Warming increases soil respiration in a carbon-rich soil without changing microbial respiratory potential

Increases in global temperatures due to climate change threaten to tip the balance between carbon (C) fluxes, liberating large amounts of C from soils. Evidence of warming-induced increases in CO2 efflux from soils has led to suggestions that this response of soil respiration (RS) will trigger a positive land C–climate feedback cycle, ultimately warming the earth further. Currently, there is little consensus about the mechanisms driving the 10 warming-induced RS response, and there are relatively few studies from ecosystems with large soil C stores. Here, we investigate the impacts of experimental warming on RS in the C-rich soils of a Tasmanian grassy sedgeland, and whether alterations of plant community composition or differences in microbial respiratory potential could contribute to any effects. In situ, warming increased RS on average by 28% and this effect was consistent over time and across plant community composition treatments. In contrast, warming had no impact on microbial 15 respiration in incubation experiments. Plant community composition manipulations did not influence RS or the RS response to warming. Processes driving the RS response in this experiment were, therefore, not due plant community effects and are more likely due to increases in belowground autotrophic respiration and the supply of labile substrate through rhizodeposition and root exudates. CO2 efflux from this high-C soil increased by more than a quarter in response to warming, suggesting inputs need to increase by at least this amount if soil C stocks 20 are to be maintained. These results indicate the need for comprehensive investigations of both C inputs and losses from C-rich soils if efforts to model net ecosystem C exchange of these crucial, C-dense systems are to be successful.


140
Air temperature at 5 cm height and soil temperature at 5 cm depth in each plot was logged continuously with iButton dataloggers. Over the entire five year period, the warming treatment increased air temperature 5 cm above the soil surface by 1.56C (P<0.004) and soil temperature at 5 cm depth by 1.29C (P<0.001).

145
A 50 mm length of 100 mm diameter PVC pipe was inserted into the soil to a depth of 2 cm, extending 3 cm above ground height, within the centre 0.25 m2 of each plot for soil respiration measurements. Soil respiration was measured with a CO2/H2O infrared gas analyser (IRGA) (Licor, model LI-6400) with attachment of a Licor 6400-09 soil chamber, which attached to PVC pipes. Soil respiration was measured in situ monthly from August 2017 to June 2018. On each occasion, three measurements of in situ soil respiration, defined as the CO2 efflux 150 rate, were made in each plot. The average value of these three measurements was used in subsequent analyses.
Soil temperature and moisture in each plot were measured at the exact same time as the soil respiration measurements on each occasion. Soil temperature was measured with a soil thermocouple probe (LiCor 6000-09TC) attached to the LI-6400. Volumetric soil water content (SWC) was estimated at 5 locations in each plot using a hand held TDR probe at 0-5cm depth. The 5 separate measurements of SWC where then averaged to 155 obtain one SWC value per plot on each measuring occasion. https://doi.org/10.5194/bg-2020-144 Preprint. Discussion started: 25 May 2020 c Author(s) 2020. CC BY 4.0 License.
Six randomly placed soil samples, amounting to a total of approximately 25-30 g fresh weight, were collected from each plot using a 1.5 cm diameter hand corer to a depth of 5 cm below ground level, twice throughout the year. Samples were collected on the 02/03/18, representing the end of summer, or growing season soil, and on the 25/06/18, representing winter soils.

Laboratory incubations
Soil cores collected in situ were immediately placed on ice for return to the laboratory, where they were refrigerated (4°C) overnight. The following day, the samples were composited at the plot level and sieved through a 4 mm sieve for one minute to remove leaves and large roots. A 10 g fresh-weight sub-sample was removed and 165 oven dried from each composite sample for the determination of total soil C. Each subsample was ground to a powder in a Retsch Mixer Mill (MM200, Retsch GmbH, Haan) and then C content was analysed by combustion in a Perkin Elmer 2400 Series II Elemental Analyser (Perkin Elmer Australia, Melbourne). The remaining soil was used immediately for laboratory incubations to determine microbial respiration, as detailed below.
Microbial respiration as a function of temperature was determined by incubation using soils sampled in the Silver

170
Plains warming experiment at the end of summer and in mid-winter 2018. For each plot, three replicate samples weighing four to eight grams from the composite sample were placed in 100 mL specimen jars, each of which was incubated at a different temperature. Each sample was wetted to bring them to 90% of field capacity for winter soils and 60% of field capacity for summer soils to represent prevailing soil moisture conditions in each respective season. Once water was added to all soil samples, specimen jars were placed in 500 ml preserving jars 175 with tightly fitting lids containing a septum to allow gas headspace samples to be collected by syringe. Jars were stored in dark incubation cabinets at temperatures at one of 10, 17 or 25C, with one sample from each plot at each temperature. Headspace gas of jars were sampled (20 ml) using a syringe on days 1,2,4,5,7,9,12,15,19,23,29,35,49,56,63. After extracting samples from each jar, headspace samples were analysed for CO2 concentration, representing soil respiration, and microbial respiratory potential was thus defined as the rate of CO2 180 release. To analyse headspace gas, samples were injected directly into an infrared gas analyser (LI-6262, Li-Cor, Lincoln, NE). After measurements were taken and analysed, jars were ventilated for 20 minutes and headspace gas equilibrated with atmospheric air. Following this, lids were replaced and headspace gas was sampled and analysed again to obtain starting CO2 concentration for each jar. C mineralisation over the sample period was calculated from the increase in headspace CO2 concentration.

185
Total C mineralisation over the entire incubation period was simply the sum of the amount of C mineralised over each sample period. Daily C mineralization results (dC/dt) were analysed using non-linear curve fitting routines in R (version 3.4.3, R Core Team, 2017), with a single pool plus constant model (Pendall et al., 2011) to estimate the size of the labile C pool (Ca), the intrinsic decay constant of the labile pool (k), and the intrinsic decay constant of the stable C pool (Y0): https://doi.org/10.5194/bg-2020-144 Preprint. Discussion started: 25 May 2020 c Author(s) 2020. CC BY 4.0 License.

Data analysis
Field soil respiration rates were analysed using a 2-factor repeated measures ANOVA with warming and removal as the fixed factors. Since soil temperature (TS) and SWC are known controllers of RS and varied substantially over the year, we also analysed field RS with a 2-factor ANCOVA with TS and SWC and the interaction between TS and SWC as covariates. Treatment means were calculated as least-squares means using the lsmeans package to account for the influences of covariates (Russel V. Lenth, 2016). Treatment effects on SWC and TS were analysed using 2-factor repeated measures ANOVA exactly as for RS.

200
Because there was a significant influence of warming on RS, we created a separate model of the influence of SWC and TS on in situ RS for warmed and unwarmed plots. Since the respiration temperature relationship is best described by an Arrhenius-type function (Fang & Moncrieff 2001), we used multiple regression techniques to fit an exponential relationship to RS and SWC, TS and the interaction between TS and SWC. Such a non-linear relationship fitted the observed data far better than a linear model, as compared by the Akaike information criterion 205 corrected for finite sample size.
Total cumulative CO2 emitted in laboratory incubations, Ca, k, and Y0 for each season were compared using threefactor analysis of variance ANOVA for both summer and winter soils with incubation temperature, warming and species removal as fixed factors, including all interactions. Seasonal differences were also analysed using fourfactor ANOVA, with season also included as a fixed factor along with warming effect, removal and incubation 210 temperature.
All statistical analyses were carried out in R (version 3.4.3). Data were checked for heteroscedasticity and normality and the required transformations were made using the Box Cox power and logarithmic transformations.
Significant treatment effects were further analysed using Tukey's HSD post hoc comparisons.

215
3 Results indicating that plant species removal did not alter the influence of the warming treatment.

Relationships between environmental factors and CO2 efflux
Both soil temperature (F1,33= 33.62, P<0.001) and SWC (F1,33= 5.95, P=0.02) were strong controllers of soil CO2 efflux over the year at Silver Plains (Fig. 2). However, treatment effects on these abiotic factors alone were 260 insufficient to explain the higher C efflux in warmed plots, as ANCOVA indicated that the warming treatment still induced significant increases in CO2 efflux when variation in soil T and SWC were accounted for (F1,33= 44.83, P<0.001). Thus, the warming treatment increased soil CO2 efflux independently of its effects on soil temperature and SWC (Fig. 2). Across the whole year LS mean CO2 efflux rates for ambient soils was 6.07 (C.I=5.69,6.45) µmol CO2 m-2 s-1 but 8.48 µmol CO2 m-2 s-1 (CI=8.09,8.86) for warmed soils, amounting to a 265 warming-induced increase of 28% at a common soil temperature and SWC. As CO2 efflux measurements spanned a large variation in both soil T and SWC, it was possible to discern a trend whereby the stimulation of C efflux by warming became more pronounced as soil temperature increased (Fig. 2)

Models of CO2 efflux
As ANCOVA indicated that soil CO2 efflux at Silver Plains was significantly influenced by soil temperature, SWC and a strong warming effect, the relationship between these covariates and CO2 efflux could be estimated separately for ambient and warmed treatments. First a general regression model of CO2 efflux was fit and selected 280 using model selection based on AICc. The most parsimonious and accurate model was one which included soil temperature (TS), SWC, and a SWC x TS interaction term (Int.term).
This model was then fit independently to ambient and warmed plots using the relative coefficient values, with 89% of the variance in CO2 efflux explained in warmed plots Eq.
(2) and 82% in ambient plots Eq. (3). in which it does so is also dependent upon the SWC (Fig. 3A and B). Thus, the impact of experimental warming on soil CO2 efflux was greatest in warm (Ts>15°C) relatively dry conditions (SWC<30%; Fig. 4).

Total C mineralisation
To determine whether experimental treatments altered potential microbial respiration, soil samples were collected in summer and winter for laboratory incubations. These incubations allowed the temperature sensitivity of soil respiration, the size of the labile C pool (Ca) and its decay constant (k) to be assessed, as well as estimating the 310 decay constant of the more resilient stable C pool (Y0) to be assessed in constant, optimal conditions. From soils collected in summer, the total amount of C mineralised increased substantially as an effect of incubation temperature, however there were no effects of either the warming or removal treatments. On average, soil incubated at 17°C for two months emitted 48% more C than at 10°C, and a further 22% at 25°C (F2,82=80.9, https://doi.org/10.5194/bg-2020-144 Preprint.

Labile C
In summer soil, incubation temperature significantly increased the size of Ca on average by 50% from 10°C to 17°C, and by a further 18% at 25°C (P<0.001) (Fig. 6A). There were no treatment effects on the size of Ca. Winter soil incubations reflect similar results to those for summer soils, with a 27% increase in Ca pool size from 10°C 330 to 17°C, and a further 27% increase to 25°C (P=0.001). As with summer soil there were no treatment effects.
Overall, season had no effect on Ca, however incubation temperature increased Ca across the two seasons of 36% from 10°C to 17°C and a further 24% at 25°C (P<0.001).
The intrinsic decay constant of the labile pool (k) in summer soil was not affected by incubation temperature (F2,82=0.39, P=0.68), the warming (F1,82=0.06, P=0.8) , or removal treatments, i.e. neither dominant nor random 335 biomass removal (F2,82=0.31, P=0.73), was significantly influenced by an interaction between warming and species removal (F2,82=3.14, P=0.05) (Fig. 6C). In ambient plots, removing the dominant species tended to increase k, however, in warmed plots, the opposite occurred. Post hoc analysis revealed the greatest differences in k were observed between warmed x no removal and warmed x dominant removal plots, and warmed x dominant removal and ambient x dominant removal plots. In winter, there were no treatment or incubation temperature 340 effects on k, however k was on average 42% greater in summer (F1,196=201.09, P<0.001).

Intrinsic decay constant of the stable C pool
From summer soil, the size of the stable C pool (Y0) also increased significantly (F2,82=78.01, P<2-16) as a function of incubation temperature with an average increase of 47% from 10°C to 17°C, and a further 20% at 25°C (Fig.  6C). There were no treatment effects on the Y0 of summer soil. For winter soils, responses to treatments were similar to those of summer soils. There were no treatment effects, but incubation temperature increased Y0 on average by 27% from 10°C to 17°C, and a further 28% at 25°C (F2,112=45.9, P<0). Overall Y0 was 39% higher in summer than in winter (F1,196=137.61, P<0.001), and incubation temperature also significantly increased Y0 overall, with on average a 38% increase from 10°C to 17°C, and a further 23% at 25 °C (F1,196=107.28, P<0.001),

350
however there were no treatment effects.

Proportion of total C that was labile
From summer soil, the proportion of total C that was from Ca was only affected by incubation temperature with on average a 49% increase from 10°C to 17°C, and a further 22% increase when incubated at 25°C (F2,82=77.73,

355
P<0.001; Fig. 6D). There were no treatment effects. Similarly, in winter, the proportion of total C that was Ca increased only as a function of increasing incubation temperature, with on average a 24% increase from 10°C to 17°C and a further 27% at 25°C (F2,112=22.19, P<0.001). Overall, the proportion of total C that was Ca, increased substantially as a function of incubation temperature (F2,196=67.94, P<0.001) with a 35% increase from 10°C to 17°C, and a further 25% increase at 25°C, however there were no overall treatment effects.

Total soil C content
Overall, irrespective of removal treatment, total soil C % averaged 19.2 ± 0.4 (P<0.001). C % was 18.7 ± 0.7 and 19.7 ± 0.6 in ambient and warmed soils respectively, however there were no significant treatment effects .

375
All incubation results were also analysed per gram of soil C but results were essentially identical to those expressed per gram of soil dry weight, above.

Discussion
The main aim of this study was to investigate whether warming increases RS in situ, and whether any observed 380 treatment effects were due to an increased ability of the soil microbial community to mineralise SOC. Additionally, we investigated whether manipulating plant community composition affected the RS response to warming. Results removal interaction that influenced the decay constant of the soil labile C pool (k). Overall, the results from this study suggest that as there was no change in microbial respiratory potential, the observed increase in soil respiration in situ was largely an effect of altered plant activity in warmed plots.

395
Warming increased RS in situ over the course of the sampling period from November 2017 to June 2018. This increase in CO2 efflux observed from soils in situ encompasses the response of both microbial (heterotrophic) respiration, and plant root (autotrophic) respiration, and amounted to an average increase in soil C efflux of 28%.
The observed increase of RS in response to warming is in line with multiple other studies, although most of these focus on soils in the low-to mid-range of soil C stocks and in northern hemisphere locations (Lu et al., 2013;van 400 Gestel et al., 2018). There are 4 possible mechanisms whereby could have increased in RS: 1.) Increased temperature sensitivity of RS; 2.) Influence through change in plant community composition; 3.) Enhanced substrate supply through SOM and 4.) Plant induced alteration to soil microhabitat. The substantial RS response to warming could be due to one or a combination of these processes and determining which were likely to be involved has significant ramifications for our ability to predict future soil C dynamics.

Increased temperature sensitivity of RS
One of the proposed mechanisms behind the increased RS response to warming, and subsequent loss of soil C stores is an increase in the temperature sensitivity of RS, i.e. increased decomposition of SOM (Kirschbaum, 1995). This response, mainly attributed to an increase in enzyme kinetics with temperature, is linked strongly to 410 substrate availability (Davidson and Janssens, 2006). At Silver Plains, the overall significant increase in RS rates from warmed plots in situ implied that the temperature sensitivity of RS was higher under warming. The highest RS rates were recorded during the growing season in spring and summer, suggesting primary productivity, microbial activity and environmental factors such as precipitation are likely to substantially influence respiration rates (Almagro et al., 2009). However, despite the strong dependence of RS on soil water content and soil 415 temperature, warmed plots had higher rates of C efflux from the soil under particular combinations of soil temperature and moisture (Section 3.1.5, Fig. 3). The restrictive effect of high soil water content and low soil temperature on RS observed in this study is widely documented and due to the creation of anoxic conditions limiting microbial access to substrate (Schimel et al., 1994;Syed et al., 2006;Sierra et al., 2015). Hence the observed effect of soil water content and soil temperature on RS was anticipated, however the degree to which 420 https://doi.org/10.5194/bg-2020-144 Preprint. Discussion started: 25 May 2020 c Author(s) 2020. CC BY 4.0 License.