Introduction

Drylands, which cover 45% of the world’s land area, store 32% of all terrestrial soil organic carbon (SOC) (Prăvălie 2016; Plaza et al. 2018), and play a dominant role in key aspects of global carbon (C) cycling. For example, drylands may dominate the trend and the interannual variability in atmospheric CO2 concentrations and the terrestrial C sink (Poulter et al. 2014; Ahlström et al. 2015). Wetting–drying cycles (whereby dry soils are wet by a precipitation event and subsequently dry) are common in drylands and may result in large and rapid soil CO2 emission pulses, a phenomenon known as the Birch effect (Birch 1958). The effects of wetting–drying cycles on soil respiration may be amplified by or mitigated by changes in soil C availability and other climate change factors (Barnard et al. 2020). However, how wetting–drying cycles affect dryland soil C cycling physically, chemically, and biologically is far from clear. This knowledge gap may severely constrain predictions of dryland contributions to atmospheric CO2 variability and the soil C sink under future climates.

In order to improve our capacity to predict soil respiration in a changing world, it is essential to simultaneously determine the physical, chemical, and biological mechanisms underlying the rewetting effect, because the controls may intensify or offset each other. Physical processes associated with rewetting, such as the slaking of soil aggregates, can expose protected soil organic matter to microbial consumption. This subsequently enhances the biological mechanisms of soil C loss (Lado-Monserrat et al. 2014). Similarly, rewetting cycles change the kinetics of sorption–desorption reactions (Patel et al. 2021), the chemical profile of soil organic matter, and its subsequent processing by decomposers. Concentrations of bioavailable SOC in the soil are strongly affected by precipitation events, which enhance inputs of soluble organic matter from the litter layer (Austin 2011; Liu et al. 2018). Fresh plant-derived C (leachate) can prime the microbial decomposition of older organic matter (Kuzyakov 2002) by enhancing the amount of energy-rich C available, leading to net soil C loss. By contrast, precipitation-driven organic matter inputs could enhance the soil C sink if they are immobilized within the microbial biomass (Liang et al. 2017).

Biological responses to rewetting, mediated by soil microbes, are ultimately responsible for most soil C losses and gains. Microbes are the engines of soil C loss, producing the enzymes which catalyze the decomposition of SOC, and mineralizing labile organic matter to CO2. At the same time, the microbial biomass is an important precursor to the formation of mineral-associated organic matter (Liang et al. 2017). Microbes also influence the physical structure of the soil and thereby C cycle responses to subsequent climatic events. For example, soil fungi can stimulate soil aggregate formation through hyphal enmeshment (Tisdall 1994), which may compensate for soil aggregates that are disrupted by wetting–drying cycles.

The above-mentioned interactions among physical, chemical, and biological processes may explain why the magnitude of the rewetting effect is dependent upon the moisture status of the soil prior to a wetting event, and changes in soil C availability that may accompany soil rewetting. Such complex interactions among C availability and the timing and magnitude of precipitation events were captured by a previous field climate manipulation experiment in a dryland ecosystem on the Colorado Plateau, where we monitored hourly soil respiration in in situ chambers for a full year in warmed and unwarmed plots that were amended with Atriplex confertifolia litter or biological soil crusts in a full factorial design (Waring et al. 2021). The field study found that natural precipitation events stimulated soil respiration across treatments, in accordance with many previous studies (Xiang et al. 2008; Lado-Monserrat et al. 2014; Waring and Powers 2016; Darrouzet-Nardi et al. 2018). However, the magnitude of the Birch effect was much greater under leaf litter than biocrust (Waring et al. 2021). The magnitude of rewetting CO2 pulses was also correlated with the magnitude of soil moisture change during a rainfall event, but this relationship weakened during the drier part of the year. Identifying the precise mechanisms underlying this pattern is challenging in situ, because of the difficulty in disentangling the effects of mean soil moisture content from its variability over a given window of time (Fierer and Schimel 2002).

Many previous studies have compared respiration rates in soils held at a stable moisture content to rates in soils that received additional pulses of moisture. It is not surprising that the soils that received greater moisture inputs overall showed greater respiration rates. The more important question is: In soils incubated at the same mean moisture content, does increased variability enhance or inhibit respiration? Because our previous in situ study (Waring et al. 2021) could not reveal the mechanisms underpinning these patterns, we conducted a complementary laboratory incubation experiment with soils harvested from the same field experiment. To assess how soil moisture content and temporal fluctuation affect soil C cycling, we independently manipulated the mean and variability of soil moisture, tracking soil respiration rate changes both during and after rewetting cycles. To quantify the effects of prior climate regimes on soil respiration patterns, we also collected soil samples from both warming and control treatments at the field experiment. In the field, soils in the warming treatment were on average 3 °C warmer, tended to be wetter (likely due to reduced plant cover; Winkler et al. 2019), and experienced larger fluctuations in moisture content (Darrouzet-Nardi et al. 2018). We then quantified how our laboratory treatments affected soil physical properties (aggregate size distribution), chemical properties (labile C availability), and biological properties (microbial biomass, enzyme activities, priming effect, and microbial community composition), and we attempted to link these to observed patterns of soil respiration.

Drylands are exquisitely sensitive to perturbations and are generally slow to recover post disturbance because of their limited resource availability (Poulter et al. 2014; Steven et al. 2021); however, little is known about whether wetting–drying cycles have legacy effects on soil respiration and if so, how long these legacy effects last. Without this information, we may inaccurately estimate and/or predict the effects of wetting–drying cycles on dryland soil C cycling. To address this knowledge gap, we divided the incubation into a 12-week rewetting phase (with six 14-d wet-dry cycles) followed by a 4-week recovery phase, where the soils were held at a constant moisture content. Legacy effects were determined from differences in respiration rates during the recovery period between samples exposed to wet-dry cycles versus those that had been held at constant moisture throughout the two phases.

Our experimental manipulations enabled us to answer three research questions: (1) how do wetting–drying cycles affect dryland soil respiration physically (via affecting soil aggregate formation), chemically (via affecting soil dissolved C), and biologically (via affecting soil fungal communities, soil enzyme activities, soil microbial biomass, and soil priming effects)? (2) are the effects of dryland wetting–drying cycles on soil respiration modified by soil C availability, mean soil moisture content, and/or prior climate conditions? and (3) do legacy effects of wetting–drying cycles on soil respiration exist in drylands after moisture cycles have stabilized? We hypothesized that (1) wetting–drying cycles should stimulate soil respiration in drylands by disrupting soil aggregates (physical mechanism), changing the pools of C to which microbes have access (chemical mechanism), and altering soil microbial community physiology and structure (biological mechanism); (2) wetting–drying cycles and higher mean soil moisture should trigger larger respiration pulses when additional labile C is available, because soil microbes are C-starved in drylands (Schimel 2018); and (3) effects of wetting–drying cycles on soil respiration should persist even after the wetting fluctuations cease, since drylands may have low recovery capacity to extreme climate change (Steven et al. 2021).

Methods

Site description

We sampled soils from a study site located in a semiarid ecosystem on the Colorado Plateau near Castle Valley, Utah, USA (38° 38′ 4″ N, 109° 24′ 38″ W; 1310 m above sea level). Mean annual temperature and precipitation are 13 °C and 269 mm, respectively. The soils are classified as Rizno fine sandy loam (Loamy, mixed, superactive, calcareous, mesic Lithic Ustic Torriorthents), which are shallow, well-drained soils derived from sandstone, siltstone, and limestone. The vegetation is a mix of grasses (Achnatherum hymenoides and Plueraphis jamesii) and shrubs (Atriplex confertifolia and Gutierrezia sarothrae). This site maintains a climate manipulation experiment that began in 2005 and that alters multiple aspects of climate (Wertin et al. 2015). A field experiment using the infrastructure of the climate experiment was conducted to ask questions about interactions between climate and C source, and started at this site in August 2017 (Waring et al. 2021). The field experiment included a fully factorial combination of two treatments: (1) warming (3 °C above ambient conditions via infrared radiant heat lamp vs. control) (Wertin et al. 2015); and (2) C input (soils topped with living biocrust vs. those topped with Atriplex confertifolia leaf litter); our study only used warmed and control soils from the latter C input treatment because the greater effect of Atriplex confertifolia leaf litter than living biocrust on soil respiration was found at the field site (Waring et al. 2021). To help quantify the incorporation of new C inputs into the soil through time, the 2017 field experiment took soils from the site and washed them with concentrated sodium hypochlorite solution to remove soil organic matter (Mikutta et al. 2005). Treated soils had a remaining organic C concentration of 2.9 ± 0.1 mg organic C g−1 soil, representing a 28% decrease from initial values. 12 kg of C-depleted soils (6 cm depth) were then used to fill automatic CO2 flux chambers installed in situ in each plot, and approximately 400 g Atriplex confertifolia leaf litter was applied to the top 3 cm soil depth. In November 2018, we collected soils from all Atriplex chambers. We then composited soil samples from the five chambers under the control climate treatment and from the five warming chambers. The former and latter composite soil samples were used to represent control and prior warming treatments in this lab incubation experiment, respectively (Fig. S1).

Experimental design

All treatments were applied to 50 g (dry weight) of soil incubated at room temperature (20 °C) in 250 ml glass jars. The experiment was a fully factorial combination of the following treatments: (1) prior warming: previously warmed soils versus unwarmed soils; (2) soil moisture: 0.04 g H2O g−1 dry soil (20% of water holding capacity [WHC]) versus 0.08 g g−1 (40% of WHC); (3) soil moisture variability: stable versus variable (wetting–drying cycle, see below); and (4) C addition: control versus C addition (2.0 mg glucose-C g−1 dry soil for each wetting–drying cycle).

Prior to the start of the experiment, soils were sieved to 4 mm, homogenized, and conditioned for one week at the appropriate soil water content (0.04 g g−1 or 0.08 g g−1). The incubation that followed had two periods: (1) a rewetting period including six, two-week wetting–drying cycles; and (2) a recovery period that lasted four weeks, during which soil moisture was maintained at a stable value in all microcosms. With 16 unique treatment combinations and 5 replicates per treatment, there were 80 microcosms in total.

During the rewetting period, ‘stable’ or ‘variable’ moisture treatments were imposed. For the variable soil moisture treatment, soils were subject to two-week wetting–drying cycles. During each cycle, soils in the variable treatment were rewetted, then allowed to dry sufficiently to achieve a time-weighted mean moisture equal to one of the mean soil moistures maintained in the stable moisture treatment (0.04 g g−1 or 0.08 g g−1). To achieve an average of 0.04 g g−1 moisture content in microcosms assigned to the low, fluctuating moisture treatment regime, microcosms were rewetted to 0.16 g g−1 gravimetric soil moisture (80% of WHC), then allowed to dry to 0.003 g g−1 soil moisture (1.5% of WHC), which we accomplished by removing the jar lid for 168 h. The jar lid then was closed loosely until the next wetting–drying cycle (Fig. S2a). Similarly, microcosms assigned to the high, fluctuating moisture regime were rewetted to 0.16 g g−1 soil moisture (80% WHC), dried to 0.062 g g−1 soil moisture (31% of WHC) by removing the jar lid for 92.9 h, and then loosely capped (Fig. S2b). To aid in the interpretation of our results, we calculated time-weighted mean soil water potentials under stable and variable soil moisture treatments following Saxton et al. (1986). We found that, while time-weighted mean soil water contents were the same, time-weighted mean water potentials were substantially lower under wetting–drying cycles than stable soil moisture treatment (− 15.0 MPa vs. − 3.1 MPa in the 0.04 g g−1 treatment, and − 0.4 MPa vs. − 0.2 MPa in the 0.08 g g−1 treatment). For the stable soil moisture treatments, jar lids were kept closed loosely to avoid soil evaporation during the rewetting period, maintaining soil moisture at a constant level. During the recovery period, all microcosms were maintained at a stable (either 0.04 g g−1 or 0.08 g g−1) moisture content for the duration.

To manipulate C inputs, at the beginning of each wetting–drying cycle, 0.625 ml glucose solution (which is equal to 2.0 mg glucose-C g−1 dry soil) was added instead of DI water during the rewetting period; this volume was accounted for in our calculations of soil moisture so that we did not over-wet the soils when adding glucose. Glucose was added to mimic fresh leachate (e.g. dissolved C) moving from the litter layer into the soil during precipitation events. Soils in the ‘stable’ moisture treatment with added C received the same volume of glucose solution, which replaced the very small amounts of water lost to evaporation in the loosely capped jars. We did not add glucose solution to any microcosms during the recovery period. At the beginning of the last wetting–drying cycle, a 13C glucose solution (2.0 mg glucose-C g−1 dry soil at 5 atom % 13C) was added instead of the 12C glucose solution to the C input treatment to quantify priming effects (see below).

Measurements of soil respiration and soil parameters

During the course of the incubation experiment, headspace samples (10 ml) were collected with a syringe once a week after sealing jars tightly for 24 h. Specifically, the first sample was taken at the beginning of each two-week cycle (24-h) when water addition has the greatest effect on soil respiration (Canarini et al. 2017). The second sample was taken at the beginning of the second week, when soil moisture under variable soil moisture treatment had decreased to the lowest value. Therefore, the timing of the headspace samples integrated treatment-related differences in soil respiration across the entire wet-dry cycle. Headspace CO2 concentrations were analyzed using gas chromatography (GC-2016 Greenhouse Gas Analyzer, Shimadzu, Kyoto, Japan). The R package “MESS” was used to calculate cumulative soil respiration (μg CO2-C g−1 soil) in each microcosm over the entire experiment. Headspace 13CO2 concentrations and isotopic abundance were measured during the last wetting–drying cycle of the rewetting period using a G2131-i Isotope and Gas Concentration Analyzer (Picarro Inc., Santa Clara, CA, USA).

We destructively harvested subsets of the microcosms in each treatment at the end of rewetting (3/5 replicates per treatment) and recovery periods (the remaining 2 replicates per treatment). Soil samples collected during harvests were immediately used to measure soil microbial biomass C. Soil subsamples were air-dried at room temperature for measuring soil organic C and total nitrogen (N) concentrations. Additional subsamples were stored at either 4 °C for measuring soil aggregate stability or at – 80 °C for measuring soil enzyme activities and soil fungal community composition. Soil microbial biomass C was determined via a modified fumigation-extraction technique (Brookes et al. 1985), with extracts analyzed for dissolved organic C concentration on a TOC-L (Shimadzu, Kyoto, Japan). Total soil organic C and total N were quantified via an elemental analyzer (ECS 4010 Elemental Analyzer, Costech Analytical Technologies, Valencia, CA, USA), following inorganic C removal with sulfurous acid. Soil aggregate size distributions into four different classes (> 2, 2–0.25, 0.25–0.053, and < 0.053 mm) were quantified via wet sieving following (Denef and Six, 2005). The aggregate mean weighted diameter (MWD, mm) which is an index of aggregate stability was calculated as follows:

$$ {\text{MWD}}\,{ = }\,\sum\nolimits_{{\text{i = 1}}}^{{4}} {{\text{X}}_{{\text{i}}} \,{\text{W}}_{{\text{i}}} } $$
(1)

where \({\mathrm{X}}_{\mathrm{i}}\) is the mean diameter of the ith size fraction (mm), and \({\mathrm{W}}_{\mathrm{i}}\) is the proportion of the ith size fraction in the whole soil (%). Activities of extracellular enzymes (acid phosphatase, β-glucosidase, cellobiohydrolase, leucine aminopeptidase, and β-N-acetylglucosaminidase) were assayed colorimetrically for each soil sample using p-nitrophenol conjugated substrates following German et al. (2011), with absorbances measured on a microplate spectrophotometer (Spectramax M2 microplate spectrophotometer, Molecular Devices, San Jose, CA, USA). Since significant and positive correlations were found among the activities of all five soil enzymes (Fig. S3), we used β-glucosidase activity, which is sensitive to the presence of glucose, to present soil enzyme activity in the statistical analyses.

A DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) was used to extract DNA for all soil subsamples (0.25 g) from both the first and second harvests. Since the concentration of DNA extract from each soil sample was low, we combined all DNA extracts within a treatment and a harvest, respectively. As a result, we had a total of 32 samples (16 samples each for the first and second harvests). We concentrated the DNA extracts using a kit (ReliaPrep™ DNA Clean-Up and Concentration System). DNA extracts were then pooled in equimolar concentrations and sequenced on the Illumina MiSeq platform. We focused our microbial community analyses on soil fungi, which are thought to have a greater capacity of resistance to wetting–drying cycles (Barnard et al. 2013), and therefore play a dominant role in the soil biogeochemistry of drylands (Collins et al. 2008). Primers ITS1f-ITS2 were used to amplify fungal marker genes following Earth Microbiome Project protocols, and we used the QIIME 2 (version 2020.11.0) to analyze the sequencing data (Bolyen et al. 2019). The DADA2 algorithm was used to denoise sequences and to produce putatively error-free amplicon sequence variants (ASVs) that lacked chimeras. A 99% identity threshold was used for the OTU clustering of the ASVs. A Naïve Bayesian classifier was trained on the UNITE (version 8.2) training dataset to assign fungal taxonomy (Põlme et al. 2020). In order to remove non-fungal OTUs from the dataset, a taxonomy-based filtering procedure was applied. Before determining α-diversity for each treatment by using package “vegan” in R, fungal sequences were rarified to 12,000 per sample. Finally, package “ape” in R was used to create a random phylogenetic tree based on representative OTU sequences, and then we used package “picante” to calculate phylogenetic diversity in each treatment group and net relatedness index (NRI) under each treatment combination. NRI is a measure of the average phylogenetic distance among taxa in a community, with positive values indicating clustering and negative values indicating phylogenetic over-dispersion.

Determination of priming effects

We used isotope data from the final experimental harvest to quantify the priming effect (μg CO2-C g−1 soil) using Eq. (2) (Chen et al. 2019):

$$ {\text{Priming}}\,{\text{effect}}\,{ = }\,{\text{C}}_{{{\text{treat}}}} \; \times \;\left( {1 - {{\left( {{\text{at\% }}_{{{\text{treat}}}} - {\text{at\% }}_{{{\text{control}}}} } \right)} \mathord{\left/ {\vphantom {{\left( {{\text{at\% }}_{{{\text{treat}}}} - {\text{at\% }}_{{{\text{control}}}} } \right)} {\left( {{\text{at\% }}_{{{\text{glucose}}}} - {\text{at\% }}_{{{\text{SOM}}}} } \right) - {\text{C}}_{{{\text{control}}}} }}} \right. \kern-0pt} {\left( {{\text{at\% }}_{{{\text{glucose}}}} - {\text{at\% }}_{{{\text{SOM}}}} } \right) - {\text{C}}_{{{\text{control}}}} }}} \right) $$
(2)

where \({\mathrm{C}}_{\mathrm{treat}}\) and \({\mathrm{C}}_{\mathrm{control}}\) are the total C respired (μg CO2-C g−1 soil) from soils in the C input and non-C input treatments, and \({\mathrm{at\%}}_{\mathrm{treat}}\) and \({\mathrm{at\%}}_{\mathrm{control}}\) are the 13C isotope signatures of CO2 respired from these same treatments. Meanwhile, \({\mathrm{at\%}}_{\mathrm{glucose}}\) and \({\mathrm{at\%}}_{\mathrm{SOM}}\) are the 13C isotope abundances of added glucose and soil organic matter, respectively. We assumed that \({\mathrm{at\%}}_{\mathrm{control}}\) and \({\mathrm{at\%}}_{\mathrm{SOM}}\) had the same value, which was equal to the 13C isotope abundance of CO2 from soil that was not amended with C.\({\mathrm{C}}_{\mathrm{treat}}\), \({\mathrm{C}}_{\mathrm{control}}\), \({\mathrm{at\%}}_{\mathrm{treat}}\), and \({\mathrm{at\%}}_{\mathrm{control}}\) were measured 1 and 13 d after adding 13C-glucose, while measurement of \({\mathrm{at\%}}_{\mathrm{SOM}}\) was only made 13 d after adding 13C-glucose. Since the priming effect did not change significantly when we used data from 1 and 13 days post-13C-glucose addition, we reported the average priming effect from these two timepoints.

Data calculation and statistical analyses

A repeated measures linear mixed effects model was conducted to quantify the individual and interactive effects of the four experimental treatments and two incubation periods on soil respiration. We conducted ANOVA analyses to examine the effects of these same treatments on variables measured at the experimental harvest: MWD, soil enzyme activity, soil microbial biomass C, soil organic C and total N concentrations, the magnitude of the soil priming effect, phylogenetic diversity (PD), and net relatedness index (NRI). For the linear mixed models, microcosm identity was modeled as a random effect, and R package “lmertest” was used to estimate P values using Satterthwaite’s approximation. Pearson’s correlations were used to determine relationships among measured soil variables. All statistical analyses were conducted in R. Soil fungal community data were visualized with non-metric multidimensional scaling (NMDS) plots, and compositional differences among treatments were assessed with a PERMANOVA test.

Results

Soil respiration

The four experimental treatments (prior warming, C addition, soil moisture content, soil moisture variation) had strong individual effects on soil respiration (Table 1 and Figs. 1 and S4). The classical ‘Birch effect’ pattern (a pulse of CO2 immediately following rewetting) was only observed in glucose-amended microcosms in the higher moisture treatment, which had been previously exposed to warming (Fig. S4), and this pattern was present only in some rewetting cycles. When compared to stable soil moisture, wetting–drying cycles decreased soil respiration by 14.5% overall (Fig. S5), and this effect was not modified by any other factors (Table 1).

Table 1 Results of ANOVA analyses on the effects of treatment factors on soil variables
Full size table
Fig. 1
figure 1

Cumulative soil respiration under control (a) and carbon input treatments (b) over the course of the incubation experiment. Note the difference in y-axis scale between panels (a) and (b). The grey solid vertical lines between weeks 12 and 13 represents the start of recovery period. Values are means, and error bars represent ± 1 standard deviation of the mean. Colors, polygon shapes, and line styles depict the varied treatments and conditions, and the key is shown in panel (a)

Full size image

However, interactions among warming, C addition, and soil moisture content were common. For example, C input enhanced soil respiration responses to soil moisture; higher soil moisture increased soil respiration by 1.0% in soils without C input versus 79.4% in soils with C input (Fig. 2a). Prior warming modified responses to new C inputs, as previously warmed soils had 14.4% higher respiration than control soils, but 35.3% higher respiration in soils amended with glucose (Fig. 2b). Moreover, higher soil moisture increased soil respiration by 38.3% in previously unwarmed soils, versus 66.0% in previously warmed soils (Fig. 2c). The effects of soil moisture variability did not vary between the rewetting and recovery periods (Table 1), indicating a potential legacy effect of rewetting events. Specifically, when compared to the stable soil moisture regime, rewetting decreased average soil respiration by 15.3 and 7.3% during rewetting and recovery periods, respectively (Fig. S5), despite cessation of moisture fluctuations during the recovery.

Fig. 2
figure 2

Interactive effects of global change factors on mean soil respiration rates across all measurement time points during both rewetting and recovery periods. Control and carbon input represent soils without and with glucose addition during the experiment, respectively. Control and warmed represent soils that were under ambient air temperature and had been previously warmed in situ (for 15 months), respectively. Low and high moisture treatments represent soils that received 0.04 g g−1 and 0.08 g g−1 gravimetric soil moisture, respectively. Values are means, and error bars represent ± 1 standard error of the mean. See Table 1 for statistical results

Full size image

In addition to examining overall soil respiration rates, we examined the source of C that was respired (background organic matter vs. new C inputs), in other words, the priming effect (Table 1). Wetting–drying cycles diminished priming effects (relative to stable soil moisture regimes), but only under low soil moisture content (Fig. 3). Prior warming significantly increased the soil priming effect by 66.8% when soil moisture content was high, but the difference in the soil priming effect between prior warming and control became insignificant under low soil moisture content.

Fig. 3
figure 3

Interactive effects of wetting–drying cycles and soil moisture on the soil priming effect. Stable and variable represent soils with stable soil moisture during the course of the experiment and soils that experienced rewetting cycle each two weeks, respectively. Control and warmed represent soils that were under ambient air temperature and had been previously warmed in situ (for 15 months), respectively. Low and high moisture treatments represent soils that received 0.04 g g−1 and 0.08 g g−1 gravimetric soil moisture, respectively. Values are means, and error bars represent ± 1 standard error of the mean. See Table S1 for statistical results

Full size image

Soil physical and chemical properties in response to treatments

Soil aggregate size (measured as MWD) responded to three-way interactions among C input, soil moisture content, and prior warming; and among C input, soil moisture content, and soil moisture variability (Table 1). In general, MWD was greatest in soils with high soil moisture, added C input, and a stable, non-fluctuating soil moisture regime (Fig. 4d). In addition, while the C input treatment had by far the largest effect on dissolved organic C concentrations, there were also differences between the stable and variable soil moisture treatments (Table 1). However, these emerged only at low soil moisture contents and under glucose amendment, where dissolved C accumulated in the variable moisture treatment (Fig. S6). In summary, when labile C was not limiting, the moisture regime that favors aggregate development (high, stable) is the opposite of the regime under which dissolved C accumulates (low, variable). We also found total organic C concentration responded to the two-way interaction among C input and soil moisture content (with a similar pattern for total N), although C input had by far the largest absolute effect (Table 1).

Fig. 4
figure 4

Interactive effects of global change factors on mean weighted diameter (MWD, an index of aggregate stability) during both rewetting and recovery periods. Control and carbon input represent soils without and with glucose addition during the experiment, respectively. Control and warmed represent soils that were under ambient air temperature and had been previously warmed in situ (for 15 months), respectively. Low and high moisture treatments represent soils that received 0.04 g g−1 and 0.08 g g−1 gravimetric soil moisture, respectively. Stable and variable represent soils with stable soil moisture during the course of the experiment and soils that experienced rewetting cycle each two weeks, respectively. Values are means, and error bars represent ± 1 standard error of the mean. See Table S2 for statistical results

Full size image

Soil biological properties in response to treatment

Soil β-glucosidase activity was not significantly affected by rewetting, C input, and previously warming treatments individually, but was stimulated by high soil moisture content (Table 1). In addition, Soil β-glucosidase activity was significantly higher during the recovery than the rewetting period. Meanwhile, only C input showed a significant and positive individual effect on soil microbial biomass C (Table 1).

The composition of soil fungal communities was significantly affected by C input and prior warming (Figs. 5 and S7). In particular, the percentage of Ascomycota was increased by C input but was decreased by prior warming (Fig. 5). The relative abundance of fungal phyla was not significantly affected by soil moisture content, soil moisture variability, or experimental period (Fig. S7). Phylogenetic diversity was significantly affected by warming and period (Table 1). More specifically, warming increased phylogenetic diversity, and higher phylogenetic diversity was found in rewetting than recovery period. The net relatedness index was not significantly altered by any treatments.

Fig. 5
figure 5

NMDS of soil fungal communities (a) and the relative abundance of fungal phyla (b) with the carbon input treatment. Control and carbon input represent soils without and with glucose addition during the experiment, respectively

Full size image

Correlations between soil respiration and soil parameters

According to Pearson’s correlation analysis, mean soil respiration rate was positively correlated with soil priming effect (p < 0.001) during the rewetting period (Fig. S8). No significant relationships among mean soil respiration rate, MWD, soil microbial biomass, and soil β-glucosidase activity were found.

Discussion

Our dryland soil incubation experiment provided mixed support for our three hypotheses. After controlling for differences in mean soil moisture, wetting–drying cycles decreased the time-integrated soil respiration rate, which is in direct contrast with our first hypothesis. Second, we found that C availability amplified the positive effects of soil moisture on respiration, but did not affect rewetting dynamics, providing only partial support for our second hypothesis. Third, as we expected under our third hypothesis, a legacy effect of wetting–drying cycles on soil respiration was found. It should be noted that this was a lab experiment with heavily modified soils; we used this design to assess mechanisms underlying rewetting responses, but we may not have captured phenomena occurring in intact soils in situ. Below, we discuss the biogeochemical mechanisms underlying these patterns.

Wetting–drying cycles and their influence on soil respiration

Previous studies have found that wetting–drying cycles are associated with pulses of soil respiration, presumably driven by enhanced microbial consumption of osmolytes, osmoprotectants such as trehalose, or organic matter released from organo-mineral desorption or disrupted aggregates (Fierer and Schimel 2002; Schimel 2018; Barnard et al. 2020; Slessarev et al. 2020; Patel et al. 2021). However, we found that wetting–drying cycles instead decreased soil respiration, which is contrary to our first hypothesis. Wetting–drying cycles did not affect enzyme activity, soil microbial biomass, or microbial community composition, in line with a few other studies (Daou et al. 2016; Canarini et al. 2017). Birch effects were only observed in microcosms with variable, high moisture, and C amendment. The temporal dynamics of soil water content in the high versus low moisture treatments may explain these surprising results. For microcosms in the high (0.08 g g−1) moisture treatment, the change in soil water potential during the course of a drying-rewetting cycle may have been insufficient to trigger a microbial drought stress response, despite substantial variation in soil water content. Meanwhile, for microcosms in the low (0.04 g g−1) moisture treatment, we suggest that the respiratory patterns we observed may have reflected physiological adjustments of soil microbes to transient changes in soil moisture regime in a highly resource-limited environment. Immediately following rewetting, soil moisture was extremely high (80% of WHC), potentially reducing oxygen for soil microbes. To maintain the same mean moisture content as observed in the ‘stable’ moisture treatment, the soils were then dried to 1.5% WHC, presumably inducing severe drought stress. The relatively brief window of time during which moisture conditions were optimal could have contributed to the slow soil C turnover rate, as microbes in the low, variable moisture treatment experienced an average soil water potential incompatible with metabolic activity (Manzoni et al. 2012). Thus, soil respiration rate clearly does not depend solely upon the mean soil moisture over some time interval, but rather the temporal dynamics of moisture—the amplitude of the change, the time period over which a shift between dry and wet conditions occurs, and the relationship between soil water content and soil water potential. Soil C losses would be incorrectly predicted if soil moisture variability is ignored, including the amplitude, peak, and period of water cycles. Moreover, the only microcosms which displayed ‘pulses’ of CO2 emission were those which received glucose amendments, indicating the great role of carbon availability in mediating the Birch effect.

Effects of multiple global changes on soil respiration and parameters

Although the effects of wetting–drying cycles on respiration were not modified by the overall availability of labile C and water, other soil parameters did respond to treatment interactions, suggesting multifactor global changes may ultimately affect soil C cycling in the long term. We saw the largest increase in aggregate formation in soils with the addition of glucose, high water content, and stable soil moisture regimes, which are conditions likely to favor microbial growth. We found that the C input treatment increased soil microbial biomass C and relative abundance of Ascomycota, which have a higher capacity to decompose labile C (e.g., the glucose that we added as C input in the present study) (Treseder and Lennon 2015). Soil fungi play an important role in aggregate formation through production of fungal hyphae, which enmesh soil particles, and secretion of extracellular polysaccharides, which bind them together (Tisdall 1994). Moreover, under conditions of sufficient C and moisture availability, aggregate stability was greater under stable moisture regimes. This might be attributed to the role of wetting–drying cycles in disrupting soil aggregates. Meanwhile, soil dissolved organic C accumulated when soil moisture was low and variable, suggesting microbes were not able to take advantage of labile C amendments due to the aforementioned osmotic stress.

An interaction between moisture content and moisture variability also determined the extent to which microbes utilized ‘background’ soil organic matter versus fresh C inputs. Fresh C inputs did not stimulate respiration to the same extent under wetting–drying cycles versus stable soil moisture regimes, at least when soil moisture was already low. This might be partially explained by the lesser enzyme activity in drier soils, as soil enzymes catalyze the decomposition of old particulate SOC (Kuzyakov et al. 2000; Kuzyakov 2010), and lower soil enzyme activities are often associated with lower soil priming effects (Chen et al. 2014). Perhaps the greater degree of osmotic stress associated with wetting–drying cycles suppressed co-metabolism of more slowly cycling, recalcitrant organic matter. Taken together, the C cycle responses we observed suggest that the impacts of soil moisture regime (mean and variability) not only affect short-term CO2 fluxes, but also longer-term dryland soil C storage.

Finally, we found that labile C inputs had a greater impact on soil respiration than soil moisture content, rewetting, and prior warming, and the C amendment often amplified the effects of other treatments. This suggests substrate availability was the most influential driver of soil respiration in these severely C-depleted soils, which are typical of many drylands (Schimel 2018; Choi et al. 2022). Lower respiration under C limitation was aggravated further by low soil moisture, which inhibits organic matter diffusion (Schimel 2018), consistent with the patterns observed in our earlier field and modelling experiments (Waring et al. 2021).

The legacy effects of wetting–drying cycles on soil C cycling

We found that soils that experienced wetting–drying cycles exhibited lower respiration rates even after moisture fluctuations ceased, suggesting that these cycles may have legacy effects on soil C cycling. However, we quantified soil biogeochemical responses for only 28 days following the cessation of rewetting cycles, so the potential for longer-term recovery remains unclear. Although we did not find significant effects of wetting–drying cycles on the size of the microbial biomass or soil fungal community composition, consistent with a previous study (Barnard et al. 2013), these legacy effects may reflect physiological adaptations of the microbial biomass to wetting–drying cycles, such as changes in C use efficiency, osmolyte production, or dormancy. Phylogenetic diversity decreased once rewetting stopped in the present study, indicating that the disturbance induced by extreme events (rewetting here) could help maintain microbial diversity. Interestingly, soils which had experienced warming in the field in the year prior to this lab experiment were often more responsive to elevated soil moisture and C availability, and were the only soils to exhibit Birch-like pulses of respiration on rewetting. The composition of fungal communities in these soils was also quite distinct from those with no previous history of warming. The positive effect of prior warming on phylogenetic diversity indicates that the changes in soil carbon quality (e.g. high to low quality) caused by warming (Melillo et al. 2017) might further increase phylogenetic diversity to decompose diverse soil carbon types. This suggests longer-term functional legacies of climate regime, potentially mediated by compositional shifts in microbial communities.

Conclusion

Although wetting–drying cycles are common in drylands, the mechanisms underlying their effects on soil C cycling are poorly understood. We found that wetting–drying cycles decreased soil aggregate stability, but the intensified osmotic stress associated with these cycles inhibited soil respiration. Our findings suggest that the underlying drivers of the Birch effect are complex, involving interactive physical, chemical, and biological mechanisms, and that past climate regimes may influence the responsiveness of soil respiration to moisture pulses. The results also underscore the importance of considering multiple types of control over soil C storage and flux (Waring et al. 2020), particularly in drylands where resources are so variable and where interacting C cycling controls are relatively poorly understood.