −1 ) to a daily flux (24 h) and multiplied it by the average daily area of each hydrologic state. The resulting average flux per wetland per day was multiplied by 183 days to approximate the average seasonal flux for each hydrologic state.

Hydrology
The average percentage of wetland surface area in the W→W hydrologic state decreased from April (40%) through September (14%) while the percentage of wetland surface area in the D→D hydrologic state increased from 53% to 83% (Figure 2). The percentage of W→D wetland surface area peaked in July (9%) as wetland water levels receded. The percentage of D→W wetland area was 3% from April through June and 1% from July through September (Figure 2). significance was inferred using α = 0.05. Furthermore, we calculated average monthly flux for each hydrologic state, as well as an overall flux for all states combined. Lastly, we estimated an average cumulative seasonal (April-September) flux (CH4, N2O, and CO2) for each hydrologic state that incorporated the surface area of each state. To do this, we calculated average surface area (m 2 ) of each hydrologic state based on daily wetland totals; if a hydrologic state did not occur for a given wetland/date combination a value of zero was assigned for that state. We then scaled the average GHG flux (LSM; mg CH4/CO2 or μg N2O m −2 hr −1 ) to a daily flux (24 h) and multiplied it by the average daily area of each hydrologic state. The resulting average flux per wetland per day was multiplied by 183 days to approximate the average seasonal flux for each hydrologic state.

Hydrology
The average percentage of wetland surface area in the W→W hydrologic state decreased from April (40%) through September (14%) while the percentage of wetland surface area in the D→D hydrologic state increased from 53% to 83% (Figure 2). The percentage of W→D wetland surface area peaked in July (9%) as wetland water levels receded. The percentage of D→W wetland area was 3% from April through June and 1% from July through September (Figure 2).     (Figure 3c). When combined, CO 2 flux from the D→D and W→D states was three times greater than flux from the W→W and D→W states. There were no differences in CO 2 flux between the D→D and W→D or between W→W and D→W states (Figure 3c).  (Figure 3c). When combined, CO2 flux from the D→D and W→D states was three times greater than flux from the W→W and D→W states. There were no differences in CO2 flux between the D→D and W→D or between W→W and D→W states (Figure 3c).  states differed. Area-adjusted cumulative N 2 O and CO 2 flux, however, displayed dissimilar patterns where the relative contribution from the W→D state was reduced when the area of the hydrologic states was considered (Figure 3b-c,e-f). Based on the area-adjusted cumulative fluxes, the W→D and D→W hydrologic states collectively accounted for 9.0% of CH 4 flux, 20.1% of N 2 O flux, and 7.5% of CO 2 flux, while accounting for a combined 7.0% of wetland surface area. The D→W (1.8% of wetland surface area) and W→D (5.2% of wetland surface area) hydrologic states accounted for 1.2% and 7.8% of CH 4 flux, respectively (Figure 3d), while the W→W state (26.5% of wetland surface area) accounted for 86.5% of CH 4 flux. For N 2 O, the D→W and W→D hydrologic states accounted for 0.3% and 19.7% of N 2 O flux (positive flux only), respectively, while the D→D state (66.5% of wetland surface area) accounted for 79.9% of total N 2 O emissions (Figure 3e). For CO 2 , the D→W and W→D hydrologic states accounted for 0.7% and 6.8% of CO 2 flux, respectively, while the D→D and W→W states accounted for 81.0% and 11.5%, respectively (Figure 3f).

Greenhouse Gas Fluxes
Average monthly fluxes of CH 4 , N 2 O, and CO 2 generally increased from April through July and declined thereafter (Figure 4a-c). Monthly CH 4 fluxes, however, increased throughout the year for the W→W, W→D, and D→W states (Figure 4a). Similarly, monthly N 2 O fluxes from the W→D state remained relatively high throughout the year (Figure 4b). N 2 O fluxes from the D→D state, although lower than those from the W→D state, were also relatively stable throughout the year. Monthly CO 2 fluxes from all hydrologic states displayed seasonal variability similar to the overall monthly average, and the drier states (D→D, W→D) generally displayed greater average fluxes than the wetter states (W→W, D→W) (Figure 4c).

Discussion
As the hydrologic condition of a wetland changes over time, so does the spatial distribution of GHG fluxes. Through examination of a long-term database, we demonstrated seasonal variability in PPR wetland hydrology and revealed distinct patterns in GHG fluxes among four hydrologic states (Figure 3a-c). Fluxes of CH4 and N2O were strongly affected by the previous hydrological state and

Discussion
As the hydrologic condition of a wetland changes over time, so does the spatial distribution of GHG fluxes. Through examination of a long-term database, we demonstrated seasonal variability in PPR wetland hydrology and revealed distinct patterns in GHG fluxes among four hydrologic states (Figure 3a-c). Fluxes of CH 4 and N 2 O were strongly affected by the previous hydrological state and displayed a lag effect, while CO 2 was minimally affected by transitions between wet and dry states. These findings demonstrate the importance of considering hydrologic variability or state (e.g., W→W, W→D) when collecting GHG samples, estimating fluxes, or interpreting GHG data of dynamic wetland systems.

CH 4
It has been well established that CH 4 flux from wetlands is influenced by, among other things, soil saturation and temperature, and that seasonal flux often peaks during mid-summer when conditions are optimal for the production of CH 4 [6,11,16,25,28,29]. Correspondingly, overall average monthly CH 4 fluxes (solid line in Figure 4a) presented as part of this study generally increased from April through July and decreased nearly 30% from July to September (Figure 4a). These qualitative trends, which are based on average wetland fluxes for the entire wetland, suggest that CH 4 production and flux decline as the season progresses and abiotic (e.g., temperature, soil moisture) and biotic (e.g., vegetation) conditions change. Yet, when CH 4 fluxes are examined in relation to hydrologic state or condition, it is apparent that fluxes can remain high throughout the season in W→W, D→W, or W→D locations (Figure 4a).
It was not surprising that the highest and lowest average CH 4 fluxes (Figure 3a) were associated with the W→W and D→D states as methanogenesis occurs in water-logged, anoxic environments. The contribution from the W→D state represents an additional source of CH 4 that could be excluded from estimates or models that do not capture this seasonally-transient area of PPR wetlands. In many instances, soils of W→D areas remain saturated just a few centimeters below the surface [11,13], providing conditions suitable for the production and flux of methane. The loss of the diffusive barrier associated with ponded water also may promote CH 4 flux from the recently dried soils [30,31], which also can absorb solar radiation and be warmer than inundated soils. There is also less opportunity for CH 4 consumption by methanotrophy as the gas does not have to diffuse through the oxygenated water column prior to being emitted to the atmosphere [32,33]. The finding of lower CH 4 fluxes from the D→W state, compared to the W→W state, suggests that there may be a lag ramp-up period in the production of CH 4 once a non-ponded site is inundated [34]. This potential lag may be due to availability of oxidized alternative electron acceptors to fuel microbial activity, or CH 4 could be consumed in an oxidized soil layer between the deeper, anoxic soils and the overlying water column of these newly-flooded locations [32,35].

N 2 O
Previous research has demonstrated seasonal variability in N 2 O flux, although N 2 O fluxes tend to be more episodic than CH 4 or CO 2 [6, 16,25,26]. N 2 O flux from wetlands in a cropland setting has also been shown to be greater than those associated with grasslands [6, 25,36]. The observation of intermittent N 2 O fluxes has been associated with factors such as weather (precipitation, spring thaw), fertilizer application, soil moisture, and carbon availability [22,[36][37][38]. Based on the long-term data analyzed for this study, overall N 2 O fluxes were relatively low earlier in the year when temperatures are relatively low and a larger proportion of wetlands are inundated compared to summer and fall (Figures 2 and 4b). These early-season conditions (e.g., ponded water, anoxic sediments) typically are not optimal for denitrification and subsequent N 2 O flux [39]. When examined in relation to hydrologic condition, however, it is clear that the greatest N 2 O flux rates are associated with the W→D transition state, regardless of time of year (Figure 4b). These recently dried areas, which often are characterized by partially-saturated soils, can provide optimal conditions for N 2 O production [8]. Process-based literature suggests that N 2 O production through denitrification peaks when soil moisture is at, or slightly exceeds, field capacity (~60% water-filled pore space; WFPS) [39,40]. Correspondingly, Pennock et al. [8] and Gleason et al. [25] reported peak N 2 O emissions from PPR wetlands when soil WFPS declined below 60%. Results of this study also showed that the D→D state can be important contributor to N 2 O flux, despite the aerobic conditions that often characterize these areas. Nitrification has shown to be a contributor to N 2 O flux from such environments, especially in areas characterized by high mineralization rates [22,36,39,41].

CO 2
Results of this study revealed typical seasonal trends in CO 2 flux [25,42,43] that corresponded to trends in CH 4 and N 2 O; seasonal fluxes generally increased until July and decreased thereafter (Figure 4c). Previous studies of PPR wetlands have revealed seasonal relations between wetland CO 2 flux and soil moisture [21,25,26]. Daniel et al. [44] reported no difference in CO 2 flux along a hydrologic gradient of Playa wetlands, but the highest mean values were from the drier areas near the wetland edge. Accordingly, we anticipated differences in CO 2 flux between the W→W, D→D, and transition (W→D, D→W) states due to the associated soil properties (e.g., oxygen, soil moisture), which regulate microbial activity. For example, we assumed that the W→D state was relatively saturated and anoxic given the continued flux of CH 4 and high flux of N 2 O; thus, differences in CO 2 flux were expected between the D→D and W→D hydrologic states. However, unlike CH 4 and N 2 O, there was no effect of hydrologic transitions on CO 2 . Rather, the D→W and W→W states had similar CO 2 fluxes, as did the W→D and D→D states (Figure 3c). These findings demonstrate that when soils transition from wet to dry or from dry to wet, the associated effect on CO 2 flux is relatively quick (<2 weeks). These observations are likely due to a combination of factors associated with inundation, including reduced rates of decomposition, uptake of dissolved CO 2 into surface waters, and lower diffusive flux rates compared to rates from exposed soils and vegetation. Ultimately, the presence or absence of ponded water appears to be a primary factor associated with wetland CO 2 flux.

Extrapolating Fluxes
GHG flux data are often presented by landscape positions or sample zones due to the inherent differences among them. However, overall estimates for the entire wetland are often preferred by flux modelers, land managers, and policy makers [45][46][47][48]. When calculating GHG flux at the wetland level, considering the distribution of the samples (e.g., vegetation zones, landscape positions, ponded or non-ponded) and the representative surface area of the hydrologic states, should result in improved estimates. For example, projecting flux estimates from the upper landscape positions of a wetland catchment to the entire catchment could overestimate CO 2 flux and underestimate CH 4 flux [23]. Conversely, CH 4 flux could be overestimated by projecting flux estimates from the ponded portion of a wetland to the entire basin, especially for wetlands that partially or completely dry throughout the ice-free season [49]. Similarly, basin-level flux estimates based on the ponded and non-ponded portions of a wetland, but not the transition areas (e.g., D→W or W→D), could affect GHG flux [8,22]. For example, excluding the transition states in CH 4 and N 2 O mean flux calculations results in 15% and 31%, respectively, lower estimates than those that included all the hydrologic states. Therefore, it is important to adjust estimates according to the proportion of the wetland that each hydrologic state represents, including the hydrologic transition area.
While the number and scope of wetland GHG studies is expanding in the PPR, many of the initial studies were aimed at developing methodologies, examining processes, and identifying the important abiotic and landscape factors associated with the GHG balance of wetlands. Moving forward, policy and decision makers require regional estimates detailing the primary sources and sinks of GHGs [48]. Although various remote-sensing technologies are being developed to identify and describe wetlands [50][51][52], many PPR studies rely on existing spatial databases to identify wetlands and estimate wetland areas at various spatial scales. Commonly-used databases in the U.S. include the U.S. Fish and Wildlife Service's National Wetland Inventory (NWI) and the U.S. Geological Survey's National Land Cover Database (NLCD) [53][54][55]. Wetland areas provided by these databases represent a snapshot in time and are adequate for providing coarse estimates; however, they do not capture the seasonal and annual variability in wetland hydrology, which should be taken into consideration to refine flux estimates. Creed et al. [23] suggested that the coupling of digital terrain analyses with remote-sensing techniques could enhance our ability to characterize dynamic hydrologic parameters such as soil moisture. For example, satellite detection of surface water using Landsat or Synthetic-aperture radar imagery (e.g., Dynamic Surface Water Extent, Global Surface Water Explorer) are available at 8-to 16-day time steps, thus providing an opportunity to generate wetland maps that capture dynamic inter-and intra-annual hydrologic transitions [56][57][58].

Conclusions
Investigations of wetland hydrology, biotic communities, and water chemistry have demonstrated the importance of considering the short-and long-term hydrologic variability of PPR wetlands when analyzing data and interpreting results [13,[59][60][61]. Previous research has demonstrated variation in GHG fluxes among landscape positions of PPR wetlands [16,[21][22][23], and with this study we demonstrated the importance of considering dynamic hydrologic states when assessing GHG fluxes from seasonally-ponded wetlands. Sample designs and flux projections that account for hydrologic states should result in improved estimates of wetland GHG flux. For researchers aiming to generate GHG budgets of wetlands, our findings suggest that more frequent sampling of individual wetlands (at the expense of more wetlands with fewer samples) would be necessary for an accurate assessment of CH 4 and N 2 O, while fewer samples may be sufficient for CO 2 . Future research should include integration of intra-annual, within-wetland hydrology to improved GHG flux estimates for seasonal wetlands.