Carbon turnover and microbial activity in an artificial soil under imposed cyclic drainage and imbibition

Water table fluctuations generate temporally and spatially dynamic physicochemical conditions that drive biogeochemical hot spots and hot moments in the vadose zone. However, their role in the cycling of soil C remains poorly known. Here, we present results from unvegetated column experiments filled with 45 cm of artificial soil containing 10% humus, and inoculated with a natural microbial extract. In one series of three replicate columns, five cycles, each consisting of a 4‐wk drainage followed by a 4‐wk imbibition period, were imposed, whereas in a second series, the water table remained static. Depth‐resolved O2 concentration profiles and headspace CO2 effluxes were markedly different between the two regimes. In the fluctuating regime, drainage periods yielded 2.5 times greater CO2 effluxes than imbibition periods. At the end of the experiment, the fluctuating water table columns exhibited a distinct zone of organic C (OC) depletion in the depth interval of 8–20 cm that was not observed under the static regime. Although this zone showed elevated levels of adenosine triphosphate (ATP), the microbial biomass was actually lower than at the corresponding depth interval of the static regime. A vertically stratified microbial community established in all columns that depended on oxygenation with depth. The 16S ribosomal RNA (rRNA) gene analyses showed a slightly higher diversity in the soil exposed to moisture fluctuations, but there was no clear difference in major taxa and microbial community composition between treatments. These results thus suggest that the localized enhancement of OC degradation induced by the water table fluctuations was driven by a more active, rather than a more abundant or compositionally very different, microbial community.

Soil moisture is a primary factor controlling the delivery of O 2 along the soil profile, which is in turn linked to the availability and utilization of other electron acceptors and, consequently, to the rate and mechanisms of soil organic C (OC) degradation (Keiluweit, Nico, Kleber, & Fendorf, 2016;Moyano, Manzoni, & Chenu, 2013). Therefore, understanding how subsurface soil biogeochemical functioning responds to periodic variations in oxygenation mediated by groundwater table fluctuations is essential for predicting the fate and turnover of soil organic matter and the associated soilatmosphere gas exchanges (Klüpfel et al., 2014;Moyano et al., 2013).
In a fluctuating water table riparian-soil column experiment, Rezanezhad et al. (2014) found that transient redox conditions enhanced microbial oxidation of soil OC, and that water table drawdown yielded peak CO 2 emissions. Moisture content and redox conditions influence both microbial activity and the development of microbial communities in soils by inducing water potential stress and modulating the delivery of energy yielding electron acceptors (Jost et al., 2011). Porenetwork connectivity also affects the interactions within bacterial communities because of different abilities to use electron acceptors and metabolize substrates under varying redox potentials (Dassonville & Renault, 2002;Vos, Wolf, Jennings, & Kowalchuk, 2013). Pett-Ridge et al. (2006) and DeAngelis, Silver, Thompson, and Firestone (2010) assessed the effects of redox conditions on the microbial community composition of a humid tropical soil by carrying out incubation experiments under oxic, anoxic, and fluctuating (oxic-anoxic) redox conditions. Their results showed that the microbial biomass concentration and community composition remained stable under redox fluctuating conditions. However, the biomass concentrations decreased and the microbial composition changed under both the static oxic and static anoxic conditions. These findings thus imply that frequent depletion and resupply of O 2 may enhance microbial tolerance to changing environmental conditions.
Soil incubation experiments are often carried out with natural soils. The latter are inherently complex matrices, making it often difficult to identify all the factors and interactions controlling soil C cycling. The use of model matrices, such as artificial soil mixtures, provides an alternative approach for answering specific research questions. With an artificial soil, the initial conditions, including the composition and structure of the soil matrix, are well defined and artifacts related to heterogeneity are minimized (Pronk et al., 2017). Previous experiments with artificial soils have helped elucidate the specific roles of mineral components in soil development, the formation of organo-mineral associations, and microbial community changes Pronk, Heister, Ding, Smalla, & Kögel-Knabner, 2012).
Here, we present results from soil column incubations in which columns packed with a well characterized artificial soil as starting material were subjected to cycles of drainage and imbibition, whereas in identical control columns, the water table was held at a fixed position. The soil columns were spiked with a microbial inoculum isolated from a natural soil environment. The responses to the imposed hydraulic regimes were monitored by measuring greenhouse gas effluxes, depth-dependent redox potentials, and dissolved O 2 concentrations, and by periodically sampling and analyzing the pore-water chemistry. At the end of the incubations (329 d), the cores were sliced and the depth distributions of OC, microbial biomass, adenosine triphosphate (ATP) concentrations, and the microbial community composition were determined. The results shed new light on the response of the microbial community to dynamic oxygenation and moisture conditions, as well as the consequences for OC turnover.

Artificial soil columns
Six clear hard acrylic columns, with 7.5-cm i.d. and 60-cm length, were each packed with 45 cm of a homogeneous artificial soil mixture at an initial average bulk density of 1 g cm −3 , leaving a headspace of ∼660 cm 3 . Similar to a previously described method (Pronk et al., 2012), the artificial soil was prepared by dry mixing 77.5% quartz sand (3Q-ROK, US Silica), 10% montmorillonite (Ceratosil WG, Süd Chemie), 2.5% goethite (Bayoxide EF 200,Lanxess), and 10% organic matter collected from the H-horizon humus of a mixed deciduous forest soil at the rare Charitable Research Reserve, Cambridge, ON, Canada. The organic matter was air dried, homogenized, sieved (<2 mm), and used without sterilization. Dibasic calcium phosphate (0.8 mg g −1 ) was added as a P supplement. After mixing, the soil was partially wetted with synthetic groundwater and inoculated with the water extractable microbial community extracted from the same humus at a ratio of 6 g humus kg −1 soil (Pronk et al., 2012). Synthetic deoxygenated groundwater was prepared to match the major ion chemistry of the local groundwater at the rare site where the organic matter was collected: KHCO 3 (0.5 mM), CaCl 2 (1 mM), and MgCl 2 (1 mM). The OC content measured on the initial artificial soil mixture was 35 ± 3 mg g −1 , the molar C/N ratio was 36, and the soil water pH was 7.5.

Incubation regimes
Two sets of triplicate columns were incubated under two contrasting treatments: static and fluctuating. The columns are hereafter referred to as C1, C2, and C3 for the static (control) columns and F1, F2, and F3 for the fluctuating columns. The hydraulic regimes were imposed as shown in Figure 1 and described in detail in Rezanezhad et al. (2014). Briefly, a programable multichannel pump regulated the water level in a hydrostatic equilibrium reservoir by flowing water in (out) from (to) a water storage reservoir. A hydrostatic equilibrium reservoir was hydraulically connected to the three replicate soil columns in each of the two hydraulic treatments. Raising or lowering the water level in the hydrostatic equilibrium column then caused the water level in the soil columns to either rise or drain.
The water table in the static columns was held at a depth of 20 cm below the soil surface (bss) for the entire duration of the experiment. Some water was added to the equilibrium column on a weekly basis to compensate for evaporative loss. In the fluctuating water table treatment, five consecutive drainage and imbibition cycles were imposed. Imbibition was imposed by instantaneously increasing the water level in the equilibrium column to the height of the soil surface Core Ideas • Artificial soil columns were incubated with natural soil microbial inoculum. • Replicates were subjected to cycles of drainage and imbibition. • Drainage after water-logged conditions led to pronounced pulses of CO 2 efflux. • Incubation treatment yielded distinct zone of OC depletion in the fluctuation zone. • Enhanced turnover was driven by more active and not more diverse biomass community.
(0 cm) and maintaining that level for 28 d. Thus, water flowed into the soil-packed columns from the equilibrium reservoir ( Figure 1). During drainage, the water level was lowered from 0 to 40 cm bss by allowing free drainage of the soil columns into an outflow reservoir. The drainage volume was measured and samples were collected for chemical analysis. When the water level in the equilibrium reservoir reached 40 cm bss, it was held there for 28 d until the next cycle. The columns were incubated in the dark at 23.6 ± 1.6 • C. The headspace of the soil columns was flushed with watersaturated air through the upper ports of the columns to minimize evaporative loss, whereas the hydrostatic equilibrium reservoirs were sparged continuously with Ar gas to ensure anoxic conditions. Columns C1 and F1 were sacrificed 99 d after the start of incubation, after the completion of the first drainage and imbibition cycle. Columns C2, C3, F2, and F3 were incubated for 329 d before being sampled sacrificially.

In situ sensors and pore water sampling
The columns were provided with regularly spaced ports for installing redox potential (E h ) and O 2 sensors and extracting pore water along the depth profile of the columns (Figure 1). Pore water samples were periodically collected with ceramic MicroRhizon samplers (5-cm length, 2.5-mm diam., with a filter pore size of 0.15 μm; CSS5 MicroRhizon, #19.21.23F). The ceramic samplers were installed at 5-, 15-, 25-, and 35cm depth bss in all soil columns. Up to 5 ml of pore water was extracted every 2 to 4 wk using a vacuum pump set at −100 mbar. However, in the static water table columns, little or no pore water could be extracted at depths above the water table.
High-resolution E h measurements were carried out at 5-min intervals with glass-tip microelectrodes that were combined with an open-ended gel-stabilized reference electrode that was in contact with the bottom inflow and outflow of the fluctuating (F), comprises three replicates. All replicate soil-packed columns are hydraulically connected to an equilibrium water reservoir (equilibrium column) that controls the pressure head at the lower boundary of the soil columns and enables the transitions from drainage (drop in equilibrium reservoir level) to imbibition (rise in reservoir level) and vice versa soil column (10 μM tip diameter, Unisense). The E h electrodes were installed in column C1 at depths of 3 and 37 cm bss, and in column F2 at depths of 3, 10, 21, and 37 cm bss. Dissolved O 2 was measured every 3 h using an in-house developed luminescence-based optode sensor consisting of fiber optic cables with Pt(II) meso-tetra(pentafluorophenyl)porphine (PtTFPP) sensing and imaging tips (Badocco, Mondin, & Pastore, 2012). The sensing tips were inserted at 2. 5, 8.5, 14.5, 20.5, 26.5, and 32.5 cm bss in columns C2 and F2. The uncoated imaging ends were placed in front of a digital single-lens reflex (DSLR) camera with blue light-emitting diode (LED) light of 447.5-nm wavelength (Larsen, Borisov, Grunwald, Klimant, & Glud, 2011). The LED illumination provided an excitation light source to trigger the emission of PtTFPP luminescence. The emitted light was photographed and the light intensity was analyzed using ImageJ software (Rasband, 2015). The Stern-Volmer (SV) equation relates the intensity to the O 2 concentration: where I 0 denotes the emitted light intensity in the absence of O 2 , I is the light intensity in the presence of O 2 , [O 2 ] is the concentration of O 2 , and K ′ SV is the SV constant (Badocco et al., 2012).
Columns C3 and F3 were also fitted with nonpolarizing Ag-AgCl potential electrodes and Ag-AgCl current coil electrodes for spectral induced polarization (SIP) and electrodic potential (EP) measurements ( Figure 1). The acquisition of the geoelectrical signals is part of a separate study. Because the SIP and EP records are not essential to the interpretations of the biogeochemical and microbiological data, they are not presented here.

Aqueous analyses
The extracted pore water and drainage water samples were subsampled for analysis of dissolved OC (DOC), SO 4 2− , Fe 2+ , electrical conductivity (EC), and pH. The DOC measurements were performed on samples acidified by adding 20 μl of 1 M HCl into 1 ml of sample, then analyzed via the nonpurgeable OC method on a total C analyzer (Shimadzu TOC-LCPH/CPN, method detection limit = 0.011 mM). For analysis of Fe 2+ with the ferrozine method of Viollier, Inglett, Hunter, Roychoudhury, and Van Cappellen (2000), subsamples were immediately placed in an anaerobic chamber (Coy, <0.0014 mg L −1 O 2 in N 2 with 3% H 2 ). Analysis of SO 4 2− was done by ion chromatography (Dionex ICS-5000; method detection limit = 0.27 μM). Both pH and EC were measured using a handheld probe (Horiba B-213 LAQUATwin pH/EC). Because unsaturated conditions and filter membrane clogging occasionally impeded extraction of the full pore water volume, not all the analyses could be performed at all sampling times. This was particularly true for the upper parts of the columns above 20 cm bss.

Headspace gas fluxes
The headspace of columns C1, C2, F1, and F2 was sampled every 1-2 d by connecting the two lateral headspace ports to an automated multiplexer CO 2 flux measurement system ( Figure 1; LI-8100, LI-COR Biosciences). The measurement setup and procedure are outlined elsewhere (Rezanezhad et al., 2014). Briefly, the CO 2 efflux rate was obtained from the rate of concentration increase within the headspace (dCO 2 /dt, μmol m −2 s −1 ), following the method of Davidson, Savage, Verchot, and Navarro (2002). Effluxes of CH 4 and N 2 O were determined weekly. Samples were collected at 0 and 30 min after sealing the headspace. Gas analyses were carried out on a gas chromatograph (Model GC-2014, Shimadzu) equipped with an electron capture detector (ECD) and flame ionization detector (FID). The detection limit was 0.28 μg L −1 for CH 4 and 0.04 μg L −1 for N 2 O.

Soil analyses
Solid-phase analyses were performed on soil material retrieved from columns C1 and F1 after 99 d of incubation, and for the other columns at the end of the incubation (329 d). The soil was extruded from the columns using a lifting jack and sliced every 2 cm as described in Rezanezhad et al. (2014). Prior to homogenizing a soil slice, it was subsampled for gravimetric determination of bulk density (ρ b ) and water content determination by oven drying of a known volume at 105 • C for 24 h. The remaining material was then homogenized and subsampled for microbial biomass C (MBC) and ATP measurements, as well as DNA sequencing. Additional 10-g subsamples were frozen at −20 • C, freeze dried, ground to fine powder, and analyzed for total OC, inorganic C, and N concentrations using a CHNS Carbo Erba analyzer (method detection limit = 0.1% dry).
The concentrations of ATP were determined in all columns at eight specified depths. For extraction and analysis, ∼2 g of soil were flash frozen in liquid N 2 and stored at −20 • C. The methods of Jenkinson and Oades (1979) and Redmile-Gordon, White, and Brookes (2011) were used for extracting ATP from the soil samples. In our adaptation of the methods, HEPES buffer [4-(2-hydroxyethyl)-1piperazineethanesulfonic acid] replaced the arsenate buffer; we found that 20 mM HEPES adequately buffered the samples to pH 7 without interfering with the luminescence measurements. Briefly, duplicate samples were extracted by adding TIP reagent (Redmile-Gordon et al., 2011; TIP = 0.92 M trichloroacetic acid, 0.59 M imidazole, and 0.47 M Na 2 HPO 4 ) at a solid/liquid ratio of 0.1 ml TIP g −1 soil, then sonicated for 2 min, placed on ice for 8 min, and filtered (Whatman 42, 2.5-μm pore size). The extract was flash frozen and stored at −20 • C until analysis. Concentrations of ATP were determined through triplicate luminescence measurements on a FlexStation 3 multi-mode microplate reader (Molecular Devices) in multiwell-plate format using a cell viability assay kit (BacTiter-Glo microbial cell viability assay). Following Apostol, Miller, Ratto, and Kelner (2009), limits of quantification and detection of 0.03 and 0.01 nmol g −1 , respectively, were determined from the standard deviations obtained on extracted samples of autoclaved artificial soil.
The chloroform fumigation method was used to extract MBC from soil subsamples (Vance, Brookes, & Jenkinson, 1987). Approximately 2 g of soil was extracted with 0.5 M K 2 SO 4 , either after 24-h exposure to chloroform or without fumigation. The DOC concentration of the extracts (both fumigated and nonfumigated sample extracts) were analyzed, and the extractable MBC concentrations were calculated from the difference between the DOC concentrations in the fumigated and nonfumigated samples. Note that the determination of MBC was only performed on the columns sampled at the end of the incubations.
Paired-end reads were assembled using the paired-end assembler for Illumina sequences (PANDAseq version 2.8; Masella, Bartram, Truszkowski, Brown, and Neufeld, 2012) using a quality threshold of 0.9 and minimum overlap of eight bases. Assembled reads were analyzed using Quantitative Insights Into Microbial Ecology (QIIME version 1.9.0; Caporaso et al., 2010), managed by automated exploration of microbial diversity (AXIOME version 1.5; Lynch, Masella, Hall, Bartram, and Neufeld, 2013). Sequences were clustered into operational taxonomic units (OTUs) with UPARSE (Edgar, 2013) at 97% identity and aligned with the Python Nearest Alignment Space Termination tool (PyNAST version 1.2.2; Caporaso et al., 2010). Representative sequences were classified using the Ribosomal Database Project (RDP version 2.2; Wang, Garrity, Tiedje, and Cole, 2007) with a confidence threshold of 0.8 against the Greengenes database (McDonald et al., 2012). The AXIOME software generated principal coordinate analysis (PCoA) (Gower, 1966) ordinations using Bray-Curtis dissimilarities (Bray & Curtis, 1957). Permutational multivariate analysis of variance (PER-MANOVA) (Anderson, 2001) was used for statistical testing of group similarities. The PERMANOVA was conducted with the adonis function in the vegan package (version 2.4-4) in R (version 3.4.0) with 5,000 permutations. Indicator species analysis (Dufrêne & Legendre, 1997) was conducted using a stringent indicator value threshold of 0.9, sequence abundance threshold of 300, and a p value cutoff of <.001.

RESULTS AND DISCUSSION
In what follows, we discuss the results obtained on the columns that underwent the drainage-imbibition cycles and compare them with the results under static water table conditions. The full set of results for the fluctuating and static water table columns can be found in the Supporting Information. The discussion further focuses on the data collected between 100 and 329 d. Visual inspection of the pore water O 2 concentration and redox potential data series under fluctuating water table indicated that the soil biogeochemical system required at least 10 wk to relax from its initial conditions. We therefore assumed that beyond 100 d, the observed differences between the dynamic and static water table soil columns primarily reflected the differences in the hydraulic regimes.

Pore water geochemistry
The time series O 2 concentrations measured at five different depths for the fluctuating water table are presented in Figure 2, along with pore water pH and DOC measurements and headspace CO 2 fluxes. Note that the figure shows the pore water pH and DOC data for both columns F2 and F3, but only the O 2 concentrations and CO 2 fluxes measured in column F2. The complete time-series pore water geochemical data for DOC, SO 4 2− , and Fe 2+ , for both dynamic and static water table regimes, are presented in the Supporting Information (Supplemental Figures S1, S2, and S3).
Pore water pH measured at 15 cm bss in both dynamic water table columns F2 and F3 exceeded pH 8 and showed a slight upward trend after 200 d ( Figure 2b). The concentrations of DOC clearly recorded the drainage and imbibition periods with variations of up to 8 mM ( Figure 2c). During drainage, the DOC concentrations fell, whereas they increased during imbibition. The DOC data also showed good reproducibility between the duplicate columns F2 and F3. Emissions of N 2 O were below detection, and those of CH 4 were low, mostly falling between 0.01 and 0.4 μmol m −2 s −1 (see Supplemental Section S4).

Oxygen and redox dynamics
During the three complete drainage and imbibition cycles after Day 100, systematic and consistent variations in the O 2 concentrations were observed in column F2 (Figure 2a). Drainage periods resulted in sharp increases in the O 2 concentrations at depths of 8.5, 14.5, and 20.5 cm bss. At 8.5 and 14.5 cm, the O 2 concentration increases coincided closely with the onset of drainage. The O 2 concentrations asymptotically approached their maximum value (∼0.22 mM) within 3-4 d and then remained constant for the remainder of the drainage period. At 20.5 cm, the initial rapid increase of the O 2 concentration was delayed by ∼3 d relative to the shallower depths, then slowly rose from 0.12 to 0.20 mM throughout the drainage period. Imbibition periods resulted in a staggered drop in the O 2 concentrations, beginning ∼4 d after the start of imbibition at 20.5 cm bss and up to 19 d at 8.5-cm depth.
The observed hysteresis in the response of the O 2 concentrations to drainage and imbibition (i.e., a more lagged response after drainage, with changes occurring well into the imbibition period) likely reflects the hysteresis in the moisture response due to the shrinking and swelling of the 10% montmorillonite in the artificial soil (Saiyouri, Hicher, & Vadose Zone Journal F I G U R E 2 Timeseries results for (a) O 2 along the soil depth profile, (b) pore water pH, (c) pore water dissolved organic C (DOC), and (d) headspace CO 2 efflux, measured in the fluctuating treatment, column F2. Optodes for O 2 measurement were only installed in column F2. Alternating gray and white rectangles signify periods of imbibition and drainage, respectively. Porewater data (pH and DOC) are shown for the fluctuation zone (at 15 cm), that is, the depth interval that experienced the most pronounced variations in O 2 . (d) Integrated CO 2 emission (g C) calculated by integrating the CO 2 flux curves over each period-imbibition (IMB, blue) and drainage (DRN, pink)-is plotted under the flux curve, from Day 100 to the end of the experiment Tessier, 2000;Xie, Agus, Schanz, & Kolditz, 2004). In fact, because of the high water retention capacity, the uppermost soil still retained ∼30% water saturation at the end of the 4-wk drainage periods (data not shown). At depths of 26.5 and 32.5 cm bss, the O 2 concentrations were below 0.05 mM with no visible imprint of the fluctuating water table. In the static water table column C2, the O 2 levels showed little temporal variability (Supplemental Figure S4a); the concentrations hovered around 0.25 mM at 8.5 and 14.5 cm bss, then dropped to <0.075 mM at 20 cm and deeper.
The redox potential (E h ) exhibited trends consistent with those of O 2 (Supplemental Figure S5a). The E h data series, however, was incomplete due to the limited electrode lifetime of around 6 mo. Therefore, largely based on the measured O 2 dynamics under fluctuating water table, we define three depth zones in the fluctuating water table columns to facilitate the presentation and discussion of the results. The oxic zone (0-8 cm) is characterized by the presence of O 2 levels in excess of 0.2 mM during at least two-thirds of the time under the dynamic water table regime. By contrast, the anoxic zone (below 20 cm) continuously experienced O 2 levels below 0.05 mM. The fluctuation zone (8-20 cm) is the depth interval between the oxic and anoxic zones.

Carbon dioxide emissions
Under the dynamic water table regime, a pronounced pulse of CO 2 was emitted shortly after the onset of each drainage period (Figure 2d). Similar CO 2 pulses have been observed in previous soil incubation experiments (Ebrahimi & Or, 2018;Rezanezhad et al., 2014;Wang, Bogena, Vereecken, & Bruggemann, 2018). After the pulse release, the CO 2 flux dropped to a near-constant level during the final 2-3 wk of the drained period. Both the maximum and final CO 2 flux values decreased with each successive drainage period. The onset of an imbibition period was accompanied by an initial drop in CO 2 flux, followed by a more gradual decline to final values significantly below those observed during the drainage periods.
Compared with the results for column F2, the CO 2 fluxes measured in the static water table column C2 exhibited a slow, monotonic decrease over the 100-329 days of incubation, with values intermediate between those of the drainage and imbibition periods under the fluctuating water table regime (Supplemental Figure S4b). With the exception of the initial CO 2 pulses, the magnitudes of the CO 2 fluxes measured in the fluctuating and stable water table columns fell within the typical range of measurements reported for temperate forest soils over the growing season (1-6 μmol CO 2 m −2 s −1 , e.g.; Epron, Farque, Lucot, & Badot, 1999;Hanson, Wullschleger, Bohlman, & Todd, 1993;Wang et al., 2013). However, the artificial soils were incubated at 23.6 ± 1.6 • C, on average, 5-10 • C warmer than shallow soils during a temperate growing season. Thus, a lower incubation temperature would have resulted in fluxes likely close to the lower limit of the reported range.
Drainage-enhanced CO 2 fluxes are expected because of the concomitant influx of O 2 , which stimulates aerobic respiration, in part coupled to the oxidation of soluble fermentation compounds produced under the anoxic conditions of the preceding period of imbibition (Huang et al., 2019;Stegen et al., 2016). The production of fermentation compounds is in line with the observed increase in DOC concentration, at 15 cm, during each imbibition period (see Supplemental Figure S1), which is expected to stimulate CO 2 production (Cleveland, Wieder, Reed, & Townsend, 2010). Our interpretation is consistent with the observed variations in DOC at 15-cm depth in column F2, which suggest consumption of dissolved organic substrates during drainage that were accumulated in the previous imbibition period (Figure 2c). Prolonged anaerobic conditions may also have led to the release of mineral-protected C after Fe reduction (Huang & Hall, 2017). Fluctuations in pore water Fe 2+ concentrations (see Supplemental Figure S3) provide evidence that Fe reduction took place in the columns, at least during some imbibition periods. Furthermore, the initial CO 2 pulse at the start of the drainage periods may additionally reflect the rapid degassing of previously produced CO 2 to the air-filled pore space created by the lowering of the water table (Oertel, Matschullat, Zurba, Zimmermann, & Erasmi, 2016;Wang et al., 2018). In contrast, decreased CO 2 production during imbibition arises from kinetically and/or thermodynamically constrained C decomposition during O 2 limitation (Keiluweit et al., 2016). Additionally, during the water-logged conditions of the imbibition periods, emission of anaerobically produced CO 2 would be limited by slow diffusion through the pore water network.
The CO 2 emissions integrated over the drainage periods were, on average, 2.5 times higher than those of the imbibition periods (Figure 2). Furthermore, the integrated CO 2 efflux of 0.56 g C during the first drainage period decreased to 0.27 g C for the last drainage period. The integrated CO 2 emissions during the imbibition periods also decreased, albeit more gradually, from 0.21 to 0.15 g C. Vogel et al. (2014) similarly observed gradually declining CO 2 effluxes in soil incubations lasting 562 d. These authors also showed that the CO 2 production could be reactivated by adding fresh organic substrates. The decreasing temporal trends in CO 2 emissions in both fluctuating and static water table columns are therefore ascribed to the progressive depletion of the finite pool of degradable soil organic matter. Contrary to natural soil systems, there is no resupply of new organic substrates in the column experiments, thus causing CO 2 production to slow down with time (Pronk et al., 2017).

Organic carbon turnover
The distributions of OC, MBC, and ATP measured at the end of the incubation period (329 d) are presented for both the fluctuating and static water table regimes in Figure 3. The total integrated OC losses in the columns were calculated by subtracting the final OC concentrations from the initial OC content of the artificial soil (35 ± 3 mg g −1 ). Because of the fairly high OC content of the initial soil, the OC losses were relatively small. Nonetheless, the estimated OC loss relative to the total initial C stock in the fluctuating water table column (8.7 ± 0.9%) was higher than in the static water table column (7.7 ± 0.9%). Most of the difference in OC loss between the two water table regimes was localized within the 8-to 20-cm depth interval-that is, the fluctuation zone defined in Section 3.2 (Figure 3a). A similar mid-column zone of OC depletion under a fluctuating water table has been observed in a previous soil column incubation experiment (Rezanezhad et al., 2014). In this earlier study, the localized depletion of OC was ascribed to enhanced degradation of soil organic matter under the alternating oxic-anoxic conditions driven by the imposed dynamic hydraulic regime.
The total OC loss in the fluctuating water table columns F2 and F3, integrated over the entire duration of the experiment (1-329 d), was estimated from the difference between the OC distributions measured at the end and the initial OC concentration of the homogeneous soil mixture. The resulting OC loss was 5.4 ± 0.7 g C. For comparison, the integrated CO 2 efflux over the 329 d of incubation amounted to 3.9 ± 0.4 g C, or 72% of the OC loss. Note that the reported ± error of integrated CO 2 production was computed based on a 10% overestimation of closed-chamber methods, after Rayment (2000). Figure 3b compares overall OC loss to integrated outflow drainage and gas efflux. Drainage DOC and dissolved inorganic C (DIC) Vadose Zone Journal F I G U R E 3 Depth profiles of (a, d) organic C (OC), (b, e) microbial biomass C (MBC), and soil (c, f) adenosine triphosphate (ATP) at the end of the incubation (t = 329 d). Top-row panels correspond to the fluctuating water table replicates, and bottom-row panels correspond to those of the static water table treatments. Blue rectangles denote the fluctuation zone interval (8-20 cm) in the fluctuating treatment columns, and gray intervals denote the static water level in columns C2 and C3. Prior to packing, the OC content of the homogenized artificial soil mixture was 35 ± 3 mg g −1 fluxes were computed based on the average concentration of DOC and DIC measured in fluctuating-column outflow samples during each drainage period and the corresponding total discharged volume (on average, 110 ml per drainage period). Both the DOC and DIC drainage fluxes were two orders of magnitude lower, each accounting for only 2% of the OC loss. The results thus imply that most OC was removed from the soil system via the efflux of CO 2 , whereas leaching of DIC and DOC were minor C sinks. Nonetheless, ∼24% of the OC loss was not accounted for.
In addition to CO 2 , anaerobic degradation of OC can generate CH 4 . The measured CH 4 emissions, however, were very low, with values, after the initial 100 d of incubation, never exceeding 0.4 μmol m −2 s −1 , despite the water-logged anoxic conditions during the imbibition periods (weekly CH 4 flux measurements are presented in the supporting information, Supplemental Figure S6). Integrated over the entire time of the experiment, the efflux of CH 4 only represented 1.4% (0.08 ± 0.03 g C) of the OC loss (Figure 3b). Note that the weekly sampling schedule likely yields an underestimation of total CH 4 emissions due to missed pulses. The suppression of CH 4 emissions during the imbibition periods was probably due, at least in part, to anaerobic respiration coupled to the reduction of inorganic and organic terminal electron acceptors. Dissimilatory Fe(III) and SO 4 2− reduction are possible respiration pathways, given the presence of goethite in the artificial soil mixture and the detection of measurable pore water SO 4 2− levels in the oxic and fluctuation zones (Supplemental Figure S2). Furthermore, natural organic matter may possess significant electron accepting capacity and support anaerobic production of CO 2 without concomitant production of CH 4 (Aeschbacher, Vergari, Schwarzenbach, & Sander, 2011;Gao, Sander, Agethen, & Knorr, 2019;Keller & Takagi, 2013), whereas the high relative abundance of CH 4oxidizing organisms (i.e., Crenothrix spp.) at the oxic-anoxic interface likely also limited CH 4 efflux (see Section 3.5).
Further work will be needed to identify and quantify all the various biotic and abiotic processes that control the fate of OC and the emissions of CO 2 under dynamic drainage and imbibition. These include the adsorption of DOC (Kothawala, T A B L E 1 Sample diversity in the 0-to 8-cm, 8-to 20-cm, and >20-cm depth intervals, after 99 and 329 d of incubation. Under a fluctuation water table, these depth intervals correspond to the oxic, fluctuation, and anoxic zones (see Section 3.2), respectively. In the static water columns, the depth of the water table is held constant at 20 cm below the soil surface. Error margins indicate the SD based on all samples from the corresponding depth interval (on average, 15 samples per depth). A two-sample t test was used to compare the average operational taxonomic unit (OTU) abundance between constant and fluctuating columns at the same depth interval Moore, & Hendershot, 2008), as well as the formation of carbonate mineral phases at the alkaline pore water pH values measured ( Figure 1) and with Ca 2+ and Mg 2+ cations present in the artificial groundwater solution or Fe 2+ produced by the dissimilatory reduction of goethite in the artificial soil matrix coupled with OC oxidation (Fredrickson et al., 1998). In addition, the closed chamber CO 2 flux calculations may underestimate the actual fluxes. Rayment (2000) found that closed chamber flux measurements yield an underestimation of 10% because the volume of air-filled pore space is not accounted for in the total volume for the flux calculation. Nevertheless, the available data imply that the observed OC loss during the experiments was mainly the result of microbially mediated degradation of the forest soil humus.

Microbial community: Composition
The number of OTUs measured after 99 d and at the end of the experiment on dynamic and static water table columns are summarized in Table 1. The actual OTUs are shown in Supplemental Figures S7 and S8, for Days 99 and 329, respectively. The lowest numbers of OTUs were found in the deeper part of the columns, and the highest were found in the upper part. With the exception of the deepest samples collected after 99 d, the OTU counts were systematically higher in the dynamic compared with the static water table regime. The 8-to 20-cm depth interval exhibited the largest differences in OTU counts between the fluctuating and static water table regimes. In the fluctuating water table columns, the numbers of OTUs recovered from the fluctuation zone after 329 d were similar to those in the oxic zone, but significantly larger than in the anoxic zone. Thus, judging from the OTU counts, the fluctuating water table yielded a relatively higher microbial diversity than the static water table, particularly in the 8-to 20-cm depth interval (Table 1). The 16S rRNA gene sequencing of 113 samples generated 2,652,681 assembled paired-end reads. After 99 and 329 d, the largest differences in microbial community composition were observed between the uppermost and deepest sections of the soil columns, under both fluctuating and static water table (Supplemental Figures S7 and S8). The differences between depth intervals were, in fact, more pronounced than between the two water table regimes, with the same taxa dominating in the same depth intervals. In both water table regimes, the most abundant gene sequences in the 0-to 8-cm soil depth were affiliated with Sinobacteraceae, Hyphomicrobiaceae, and Cytophagaceae, whereas OTUs affiliated with Crenotrichaceae (Crenothrix spp.) and Hyphomicrobiaceae (Devosia spp.) were most abundant in the 8-to 20-cm depth interval. Known aerobic organisms of Devosia spp. and Cytophagaceae were detected in the columns from 0 to ∼16 cm before decreasing in abundance at greater depths. Crenothrix spp., which are known CH 4oxidizing organisms, were most abundant between 16-and 20-cm depths in all the columns. In the static water table columns, up to 16% of reads were affiliated with Crenothrix spp., and up to 18% were affiliated in the fluctuating water table columns. These high Crenothrix OTU counts may help explain the low CH 4 emissions measured during the experiment. At depths below 20 cm, OTUs affiliated with Bacteroidales, Clostridia, and Ignavibacteriales dominated in all columns at both time points (99 and 329 d).
In a previous experiment using the same column design, via denaturing gradient gel electrophoresis (DGGE), Rezanezhad et al. (2014) reported a lack of a pronounced differentiation in the microbial community structure between static and fluctuating water table conditions. We expected that the sequencing method used in the present study would be better suited to shed light on microbial community differences in response to a dynamic moisture regime. The microbial community in the fluctuation zone exhibited a higher OTU than its static counterpart (Table 1), indicating a slightly higher diversity of the microbial community present. However, the 16S rRNA gene analysis showed that, in corresponding depth intervals, the same major taxa dominated under both water table regimes.

Vadose Zone Journal
This indicates that although the moisture fluctuations may have created additional niches supporting a higher community diversity, those differences do not seem to strongly affect overall microbial community structure.
By contrast, some earlier research with batch incubations found that shifts in soil aeration resulted in changes in microbial community diversity, composition, and functionality (Pett-Ridge & Firestone, 2005;Picek, Šimek, & Šantrůčková, 2000), linked closely to redox-dependent N cycling (Pett-Ridge, Petersen, Nuccio, & Firestone, 2013;Randle-Boggis, Ashton, & Helgason, 2018). The latter was likely much less important in our experiment, because of the lack of N in the artificial groundwater and the high C/N ratio of the humus.
The presence of expandable clay and 10% humus yielded an artificial soil with a high water-retention capacity. Visual observation showed that during the entire drainage periods, the soil retained a wet aspect. Thus, even in the upper part of the soil, the microbial community was never exposed to complete desiccation, which in turn may have enabled the persistence of anaerobic microsites (Keiluweit, Wanzek, Kleber, Nico, & Fendorf, 2017;Sexstone, Revsbech, Parkin, & Tiedje, 1985). Furthermore, the initial microbial community was enriched from forest humus collected directly from a riparian zone and was likely already well adapted to variable redox conditions (Evans & Wallenstein, 2014). We therefore propose that the high OTUs in the fluctuation zone reflect the combination of a finite water retention, the persistence of local anoxic microsites during the periods of drainage, and the periodic ingression of O 2 . How these conditions affect the activity of the microbial community is assessed in the section below.

Microbial community: Biomass and activity
The turnover of MBC occurs over months and can be stabilized over significantly longer time scales (Miltner, Bombach, Schmidt-Brücken, & Kästner, 2012), whereas ATP is far less stable, providing a measure of metabolic activity on a timescale of days (Blagodatskaya & Kuzyakov, 2013). Thus, ATP provides microbial community activity information over a short period of time prior to sampling, whereas microbial biomass integrates longer term information and includes both active and dormant biota (Mellage et al., 2015;Stolpovsky, Fetzer, Van Cappellen, & Thullner, 2016).
The MBC concentrations were highest at the top of the columns and lowest in the bottom parts of both hydraulic regimes (Figures 3b and 4e). The concentrations in the aerobic zones of the columns fall toward the lower end of the range of previously reported MBC in soils with similar OC content (Blume et al., 2002). In the anaerobic zones, energetic limitation of microbial growth and activity in the absence of O 2 results in lower biomass C concentrations (Inubushi, Brookes, and Jenkinson, 1991). In the static water table columns, biomass was relatively constant above the water table (∼300 μg MBC g −1 soil) and dropped abruptly to <100 μg MBC g −1 soil below 20 cm (Figure 3e). In the fluctuating water table column, the biomass decreased gradually between 8 and 20 cm bss (Figure 3b). Therefore, in contrast with what might be expected, the zone of highest OC depletion did not coincide with a larger microbial population. In fact, at the depth where the lowest OC was measured in columns F2 and F3 (Figure 3a), roughly 13 cm bss, the biomass C concentration was only one-third of that at the same depth in the static water table columns C2 and C3.
The ATP concentrations in columns C2 and C3 gradually decreased with depth reaching levels near the limit of quantification below the water table (Figure 3f). The depth distributions of ATP under the fluctuating water table regime were markedly different (Figure 3c). In both columns F2 and F3, a peak in ATP was measured within the 8-to 20-cm fluctuation zone-that is, the depth interval with relatively low microbial biomass (Figure 4a) but the highest OC depletion (Figure 3a). Above 8 cm, columns F2 and F3 exhibited quite different ATP concentrations, with low values in F2 and relatively high values in F3. The reason for this difference is unknown. At depths >20 cm, the ATP concentrations in F2 and F3 dropped to levels comparable with those observed below 20 cm in the static water table columns. The ATP concentrations of the oxic portions of columns F3, C2, and C3 were at the lower end of the range reported for aerobic soils (Lin & Brookes, 1996;Qiu, Chen, Luo, Xu, & Brookes, 2016;Redmile-Gordon et al., 2011).
The ATP concentrations in Figures 4c and 4f are expressed per unit mass of soil. Alternatively, the ATP concentrations can be normalized to the MBC (ATP/MBC), as shown in Figure 5 (Panels a and c). The latter normalization yields biomass-specific activities with a peak value of 3.9 μmol ATP g −1 MBC in the fluctuation zone of the dynamic water table columns (Figure 5a). By contrast, ATP/MBC values in the static hydraulic treatment were highest near the surface (Figure 5c, ∼2.6 μmol ATP g −1 MBC) and decreased steadily with depth. Within the 8-to 20-cm interval of the stable water table columns, ATP/MBC values were three times lower than in the same depth interval of the fluctuating water table columns. Together with the similar microbial community compositions across both water table regimes (Section 3.5), these results confirm the findings of Rezanezhad et al. (2014), who concluded that the preferential OC depletion under fluctuating water table conditions "primarily reflect differences in the activity of various components of the microbial community rather than major changes in community composition." The ATP maximum in the 8-to 20-cm zone of the fluctuating water table columns observed at the end of the experiment had not yet developed after 99 d, at which point the ATP profiles in the stable and fluctuating water table columns Carbon-loss [g-C] (b) (a) F I G U R E 4 (a) Proportion of organic C (OC) loss in different depth intervals. The proportions are computed by subtracting the depth-dependent OC concentrations measured at the end of the incubation (t = 329 d) from the initial concentration in the homogenized artificial soil. Green bars are for the static water table treatment, and orange bars for the fluctuating water table treatment. Note that the 8-to 20-cm depth interval corresponds to the fluctuation zone in the fluctuating water table treatment. In the static water table columns, the water table is positioned at 20 cm below the soil surface. In both hydraulic treatments, the shallower and deeper intervals remained primarily aerobic and anaerobic, respectively. (b) The integrated OC loss, over the entire depth profile of fluctuating column F2, is compared with the outflow losses of C as dissolved organic C (DOC) and dissolved inorganic C (DIC) during drainage periods (dDOC, and dDIC, respectively), and the integrated CO 2 efflux (eCO 2 ) and CH 4 efflux (eCH 4 ) from the start to the end of the incubation experiment were rather similar (Figures 5b and 5d). This suggests that during the first 99 d of incubation, the microbial populations were still adapting to the prevalent redox conditions, consistent with the OTU profiles at Day 99 that already showed a differentiation in community composition between the top and bottom of the soil cores (Supplemental Figure S6). Overall, the microbial community structure and activity in our experimental soil columns responded on timescales that are in line with previous studies that report soil microbial community development timescales on the order of 3-12 mo Ding et al., 2013).
The enhanced OC turnover in the fluctuation zone of the dynamic water table columns was likely carried out by active cells that remained in a high-energy state in order to continuously adjust to changing O 2 availability and soil moisture, rather than relying on the reactivation of dormant cells (De Nobili, Contin, Mondini, & Brookes, 2001). Sardinha, Müller, Schmeisky, and Joergensen (2003) argue that environments inhabited by less "specialized" microbial communities result in more competition for resources between microbial species. Cells in a metabolically diverse community will therefore maintain a higher ATP to microbial biomass ratio as a survival strategy (De Nobili et al., 2001;Pfeiffer, Schuster, & Bonhoeffer, 2001). High levels of ATP production, however, are energy expensive and may translate into less growth (Fontaine, Mariotti, & Abbadie, 2003) while still requiring OC oxidation to meet maintenance requirements (Kleerebezem & Van Loosdrecht, 2010). The ATP/MBC profiles measured in our study similarly suggest that increased OC degradation under the dynamic redox conditions of the fluctuating water table soil columns was due to a more active, not a larger, microbial community.

SUMMARY AND CONCLUSIONS
Water table fluctuations create spatially and temporally variable conditions in the soil environment. To identify potential impacts of these variable conditions on soil C dynamics, as well as the structure, abundance, and activity of the soil microbial population, we performed an 11-mo-long experiment on columns filled with an artificial soil matrix and inoculated with a microbial community from a riparian soil of a cold temperate deciduous forest. In one set of columns the water table remained at a fixed depth; in the other, drainage and imbibition cycles were imposed. Within the first 99 d, differences in the microbial community composition developed between the top and bottom portions of the columns, but little differentiation was observed between the static vs. fluctuating water table regimes. By the end of the experiment (329 d), however, taxonomic differences in the mid-column depth intervals (8-20 cm) had appeared between the two hydraulic regimes.
Under the fluctuating water table regime, the 8-to 20-cm interval (the fluctuation zone) experienced large oscillations in redox potential, pore water concentrations of O 2 , redox active constituents (Fe 2+ , SO 4 2− ), and DOC. The efflux of CO 2 further showed large variations in response to the water table fluctuations. At the end of the experiment, the fluctuation zone exhibited a distinct depletion of OC, compared with the upper (oxic zone) and lower (anoxic zone) soil layers. The water table fluctuations also yielded a somewhat higher microbial diversity in the 8-to 20-cm interval, although under both hydraulic regimes the same taxa dominated at middepths.
The mid-depth OC depletion in the fluctuating water table columns did not coincide with higher MBC concentrations, but rather with peak values of the ATP/MBC ratios that were two to three times higher than in the same 8-to 20-cm depth interval of the static water table columns. These findings therefore imply that the enhanced OC degradation associated with the highly variable redox conditions in the fluctuation zone was driven by a metabolically more active soil microbial population, not a compositionally distinct or more abundant microbial population. Further work will be needed to comprehensively delineate the biogeochemical mechanisms and ecological interactions that are responsible for the higher soil OC turnover under fluctuating water table conditions, and to determine their overall impact on C cycling and fluxes in realworld soil systems.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

ACKNOWLEDGMENTS
The work of G.J. Pronk was funded by a Research Fellowship of the German Research Foundation (DFG,PR1550/1-1). Additional support was received from the Canada Excellence Research Chair (CERC) program in Ecohydrology, the Waterloo-Technion University Cooperation Programme, Natural Sciences and Engineering Research Council (NSERC) Strategic Partnership Grant (STPGP494652-16), the Global Water Futures (GWF) funded under the Canada First Excellence Research Fund (CFREF), and the Training Toward Environmentally Responsible Resource Extraction (TERRE)-Collaborative Research and Training Experience program and courses. The authors thank Marianne Vandergriendt, Linden Fairbairn, and Riley Mills for their assistance with laboratory analyses. We also thank the rare Charitable Research Reserve (Cambridge, ON, Canada) for access to the field site where some samples were collected.