Response of Black Ash Wetland Gaseous Soil Carbon Fluxes to a Simulated Emerald Ash Borer Infestation

The rapid and extensive expansion of emerald ash borer (EAB) in North America since 2002 may eliminate most existing ash stands, likely affecting critical ecosystem services associated with water and carbon cycling. To our knowledge, no studies have evaluated the coupled response of black ash (Fraxinus nigra Marsh.) wetland water tables, soil temperatures, and soil gas fluxes to an EAB infestation. Water table position, soil temperature, and soil CO2 and CH4 fluxes were monitored in nine depressional headwater black ash wetlands in northern Michigan. An EAB disturbance was simulated by girdling (girdle) or felling (ash-cut) all black ash trees with diameters greater than 2.5 cm within treated wetlands (n = 3 per treatment). Soil gas fluxes were sensitive to water table position, temperature, and disturbance. Soil CO2 fluxes were significantly higher, and high soil CH4 fluxes occurred more frequently in disturbed sites. Soil CH4 fluxes in ash-cut were marginally significantly higher than girdle during post-treatment, yet both were similar to control sites. The strong connection between depressional black ash wetland study sites and groundwater likely buffered the magnitude of disturbance-related impact on water tables and carbon cycling.


Introduction
Disturbance events are known to impact the natural function of wetland ecosystems by altering nutrient, carbon, energy, and hydrologic fluxes [1,2]. Recently, emerald ash borer (EAB) (Coleoptera: Buprestidae, Agrilus planipennis Fairmaire) was introduced as an invasive pest to ash (Fraxinus spp.) forests in North America [3]. Hundreds of millions of ash trees were estimated to have been killed in North America within the decade following the initial detection of EAB in southeastern Michigan in July 2002 [4,5]. The expansion of EAB results in significant economic costs [6,7] and causes significant perturbation to forest ecosystems [5,8,9], and will likely continue to decimate existing ash stands throughout North America.
Forested wetlands sequester large quantities of carbon from the atmosphere [10], and their carbon storage capacity is largely a result of prolonged periods of elevated water table position Forests 2018, 9,324 3 of 17 CH 4 once produced are highly sensitive to water table position [34]. Specific physical and chemical conditions need to be present for methane production to occur. Among the most important of these is the combined depletion of oxygen and absence of other, more efficient electron acceptors (e.g., NO − 3 , Fe 3+ , Mn 4+ , and SO 2− 4 ) that must be present before methanogenesis and CO 2 reduction can proceed [28,35]. The initiation or rate of methane production has also been shown to be sensitive to pH [36], temperature [15], and substrate quality and availability [19,37].
Due to the co-occurrence of methanogens and methanotrophs in peatland soils, water table position is the single most influential physical parameter that regulates the quantity of CH 4 fluxes from the soil surface. Several studies have evaluated the response of CH 4 fluxes in peatlands to temperature and water table conditions [33,[38][39][40]. Yet, to our knowledge, no study has examined the relationship between water table, soil temperature, and soil CO 2 and CH 4 fluxes in deciduous forested peatlands, nor has any study evaluated the response of gaseous soil carbon fluxes to a simulated invasive pest disturbance in deciduous forested peatlands. Temperature-dependent incubation experiments of CH 4 production in cores extracted from hummock, lawn, and hollow locations in a peatland in southwestern Scotland revealed that CH 4 production was similar among sites, yet in situ CH 4 fluxes were highly variable and closely associated with the position of the water table [33]. The response of water tables and soil temperatures to an EAB-induced disturbance in black ash (Fraxinus nigra Marsh.) wetlands will likely alter CH 4 production processes, and ultimately the amount of CH 4 emitted from the soil surface.
This study examines the relationship between, and the coupled response of, water table, soil temperature, and gaseous soil carbon fluxes within the context of the previously established altered hydrologic regime in depressional black ash wetlands. The specific objectives of this study were to (1) characterize the relationships between water table position, soil temperature, and soil CO 2 and CH 4 fluxes; and (2) evaluate the responses of soil temperature and gaseous soil carbon fluxes to disturbances in black ash wetlands. Ultimately, this study will augment the currently limited knowledge of carbon cycling processes in deciduous forested peatlands and examine the ecosystem-level implications of a looming emerald ash borer infestation. We hypothesize that (1) water table position and soil temperature will be major influences on soil CO 2 and CH 4 fluxes from these soils; (2) soil temperature will increase because of increased insolation in disturbed sites; (3) soil CO 2 fluxes will increase with the increased soil temperature in disturbed sites; and (4) soil CH 4 fluxes will be higher because of the expected increased temperatures and elevated water tables detected in response to a simulated EAB infestation.

Study Area
The Ottawa National Forest (ONF) is located in the western portion of the upper peninsula of Michigan, USA ( Figure 1). The ONF contains approximately 525 km 2 of wetlands within 4000 km 2 of mixed-hardwood forested area. Based upon US Department of Agriculture Forest Service Forest Inventory Analysis data, it is estimated that 70% of the 180 km 2 deciduous forested wetlands within the ONF are dominated by black ash [41].
Black ash wetlands in the ONF commonly occur as landform depressions surrounded by mixed-hardwood upland tree species such as sugar maple (Acer saccharum L.) and white pine (Pinus strobus L.). Other tree species, such as red maple (Acer rubrum L.), yellow birch (Betula alleghaniensis Britt.), balsam fir (Abies balsamea L. (Mill)), and white cedar (Thuja occidentalis L.), are common, though typically at low abundance within these black ash-dominated wetlands [42]. Shrub, fern, and sedge species are abundant but unevenly distributed in the understory.
Black ash wetland soils in this study were composed of woody peats. The study site soils were, on average, 75% organic matter and 47% organic carbon [17]. Peat thicknesses in the study sites ranged from 40 cm to greater than 690 cm, with an average peat depth of 140 cm. A clay lens or a poorly sorted clay loam was commonly detected at the bottom of the peat layer. Black ash wetlands in the ONF are also characterized by seasonally inundated conditions and commonly have discernible surface drainage outlet channels. Most soils in the ONF originate from glacial landforms and may exist as thin horizons overlying near-surface bedrock [43]. were each subdivided into three treatment sites (diamonds). Regional precipitation data were obtained from the National Oceanic and Atmospheric Administration's National Centers for Environmental Information (NOAA-NCEI) [44] station in Ironwood, Michigan (star).

Experimental Study Design
Nine study sites within the Ottawa National Forest, ranging in size from 0.25 to 1.25 ha, were established in 2011 in isolated landform depressions within first-order watersheds [17] (Figure 1). Black ash basal area ranged from 26% to 85% of total basal area among study sites (mean = 66%), and black ash was always the most abundant overstory tree [42]. Treatments were assigned randomly to 3 sites in 3 geographically distinct locations (n = 3 per treatment). Each distinct location included 1 untreated control, 1 girdle treatment, and 1 ash-cut treatment. A 1-year baseline data-collection period (2012) preceded the 2-year posttreatment monitoring period (2013-2014) of the study. The girdle treatment was used to simulate an active EAB infestation, wherein near-complete stand mortality could take 3-4 years [45]. In the girdled sites, all black ash trees greater than 2.5 cm in diameter at a height of 1.37 m had the bark, cambium, and phloem removed in a 15-30 cm tall circumferential band. The ash-cut treatment was used to simulate post-disturbance ecosystem conditions that might occur following complete loss of black ash from the overstory. In the ash-cut sites, all black ash trees greater than 2.5 cm at a height of 1.37 m were cut with chainsaws within 1.2 m of the soil surface during February 2013 and left on site. Treatments were applied between November 2012 and February 2013.

Instrumentation and Monitoring
Three random locations within each wetland site were determined with ArcGIS (Esri, Redlands, CA, USA), and were located at least 8 m from the wetland boundary. Two 13 cm tall gas collars constructed of 25.4 cm diameter PVC pipe were co-located 6 m apart and driven into the soil to a depth of 5 cm at each random location (n = 6 per site). Portable static dark chambers were used to measure both CO2 and CH4 fluxes [46,47], and were constructed of 25.4 cm ID PVC slip caps. A rim of closed cell foam was adhered to a PVC ring mounted within the chamber to ensure that there was an airtight seal between the gas collar and the chamber. The dark static chambers used to measure CH4 fluxes were equipped with a small fan to mix the chamber headspace [40]. When water table elevations were each subdivided into three treatment sites (diamonds). Regional precipitation data were obtained from the National Oceanic and Atmospheric Administration's National Centers for Environmental Information (NOAA-NCEI) [44] station in Ironwood, Michigan (star).

Experimental Study Design
Nine study sites within the Ottawa National Forest, ranging in size from 0.25 to 1.25 ha, were established in 2011 in isolated landform depressions within first-order watersheds [17] (Figure 1). Black ash basal area ranged from 26% to 85% of total basal area among study sites (mean = 66%), and black ash was always the most abundant overstory tree [42]. Treatments were assigned randomly to 3 sites in 3 geographically distinct locations (n = 3 per treatment). Each distinct location included 1 untreated control, 1 girdle treatment, and 1 ash-cut treatment. A 1-year baseline data-collection period (2012) preceded the 2-year posttreatment monitoring period (2013-2014) of the study. The girdle treatment was used to simulate an active EAB infestation, wherein near-complete stand mortality could take 3-4 years [45]. In the girdled sites, all black ash trees greater than 2.5 cm in diameter at a height of 1.37 m had the bark, cambium, and phloem removed in a 15-30 cm tall circumferential band. The ash-cut treatment was used to simulate post-disturbance ecosystem conditions that might occur following complete loss of black ash from the overstory. In the ash-cut sites, all black ash trees greater than 2.5 cm at a height of 1.37 m were cut with chainsaws within 1.2 m of the soil surface during February 2013 and left on site. Treatments were applied between November 2012 and February 2013.

Instrumentation and Monitoring
Three random locations within each wetland site were determined with ArcGIS (Esri, Redlands, CA, USA), and were located at least 8 m from the wetland boundary. Two 13 cm tall gas collars constructed of 25.4 cm diameter PVC pipe were co-located 6 m apart and driven into the soil to a depth of 5 cm at each random location (n = 6 per site). Portable static dark chambers were used to measure both CO 2 and CH 4 fluxes [46,47], and were constructed of 25.4 cm ID PVC slip caps. A rim of closed cell foam was adhered to a PVC ring mounted within the chamber to ensure that there was an airtight seal between the gas collar and the chamber. The dark static chambers used to measure CH 4 fluxes were equipped with a small fan to mix the chamber headspace [40]. When water table elevations exceeded 12 cm above the ground surface (6.3% of samples), high-density polyethylene extensions 50 cm long and 28 cm in diameter were used to connect the soil gas collar with the dark chamber. CO 2 and CH 4 fluxes were measured at approximately monthly intervals during snow-free seasons (June-October) from 2012-2014. All measurements were collected on calm days between 11:00 and 17:30 h (EDT) to reduce insolation variability. Measurements were not collected if more than 6.4 mm of precipitation fell within the previous 8 h. All 18 gas collars within 1 geographically distinct location were measured on the same day, and all sites were measured within a 4-day period each month.
An infrared gas analyzer (IRGA) (PP Systems EGM-4, PP Systems International Inc., Amesbury, MA, USA) was used with a static dark chamber, equipped with a small fan to mix the chamber headspace, to measure CO 2 fluxes [39,40]. CO 2 concentrations were logged by the IRGA at 2-3 s intervals for a total of 124 s. The slope of CO 2 concentration versus time (t, s) was estimated by linear regression [48]. Portions of nonlinear responses that occurred within the first 80 s of the measurement period were manually removed prior to calculating the slope. CO 2 fluxes were then estimated according to Equation (1): where dCO 2 dt is the slope from the regression, P is the atmospheric pressure (Pascal), V is the volume of the headspace within the chamber (m 3 ), A is the area of the soil encircled by the gas collar (m 2 ), R is the gas constant (J K −1 mol −1 ), and T is the air temperature (K). All dark chamber headspace volume estimates used in Equation (1) were corrected for water table position and extension headspace volume when necessary.
Gas-tight syringes were used to remove five 20 mL gas samples at consecutive 10 min intervals from each gas chamber [46]. Gas samples were immediately injected into pre-evacuated 6 mL EXETAINER ® glass vials (Labco Ltd., Lampeter, Wales, UK), and the septa were coated with 100% silicone for added leakage protection. Gas samples were analyzed within 30 days using gas chromatography (Varian 3800, Agilent Technologies Inc., Santa Clara, CA, USA) at Michigan Technological University's Hydrology and Wetlands Laboratory. As a quality-control measure, concentration standards were injected into EXETAINER ® vials in the laboratory, transported to study sites during sampling excursions, and analyzed alongside field samples. The slope of the CH 4 concentration versus time response was calculated using least-squares regression as described for CO 2 above, and the slopes of the CH 4 responses ( dCH dt ) were used in place of dCO 2 dt in Equation (1) to estimate CH 4 fluxes [15]. Gaseous soil carbon fluxes were reported as mg m −2 day −1 of CO 2 and CH 4 molecules. Approximately 8% of samples with nonlinear slopes were flagged as possible human-induced ebullition events and were not included in the analysis.
Water levels were manually measured during monthly collection intervals and were recorded at 15 min intervals (Levelogger Junior M5, Solinst Canada Ltd., Georgetown, ON, Canada) in a steel monitoring well in each wetland [17]. The water level relative to the wetland soil surface (WL mw , (cm)) was calculated for each gas collar (W L col , (cm)) using total station surveyed elevations [17] at gas collars and metal monitoring wells, respectively. Soil temperature was measured within the gas collar 5 cm below the ground surface (ST 5cm , • C) immediately following each CO 2   were subsequently installed 5 cm beneath the soil surface and within PVC solar shields located 1 m above the soil surface near the wetland monitoring well from June through November of 2014.

Statistical Analysis
Linear mixed-effects models (LMMs) were fit with the restricted maximum likelihood procedure in R [48,49] to assess the impact of water level, temperature, and treatment on soil CO 2 and CH 4 fluxes. Zero-concentration data (n = 400) from gas chromatograph calibration curves were used to determine the 0.36 ppm instrument detection limit for CH 4 [50]. Based upon the 40 min collection period, the lower detectable soil CH 4 flux limit was determined to be 1.25 (mg m −2 day −1 ). Approximately 20% of soil CH 4 flux estimates were below the lower detection limit. Distribution parameters of the natural-log transformed CH 4 flux data were calculated using maximum likelihood estimation to adjust for the censored values. Estimates for censored values were then imputed using random quantiles below the detection limit drawn from the fitted distribution as replacement values [51,52]. Following natural log-transformation of estimates of CO 2 and CH 4 from Equation (1), zero-mean assumptions of normality and homogeneity of variance were determined to be acceptable based on visual examination of residual plots for all linear mixed-effects models [48,49]. LMM methods were used to determine significance [53], the Tukey method was used to adjust p-values for all pairwise comparisons [54], and degrees of freedom were calculated according to the Kenward-Roger approximation [55]. Significance was reported where p < 0.05, and marginal significance was reported where p < 0.1. Following the transformation of CO 2 and CH 4 fluxes, the back-transformed regression equations took the form of Equation (2): where β• is the back-transformed model intercept coefficient, and β 1 is the model slope coefficient. Logged temperature and water level data from 2014 were used to test the response of maximum daily soil temperatures (ST max ) to maximum daily air temperatures (AT max ) and W L mw . The influence of W L mw on the relationship between ST max and AT max was evaluated by comparing a two-level LMM (AT max ·WL mw ) with a single-level LMM (AT max ), where AT max and W L mw were used as fixed effects, and site and day were used as crossed random effects. Additionally, the response of ST max to AT max , W L mw , and treatment during the posttreatment study period was tested with a three-level LMM where AT max , W L mw , and treatment were used as fixed effects, and site and day were used as crossed random effects.
LMM selection criteria, including the conditional variance (r 2 c ) and marginal variance (r 2 m ), the Akaike information criterion (AIC), the residual proportion change in variance (PCV), and residual standard error (RSE) were used to evaluate model performance [53,56]. Multivariate LMM response surfaces were predicted [57] by generating slopes for four water level positions: 10 cm above, 10 cm below, 30 cm below, and 50 cm below the wetland soil surface. Observations were binned with equidistant spacing between the four water level positions and used for visual interpretation only.
The influence of W L col on the relationship between soil gas flux and ST 5cm was evaluated by comparing a two-level LMM (ST 5cm ·WL col ) with a single-level LMM (ST 5cm ) for CO 2 and CH 4 , respectively (Table 1). Both ST 5cm and W L col were used as fixed effects, and correlated random intercepts and slopes were generated for each collar nested within site. Posttreatment CO 2 and CH 4 fluxes and responses to ST 5cm and W L col were modeled with a three-level LMM using ST 5cm , W L col , and treatment as interacting fixed effects with a correlated random intercept and slope for each collar nested within site. Annual differences in CO 2 and CH 4 soil fluxes by treatment were examined using a four-level LMM with ST 5cm , W L col , treatment, and year as interacting fixed-effects terms, with a correlated random intercept and slope for each collar nested within site. A four-level LMM was constructed to compare CO 2 fluxes between treatments and study periods using ST 5cm , W L col , treatment, and study period (pre-or posttreatment) as interacting fixed effects, with a correlated random intercept and slope for each collar nested within site. Finally, mean untransformed soil gas fluxes were calculated for each site during each sampling event, and Kruskal-Wallis test was used to compare between-treatment means within each year. Table 1. Model evaluation statistics for soil temperature, CO 2 flux, and CH 4 flux linear mixed models (LMMs; Figure 2). Residual standard error (RSE) was reported for soil temperature for soil CO 2 and CH 4 flux LMMs. AIC, Akaike information criterion; PCV, proportion change in variance.  Due to the limited number of CO2 flux observations collected when soil temperatures were less than 10 °C during the pretreatment study period and the absence of pretreatment CH4 flux observations, the pretreatment response slope estimates were either not reliable or nonexistent and therefore were excluded from the three-level LMM. The girdle, ash-cut, and control response slopes for soil CO2 flux (0.10, 0.08, and 0.10 mg m −2 day −1 °C −1 ) and soil CH4 flux (0.02, 0.04, and 0.06 mg m −2 day −1 °C −1 ) were not significantly different from one another during the posttreatment study period. The mean CO2 flux detected in ash-cut sites was marginally significantly higher than the controls (p = 0.094), and the mean CH4 fluxes detected in the girdle and ash-cut sites were not significantly different (p > 0.1) than the control sites during the posttreatment study period (Figure 3).

Regional Climate and Soil Temperature
Precipitation data for Ironwood, Michigan [44], indicated that the 83 cm of precipitation received in the ONF during 2012 was below the 30-year average of 89 cm, whereas the 127 cm and 111 cm received during 2013 and 2014, respectively, were above the 30-year average. The ONF received approximately 150% more snow-water equivalent (SWE) precipitation during winter 2012-2013 and 100% more SWE during winter 2013-2014 compared to winter 2011-2012. In addition, the ONF received approximately 25% more rain during 2013 and 15% more rain during 2014 compared to 2012. A comparison of model selection criteria indicated that the full LMM (AT max ·WL mw ) for ST max performed better than the null model (AT max only) by improving the relative quality (AIC), increasing the residual proportion change in variance (PCV), and decreasing the residual standard error (RSE). The full LMM explained 0.98 of the conditional variance (r 2 c ) and 0.73 of the marginal variance (r 2 m ) ( Table 1). The 0.88 ( • C • C −1 ) slope predicted for the ST max response to AT max when W L mw was farthest beneath the soil surface was larger than the 0.39 ( • C • C −1 ) slope predicted when W L mw was above the soil surface ( Figure 2). The mean daily soil temperatures of 13.2 and 13.6 • C in the girdle and ash-cut sites, respectively, were both greater than the 12.7 • C soil temperature detected in the controls, although these differences were not significant ( Table 2). Significant between-treatment mean differences are indicated by (a) and (b) at p < 0.05, where ( ) indicates detection of marginally significant between-treatment mean differences at p < 0.1.

Soil CO 2 and CH 4 Fluxes
As with soil temperature, a comparison of model statistics indicated that the full LMM (ST 5cm ·WL col ) performed better than the null LMM (ST 5cm only) for CO 2 and CH 4 , respectively, by increasing the r 2 c and r 2 m , increasing the PCV, and decreasing the RSE ( Table 1). The predicted slope of the CO 2 flux response to soil temperature was 0.18 (mg m −2 day −1 • C −1 ) when W L col was farthest beneath the soil surface, and this value was larger than the 0.06 (mg m −2 day −1 • C −1 ) response slope predicted when W L col was above the soil surface ( Figure 2) [57]. Conversely, the predicted slope of the CH 4 flux response to soil temperature was 0.05 (mg m −2 day −1 • C −1 ) when W L col was above the soil surface, and this value was larger than the 0.006 (mg m −2 day −1 • C −1 ) response slope predicted when W L col was farthest below the soil surface ( Figure 2).
Soil temperature (ST 5cm ) and water level (WL col ) were used to predict soil CO 2 and CH 4 fluxes for each treatment during the posttreatment study period (Figure 3). The CO 2 flux LMM had a 0.52 r 2 c and 0.35 r 2 m and a residual standard error of 2.1 mg m −2 day −1 , whereas the CH 4 flux LMM had a 0.37 r 2 c and 0.24 r 2 m and a residual standard error of 0.9 mg m −2 day −1 . Due to the limited number of CO 2 flux observations collected when soil temperatures were less than 10 • C during the pretreatment study period and the absence of pretreatment CH 4 flux observations, the pretreatment response slope estimates were either not reliable or nonexistent and therefore were excluded from the three-level LMM. The girdle, ash-cut, and control response slopes for soil CO 2 flux (0.10, 0.08, and 0.10 mg m −2 day −1 • C −1 ) and soil CH 4 flux (0.02, 0.04, and 0.06 mg m −2 day −1 • C −1 ) were not significantly different from one another during the posttreatment study period. The mean CO 2 flux detected in ash-cut sites was marginally significantly higher than the controls (p = 0.094), and the mean CH 4 fluxes detected in the girdle and ash-cut sites were not significantly different (p > 0.1) than the control sites during the posttreatment study period (Figure 3). LMM. The girdle, ash-cut, and control response slopes for soil CO2 flux (0.10, 0.08, and 0.10 mg m −2 day −1 °C −1 ) and soil CH4 flux (0.02, 0.04, and 0.06 mg m −2 day −1 °C −1 ) were not significantly different from one another during the posttreatment study period. The mean CO2 flux detected in ash-cut sites was marginally significantly higher than the controls (p = 0.094), and the mean CH4 fluxes detected in the girdle and ash-cut sites were not significantly different (p > 0.1) than the control sites during the posttreatment study period (Figure 3). ) for each treatment, with shaded 95% confidence intervals [57]; (B) Predicted soil CH4 flux response to soil temperature and water level ( ) for each treatment, with shaded 95% confidence intervals. was embedded, using a correlated random intercept and slope, within each LMM as an interacting fixed effect. Significant differences between treatment means are indicated by (a) and (b) at p < 0.05, where (′) indicates detection of marginally significant between-treatment means differences at p < 0.1. Figure 3. (A) Predicted soil CO 2 flux response to soil temperature and water level (W L col ) for each treatment, with shaded 95% confidence intervals [57]; (B) Predicted soil CH 4 flux response to soil temperature and water level (W L col ) for each treatment, with shaded 95% confidence intervals. W L col was embedded, using a correlated random intercept and slope, within each LMM as an interacting fixed effect. Significant differences between treatment means are indicated by (a) and (b) at p < 0.05, where ( ) indicates detection of marginally significant between-treatment means differences at p < 0.1.
There were no significant differences in mean soil CO 2 flux detected among treatments during 2012 (pretreatment), whereas the 6599 mg m −2 day −1 ash-cut mean soil CO 2 flux was marginally significantly higher (p = 0.076) than the 3193 mg m −2 day −1 control mean soil CO 2 flux during 2013, and the 6722 and 6835 mean soil CO 2 fluxes in the girdle (p = 0.042) and ash-cut (p = 0.042), respectively, were significantly higher than the 2979 mg m −2 day −1 control mean soil CO 2 flux during 2014 (Figure 4). Neither the girdle nor the ash-cut soil CH 4 fluxes were significantly different than the control during 2013 and 2014 (Figure 4). When the entire pre-and posttreatment periods were considered, the 5625 and 6545 mg m −2 day −1 mean soil CO 2 fluxes detected in the girdle (p = 0.096) and ash-cut sites (p = 0.046), respectively, were significantly higher than the 3038 mg m −2 day −1 detected in the control sites during the posttreatment study period ( Table 2). The nonparametric Kruskal-Wallis test shows that higher CO 2 fluxes occurred more frequently in the ash-cut than the controls (p = 0.085) ( Table 2). The 11.9 mg m −2 day −1 ash-cut mean soil CH 4 flux was almost twice as high as the 6.6 mg m −2 day −1 control mean soil CH 4 flux, and marginally significantly (p = 0.079) higher than the 2.4 mg m −2 day −1 girdle mean soil CH 4 flux during the posttreatment study period (Table 2). While the mean CH 4 fluxes detected in girdle and ash-cut sites were similar to controls during the posttreatment period, the nonparametric Kruskal-Wallis test shows that higher CH 4 fluxes occurred more frequently in the ash-cut sites than the controls (p = 0.026) ( Table 2). Within-treatment soil CO 2 flux comparisons were not considered due to the limited number of flux observations collected when soil temperatures were less than 10 • C during the pretreatment study period, and no pretreatment CH 4 flux data were available.
CH4 fluxes detected in girdle and ash-cut sites were similar to controls during the posttreatment period, the nonparametric Kruskal-Wallis test shows that higher CH4 fluxes occurred more frequently in the ash-cut sites than the controls (p = 0.026) ( Table 2). Within-treatment soil CO2 flux comparisons were not considered due to the limited number of flux observations collected when soil temperatures were less than 10 °C during the pretreatment study period, and no pretreatment CH4 flux data were available.  , where ( ) indicates detection of marginally significant between-treatment mean differences at p < 0.1. Boxplot lines represent median, 25%, and 75% quantiles, while whiskers represent quantiles ± 1.5 · interquantile range, respectively [58].

Soil Temperature and Water Level
Wetland water level (W L mw ) significantly altered the response of soil temperature (ST max ) to air temperature (AT max ; Figure 2). When W L mw was lower and the soil presumably drier, ST max more closely paralleled AT max . In contrast, when W L mw was higher (near or above surface), ST max was less correlated to AT max . The combined influence of the higher specific heat capacity of water and the consistent groundwater discharge into black ash wetlands [17] likely decreased the sensitivity of ST max to AT max when soils were saturated. Furthermore, soil temperature response to treatment was likely mitigated by the consistent influx of groundwater because of the expected increase in magnitude of latent heat exchange and reduction in magnitude of sensible heat exchange between soil and atmosphere [59].
The wetter conditions (Table 2) and higher wetland water levels during the posttreatment period [17] may have buffered the magnitude of ST max response to disturbance. Because of the wetter-than-average posttreatment water years and the resultant elevated water tables, the average soil temperatures in girdle and ash-cut sites were not significantly greater than the controls, despite the presumed increase in canopy openness and insolation. We theorized that the increased amount of insolation likely received by soils in the disturbed study sites would have a greater influence on the response of ST max to AT max during drier periods. It is likely that average ST max differences would have been larger during the posttreatment study period under drier conditions.

Responses of Soil CO 2 and CH 4 Fluxes to Disturbance
The water level position relative to the wetland soil surface (W L col ) regulated the response of soil CO 2 and CH 4 fluxes to soil temperature (ST 5cm ; Figure 2), and co-varied soil CO 2 and CH 4 flux responses to ST 5cm and W L col have been observed in numerous wetland types [15,19,60]. Soil CO 2 fluxes tended to be higher and increase more rapidly in response to increased temperature when W L col was below the wetland soil surface, whereas soil CH 4 fluxes tended to be higher and increase more rapidly in response to increased temperature when W L col was at or above the wetland soil surface. The magnitude of observed soil CO 2 and CH 4 fluxes from depressional black ash wetlands was similar to fluxes observed in wetlands under field and laboratory conditions [15,[61][62][63].
Given that both the magnitude of soil CO 2 and CH 4 fluxes and the correlation of ST max to AT max were controlled by wetland water levels, it is likely that interannual climatic conditions also affect the relative magnitude of mean annual soil CO 2 and CH 4 fluxes. Specifically, mean annual soil CO 2 fluxes may be higher during drier periods and lower during wetter periods. Conversely, soil CH 4 fluxes would be expected to be higher during wetter periods and lower during drier periods. Therefore, it is possible that the soil CO 2 flux response to disturbance was buffered and the soil CH 4 flux response was enhanced because of the wetter-than-average conditions during the posttreatment study period [17]. Black ash wetland water levels tended to be higher during the two wetter water years [17], and soil temperature tended to be less responsive to air temperature when water levels were higher ( Figure 2). Moreover, wetland soil CO 2 fluxes are typically lower when water tables are above the surface compared to fluxes measured during lower water table positions [19,62] Therefore, soil CO 2 fluxes from disturbed sites may have been higher and possibly more dissimilar than control sites if drier climatic conditions had occurred during the two posttreatment monitoring years.
The magnitude of soil CO 2 fluxes was significantly higher in disturbed sites, and high soil CH 4 fluxes occurred more frequently in the ash-cut than the control during the posttreatment study period (Figures 3 and 4; Table 2). As evidenced by the similarity among slopes, the sensitivity of soil CO 2 fluxes to environmental variables (ST 5cm and W L col ) did not change following disturbance ( Figure 3). Therefore, the higher soil CO 2 fluxes from disturbed sites were not caused by an altered sensitivity to environmental variables. Similarly, soil CH 4 response slopes for the three treatments were not significantly different from one another during the posttreatment study period (Figure 3), and the significantly more frequent occurrence of high CH 4 fluxes in ash-cut sites may not have been caused by an altered sensitivity to environmental variables.
It is possible that the increased rate of heterotrophic respiration from disturbed soils stemmed from an alteration in the organic chemical composition of peatland soil [64]. Carbon mineralization rates were positively correlated with labile carbon compounds and negatively correlated with recalcitrant carbon compounds in a northern Minnesota peatland [65]. Increased rates of CH 4 production were detected following the addition of glucose to lignin-rich peat in northern North American peatlands matter [66], and CO 2 and CH 4 production was significantly affected by pH and C:N ratios in Dutch peat grasslands [67]. Fine root respiration may also account for a substantial proportion of soil CO 2 fluxes, and increased soil CO 2 fluxes have been attributed to turnover and decomposition of fine roots [68][69][70]. Black ash trees have a shallow, fibrous root system [71], and the majority of this fine root system was likely available for decomposition following disturbance. Unlike the rapid infilling from co-occurring tree species that occurred following EAB infestation in green ash (Fraxinus pennsylvanica Marsh.) stands in northwest Ohio [72], neither overstory or understory stems of co-occurring tree species within these black ash wetland study sites responded positively following treatment [42]. Therefore, the increased magnitude of CO 2 fluxes (Figure 4) from disturbed black ash wetland soils may be at least partially attributed to the transient large influx of decomposable fine roots and an associated response in microbial activity and biomass [63,73], and not likely the result of increased autotrophic respiration from the release of co-occurring tree species. Inversely, the loss of autotrophic root respiration from mature individuals may lead to a decrease in soil CO 2 flux following disturbance [74,75]. A lack of response of inorganic soil nitrogen availability following treatment in these sites during this study period [76] supports the explanation that increased microbial activity drove the observed CO 2 response.
The study sites were consistently connected to mineral-rich groundwater during the study period [17], and this connection likely influenced carbon mineralization dynamics within depressional black ash wetland soils. The relative standard error of soil CH 4 fluxes (18.0%) was considerably larger than the relative standard error of soil CO 2 fluxes (<0.1%; Table 2). Because the highest rates of methane oxidation are expected near oxic/anoxic interfaces [28], the large variation of soil CH 4 fluxes that occurred when wetlands were saturated and/or inundated (Figure 3) was not likely caused by increased rates of methanotrophy within the peat profile. This further suggests that methanogenesis may have been constrained by the availability of electron acceptors [77], inhibited by chemical compounds [28] or low pH [78], or transported to the atmosphere through vegetation with arenchyma [79,80].
When compared to control sites, and because of the warmer and wetter conditions observed in the disturbed sites after treatment, we expected to observe higher posttreatment mean soil CH 4 fluxes in ash-cut and girdle sites. However, the persistent connection with mineral-rich groundwater during the wetter posttreatment study period may, in fact, have buffered the relative magnitude of CH 4 flux response to disturbance by inhibiting methanogenesis. Specifically, it is likely that study sites were consistently being replenished with more efficient electron-accepting redox species such as Fe 2+ , Mn 4+ , or SO 2− 4 during the wetter posttreatment period, which effectively inhibited methanogenesis. It is also possible that methane production within the soil increased following disturbance, but increased CH 4 fluxes were not detected from the soil surface due to plant-mediated transport.
The large increase of herbaceous cover following disturbance in these sites may have limited natural woody regeneration recruitment to larger seedling size classes or saplings [42], and canopy removal has been shown to convert black ash-dominated wetlands to herbaceous or scrub-shrub-dominated wetlands [81]. As a result, canopy replacement through natural regeneration may take several decades [82], where planting mitigation efforts will influence the tree species composition, likely decrease the duration of canopy recovery, and help maintain ecosystem function following disturbance [83][84][85]. Though the hydrogeomorphic setting is distinctly different from the black ash wetlands considered here, these findings may have important implications for potential effects in other ash-dominated wetlands such as those dominated by green and pumpkin ash (Fraxinus profunda Bush). The contribution of herbaceous vegetation to soil respiration following disturbance may be substantial [86], and the contribution of the herbaceous community to autotrophic soil respiration may have increased, given the increased amount of herbaceous cover detected in these sites post treatment [42]. In addition, the increased magnitude of heterotrophic respiration associated with the elevated soil temperatures and transient large influx of decomposable fine roots likely contributed to the significantly higher girdle and ash-cut mean soil CO 2 fluxes during the posttreatment period when compared to controls. Warmer and wetter conditions are expected in the upper Great Lakes region under elevated atmospheric CO 2 conditions, which may also affect soil CH 4 fluxes in EAB-affected wetlands [87]. Therefore, it is likely that successful planting mitigation efforts will effectively decrease the duration of EAB infestation-induced increases of soil CO 2 fluxes and buffer the losses of soil C storage within depressional black ash wetlands.

Conclusions
Control, girdle, and ash-cut treatments were applied to three sites and used to simulate an EAB disturbance. Water level, soil temperature, and soil CO 2 and CH 4 fluxes were monitored during preand posttreatment study periods. Mean soil temperatures were higher in girdle and ash-cut sites compared to the controls. The relative water table position influenced the degree of soil temperature sensitivity to air temperature, where soil temperature more closely paralleled air temperature when water levels were below the wetland soil surface. Consequently, the wetter-than-average posttreatment climatic conditions [17] likely inhibited the magnitude of soil temperature response to disturbance where average or drier-than-average climatic conditions may have yielded even larger differences between treated and untreated sites.
Soil CO 2 fluxes were significantly higher and higher soil CH 4 fluxes occurred more frequently in disturbed sites, but mean soil CH 4 fluxes were similar in disturbed and undisturbed sites. Soil temperature and wetland water levels controlled soil CO 2 and CH 4 fluxes, where warmer and drier conditions yielded higher soil CO 2 fluxes, and warmer and wetter conditions yielded higher soil CH 4 fluxes. The relationship between soil CO 2 and CH 4 fluxes and soil temperature and water level was not influenced by a simulated EAB disturbance. The combination of higher mean soil temperatures, increased herbaceous cover, and the transient large influx of decomposable fine roots following canopy suppression may have contributed to the higher soil CO 2 fluxes detected in disturbed sites. Conversely, the consistent connection with mineral-rich groundwater flow likely constrained methanogenesis, and consequently reduced the relative magnitude of soil CH 4 flux response to a simulated EAB disturbance.
Black ash wetland soil carbon mineralization processes were sensitive, but also showed resilience to EAB-induced impacts. Both CO 2 and CH 4 fluxes were regulated by soil temperature and water level position. The consistent connection with groundwater flow systems and wetter-than-average weather conditions during the posttreatment period likely buffered the relative magnitude of soil CH 4 response to disturbance.