Inter-Annual Variability of Area-Scaled Gaseous Carbon Emissions from Wetland Soils in the Liaohe Delta, China

Global management of wetlands to suppress greenhouse gas (GHG) emissions, facilitate carbon (C) sequestration, and reduce atmospheric CO2 concentrations while simultaneously promoting agricultural gains is paramount. However, studies that relate variability in CO2 and CH4 emissions at large spatial scales are limited. We investigated three-year emissions of soil CO2 and CH4 from the primary wetland types of the Liaohe Delta, China, by focusing on a total wetland area of 3287 km2. One percent is Suaeda salsa, 24% is Phragmites australis, and 75% is rice. While S. salsa wetlands are under somewhat natural tidal influence, P. australis and rice are managed hydrologically for paper and food, respectively. Total C emissions from CO2 and CH4 from these wetland soils were 2.9 Tg C/year, ranging from 2.5 to 3.3 Tg C/year depending on the year assessed. Primary emissions were from CO2 (~98%). Photosynthetic uptake of CO2 would mitigate most of the soil CO2 emissions, but CH4 emissions would persist. Overall, CH4 fluxes were high when soil temperatures were >18°C and pore water salinity <18 PSU. CH4 emissions from rice habitat alone in the Liaohe Delta represent 0.2% of CH4 carbon emissions globally from rice. With such a large area and interannual sensitivity in soil GHG fluxes, management practices in the Delta and similar wetlands around the world have the potential not only to influence local C budgeting, but also to influence global biogeochemical cycling.


Introduction
Wetlands are particularly good locations for sequestering atmospheric carbon (C) through the uptake, transformation, and storage of CO 2 into plant biomass [1][2]. While the vegetation in many wetlands have high rates of CO 2 uptake from photosynthesis, anaerobic soils reduce the rate of organic matter decomposition associated with live and dead root fractions, litter and woody debris deposited within the soil, and particulate organic C incorporated into the soil [3][4][5]. This balance is favorable for storing organic C, and thus potentially providing a natural atmospheric filter for CO 2 -based greenhouse gas emissions (GHG) while simultaneously immobilizing C over long time periods. Indeed, high net primary productivity coupled with reduced decomposition of soil-associated C has jettisoned wetlands to the forefront of scientific curiosity and C legislation as the global atmospheric C pool rises and mitigation is explored [6].
Increases in CO 2 concentrations in the atmosphere are driven by increased cement production, fossil fuel emissions, and land use change [7], such that concentrations are increasing by 1.9 ppm/year, which equates to perennial increases of 8.6 Pg C/year [8] (1 Pg = 10 3 Tg = 10 6 Gg). Given that soil CO 2 emissions are approximately 68 Pg C/year [9], reducing annual soil CO 2 emissions from wetlands offer a potential mechanism to help offset atmospheric CO 2 loading. However, specific CO 2 -reduction management regimes must be identified, targeted, and prescribed on a large scale. Also, management regimes must not facilitate emissions of more deleterious gases, such as CH 4 . In fact, despite a much smaller increase in CH 4 emissions in recent years [7], CH 4 still accounts for 25% of the current global warming trend [10]. Wetlands account for approximately 20-25% of the global CH 4 emissions [11], and scientists and managers are facing difficulties amalgamating the vast number of studies targeting GHG emissions in wetlands [2,12] to identify better hydrologic, vegetation, and soil management strategies on scales that make a difference.
From a C balance perspective, CO 2 is the primary molecular source of C into and out of the atmosphere surrounding wetland environments, nearly always being more important than CH 4 [12][13]. However, because CH 4 has a greater ability to contribute to global warming than CO 2 , by a factor of 32 [14], a good proportion of wetland studies focus intently on CH 4 [15][16][17][18]. Thus, while wetlands do not emit greater amounts of CH 4 to the atmosphere than CO 2 [12], an incremental change in CH 4 emissions has a disproportionately stronger influence on global warming than a similar shift in CO 2 [12,19]. Accordingly, a primary driver of emissions among GHGs over time is land use change and management [20][21].
Shifts in agricultural production of specific crops and irrigation strategies have become a focal point for GHG research in some regions of the world [22][23]. For example, in the Sacramento-San Joaquin Delta of California, USA, agricultural sites managed as drained (pasture, corn field) served as net ecosystem sources of C, emitting 341 g C/m 2 /year of CO 2 and 11 g C/m 2 /year of CH 4 , while agricultural sites managed as flooded (rice paddy, restored wetlands) served as net ecosystem sinks of C, taking up 397 g C/m 2 /year of CO 2 while simultaneously releasing a greater proportion of CH 4 under this hydrologic regime (ranging from 39 to 53 g C/m 2 /year) [24]. CO 2 and CH 4 emissions from rice paddy and restored wetlands can vary widely among location based on a number of environmental factors, but often related to water table management [24][25][26][27]. Here-in, and for this reason, we focus our current study on gaseous soil C fluxes from different wetland types in the Liaohe Delta of Northeast China to add to a growing body of research that scales assessments spatially [13,28].
Both the scale of land-use in the Liaohe Delta and year-to-year variation in CO 2 and CH 4 fluxes from specific wetlands have the potential to influence regional-and global-scale C cycling [29]. The Liaohe Delta encompasses 5922 km 2 of natural or managed lands, surrounded or bisected by only 678 km 2 of towns and roads [30]. Of those natural or managed lands, 55.5% (or 3287 km 2 ) are rice paddy (Japonica variety, Oryza sativa: 2465 km 2 ), reed (Phragmites australis: 786 km 2 ), seablite (Suaeda salsa: 32 km 2 ), or mixed communities of reed and seablite (4 km 2 ). While rice paddy encompasses the greatest agricultural area in the Delta, over 772 km 2 of Phragmites are managed (compared with 14 km 2 unmanaged) and harvested annually for paper production [30]. S. salsa marshes make up the smallest percentage but soils are often associated with high organic C and nitrogen concentrations [31]. Rice paddy and P. australis wetlands both typically serve as sinks for CO 2 , but emit CH 4 [24][25]32], while this course is less clear for S. salsa marshes [33][34][35][36].
Much uncertainty arises when measuring soil CO 2 and CH 4 fluxes infrequently or in singular years [26,[34][35], if the goal of such assessment is to upscale to larger areas and over multiple years. Inter-annual and seasonal variations in soil CO 2 and CH 4 fluxes are large in coastal wetlands of Northeast China [36][37][38]. Indeed, we recognized this in a previous study [38], which focused on linking environmental variables (e.g., salinity, soil temperature, water table depth, plant biomass) from a single year to emissions of CO 2 and CH 4 from P. australis marshes, S. salsa marshes, and rice paddy wetlands of the Liaohe Delta. Here we expand this research to span three very different years from the perspective of hydrology (2012-2014). We had two primary objectives. First, we wanted to know how much C is being lost annually from the soils of these habitat types on an areal basis as assessed using standard discrete sampling techniques. Second, we wanted to assess and discuss drivers of inter-annual variation in emissions of CO 2 and CH 4 from three primary wetland types in the Liaohe Delta. While this study does not measure total C balance, we focused intensely on soil C emissions, which are usually the most variable and uncertain component of the C cycle. The linkages among soil C emissions, vegetation type, hydrologic management, season, and spatial coverage of habitats are described across a deltaic region potentially large enough in extent to influence the global C budget.

Results
Inter-annual variability of CO 2 and CH 4 fluxes CO 2 fluxes varied from year-to-year, especially for Phrag2 and Rice (Fig 1A). Significant site by date interactions highlight this variability (Table 1). For example, while peak soil CO 2 fluxes were highest in Rice in August of 2012 (1831 mg CO 2 /m 2 /h) and July of 2014 (1937 mg CO 2 /m 2 /h), overall lower fluxes prevailed from that wetland type in 2013 (<937 mg CO 2 /m 2 / h). Capacity for Phragmites wetlands to emit CO 2 from the soil was demonstrated strongly on Phrag2 in 2013 for a single period in June (3339 mg CO 2 /m 2 /h, Fig 1A), and was otherwise fairly consistent among years. CO 2 fluxes among sites differed from each other on 17 of the 19 dates assessed. Though variable, on average CO 2 fluxes from Phrag2 (780 mg CO 2 /m 2 /h) were consistently higher (P < 0.05) than Suaeda1 (335 mg CO 2 /m 2 /h), Suaeda2 (402 mg CO 2 /m 2 /h), Phrag1 (476 mg CO 2 /m 2 /h), and Rice (500 mg CO 2 /m 2 /h) ( Table 2, Fig 2). A potential driver of inter-annual variability for CO 2 fluxes from specific sites was hydrology; i.e., Phrag2 and Rice (and the wider region) were flooded for a longer duration in the 2013 growing season from atypical river flooding (Fig 3). Along with water table depth fluctuations, seasonal differences in warming of soils and fluctuations in salinity from year-to-year also influence CO 2 fluxes on these sites (Fig 4).
Significant site by date interactions were noted for CH 4 fluxes as well (Table 1). In the year receiving more persistent flooding (2013), CH 4 emissions were 16 times higher from Phrag2 and 6 times higher from Rice than in the other two years ( Fig 1B). However, CH 4 flux increases were not statistically related to water table depth that year (P > 0.5), in contrast to overall correlations among years (Fig 4), suggesting interactive influences with other site variables in 2013. CH 4 fluxes differed among sites on 15 of the 19 dates assessed. Despite the site x date interaction, CH 4 emissions from Phrag2 (10.4 mg CH 4 /m 2 /h) were consistently higher (P < 0.05) than emissions from Phrag1, Suaeda1, and Suaeda2 (mean, 0.45 mg CH 4 /m 2 /h, Table 2, Fig 2) on most dates. Variability between the replicate S. salsa sites was minimal for CH 4 , as salinity kept CH 4 emissions low at these sites. In contrast, the two replicate Phragmites sites behaved very differently; CH 4 emissions were 8 times higher from Phrag2 than Phrag1 ( Table 2). Table 2 depicts means for CO 2 fluxes and CH 4 fluxes, along with some environmental variables, as an average of the three years (see S1 Table for additional variables). Of the factors identified as influencing CO 2 and CH 4 fluxes in the first year of study [38], most remained unchanged as predictors over three years. Briefly, the overall correlations (across sites) between soil CO 2 fluxes, soil Eh, soil temperature, aboveground biomass, water table depth, and HCO 3 concentrations were positive and significant, and soil CO 2 fluxes were negatively correlated with salinity. CO 2 fluxes were driven principally by the amount of aboveground plant biomass available to route CO 2 belowground and through plant tissue, or facilitate microbial soil respiration from exudate production. Thus, to the degree that environmental variables influence plant biomass, they also influence soil CO 2 fluxes. Direct influences of salinity on CO 2 emissions were not clear; high variability in CO 2 emissions at lower salinities (<16 PSU) with reduced variability at high salinity (Fig 4) is confounded by a different vegetation type, S. salsa versus P. australis, when salinities exceeded 24 PSU. Otherwise, the capacity for CO 2 fluxes is reduced in a seemingly linear fashion with salinity. On the other hand and with the exception of one data point from Phrag2, soil temperature limited soil CO 2 flux to < 690 mg CO 2 /m 2 /h below 18°C. While this is not imposing, much of the cumulative CO 2 fluxes from all wetland types in the Liaohe Delta occurred as soil temperatures rose above 18°C to a recorded high of 29°C over three years (Fig 4).

Drivers of inter-annual variability
The overall correlations (across sites) between CH 4 fluxes and water table depth, soil temperature, and pore water HCO 3 concentrations were positive and significant; whereas, the relationship between soil CH 4 fluxes, Eh, and salinity were negative and significant. There was no significant correlation between CH 4 emission rates and plant aboveground biomass, which was a little surprising since CH 4 is often routed through vegetation [15,17]; but see Materials and Methods about cutting Phragmites. Thus for CH 4 , two environmental variables were critical and provided even clearer thresholds than seen for CO 2 flux when analyzed across sites. Similar to CO 2 , the first limiting variable for CH 4 flux was soil temperature. The soils of the Liaohe Delta freeze each year, and major fluxes of CH 4 are not promoted strongly from any wetland type until soil temperatures exceed 18°C (Fig 4). While this was previously suggested [38], the relationship is strengthened by concurrence across the three years of study versus one. The second variable is salinity. This gives way to a second threshold of 18, in that CH 4 fluxes were widely suppressed (< 1 mg CH 4 /m 2 /h) at soil salinity concentrations >18 PSU across all three years of study (Fig 4). High CH 4 fluxes were associated with soil temperature >18°C and soil salinity <18 PSU. Thus, wetland management activities in the Liaohe Delta facilitating these two conditions may simultaneously facilitate greater CH 4 emissions (as long as pore water SO 4 2availability is also low). Hourly mean (± 1 SE) soil CO 2 and CH 4 fluxes by site. a) Soil CO 2 fluxes, and b) soil CH 4 fluxes by site over three years from five wetland sites (two Phragmites australis, two Suaeda salsa, one rice) located in the Liaohe Delta, China. Means followed by the same letters are not significantly different at α = 0.05. While these site means and differences represent the general trends persisting across all months sampled, a significant site by date interaction ( Emissions of CO 2 , CH 4 , and gaseous carbon from the Liaohe Delta Approximately 2861 Gg of C is estimated to be emitted from the soils of the Liaohe Delta wetlands annually (Table 3). This value ranges from 2508 to 3285 Gg C/year depending on the year that the estimate was made, and to a lesser degree the specific representative sites used to attain the estimate (Table 3). Site selection was especially important for P. australis in 2012 and 2013 when Phrag2 had 55% and 129% higher overall C emissions than Phrag1, respectively. When summed, C emissions from CH 4 were only 1.9% of the C emissions from CO 2 ( Table 3). Emissions of C from CH 4 equated to approximately 53 Gg C/year, but ranged more broadly from year-to-year for P. australis wetlands (10-18 times) versus S. salsa or Rice. Year-to-year variability was less within habitats for CO 2 emissions, but did vary by up to a factor of 2.6 for specific among year comparisons. Emissions of C from CO 2 equated to approximately 2808 Gg C/year.

Temporal scale and variability
Variability in gaseous C emissions from wetlands as a component of the mass C balance is important to consider when determining whether specific wetlands experience net gains,   4 /year (or about 7 Tg C/year), but uncertainty around this value is up to 100% [2]. Indeed, C emissions from CH 4 were over an order of magnitude higher from Phrag1 and Phrag2 in 2013 versus 2012 and 2014, and C emissions from CO 2 from Suaeda1 and Suaeda2 in 2014 were nearly double emissions in 2012 and 2013 (Table 3).
To compound this further, GHG techniques measure vastly different things [42]; e.g., compare large dark static flux chambers (30,250 cm 2 ) incorporating vegetation (as used here) versus small static chambers (80 cm 2 ) that exclude vegetation versus eddy covariance, which measures the net ecosystem exchange of C over many hectares. While we standardize our sampling area and techniques among years, year-to-year CO 2 fluxes varied by a factor of up to 2.6 and CH 4 fluxes varied by a factor of up to 18 in the Liaohe Delta within a specific wetland type. Similar trends were reported from the nearby Yellow River Delta, where complete reversals of CO 2 and CH 4 fluxes from soil uptake to efflux were documented for some wetland types (e.g., P. australis) among years [36], although the reasons were not discussed. Indeed, we can conclude that GHG assessments across multiple years are critical for determining mass C fluxes from wetlands.

Factors influencing area-scaled CO 2 and CH 4 emissions
On average, 28.0% of the C emissions from CO 2 were derived from P. australis wetlands (800 Gg C/year) and 71.2% were derived from rice (1985 Gg C/year), leaving only 0.8% associated with S. salsa wetlands (23 Gg C/year) ( Table 3). These differences are compounded mostly by the areal extent of each wetland type. Statistically, CO 2 flux from only one P. australis wetland (Phrag2) was consistently greater than the two S. salsa sites and rice when standardized over a square meter area (Table 2; Fig 2). Noteworthy among the different sites was the much greater C and nitrogen density in the soils of Phrag2 that may be influencing high CO 2 fluxes (S1 Table). Therefore, while aboveground biomass was undetermined for Phrag2 versus other sites (Table 2), the availability of C-based substrate within the soil to facilitate microbial respiration was over two times greater on Phrag2 than even Phrag1. The high proportion of soil organic C on this one P. australis site may be related to either the vegetation type itself or the particular water/harvest management influencing that site. The relative proportion of labile soil C in S. salsa soils of the Liaohe Delta was influenced greatly by the presence of vegetation [43]; in that case, bare soils versus S. salsa-vegetated tidal flats. P. australis plants are much larger than S. salsa, and have a strong ability to sequester C by maintaining high aboveground and belowground plant biomass [32]. Persistent flooding would also keep soils anaerobic and further limit decomposition; both P. australis sites were often flooded and maintained water tables above ground during sampling ( Table 2). Annual commercial harvesting of P. australis for pulp production in the Liaohe Delta compromises the role that perennial pulses of litterfall would play in facilitating nutrient recycling in this wetland type, and perhaps even upset some biogeochemical processes spatially across the Delta adding further to variation in CO 2 (and CH 4 ) emissions from P. australis.
On average, 32.4% of the C emissions from CH 4 were derived from P. australis wetlands (17.3 Gg C/year) and 67.5% were from rice (36.0 Gg C/year). What was slightly different for CH 4 versus CO 2 , was that only~0.01% of the C emissions from CH 4 was associated with S. salsa wetlands (0.006 Gg C/year) ( Table 3). For S. salsa, suppression of CH 4 was due to a combination of smaller areal extent and potentially greater SO 4 2availability in the porewater [44]; salinity concentrations were above 45 PSU at times (average of 28.5 PSU for Suaeda1, Table 2) and water tables were either maintained below ground through impoundment (Suaeda1, with the exception of 2014) or were tidal (Suaeda2) (Fig 3).
As previously suggested in a global review [16], salinities above 18 PSU also tended to limit CH 4 emissions from the Liaohe Delta; a regression superimposed on the salinity versus CH 4 flux relationship from the Liaohe Delta indicates the fit suggested previously [16], and is remarkably applicable here when applied to three years of data collection across all wetland types in the Liaohe Delta (Fig 4D). Also important is that a reduction of salinity from 7.9 to 3.1 PSU over the growing season (May-September) from the two P. australis sites and rice site in combination gave rise to a 13-fold increase in average CH 4 fluxes in 2013 versus 2012 (17.3 vs. 1.3 mg CH 4 /m 2 /h, respectively). SO 4 2delivery to soils at low salinity is associated with high spatial variability in SO 4 2suppression [16,38]; this is a three-dimensional variability in space, making larger chambers necessary for capturing net flux changes, such as these, over larger areas when salinity concentrations are low. The course of CH 4 suppression is less clear for the second S. salsa site (Suaeda2), which had a mean salinity of only 8.4 PSU. Oddly, this salinity concentration was well within the salinity ranges of both P. australis sites (7.6-8.9 PSU), yet both P. australis sites maintained high CH 4 emissions. Higher salinity in the upper soil layers would certainly influence CH 4 emissions on P. australis sites less because much methanogenesis occurs deeper where anaerobic soil layers persist and soil pore water would be fresher. CH 4 might then route from deeper-laying P. australis roots, through stem tissue, and released to the environment providing a CH 4 conduit from lower, oxygen-deficient freshwater layers that bypass salinity influence and soil layers with an active methanotrophic bacterial community [11,45]. Deep roots and rhizomes may make all the difference for P. australis, relative to S. salsa wetlands. Over the first year of study [38], low CH 4 emissions from Suaeda2 were linked to the same SO 4 2suppression mechanism observed for Suaeda1, since salinity ranged to 15 PSU at times and salinity was probably pulsed higher at other times missed by our sampling. Suaeda2 is also strongly tidal compared to all of the other sites (Fig 3), and exposed soils during low or neap tides would facilitate CH 4 oxidation to CO 2 . Fluctuating water tables may also help explain lower CH 4 at this S. salsa site (Suaeda2) as the capacity for CH 4 oxidation is greater as soils are more exposed [46]. Multi-year, area-scaled assessments that isolate P. australis or S. salsa to assess influence from additional wetlands in China are not available. However, one smaller effort from the Yellow River Delta provides some guidance [35]. There, the smaller area of P. australis wetlands assessed (88.1 km 2 in the Yellow River Delta vs. 786.0 km 2 ) and a larger area of S. salsa wetlands assessed (90.2 km 2 in the Yellow River Delta vs. 31.6 km 2 ) suggested that C emissions from the Yellow River Delta were much less (59.2 Gg C/year) than we reported from the Liaohe Delta (2861 Gg C/year). CO 2 fluxes from rice paddy soils were large across the Delta (Table 3), although CO 2 emissions from soils may be balanced by, or less than, uptake of CO 2 by photosynthesis in such a productive environment. Indeed, the same notion (i.e., > CO 2 gains vs. emissions) may hold for all three wetland types. For example, based on P. australis photosynthesis data previously reported [32], P. australis wetlands in the Liaohe Delta would fix approximately 1,600 Gg C from atmospheric CO 2 annually [30]. Based on our data, C from soil CO 2 emissions would range from 39-61% of that value across the Delta, suggesting large-scale C sequestration among P. australis wetlands in the Liaohe Delta despite large soil emissions of CO 2 . Adding CH 4 affects the balance for C by only a small amount for P. australis (add 0.24-1.9% to the percentages for CO 2 ). More quantitatively, rice paddies in California, USA had a net ecosystem uptake of 50-397 g C/m 2 /year from CO 2 [24][25]; CH 4 emissions from these same sites in California were also quite low (2.5-6.6 g C/m 2 /year [25]) to moderate (39-53 g C/m 2 /year [24]). For comparison, Liaohe Delta rice paddies registered CH 4 emissions of 4.7-27.0 g C/m 2 /year when scaled annually.
Higher CH 4 fluxes from rice may be explained, in part, by hydrologic management. Our measured fluxes of~4 mg/m 2 /h ( Table 2) were much smaller than other rice fields under continuous irrigation in China (mean ± SD, 13.6 ± 9.2 mg/m 2 /h [47]). These literature values for CH 4 emissions are 64% higher than rice cultivated under drier, intermittent irrigation (mean ± SD, 8.3 ± 7.7 mg/m 2 /h [47]). Water levels were maintained well above the soil surface for most of the active cultivating season (June to September) for all three years in the Liaohe Delta, averaging 13.4 (± 3.7 SE) cm above ground. Not all Chinese rice paddies are managed in this fashion [47]; studies have indicated that mid-season drainage of rice paddies can reduce CH 4 emissions by 36-65% [48]. Such hydrologic management is at least feasible across many hectares of the Liaohe Delta owing to the "square-land method" [30], such that individual landowners could theoretically regulate CH 4 emissions at a local scale. However, prescribing drained or moist-soil management versus persistent flooding regimes is not simple to implement based solely on univariate relationships. Rice paddies are often loaded with NO x -based fertilizers such that drainage may mitigate emissions of C from CH 4 (i.e., a rather small component of the C flux, as we show here), but greater exposure to oxygen during drainage might simultaneously facilitate denitrification of NO x and promote N 2 O emissions when denitrification is incomplete. NO x is often combined with surplus soil acetate from crop residue by chance of timing during drainage. N 2 O has an even higher radiative forcing value than CH 4 ; six times higher than CH 4 when modelled as sustained-flux global warming potentials over a 100-year time frame [19].
As we describe here-in, soil emissions are often balanced by, or are lower than, net ecosystem uptake of CO 2 in order for atmospheric C to be sequestered by wetlands. Unless wetlands are deteriorating or are unhealthy, C sequestration is a strong characteristic of wetlands, which are estimated to serve as C sinks for 0.83 Pg C/year globally [12]. Soil C emissions of CO 2 from wetlands across the Liaohe Delta were estimated as 2.8 Tg C/year (Table 3). More important, this value tended to fluctuate among years from 2.5-3.3 Tg C/year, suggesting a strong potential year-to-year influence from wetland management or from stochastic environmental fluctuations.
Adding CH 4 to this estimate makes very little difference from a C emissions perspective, affecting emissions by~0.05 Tg/year at that resolution (Table 3). However, this is not to say that CH 4 is unimportant. In fact, as we describe, wetland management that facilitates lower salinity (below 18 PSU) and a quicker seasonal return to soil temperatures of 18°C or greater (as practiced in the Liaohe Delta [30]), would influence CH 4 fluxes considerably. This was the case in 2013, when CH 4 fluxes were higher from P. australis and rice due to persistent flooding [49] and salinity reduction. A focus on radiative forcing from CH 4 [12][13]19] versus total C emissions may provide a different perspective from the Liaohe Delta. Furthermore, other studies have discovered CH 4 emissions from rice growing in Northeast China to be even higher than we reported here [47], and our sampling would have missed any pulsed CH 4 emissions due to annual thawing [50].
Estimates of global C emissions from CH 4 for natural wetlands range from 69 to 213 Tg C/ year [19,[51][52] and for rice paddies range from 25 to 30 Tg C/year [19,52], with an average of 162 Tg C/year and 27 Tg C/year for natural wetlands and rice paddies, respectively [19]. Since CH 4 is not normally taken up by wetland soils through biological activity, flux values reported here-in would approximate the true gaseous C balance for CH 4 . Caveats do apply, such as small fluxes of CH 4 into the soil due to pressure differentials [53]. For rice, we also discovered that lower pore water HCO 3 concentrations corresponded to higher CH 4 fluxes (P < 0.001, r = -0.49), suggesting that whatever is facilitating higher CH 4 fluxes (e.g., anaerobiosis) may be reducing pore water HCO 3 by influencing dissolution of CO 2 . Overall, our total estimate of 0.053 Tg C lost per year to CH 4 emissions from all Liaohe Delta wetlands assessed is also seemingly low, except that emissions from Liaohe Delta rice paddies alone make up approximately 0.2% of rice paddy CH 4 -C emissions globally. For scale, C emissions from CH 4 for Carex lasiocarpa-dominated peatlands spread out over a much larger area in China's Sanjiang Plain to the north of the Liaohe Delta was estimated to be lower, at 0.007 Tg C/year [54], than we report from rice.

Conclusions
Phragmites australis, Suaeda salsa, and rice paddy wetlands encompass an area of approximately 3,287 km 2 in the Liaohe Delta, China. This is the world's largest continuous P. australis wetland and China's third largest oil field, and the Delta produces a large percentage of the rice crop for China in a given year. Total C emissions from CO 2 and CH 4 from these wetland soils average 2.9 Tg C/year, but range from 2.5 to 3.3 Tg C/year. We surmise that hydrology by way of management (e.g., longer retention times for water held within impoundments) or natural variability (e.g., rainfall and regional flood patterns) was a primary inter-annual driver of these differences, suggesting that evaluations of greenhouse gas fluxes need to be framed over multiple years. The primary emissions of gaseous soil C were from CO 2 (~98%). While photosynthetic uptake of CO 2 would most often overwhelm CO 2 emissions from the wetland soils as they build aboveground and belowground C stores, CH 4 emissions would persist. Overall, the opportunity for higher CH 4 fluxes was associated with soil temperatures >18°C and pore water salinity <18 PSU. CH 4 emissions from rice paddy habitat alone in the Liaohe Delta represent 0.2% of total C emissions from CH 4 globally for that habitat type. With such a large area and apparently sensitive feedbacks with soil CO 2 /CH 4 fluxes on a year-to-year basis, management practices in the wetland area studied and similar wetlands around the world have the potential not only to influence local C budgeting, but also to influence global biogeochemical cycling.

Ethics statement
The Panjin Wetland Science Research Institute (Mr. Dechao Sun, director) granted permission to access sites Phrag1, Phrag2, Suaeda1, and Suaeda2, and Mr. Tiejin Li granted permission to access the Rice site within his village.

Study sites
The Liaohe Delta is located in Liaoning Province in Northeast China, and has a geomorphic connection to four rivers; the largest is the Liaohe River. The Liaohe River is 1396 km long with a drainage area of 219,000 km 2 , and contemporary agricultural and deltaic wetland area of 3606 km 2 , encompassing the world's largest reed field, expansive rice paddies, and intertidal and unvegetated wetlands [30]. Polluted river waters [55] and active oil and gas mining activity (as China's third largest oil field [56]) pose significant environmental hazards for the Delta; river water is incredibly important for wetland irrigation while industrial canals, pipelines, and oil and gas mining infrastructure have transformed the landscape. Management of wetlands involves the use of pumping stations to divert Liaohe River water to P. australis wetlands to desalinize stands, thaw soils earlier in the growing season, and buffer soils from re-freezing nightly to promote greater productivity. Indeed, this action helped to increase P. australis yield to the pulp industry by~137,000 metric tons over a 31 year period up to 1980 [30]. Local-scale hydrologic management ("square-land method") was implemented intensely both for rice and P. australis, while S. salsa marshes typically exist as natural tidal features farther down the Liaohe River but are sometime impounded. Five sites representing the three primary wetland types in the Delta were selected (Fig 5). Two sites included managed reed (Phragmites australis (Cav.) Trin. Ex Steud.) wetlands ("Phrag1" at 40°52'22.34"N, 121°36'08.89"E; "Phrag2" at 41°09'33.75"N, 121°47'42.71"E) for paper production, two sites included seablite (Suaeda salsa (L.) Pallas) wetlands (a created and semi-impounded "Suaeda1", 40°52'11.09"N, 121°36'21.72"E; a natural "Suaeda2", 40°5 7'38.62"N, 121°48'20.03"E), and one site had active rice (Oryza sativa L.) agriculture ("Rice", 41°10'38.69"N, 121°41'17.28"E). Sites were selected carefully and over many days of searching to be representative of those wetland types in the wider region. With the exception of Phrag2, soil properties were fairly consistent among sites (S1 Table).
The air temperature in the region associated with the Liaohe Delta ranges from an average low of -10.4°C in January to an average high of 27.4°C in July, with an annual average of 8°C and approximately 175 days/year frost-free [43]. The annual precipitation for the Delta is 612 mm [43]. Remarkably, the year 2013 tied with 2007 as the sixth warmest since global records began in 1850 [49]. 2013 was also warmer than both 2011 and 2012, which, though marked by cooling La Niña conditions, were 0.43°C and 0.46°C above average, respectively [49]. In addition to high temperatures in 2013, anomalous hydro-meteorological events affected northeastern China with excessive river flooding [49] with noticeable impacts to the Liaohe Delta seasonally relative to 2014 in terms of more persistent flooding on study sites, especially for Phrag1, Phrag2, and Rice.

Experimental design and GHG flux measurements
Soil CO 2 and CH 4 gas fluxes were sampled approximately monthly from June to November for Year 1 (2012), April to November for Year 2 (2013), and April to November for Year 3 (2014). Gases were collected using six, square metal frames installed permanently on 4 of 5 sites. Frames had to be moved annually to accommodate agricultural activity on one site ("Rice"). Frames had an area of 3025 cm 2 (55x55 cm), were constructed with small drain holes at the base to allow free water flow between measurement periods, and had troughs for inserting white, plastic chamber tops during sampling. Holes were plugged, troughs were filled with water, and the chamber tops were lined internally with aluminum foil to ensure that light would not penetrate into the "dark" chambers during sampling. Chamber tops were 30 cm tall, requiring that P. australis plants were cut at times; however, we limited cutting to only as much as necessary to emplace chamber tops. This practice had very little influence on CH 4 emissions when reeds were cut above standing water [45], as we practiced here. All chambers were accessed from permanent boardwalks positioned just about the soil surface.
For Year 1, gases were sampled using static flux chamber protocols [57]. Tops were emplaced and gases were extracted through rubber septa using a 15 mL syringe, and injected into pre-vacuumed 10 mL glass vials for analysis on a laboratory based gas chromatograph (GC). Samples were taken as soon as the chamber tops were emplaced, and at 20 min intervals over 60 mins. Circulating fans kept gases mixed within chambers, which were approximately 121 L in size with chamber tops emplaced. Full sampling details for Year 1 including GC information, storage and laboratory protocols were previously provided [38].
For Years 2 and 3, a portable GC (Model 915, Los Gatos Research, Mountain View, CA, USA) was used instead of a laboratory based GC in order to facilitate in-situ measurements and overcome any concerns we had in Year 1 with storing and transporting gas vials over 520 km from the Liaohe Delta to Qingdao. The chamber tops were the same as for Year 1, but septa were replaced with Tygon tubing routed to and from the portable GC. For both methods, flux rates were determined using the linear portion of fit saturation curves comparing static flux over time (Year 1) or steady state flux rate increases over time (Years 2 and 3). All samples were taken during the day and assumed to be consistent diurnally for that day, but see [58][59].

Soil characteristics
Soil cores (3 per site) were taken to a depth of 10 cm, extracted by pushing/twisting a 15-cm diameter by 1-m long metal cylinder (0.8-mm-wall) with a sharpened end into the soil with minimal compaction, and sectioned into 2 cm increments. 2-cm sections were mixed thoroughly, dried to a constant weight at 60°C, and ground. Soil bulk density, water content, and pH were determined through standard procedures, and nitrogen and carbon (total and organic) were then analyzed. Individual samples were split, with total nitrogen and total C analyzed on one section with a CHNS/O elemental analyzer (2400 Series, Perkin Elmer, Waltham, MA, USA). The second section was used to determine organic C fractions on the same elemental analyzer, but after inorganic C was removed with 4 M HCl [60]. Sections (n = 5) were averaged after analysis for each core.
Soil oxidation-reduction potentials (Eh) were determined with brightened platinum electrodes inserted to a depth of 10 cm [61], and allowed to sit for 24 h prior to measurement to Fig 5. Location of study sites and aerial distribution of habitat types sampled in the Liaohe Delta, China. Map highlights 31.6 km 2 of Suaeda salsa wetlands, 786 km 2 of Phragmites australis wetlands, and 2464.6 km 2 of rice paddy wetlands, as well as the location of our five wetland sites, including two in Phragmites australis (Phrag1, Phrag2), two in Suaeda salsa (Suaeda1, Suaeda2), and one in rice paddy (Rice). Aerial distribution data are from [30], and the shape file represents 2011 classifications (China Geological Survey).
doi:10.1371/journal.pone.0160612.g005 ensure a well-poised couple. Eh probes were referenced against calomel electrodes, and adjusted by adding 245 mV for standardization against a hydrogen electrode scale. Water level recorders (model 3001, Solinst, Georgetown, Ontario, Canada) were inserted into on-site wells during freeze-free periods, and recorded water table depth hourly. During sampling, salinity was measured from temperature-compensated conductivity on water extracted from on-site piezometers using a meter (Model 6010, Jenco Electronics, Ltd., Shanghai, China), and soil temperature was measured using manual thermometers (bi-metallic dial, H-B Instruments, Collegeville, PA, USA) inserted to a 10-cm soil depth just outside of each static flux chamber. Plant aboveground biomass was sampled monthly to coincide with gas flux measurements seven times each in 2012 and 2013 (May to November) and 4 times in 2014 (April, June, July, September) from Phrag1, Suaeda1, Suaeda2, and Rice using 55 cm × 55 cm frames (n = 6/site). Phrag2 was sampled identically when feasible; however, commercial harvesting of P. australis for pulp from that site prevented consistent biomass estimates. All vegetation within the frame was clipped at the soil surface, dried to a constant weight at 60°C, and weighed.

Statistical analysis and variability determinations
Soil CO 2 and CH 4 emissions were analyzed with ANOVA in a split-plot framework using Type IV sums of square error estimation for accounting for missing treatment combinations. Date was assigned as a whole-plot effect (repeated measures). There were a total of 19 monthly CO 2 and CH 4 flux assessments over the 3 years, but only five sites. For repeated measures analyses, the assumption of n+1>q (where n equals the sample size, i.e., number of sites, and q the number of repeated measures) was not met [62], so we nested terms to account for non-independence among repeated measures [41,63]. For significant treatment by date interactions, treatment differences were determined with Bonferroni adjustment. All data were log-transformed. The errors had a homogeneous variance and were unimodal and symmetric. Correlation analysis was used to determine whether gas fluxes and soil water table, salinity, above ground biomass, porewater HCO 3 -, Eh, or soil temperature related over a three year period.
Data were analyzed using SAS (Version 9.3, SAS Institute, Cary, NC, USA). Average annual rates and variation of CO 2 and CH 4 emissions were determined from each site for each year, and scaled assuming: (1) that mean hourly rates of CO 2 and CH 4 emissions from chambers are consistent over a measurement day, (2) that days sampled over the course of individual years (n = 5-8 times/year) are representative of the year, and (3) that no fluxes occurred when soils were frozen (December, January, February, March). Soils in the Liaohe Delta freeze solid to depths of > 0.5 m in the winter. Based on near-zero fluxes in November of every year (Fig 1), this latter assumption appears valid (but see [50,64] for CH 4 emissions). We recognize that measurements are not continuous over individual years, but we wanted to document how commonly used discrete sampling procedures can reveal inter-annual differences in important GHG fluxes related to a combination of site management and environmental factors. Mean fluxes from Phrag1/Phrag2, Suaeda1/Suaeda2, and Rice were reduced to three values each for CO 2 and CH 4 , and multiplied over area determinations from 2009 satellite imagery for the Liaohe Delta [30].
Supporting Information S1 Table. Soil characteristics from a depth of 0-10 cm at five wetland sites in the Liaohe Delta, China. Data are updated from [34] to include additional data collected in 2013 and 2014. (DOCX) S2 Table. Raw data used for interpretative purposes in "Inter-Annual Variability of Area-Scaled Gaseous Carbon Emissions from Wetland Soils in the Liaohe Delta, China". (XLSX)