Introduction

Peatlands constitute one of the largest terrestrial C-stores, accounting for at least one-third of the global soil C pool (Limpens and others 2008). Peatlands are found at all latitudes, and include brackish coastal estuaries, freshwater river deltas, tropical swamps, inland bogs, and fens (Dise 2009). Under natural, unmanaged conditions, peatlands are sinks for atmospheric CO2, because waterlogged soil conditions inhibit aerobic decomposition, favoring the accumulation of soil organic matter (Dise 2009). However, peatlands do not always exert a net “cooling” effect on the atmosphere because they also emit non-CO2 greenhouse gases (Dise 2009; Frolking and Roulet 2007). For example, peatlands are one of the largest natural sources of atmospheric CH4, a greenhouse gas approximately 25 times more effective than CO2 in absorbing long-wave radiation in the atmosphere, and responsible for 20% of current climate forcing (Forster and others 2007). Recent reports also suggest that peatlands may emit significant quantities of N2O, a gas with 298 times the global warming potential of CO2, although the dynamics and magnitude of these fluxes are poorly characterized (Jungkunst and Fiedler 2007; Repo and others 2009).

Studies of peatland greenhouse gas exchange have focussed on natural or unmanaged environments, with fewer studies investigating greenhouse gas dynamics in managed systems (Limpens and others 2008; Waddington and Price 2000). The majority of these studies have in turn concentrated on greenhouse gas fluxes from northern (that is, boreal, sub-arctic, arctic) ecosystems, rather than on temperate or tropical ones (Limpens and others 2008; Waddington and Roulet 1996; Zona and others 2009; Hendriks and others 2007). Temperate peatlands are likely to have greater overall trace gas fluxes than their northern counterparts because they experience warmer conditions and longer growing seasons (Carroll and Crill 1997; Fiedler and others 2005; Fowler and others 1995b; Hendriks and others 2007; Jungkunst and Fiedler 2007). Moreover, temperate peatlands are commonly exploited for agriculture (that is, grazing, arable crops), energy (that is, peat cutting and extraction), horticulture, and water resources, making it difficult to predict gas fluxes from these environments based on models of near-pristine northern peatlands (Charman 2002; Limpens and others 2008; Service 2007).

Past attempts to investigate greenhouse gas fluxes in managed peatlands have focussed on single gases (for example, CH4 or CO2 or N2O), or pairs of compounds (for example, CH4 and CO2) (Hendriks and others 2007; Jungkunst and Fiedler 2007; Limpens and others 2008). Nitrous oxide fluxes have been particularly neglected, even though emissions from agricultural peatlands may be substantial (Jungkunst and Fiedler 2007; Langeveld and others 1997; Regina and others 2004; Schils and others 2006; Takakai and others 2006). Nitrous oxide fluxes were largely ignored in the past because conceptual models of peatland biogeochemistry are based on N-poor northern bogs (Limpens and others 2008). However, there has been growing interest in quantifying N2O fluxes from managed peatlands, in recognition of the fact that agricultural peatlands may have enhanced N pools and cycling rates due to fertilization or manure additions by livestock, increasing the potential for N2O fluxes (Jungkunst and Fiedler 2007; Langeveld and others 1997; Regina and others 2004; Schils and others 2006; Takakai and others 2006).

Managed peatlands exhibit a high degree of both spatial and temporal variability in greenhouse gas fluxes due to dynamic patterns in soil moisture, redox, and substrate availability, driven by human modifications of peatland hydrology and vegetation (Schrier-Uijl and others 2009; Strack and Waddington 2007, 2008; Waddington and Price 2000; Chen and others 2008; Inubushi and others 2003; Hendriks and others 2007; Fowler and others 1995a; Ward and others 2007). Patterning of peatland landscapes, for example, through the introduction of drainage ditches and managed agricultural fields often drives extreme differences in the composition and magnitude of greenhouse gases fluxes by significantly altering redox dynamics at the micro- and mesotope scale (Schrier-Uijl and others 2009, 2010; Strack and Waddington 2007). Common approaches for estimating greenhouse gas fluxes include eddy co-variance measurements, which provide an integrated picture of whole-ecosystem gas exchange, or static flux chambers, which facilitate identification of within-ecosystem variability. Both measurement techniques have inherent strengths and weakness; static chamber measurements better represent spatially heterogeneous gas fluxes, whereas eddy covariance techniques yield quasi-continuous measurements that are spatially integrated (Hendriks and others 2010; Schrier-Uijl and others 2010; Smith and others 1994). Most studies have relied on single measurement techniques (for example, static flux chambers or eddy covariance) to quantify greenhouse gas exchange, leading to potentially large uncertainties in ecosystem greenhouse gas budgets (Hendriks and others 2010; Schrier-Uijl and others 2010; Smith and others 1994).

Here, we quantified greenhouse gas fluxes from a mid-latitude drained peatland in the Sacramento-San Joaquin River Delta, California, USA (hereafter simply “the Delta”), currently managed as rangeland (that is, pasture). We used a multi-scale approach that combined static flux chambers, eddy covariance (for CH4 and CO2 only), and spatially weighted upscaling; techniques that allowed us to capture the “hotspots” and “hot moments” characteristic of greenhouse gas fluxes (sensu McClain and others 2003). The Delta is the largest estuary on the Pacific coast of the Americas, and is under multiple ecological and environmental pressures, including considerable flood risk from a failing levee system and continued land subsidence due to the decomposition of peat, soil compaction, and wind erosion (Florsheim and Dettinger 2007; Mount and Twiss 2005; Service 2007). Delta peatlands are the primary conduit for urban and agricultural water for the state of California, and have experienced extensive conversion to agriculture. Prior to the California Gold Rush, the Delta consisted of over 1400 km2 of unmanaged peatland, interspersed by several hundred kilometers of natural waterways (Service 2007; Drexler and others 2009). Extensive human settlement began in the 1860s, with farmers establishing levees and drainage ditches to make Delta peatlands more suitable for the cultivation of arable crops and livestock (Service 2007). Peatland drainage and water management have led to high rates of soil C mineralization and massive land subsidence, with Delta soils now up to 10 m below sea level (Deverel and Rojstaczer 1996; Miller and others 2000).

This study explores greenhouse gas fluxes, C sequestration, and global warming potential in a managed temperate peatland. We investigated the global warming potential of a drained temperate peatland under current agricultural management practices in the Delta, and the role of different greenhouse gases (that is, CH4, CO2, N2O) in regulating overall ecosystem global warming potential. We determined how greenhouse gas fluxes vary among representative landforms and the role of spatial and temporal heterogeneity in overall ecosystem gas exchange and trace gas budgets. We also explored the role of environmental variables (for example, climate, soil moisture, and so on) in mediating gas fluxes. Findings from this study provide the first steps to answering broader, more integrative questions about the effects of human activity on soil C sequestration, greenhouse gas fluxes, and the global warming potential of managed temperate peatlands. This research also provides an empirical basis for understanding the potential contributions of these ecosystems to local and regional greenhouse gas budgets.

Methods and Materials

Study Site

Flux measurements were conducted in a peatland pasture on Sherman Island (38.04 N, 121.75 W). Observations were collected over 60 weeks, from 10 April 2007 to 28 May 2008, in an area covering approximately 0.38 km2 and managed as rangeland for over 20 years (Figure 1). Approximately 49% of the land area on Sherman Island is currently under rangeland (US Department of Agriculture 2007). The climate is Mediterranean, with rain falling predominantly during the cool winter months (November to February). Mean annual rainfall is 325 mm and mean annual temperature is 15.6°C. Historically, Sherman Island was planted with arable crops such as asparagus, corn, milo (sorghum), sugarbeet, barley, wheat, and potatoes (US Department of Agriculture 2007). The plant community at our site is dominated by two non-native, non-aerenchymatous invasive species: pepperweed (Lepidium latifolium L.), a perennial herb; and mouse barley (Hordenum murinum), an annual forage grass. Soils are classified as fine, mixed, superactive, thermic Cumulic Endoaquolls, consisting of a 25–92 cm of oxidized layer overlaying a 151–292-cm thick organic peat horizon (Drexler and others 2009). Water table depths range from 30 to 70 cm below the soil surface, and are maintained by an active system of pumps and drainage ditches (Deverel and others 2007).

Figure 1
figure 1

Remote sensing imagery of agricultural peatlands on Sherman Island. Airborne hyperspectral imager (HyMap) false-color composite (R;G;B: band 27 [near-infrared]; band 16 [red]; band 9 [green]) indicating flooding extent (dark areas) at the time of image acquisition. Overlayed are the locations of the static flux chambers, eddy covariance tower, and the tower footprint (night and day). Also represented on the image are the mapped areas (and their fractions) with different land surface wetness conditions (negligible surface water ponding [~73%], periodical inundation [~22%], perennial flooding [~5%]) used to calculate spatially weighted seasonal and annual trace gas fluxes. (Color figure online)

Full size image

Chamber Fluxes

Static chamber measurements were used to investigate the effects of microform (~1–5 m) to mesotope (~100 m–1 km) scale variations in vegetation, hydrology, and redox potential on greenhouse gas fluxes (CH4, CO2, N2O). Prior to the start of the experiment proper, we conducted a pilot study that sampled from all the major microforms and mesotopes in the study site, to evaluate overall patterns of spatial heterogeneity (ten 60-m long transects; n = 5 static flux chamber per transect; n = 50 static flux measurements). Using these preliminary data as a guide, static flux chambers were deployed in five 60-m long transects representative of the dominant microforms and mesotopic features and sampled at weekly intervals (Figure 1). These landforms were categorized as “crown” (n = 5 chambers), “slope” (n = 5 chambers), “hummock/hollow” (n = 10 chambers), or “drainage ditch” (n = 5 chambers). Crown landforms are at least 500 m2 in size and possess either a slight convexity, or no apparent slope. Rainfall and groundwater tend to drain relatively quickly from these areas (typically within 1–3 days), leading to little or no surface water ponding. Slope areas are at least 500 m2 in size, with a gentle slope (~0.001%), and typically lie down gradient from crown landforms, draining water from these higher topographic features. Hummock/hollow areas are located in lower topographic positions, and contain complexes of shallow pools and very small knolls (<10-cm high), with each feature no more than 100 m2 (typically ~25 m2) in size. These landforms are periodically inundated because of water management practices, or due to water draining from further upslope. Drainage ditches are common man-made features on farmed Delta islands and are used to remove water from the plant rooting zone to provide more suitable growing conditions for arable crops or pasture grasses (Deverel and others 2007). These landforms are perennially flooded.

Static chamber measurements were made by enclosing a 0.05-m2 area with an opaque, 2-component (that is, base and lid) vented chamber for 30 minutes; headspace samples were collected using a gas-tight syringe at five time points. Static chambers were grounded (that is, not floating) in all landforms. Water levels in the drainage ditches were sufficiently low throughout the study period (5–10-cm depth) that floating chambers were not required. Chamber bases were inserted to a depth of about 5 cm in the soil of crown, slope, and hummock/hollow landforms and to a depth of about 1 cm in drainage ditch sediment. Gas samples were stored in pre-evacuated 10-ml glass bottles sealed with Geo-Microbial Technologies septa (Geo-Microbial Technologies Inc., Ochelata, Oklahoma, USA), and analyzed for CH4, CO2, and N2O using a Shimadzu GC-14A gas chromatograph (Shimadzu Scientific Inc., Columbia, Maryland, USA), equipped with a Porapak-Q column, flame ionization detector (FID), thermal conductivity detector (TCD), and electron capture detector (ECD). The global warming potentials for CH4 and N2O were converted to CO2 equivalents by multiplying CH4 and N2O fluxes by 25 and 298, respectively (Forster and others 2007; Repo and others 2009; Frolking and Roulet 2007). These scaling factors represent the global warming potential for CH4 and N2O over a 100-year time horizon.

Carbon dioxide and N2O fluxes were calculated by applying a linear least squares regression to the chamber headspace concentration of each gas plotted against time (P < 0.05). Methane fluxes were calculated by different methods, depending on whether diffusion or ebullition was deemed to be the principal transport pathway. Diffusion was assumed to be the dominant physical transport pathway in chambers showing a linear change in CH4 concentrations over time, and fluxes calculated using a linear least squares approach (P < 0.05). Ebullition was assumed to be the dominant transport mechanism in chambers where CH4 concentrations showed either steep non-linear increases over time, or abrupt stochastic increases over time. Ebullition fluxes were calculated in one of two ways: for chambers showing steep non-linear increases, we fitted the data to a quadratic regression equation (P < 0.05), and fluxes were determined from the steep initial rise in CH4 concentrations. For chambers showing abrupt stochastic increases, fluxes were determined by calculating the total CH4 production over the entire enclosure period; a method that likely underestimates ebullition fluxes. Chamber data that did not meet any of these criteria were reported as net-zero fluxes.

Eddy Covariance

Eddy covariance measurements were employed to determine the temporal variability of spatially integrated ecosystem CH4, CO2 and H2O vapor fluxes (Baldocchi 2003; Detto and others 2010a). Nitrous oxide fluxes were only measured using the static flux chamber approach. Methane concentrations were determined using an off-axis integrated cavity output spectroscopy analyzer (Fast Methane Analyzer, Los Gatos Research, Inc, Mountain View, California, USA). Carbon dioxide and H2O vapor fluxes were measured with an open-path, infrared absorption gas analyzer (model LI-7500, LICOR, Lincoln, Nebraska, USA) (Baldocchi 2003), which was tested extensively at this study site and others to ensure optimal performance (Detto and others 2010a, b). Wind velocities and air temperature were also measured using a 3-D sonic anemometer. All measurements were acquired at 10 Hz from a tower 3.15 m above the ground and data stored on a portable field computer. Vertical fluxes were computed on 30-min average windows, and corrected for tilt angles, temperature, and water vapor fluctuation effects (Detto and Katul 2007). Gaps due to data loss and quality check filtering were filled using a trained Artificial Neural Network (Papale and Valentini 2003). Cows (~100) were an irregular presence at this site. An automated digital camera (or “cow cam”) was used to record the presence of cows in the tower footprint, omitting eddy covariance data when cows congregated near the tower base (primarily) at night, as this generated elevated flux estimates. The data reported here thus do not include the direct influence of ruminant respiration when cows were in the immediate vicinity of the tower, although more distal ruminant fluxes (that is, outside of the immediate tower precinct) were captured by our measurements.

Environmental Variables

To determine the effects of physical factors on trace gas dynamics, we measured air temperature, soil temperature, rainfall, soil moisture, and water table depth close to the tower in the crown zone. Air temperature and humidity were measured with an aspirated and shielded thermistor and capacitance sensor (Vaisala Inc, Vantaa, Finland). Rainfall was determined using a Texas Electronics (Texas Electronics Inc., Dallas, Texas, USA) tipping bucket rain gauge. Soil temperature was monitored continuously throughout the soil profile using copper-constantan thermocouples at 2, 4, 8, 16, 32, and 50 cm (n = 3 per depth). Volumetric soil moisture content (VWC; reported in percent) was measured using TDR probes (Delta-T Devices Ltd, Cambridge, UK) at 0–15, 15–30, 30–45 and 45–60 cm (n = 6 per depth). The water table was measured with a pressure sensor (GE Druck Ltd, USA) immersed in a well next to the meteorological tower. Point measurements were also collected at weekly intervals from the 0–10-cm depth during static flux chamber sampling to determine soil temperatures and soil moisture (reported either as percent VWC or percent water-filled pore space, the latter abbreviated as WFPS) in the immediate volume beneath each flux chamber. Weekly averages of soil temperature and moisture were calculated using continuous data collected in the tower footprint and point measurements collected during weekly chamber sampling.

Spatially Weighted Extrapolations of Chamber Fluxes

Two different airborne remote sensing data products of very high spatial resolution were used to determine the fractions of areas with different land surface wetness conditions on Sherman Island, including light detection and ranging (LiDAR), with a 1-m resolution, and hyperspectal imaging (HyMap) (Cocks and others 1998), with a 3-m resolution. Land surfaces were categorized as having “negligible surface water ponding” (dry drainage ditches, crown and slope), “periodical inundation” (hummock/hollow), or “perennial flooding” (wet drainage ditches). The drainage ditches were mapped on the basis of a LiDAR-derived digital elevation model (DWR 2009), with a simple decision tree classifier incorporating ground surface elevation and two topographic indices (slope, profile convexity) using the ENVI image processing environment. The area of periodic inundation was derived as flooding extent present at the time of HyMap image acquisition (Hestir and others 2008) using ISODATA unsupervised classification in the ENVI image-processing environment. The areal fractions were used to calculate spatially weighted seasonal and annual trace gas fluxes. Total daily and annual fluxes were obtained through linear interpolation of weekly static chamber measurements. Standard errors for the annual greenhouse gas budgets were calculated as follows: first, mean daily fluxes (n = 365) were calculated for each static flux chamber. Standard errors for each land surface wetness condition were subsequently determined from these mean daily fluxes. Finally, to calculate the standard error for the annual greenhouse gas budgets, spatial weightings were applied to each of the three land surface wetness conditions based on their respective areal fractions. These standard errors do not account for any uncertainty introduced by the pre-processing of the airborne remote sensing data, or through misclassification of the land surface wetness conditions.

Statistics

Statistical analyses were performed using JMP IN Version 8 (SAS Institute, Inc., Cary, North Carolina, USA). The data were log transformed whenever necessary to meet the assumptions of analysis of variance. Residuals for all analyses were checked for normality and homogeneity of variances. We used repeated-measures analysis of variance (ANOVA) to explore temporal trends in chamber fluxes and ANOVA to explore spatial patterns for normally distributed data. Kruskal–Wallis non-parametric ANOVA was used to explore spatial patterns for data that were not normally distributed. Bivariate and multiple regressions were employed to evaluate the relationship between continuous environmental variables and trace gas fluxes. Statistical significance was determined at the P < 0.05 level, unless otherwise noted. Means comparisons were performed using Fisher‘s Least Significant Difference test (Fisher’s LSD). Methane, N2O, and CO2 fluxes from static chambers were used to explore spatial and temporal patterns across landforms. We report spatially extrapolated CH4 and N2O fluxes from static chambers as an estimate of net ecosystem-scale fluxes, because soils are considered the primary source of these gases; descriptions of our methodology and error estimation are provided above. We report the net ecosystem exchange of CO2 (NEE) using the eddy covariance data, which incorporates plant uptake and respiration. We also compared static chamber estimates of soil respiration with overlapping night time ecosystem respiration (R ECO), the majority of which is derived from soils. Values are reported as means and standard errors (±1 SE).

Results

Net Ecosystem Trace Gas Fluxes

During the first year, the peatland pasture was a net source of CH4 and N2O (Figure 2), and approached net balance for CO2 with the atmosphere (near-zero NEE; Figures 2, 3). Both spatially weighted extrapolations of static flux chamber measurements and eddy covariance indicated that the ecosystem was a net source of atmospheric CH4. Spatially weighted chamber measurements, averaged across all landforms, yielded a mean daily soil flux of 25.8 ± 1.4 mg CH4-C m−2 d−1, whereas daytime eddy covariance measurements averaged 5.6 ± 0.3 mg CH4-C m−2 d−1 (Figure 2A). Annual soils emissions were estimated to be 9.5 ± 3.4 g CH4-C m−2 y−1 based on chamber measurements, or 1.6 ± 1.4 g CH4-C m−2 y−1 by eddy covariance. Soil N2O fluxes were consistently very high. Spatially weighted extrapolations of chamber fluxes averaged 6.4 ± 0.4 mg N2O-N m−2 d−1 (Figure 2C). Overall annual N2O emissions for the first year of observations were 2.4 ± 1.3 g N2O-N m−2 −1.

Figure 2
figure 2

Net fluxes of A CH4, B soil respiration, C N2O, and D CH4 and N2O expressed as global warming potential in CO2 equivalents for different landforms in the drained peatland pasture on Sherman Island. Eddy covariance observations are shown for comparison. In (A), 25 was added to the raw CH4 flux data so that negative or zero net fluxes could be log-transformed. The short-dash line within each box represents the mean, whereas the solid line represents the median. Boxes enclose the interquartile range, whiskers indicate the 90th and 10th percentiles. Lower case letters indicate statistically significant differences among means (Fisher’s LSD, P < 0.05).

Full size image
Figure 3
figure 3

Temporal variations in ecosystem CO2 fluxes (A) and CH4 fluxes from different landforms (B). Panel A shows net ecosystem exchange (NEE), ecosystem respiration (R ECO) and gross primary productivity (GPP) determined by eddy covariance measurements. Panel B shows CH4 fluxes determined by chamber measurements. Bars indicate standard errors.

Full size image

Spatially weighted chamber measurements indicated an average daily soil respiration rate of 5.9 ± 0.3 g CO2-C m−2 d−1; and were greater than soil respiration estimated from night time eddy covariance fluxes (that is, ecosystem respiration, R ECO), which yielded an average flux of 3.9 ± 0.2 g -C m−2 d−1 (Figure 2B). A large proportion of photosynthetic C uptake was lost to the atmosphere via respiration, such that NEE approached net balance with the atmosphere (Figure 3A). Eddy covariance measurements indicated that from 10 April 2007 to 9 April 2008, overall NEE was only a meagre −8.4 g C m−2 y−1; that is, close to CO2-neutrality (Figure 3A).

Spatial and Temporal Variability in Gas Exchange

Methane Emissions

Methane fluxes showed high spatial variability, but few or no temporal trends (Figures 2A, 3B; Table 1). CH4 fluxes varied significantly by a factor of up to 400 or more among landforms (Kruskal–Wallis, P<0.0001; Figure 2A; Table 1). Multiple comparison tests indicated that CH4 fluxes from drainage ditches greatly exceeded that of all other landforms by up to two orders of magnitude (466.4 ± 78.4 mg CH4-C m−2 d−1); in comparison, hummock/hollow areas emitted 9.5 ± 3.4 mg CH4-C m−2 d−1, slopes emitted 3.9 ± 0.9 mg CH4-C m−2 d−1 and crown landforms emitted 1.0 ± 0.2 mg CH4-C m−2 d−1. Methane fluxes between individual hummock and hollow microforms were not significantly different from each other in this ecosystem, unlike other peatlands (Belyea and Baird 2006; Pelletier and others 2007; Waddington and Roulet 1996), and we grouped the data from these two microforms together. Eddy covariance measurements of net CH4 exchange showed an inverse relationship with mean weekly soil temperature (r 2 = 0.32, P < 0.0001), a weak positive correlation with mean weekly VWC (r 2 = 0.20, P<0.01) and mean weekly WFPS (r 2 = 0.19, P < 0.01).

Table 1 Net Trace Gas Fluxes by Land Form
Full size table

Nitrous Oxide Dynamics

Nitrous oxide fluxes varied by a factor of 3 or more among landforms (F 3,916 = 13.6, P < 0.0001; Figure 2C; Table 1). Multiple comparisons tests indicated that the highest N2O fluxes were from crown landforms (8.7 ± 1.7 mg N2O-N m−2 d−1) and hummock/hollow (7.1 ± 1.4 mg N2O-N m−2 d−1) areas. Slopes had intermediate levels of N2O flux (5.0 ± 0.8 mg N2O-N m−2 d−1), whereas drainage ditches had the lowest emissions (2.6 ± 0.9 mg N2O-N m−2 d−1). Slopes and hummock/hollows showed little or no change in N2O emissions over time, whereas crowns and drainage ditches fluctuated due to changes in soil temperature and water-filled porosity driven by seasonality and management (F 72,916 = 3.3, P < 0.0001). N2O fluxes in crowns were positively correlated with WFPS (r 2 = 0.55, P < 0.0001; Figure 6A), weakly correlated with water table depth (r 2 = 0.25, P < 0.01), weakly correlated rainfall (r 2 = 0.15, P < 0.01), and negatively correlated with soil temperature (r 2 = 0.54, P < 0.0001; Figure 6B). Episodic rainfall in winter increased soil moisture and WFPS, increasing N2O fluxes from crowns; whereas fluctuations in WFPS due to water management practices lowered or raised N2O fluxes from drainage ditches. A multiple regression model using these drivers explained 64% (P < 0.0001) of the variability in the data. Nitrous oxide fluxes in hummock/hollows were also weakly positively correlated to water table depth (r 2 = 0.30, P < 0.0001) and rainfall (r 2 = 0.12, P < 0.01), with a multiple regression model explaining only 36% (P < 0.0001) of the overall variability in the data. Nitrous oxide fluxes from drainage ditches were positively correlated with WFPS (r 2 = 0.35, P < 0.0001), although insensitive to variations in other environmental variables. Nitrous oxide fluxes from slopes showed no response to fluctuations in WFPS, rainfall, water table depth, or temperature.

Soil and Ecosystem CO2 Fluxes

Chamber measurements of soil respiration also indicated high spatial variability (Figure 2B). Fluxes varied significantly among landforms by up to an order of magnitude (F 3,1012 = 307.3, P < 0.0001; Fisher’s LSD, P < 0.05; Figure 2B; Table 1), with crowns emitting 4.9 ± 0.2 g CO2-C m−2 d−1, slopes emitting 11.0 ± 0.2 g CO2-C m−2 d−1, hummock/hollow areas emitting 3.3 ± 0.2 g CO2-C m−2 d−1, and drainage ditches emitting 1.3 ± 0.2 g CO2-C m−2 d−1.

Unlike CH4, soil CO2 fluxes were affected by seasonal variations in soil moisture, rainfall, and temperature. Soil respiration was negatively correlated with soil WFPS when all chamber data were pooled (r 2 = 0.51, P < 0.0001; Figure 4A), as was R ECO (r 2 = 0.46, P < 0.0001; Figure 4B). Disaggregating the chamber data and analyzing the response of individual landforms indicated that slopes (r 2 = 0.61, P < 0.0001) and hummock/hollows (r 2 = 0.39, P < 0.0001) had the strongest responses to changing soil moisture, whereas crowns and drainage ditches did not appear to respond to soil moisture. Soil respiration was negatively correlated with rainfall in slope (r 2 = 0.52, P < 0.0001), hummock/hollow (r 2 = 0.23, P < 0.0001), and crown (r 2 = 0.20, P < 0.0001) landforms, although eddy covariance data showed no overall effect of mean weekly rainfall on R ECO. Tower measurements showed a strong positive relationship between R ECO and mean weekly soil temperature (r 2 = 0.77, P < 0.0001; Figure 5). Analysis of the chamber soil respiration measurements indicated that individual landforms responded differently to seasonal fluctuations in temperature. Soil respiration on slopes, for example, was much more closely correlated with temperature (r 2 = 0.52, P < 0.0001) than other landforms, such as crowns and hummock/hollows (both r 2 = 0.17, P < 0.05); whereas drainage ditches showed no temperature response whatsoever.

Figure 4
figure 4

Soil respiration, as determined by chamber measurements, plotted against soil moisture (A) and ecosystem respiration (R ECO), as determined by eddy covariance measurements, plotted against soil moisture (B). In panel A, each data point represents the mean of five replicate flux chambers. In panel B, weekly averaged values of R ECO are plotted against weekly averaged values of WFPS for the entire peatland. Bars indicate standard errors.

Full size image
Figure 5
figure 5

Ecosystem respiration (R ECO), as determined by eddy covariance measurements, plotted against soil temperature. Each data point represents weekly-averaged values of both R ECO and soil temperature. Bars indicate standard errors.

Full size image

The chamber data were subsequently analyzed using multiple regression models that included WFPS, rainfall, and temperature as driving variables. A multiple regression model using all the chamber data pooled together explained 56% of the variability in the data set (P < 0.0001). Multiple regression models applied to data disaggregated by landform indicated that WFPS, rainfall, and temperature together explained 76% (P < 0.0001) of the overall variability in the respiration rates of slopes, 56% (P < 0.0001) of the variability in hummock/hollows, and 25% (P < 0.01) of the variability in crowns.

Eddy covariance measurements of NEE, gross primary productivity (GPP), and R ECO indicated strong seasonal trends in ecosystem-scale CO2 fluxes, largely driven by changes in plant activity and modulated by fluctuations in soil respiration (Figure 3A). Net ecosystem exchange values were negative (that is, net CO2 uptake) in spring and summer 2007. Over autumn, NEE became gradually more positive, shifting toward a net CO2 source during winter. Gross primary productivity followed a similar pattern (Figure 3A), with peak GPP during spring–summer, declining GPP during autumn, and near-cessation of plant C uptake during winter. Ecosystem respiration had slightly different seasonal dynamics (Figure 3A), influenced by fluctuations in soil temperature (Figure 5). Ecosystem respiration rates remained relatively stable through spring–summer 2007, before steadily declining throughout autumn 2007, reaching an annual minima during winter (Figure 3A). Ecosystem respiration subsequently rose again during spring 2008, as soil temperature increased (Figure 3A, 5).

Net ecosystem exchange was closely correlated with GPP (r 2 = 0.72, P < 0.0001) and only weakly correlated with R ECO (r 2 = 0.10, P < 0.01), suggesting that GPP drove overall patterns of NEE. Net ecosystem exchange also showed a weak negative correlation with mean weekly soil temperature (r 2 = 0.22, P < 0.001) and a weak positive correlation with mean weekly WFPS (r 2 = 0.17, P < 0.01). Gross primary productivity was also negatively correlated with mean weekly soil temperature (r 2 = 0.60, P < 0.0001) and positively correlated with mean weekly WFPS (r 2 = 0.42, P < 0.0001).

Ecosystem Global Warming Potential

The measured global warming potential of this peatland was 530.8 ± 188.5 g CO2-C equivalents m−2 y−1, based on eddy covariance measurements of NEE and spatially extrapolated chamber fluxes of CH4 and N2O. The global warming potential of CH4 and N2O fluxes entirely offset the CO2 uptake by photosynthesis, with an equivalent of 236.8 ± 85.5 g CO2-C equivalents m−2 y−1 and 302.4 ± 168.0 g CO2-C equivalents m−2 y−1 released to the atmosphere from soil CH4 and N2O emissions, respectively.

Discussion

Methane Emissions

Methane fluxes from this ecosystem were high for arable land, contributing to regional climate warming (that is, 9.5 ± 3.4 g CH4-C m−2 y−1). This inference is independently supported by tall tower atmospheric measurements, which indicate that managed Delta peatlands are important regional sources of greenhouse gases (Zhao and others 2009). Soil CH4 fluxes for this ecosystem exceeded those for other managed temperate peatlands, including those managed for forestry (Fowler and others 1995a), grazing or dairy farming (Schrier-Uijl and others 2009; Jungkunst and Fiedler 2007), or controlled drying experiments (Strack and Waddington 2007). Drainage and management activities likely decreased CH4 emissions relative to unmanaged, “pristine” peatlands through drying of surface peats, enhancing the size of the oxic (that is, CH4-oxidizing) zone, while simultaneously diminishing the size and activity of anoxic (that is, methanogenic) horizons (Strack and Waddington 2007; Moore and Roulet 1993). Elevated soil CH4 fluxes stem from emission “hotspots” in the landscape (that is, drainage ditches) and anaerobic decomposition of the underlying peat beneath the water table. One shortcoming of our chamber-based approach is that we probably underestimated ecosystem CH4 exchange by excluding bovine fluxes. Bovine emissions are potentially quite large; using the per capita emission rates reported in the literature, we estimated that a herd of a similar size to that grazing our study site (~100 cows) could emit as much as 10–40 g CH4-C m−2 y−1 (Laubach and Kelliher 2005; Laubach and others 2008; McGinn and others 2009; Shibata and Terada 2010; Shaw and others 2007). However, bovine fluxes were difficult to estimate because cows were an irregular presence in our study site.

Discrepancies between chamber and tower measurements of CH4 flux ultimately stem from the spatial heterogeneity of CH4 sources across the landscape, combined with the size and shape of the daytime tower footprint (Detto and others 2010a). During the day, the tower footprint only captured CH4 fluxes from drier uplands immediately adjacent to the tower, and did not adequately sample from drainage ditches and hummock/hollows (that is, higher CH4-emitting landforms), which were more distally located (Figure 1). Because these drier areas were very weak or near-zero CH4 sources, the correlation between static chamber and eddy covariance measurements was very poor when we compared chamber measurements from within the daytime tower footprint against eddy covariance fluxes. Due to the complex, heterogeneous distribution of CH4-sources across the landscape and aseasonality of CH4 fluxes, the spatially weighted upscaling of chamber fluxes probably better represents ecosystem CH4 fluxes than eddy covariance measurements for this site.

Methane fluxes showed greater spatial, rather than temporal heterogeneity. Overall soil CH4 emissions were driven by aseasonal “hotspots” of biological activity. Despite the fact that drainage ditches accounted for less than 5% of the land area, they contributed more than 84% of CH4 emissions and more than 37% of ecosystem global warming potential. Fluxes from these CH4 hotspots showed little or no seasonal variability, presumably because drainage ditches are perennially wet and stably anoxic, with a relatively abundant supply of labile organic matter (Deverel and others 2007). Other CH4-emitting landforms (for example, hummock/hollows, slopes, crowns) also showed little or no temporal variability in fluxes, probably because water table depth, soil moisture, and WFPS probably did not vary enough throughout the year to cause dynamic changes in the activity of methanotrophic and methanogenic microsites (Jungkunst and Fiedler 2007). The weak inverse relationship between net CH4 fluxes and mean weekly temperature detected by the eddy covariance tower is puzzling because we did not observe a similar pattern in the chamber data. One explanation for this is that CH4 oxidation was slightly enhanced during warmer periods in the immediate footprint of the tower, decreasing net CH4 fluxes. The chamber measurements may not have detected this trend, because it was highly localized to the area immediately adjacent to the tower, rather than a more system-wide phenomena.

Nitrous Oxide Dynamics

Nitrous oxide fluxes from this peatland were large; equal to or greater than those from heavily fertilized agricultural systems and tropical forests, which are the two largest N2O sources worldwide (Matson and Vitousek 1990; Perez and others 2001). These very high N2O fluxes were probably sustained by the input of manure, agricultural run-off, and fertilizer application, all of which increase N throughput via nitrification or denitrification (Flessa and others 1998; Furukawa and others 2002; Inubushi and others 2003; Regina and others 2004; Service 2007). In addition, colonization of the site by the invasive alien pepperweed (Lepidium latifolium L.) may have also enhanced N2O fluxes because this species is known to increase soil N-turnover and throughput (Blank and Youn 2002).

Nitrous oxide fluxes were more homogeneously distributed across the landscape than CH4 fluxes, with the highest emissions coming from drier landforms. Nitrous oxide fluxes were greatest from crowns, hummock/hollows, and slopes, which together comprised more than 95% of the land area. Nitrous oxide fluxes were more responsive to changes in environmental conditions than CH4 fluxes, although N2O fluxes at the ecosystem scale were relatively aseasonal. Only crown landforms responded to seasonal fluctuations in WFPS and soil temperature. Although drainage ditches also responded to changes in WFPS, these fluctuations were driven by management practices rather than by seasonal water dynamics.

In the crowns and drainage ditches, the positive relationship between N2O fluxes and WFPS (Figure 6A) implies that denitrification was the dominant N2O-producing process in these landforms (Smith and others 1998; Bateman and Baggs 2005); an inference supported by follow-up studies using denitrification enzyme assays and stable isotope tracers to partition N2O sources (Yang and others 2011). In the crowns, the negative correlation between N2O fluxes and soil temperature (Figure 6B) probably reflects the drying effects of warmer surface soil temperatures, rather than direct temperature-inhibition of N2O production. Although crown landforms (21.1 ± 0.5°C in the 0–10-cm depth) were significantly warmer than the others (15.5 ± 0.2°C in the 0–10-cm depth), temperatures were not high enough to directly inhibit nitrification or denitrification (Barnard and others 2005; Smith and others 1998). Soil temperatures, however, were negatively correlated with WFPS, with soil temperatures above 16°C driving WFPS below 35% (r 2 = 0.49, P < 0.001); the critical moisture threshold below which N2O production from nitrification becomes increasingly substrate-limited (Bateman and Baggs 2005; Stark and Firestone 1995). Analysis of the frequency distribution of the soil temperature and moisture data indicated that surface soils were above 16°C and less than 35% WFPS for more than 75% of the year, indicating that even this high N2O production by nitrification or denitrification was probably constrained by substrate availability or by redox conditions for most of the observation period.

Figure 6
figure 6

Nitrous oxide fluxes plotted against soil moisture (A) and soil temperature (B) for crown landforms. In both the panels, four was added to the raw N2O fluxes so that net negative or zero fluxes could be log-transformed. Each data point represents the mean of five replicate flux chambers. Bars indicate standard errors.

Full size image

Soil and Ecosystem CO2 Fluxes

Although NEE for April 2007 to May 2008 approached net CO2-neutrality (Sonnentag and others 2010), the continual loss of C from this peatland has lead to massive land subsidence and soil compaction, suggesting that respiration outpaces plant C-fixation over longer time scales (Drexler and others 2009; Ingebritsen and other 2000; Miller and others 2000; Rojstaczer and Deverel 1993; Rojstaczer and Deverel 1995; Service 2007). This inference is supported by subsequent eddy covariance measurements conducted at this field site since the completion of this study, which indicates that the site is a net CO2 source over the medium- to long-term, with interannual variability in fluxes modulated by management activities (Baldocchi and others, unpublished).

Soil CO2 fluxes from this ecosystem were very high, with respiration rates that were comparable to fluxes from humid tropical forests, which have the greatest soil respiration rates globally (Raich and Schlesinger 1992). These high soil respiration rates were sustained by microbial oxidation of the underlying peat, combined with vigorous autotrophic respiration during the growing season (Deverel and Rojstaczer 1996; Miller and others 2000). Soil respiration rates were slightly higher from static chambers than soil respiration estimated from night time eddy covariance measurements, although the two were strongly positively correlated (r 2 = 0.70, P < 0.0001). This discrepancy is likely the result of CO2 storage due to nocturnal thermal stratification, or cooler temperatures lowering plant and microbial respiration at night (Baldocchi 2003).

Of the three greenhouse gases, CO2 shows the greatest seasonal variability, driven primarily by plant responses to seasonal fluctuations in soil moisture and temperature. Photoassimilation rates (GPP) determined the overall direction (that is, source or sink) and magnitude of ecosystem CO2 exchange (NEE), with peak periods of plant productivity in spring and summer leading to an overall draw down of atmospheric CO2. Soil temperature and moisture play an important role in regulating GPP during the growing season, as demonstrated by the strong negative correlation between GPP and temperature (r 2 = 0.60, P < 0.0001; that is, increasing soil temperature promoting greater C uptake), and the positive correlation between GPP and WFPS (r 2 = 0.42, P < 0.0001). During quiescent periods (that is, late autumn or winter), CO2 was emitted to the atmosphere as soil respiration gradually outpaced photosynthesis.

Soil respiration in this ecosystem was simultaneously regulated by soil moisture and temperature. Increases in WFPS following winter storms or due to water management led to reductions in soil respiration, suggesting that soil CO2 fluxes in this ecosystem may be transport-limited (Smith and others 2003; Teh and others 2005). Soil temperature, on the other hand, was positively correlated with soil respiration rates, suggesting that cooler temperatures during spring, autumn, and winter limited the overall metabolic activity of plant roots and soil microbes.

One important pathway for C loss that we did not explore in this study was aquatic C export (Deverel and others 2007; Deverel and Rojstaczer 1996). Losses of dissolved gases (that is, “gas evasion” sensu Billett and others 2004), dissolved organic C (DOC), and particulate organic C (POC) often represent a significant C loss from peatlands (Billett and others 2004; Hope and others 2001; Limpens and others 2008; Deverel and others 2007; Deverel and Rojstaczer 1996). Measurements of DOC fluxes from Delta peatlands, including Sherman Island, suggest annual soil losses on the order of 5–110 g C m−2 y−1, which roughly translates to 0.4–8% of the GPP, or 0.4–9% of R ECO on Sherman Island (Deverel and others 2007; Deverel and Rojstaczer 1996). This is a lower figure than for other temperate peatlands, where DOC exports typically account for at least 10% of ecosystem C outputs (Limpens and others 2008). However, because GPP (−1352.9 g C m−2 y−1) and R ECO (1266.6 g m−2 y−1) were so evenly matched from April 2007 to May 2008, additional C losses from aquatic exports could very well tip the balance between C inputs and outputs, shifting the ecosystem from net balance toward a net C source (Billett and others 2004; Hope and others 2001; Limpens and others 2008).

Ecosystem Global Warming Potentials

The overall global warming potential of this ecosystem (530.8 g CO2-C equivalents m−2 y−1) was high when compared to other managed temperate peatlands in Europe and North America, where global warming potentials typically range from 100 to 500 g CO2-C equivalents m−2 y−1 (Hendriks and others 2007; Langeveld and others 1997; Schils and others 2006; Strack and Waddington 2007; Frolking and Roulet 2007). The greater global warming potential of this drained peatland was probably due to a combination of warmer conditions and very high N2O fluxes. Mean annual temperature in the Delta is 15.6°C; roughly 5–6°C warmer than other temperate field sites where comparable studies have been performed (Hendriks and others 2007; Langeveld and others 1997; Schils and others 2006; Strack and Waddington 2007; Carroll and Crill 1997). Likewise, N2O fluxes at this site were as much as an order of magnitude greater than emissions from other managed temperate peatlands (Jungkunst and Fiedler 2007; Langeveld and others 1997; Regina and others 2004; Schils and others 2006), effectively shifting global warming potentials to higher values. Methane and N2O emissions offset any ecosystem C gains made by plant C-fixation throughout the year, with consistently high, aseasonal fluxes of CH4 and N2O nullifying the atmospheric cooling effects of photoassimilation, even during peak periods of plant activity.

Conclusions and Implications for Management

The high global warming potential of these temperate agricultural peatlands suggests that they are regionally important sources of greenhouse gases; an inference supported by inverse atmospheric measurements of greenhouse gas fluxes for central California (Zhao and others 2009). The high global warming potential of this peatland is ultimately driven by emissions of non-CO2 greenhouse gases (that is, CH4, N2O). These data suggest that changes in soil moisture associated with water management practices or future climate change could have important ramifications for the global warming potential of these managed peatlands. The increased abundance of persistent, perennially flooded patches in wetter years, or under less intensive drainage could greatly enhance CH4 emissions, whereas relatively modest increases in soil moisture could amplify N2O fluxes from drier landforms. Because the potential for CH4 and N2O emissions from these soils is very large, changes in soil moisture or flooding extent need only take place over a small proportion of the landscape to dramatically increase overall ecosystem global warming potential.