Contrasting Rhizospheric and Heterotrophic Components of Soil Respiration during Growing and Non-Growing Seasons in a Temperate Deciduous Forest

The contributions of heterotrophic respiration (RH) to total soil respiration (RS) for the non-growing season, growing season, and annual period are 84.8%, 60.7%, and 63.3%, respectively.Few studies have partitioned RS into its rhizospheric (RR) and heterotrophic components throughout the year in northern forest ecosystems. Our objectives were to quantify the contributions of non-growing season and heterotrophic respiration. We conducted a trenching experiment to quantify RR and RH in a temperate deciduous forest in Northeast China over two years using chamber methods. Temperature sensitivities (Q10) for RS and for RH were both much higher in the non-growing season (November to April) than those in the growing season. The Q10 for RS was higher than Q10 for RH in both seasons, indicating a higher temperature sensitivity of roots versus microorganisms. Mean non-growing season RS, RH, and RR for the two years were 94, 79 and 14 g carbon (C) m−2, respectively, which contributed 10.8%, 14.5%, and 4.5% to the corresponding annual fluxes (869, 547 and 321 g C m−2 year−1, respectively). The contributions of RH to RS for the non-growing season, growing season, and annual period were 84.8%, 60.7%, and 63.3%, respectively. Using the same contribution of non-growing season RS to annual RS, to scale growing season measurements, to the annual scale would introduce significant biases on annual RH (−34 g C m−2 yr−1 or −6%) and RR (16 g C m−2 yr−1 or 5%).We concluded that it was important to take non-growing season measurements in terms of accurately partitioning RS components in northern forests.

Soil respiration is overwhelmingly comprised of heterotrophic (R H ; respiration by microbes and soil fauna) and rhizospheric (R R ; respiration by roots, their associated mycorrhizal fungi, and other micro-organisms directly dependent on labile carbon compounds leaked from roots) components [8][9][10].R R and R H may be driven by different mechanisms [8,11], and thus respond differently to environmental factors, both at given sites [8,12,13] and globally [14,15].Partitioning R S is an important step for assessing plant physiology, C allocation, ecosystem C balance, and the climate feedback potential of changes in R S [5,[15][16][17].
Most R S measurements in northern forests are limited in the growing season [18].The lack of R S studies during the non-growing season is due not only to the misconception that there is little biological activity during this period, but also to the inherent difficulties associated with cold winter sampling over snowpack [19].However, soil respiration has been shown to continue throughout the winter in a wide variety of northern or alpine ecosystems [18, [20][21][22][23][24]. Studies which have integrated R H contribution to R S throughout an entire year or growing season show a mean value of 54% for the forest ecosystem [9].The relative contributions of R R and R H vary during the year [25][26][27][28][29][30][31][32].In cold climates, the R H contribution is commonly lower than or similar to that of R R during the growing season, and higher during the dormant periods.However, few measurements quantified R H contribution during the winter, particularly for the snow-covered period in northern or alpine ecosystems [25,27,33,34].
There are several methods for partitioning R S , each with advantages and limitations [9, 16,35,36].The trenching experiment is the most popular method used in boreal and temperate forest ecosystems [35], because of its simplicity and low cost [9].Soil respiration has been partitioned by the trenching method for six forest ecosystems in the Maoershan site during growing season [28], but no direct measurements were conducted during the non-growing season [24].In this study, the trenching approach was used throughout two successive years in a broadleaved deciduous forest.We hypothesized that using a same value of root contribution as growing season measurements throughout annual cycles would introduce significant biases in the estimating of annual R S and its components.

Experimental Design and Field Measurements
This study was conducted at the Maoershan Forest Ecosystem Research Station of Northeast Forestry University, Heilongjiang Province, Northeast China (45 • 24 N, 127 • 40 E, 400 m a.s.l.).The climate is a continental monsoon climate with a windy and dry spring, a warm and humid summer, and a dry and cold winter.The mean  annual precipitation is 629 mm, and mean annual air temperature is 3.1 • C. The mean January and July air temperatures are −18.5 • C and 22.0 • C, respectively [24].The broadleaved deciduous forest around the eddy flux tower was ~60-years old.The dominant tree species include Ulmus japonica Sarg., Fraxinus mandshurica Rupr., Betula platyphylla Suk., Populusdavidiana Dode, Juglans mandshurica Maxim., etc. [37].The soil is a type of alfisol with a parent rock of granite.
Eight 20 m × 30 m plots were set up around the eddy flux tower in August 2007.We measured R S during the growing season for all plots, and three plots (plot number 2-4) for the non-growing season.The basic stand characteristics were given in Table 1.
Table 1.Basic stand characteristics, soil organic carbon, and total nitrogen for the three plots.

SOC Density
Mg ha −1 Basal area (BA) and biomass data were from 2010.Litterfall was the mean of five 1-m 2 traps per plot [38,39] in 2010 and 2011.Fine root (diameter ≤ 2 mm) biomass (FRB) and fine root necromass (FRN) of the top 40 cm of soil were estimated by eight 10-cm diameter root cores per plot in the summer of 2008.Soil organic carbon (SOC) and total nitrogen (TN) of the top 20 cm were measured in 2016 [40].
Four 100-cm diameter trenched subplots were established in each plot by digging a trench (30 cm wide) around the outside boundary to a depth of 80 cm in October 2007.Trenches were lined with plastic sheets to prevent root entry and potential lateral CO 2 transport [41], and refilled and packed carefully with the same soil.Eight polyvinyl chloride (PVC) collars (10.2 cm inside diameter, 6 cm height) were randomly placed for R S measurements in each plot, and one collar was installed in each trenched subplot for R H measurements.
Since May of 2008, soil surface CO 2 effluxes for control plots (R S ) and trenching subplots (R H ) were measured using an LI-COR 6400 portable infrared gas analyzer (IRGA) (LI-COR 6400, LI-COR Inc., Lincoln, NE, USA) equipped with a dynamic chamber (LI-COR 6400-9, LI-COR Inc., Lincoln, NE, USA) bi-weekly to monthly during the growing season.The measurement protocol (three cycles based on a change of 5 µmol mol −1 CO 2 , subsequently averaged at each collar) was similar to that used in previous soil respiration studies at this site [28,42].Soil temperatures at 5 cm depth (T 5 ) and 10 cm depth (T 10 ) was simultaneously measured by a digital thermocouple, and the soil volumetric water content at the top 10 cm depth was detected by a Time-Domain Reflectometry (CS620, Campbell Scientific, Logan, UT, USA).
During the non-growing season, other chamber methods were adopted to measure the R S and R T because of LI-COR 6400 failure at low temperatures.A dark static chamber method was used from December 2009 to April 2010.Four steel collars (40 cm wide × 50 cm length) were installed in October 2009 in each plot, and the other four collars were installed in the four trenched subplots.A small vent in the chamber was used to balance the air pressure inside and outside the chamber [43].Before each measurement, the snow around the base was removed carefully to minimize disturbance of the snow over the base.The stainless-steel statistic chamber (40 cm wide ×50 cm length × 50 cm height) was fitted to the collarand sealed with adhesive tape.Then the snow was quickly refilled to minimize the potential CO 2 pulse from the snow hollow around the chamber base and insulate the soil from the cold air.The chamber was covered by a specific white quilt to prevent heating by solar radiation.A small fan was used to mix the gas in the chamber.Fifty mL gas samples were collected using fine needle syringes 0, 10, 20, and 30 min following placement of the chamber on the base [44].Snow depth (D S ), T 5 , and T 10 near each chamber were also measured with a ruler and a thermocouple, respectively.Samples were transferred to evacuated gaseous bags, which were stored at room temperature.The CO 2 was measured with gas chromatography (7890A GC System, Agilent Technologies, Santa Clara, CA, USA) within three days.Fluxes were calculated from the linear rate of change in gas concentration, the chamber internal volume, and soil surface area [44].
During the period of December 2010 and April 2011, the R S, R H , T 5 , and T 10 were measured using a LI-COR 8100 portable IRGA equipped with a LI-COR 8100-103survey chamber (LI-COR Inc., Lincoln, NE, USA).To avoid the lateral diffusion of CO 2 [33], deep soil collars (determined by snow depth, as did in reference 18) were inserted into the soil surface and stabilized for at least 10 min before measuring the CO 2 efflux.There were four control and four trench-treatment chambers in each plot.The D S was also measured with a ruler near each chamber of the control plot.
Different methods were used to measure soil CO 2 efflux in the growing and non-growing seasons, i.e., LI-6400 in the growing season, static chamber in the first winter, and LI-8100 in the second winter.We assume that the difference between methods could be ignored.According to Pumpanen et al. [45], the chambers could be sorted into three types: closed static chamber, closed dynamic chamber, and open dynamic chamber.The static chamber could underestimate CO 2 efflux, mainly due to the rising concentration within the chamber headspace [45].Therefore, we only used this technique in the winter, when CO 2 concentration increases were very small due to the small flux (generally <0.8 µmol m −2 s −1 ).Previous studies verified that the LI-8100 and LI-6400 gave substantially similar results [46,47].

Data Analysis
Empirical models relating R S and R H to soil temperature and water content were developed for each plot, based on discrete measurements of R S and R H , soil temperature, and soil water content (SWC), as described in Wang et al. [28,42].Logarithmic transformation of R S was needed to achieve linearity and homoscedasticity.A backward elimination procedure was performed to remove insignificant terms (α > 0.05).The form of the regression models is where ln is natural logarithm; R represents R S or R H ; T is the soil temperature; SWC is soil water content; and a, b, c, and d are regression coefficients.The temperature responses of R S and R H were regressed by the log-transformed exponential equation ln where R is R S or R H , and a and b are regression coefficients.Then the temperature coefficient (Q 10 ) was calculated as [48] Based on a comparison of models with different predictors (Supplementary Materials), we used the T 5 and SWC 10 to model the growing seasonal fluxes, and the T 5 for the non-growing season.The continuous T 5 and SWC 10 of the dataloggers were used to estimate the plot-specific T 5 and SWC 10 with simple linear models.Then the plot-specific T 5 and SWC 10 were used to drive the respiration models to get half-hour R S and R H [28,43].The seasonal and annual fluxes were the time-integrations of the half-hour R S and R H values [24].The difference in T 5 and SWC 10 between the treatments was tested by a paired t-test at the plot scale (taking the average for each plot and each year).The effect of the difference in SWC 10 between the treatments was assessed by calculating annual flux in trenched plots (R H ) with the SWC 10 values in the control and trenched plots [49].Considering the diminishing root decomposition added by the trenching treatment [35], the measurements in the trenched plots were directly considered as R H .As a result, the R R was calculated as the difference between R S and R H .
In this paper, we report only the data in the last two climatic years for the three plots (numbers 2-4) to quantify the non-growing season contribution and root contribution.One reason was to avoid noncontinuous data, and the other was to minimize the effect of decomposition of newly dead roots after trenching [35].

Seasonal Variations in Soil Respiration and Related Environmental Factors
Both air temperature (T a ) and soil temperature at 5 cm depths (T 5 ) followed a bell-shaped curve during the study period (Figure 1a).The T a varied from about −30 • C in January to a maximum of about 23 • C in June.The T 5 generally lagged to the T a by roughly one month, with a minimum (about −2 • C) in February and a maximum (about 20 • C) in August.Photosynthetically active radiation (PAR) peaked slightly earlier than T a for both growing seasons, but interannually it was lower in mid-to-late summer of 2010.During the winter months, soil volumetric water content in the 10 cm soil layer (SWC 10 ) was low (0.15 m 3 m −3 ), and increased rapidly following snowmelt in early April (2011) or mid-April (2010; about 0.4 m 3 m −3 ; Figure 1b).Snow depth (D S ) differed dramatically between the two winters.Snow depth in January was shallower for 2010, but the peak occurred in March (66 cm) for 2010 and in January (41 cm) for 2011 (Figure 1b).The R S and R H generally followed the course of T 5 (Figure 1c).The minimum of R S occurred in February for 2010 and in January for 2011 (0.11 and 0.15 µmol m −2 s −1 , respectively), and R H occurred in January for both years (0.09 and 0.06 µmol m −2 s −1 , respectively).However, R S maximized in early July while R H peaked in late July for 2010, whereas R S and R H both peaked in mid-August for 2011.The difference in R S and R H was generally followed by R S .

Responses of Soil Respiration to Soil Temperature and Water Content
Based on the log-transformed exponential model of observed RS and RH versus T5, the RS0 and RH0 (normalized respiration rate at 0 °C for RS and RH) varied between the two non-growing seasons, although RS0 was generally higher than RH0 (Figure 2).The RS0 was higher during the non-growing season between 2009 and 2010 (0.66 μmol m −2 s −1 ) than that during the following non-growing season (0.51 μmol m −2 s −1 ), while RH0 was comparable between the two non-growing seasons (0.46 versus 0.42 μmol m −2 s −1 ).The QS10 (Q10 for RS) and QH10 (Q10 for RH) were lower in the non-growing season between 2009 and 2010 than in the following non-growing season.The QS10 was lower than QH10 for both nongrowing seasons (18.54 versus 11.13 for the former, and 742.48 versus 152.93 for the latter).

Responses of Soil Respiration to Soil Temperature and Water Content
Based on the log-transformed exponential model of observed R S and R H versus T 5 , the R S0 and R H0 (normalized respiration rate at 0 • C for R S and R H ) varied between the two non-growing seasons, although R S0 was generally higher than R H0 (Figure 2).The R S0 was higher during the non-growing season between 2009 and 2010 (0.66 µmol m −2 s −1 ) than that during the following non-growing season (0.51 µmol m −2 s −1 ), while R H0 was comparable between the two non-growing seasons (0.46 versus 0.42 µmol m −2 s −1 ).The Q S10 (Q 10 for R S ) and Q H10 (Q 10 for R H ) were lower in the non-growing season between 2009 and 2010 than in the following non-growing season.The Q S10 was lower than Q H10 for both non-growing seasons (18.54 versus 11.13 for the former, and 742.48 versus 152.93 for the latter).
The R S0 and R H0 for the growing season (0.60-1.00 µmol m −2 s −1 ) were slightly higher than those for the non-growing season (0.42-0.66 µmol m −2 s −1 ), whereas Q S10 and Q H10 (2.34-3.06)were much lower than those for the non-growing season (11.13-692.29)(Figures 2 and 3).The interannual differences in R S0 and R H0 , as well as Q S10 and Q H10 for the growing season, were consistent with those for the non-growing season.For both growing seasons, R S0 was higher than R H0 and Q S10 was larger than Q H10 .The RS0 and RH0 for the growing season (0.60-1.00 μmol m −2 s −1 ) were slightly higher than those for the non-growing season (0.42-0.66 μmol m −2 s −1 ), whereas QS10 and QH10 (2.34-3.06)were much lower than those for the non-growing season (11.13-692.29)(Figures 2 and 3).The interannual differences in RS0 and RH0, as well as QS10 and QH10 for the growing season, were consistent with those for the non-growing season.For both growing seasons, RS0 was higher than RH0 and QS10 was larger than QH10.For the growing season, the statistical models of ln(RS) and ln(RH) against T5 and SWC10 explained 82.5%-91.8% of the variability (Table 2).The RS and RH were positively correlated to soil temperature and its interactions with SWC10 in all plots for both years, while showed varying correlations to SWC10.For the growing season, the statistical models of ln(R S ) and ln(R H ) against T 5 and SWC 10 explained 82.5%-91.8% of the variability (Table 2).The R S and R H were positively correlated to soil temperature and its interactions with SWC 10 in all plots for both years, while showed varying correlations to SWC 10 .
Table 2. Models of soil respiration from control plots (R S , in µmol m −2 s −1 ) and trenched plots (R H , in µmol m −2 s −1 ) against soil temperature at 5 cm depth (T 5 , in • C), and soil volumetric water content of the top 10 cm (SWC 10 , in m 3 m −3 ) for the three plots during the growing season (May to October) for 2010 and 2011.

Contribution of Rhizospheric and Heterotrophic Respiration
The contributions of R H to R S change with integration period, with the values of 84.8%, 60.7%, and 63.3% for the non-growing season, growing season, and annual period, respectively (Figure 4a).The R H contribution for the non-growing season was relatively 40% higher than that for growing season.The contributions of RH to RS change with integration period, with the values of 84.8%, 60.7%, and 63.3% for the non-growing season, growing season, and annual period, respectively (Figure 4a).The RH contribution for the non-growing season was relatively 40% higher than that for growing season.

Contribution of Non-Growing Season Soil Respiration
The cumulative RS, RH, and RR for the non-growing season range from 84 to 103 g C m −2 , from 75 to 84 g C m −2 , and from 9 to 19 g C m −2 , respectively (Table 3).The mean non-growing season RS, RH, and RR were 94, 79, and 14 g C m −2 , respectively.The mean annual RS, RH, and RR were 869, 547, and 321 g C m −2 year −1 , respectively.The mean RS, RH, and RR for the annual and non-growing season were

Contribution of Non-Growing Season Soil Respiration
The cumulative R S , R H , and R R for the non-growing season range from 84 to 103 g C m −2 , from 75 to 84 g C m −2 , and from 9 to 19 g C m −2 , respectively (Table 3).The mean non-growing season R S , R H , and R R were 94, 79, and 14 g C m −2 , respectively.The mean annual R S , R H , and R R were 869, 547, and 321 g C m −2 year −1 , respectively.The mean R S , R H , and R R for the annual and non-growing season were given in Table S2.Therefore, the non-growing season R S , R H , and R R contributed 10.8%, 14.5%, and 4.5% to the corresponding annual fluxes (Figure 4b).The non-growing season contribution of R H was twice that of R R .
Table 3. Cumulative soil surface CO 2 fluxes (± standard deviation (SD)) in the control plot (R S ), in the trenched plot (R H ), and rhizospheric respiration (R R ).The means ± SD of the three plots are given.

Response of Soil Respiration to Temperature
We found the seasonal variations of R S and R H were generally followed by that of T 5 , indicating a dominating role of temperature in respiration at the seasonal time scale.However, the different peak timing of R S and R H during the former growing season was a response to the lower mid-to-late summer PAR (Figure 1).This highlights the importance of a recent supply of assimilates directly to R R , but not to R H [11,50,51]. Recently, Zhang et al. [52] verified that changes in photosynthesis drive the seasonal soil respiration-temperature hysteresis relationship, with numerical models and 129 FLUXNET sites.These findings highlight a strong necessity of component partitioning of R S [8,11,28].
Our estimates in non-growing season R S (generally <0.8 µmol m −2 s −1 ) was well within the range of previous studies conducted in northern and subalpine forest ecosystems [24,33,[53][54][55][56]. Higher rates (>0.6 µmol m −2 s −1 ) in a temperate mixed forest in Switzerland may be a result the high soil temperature (roughly above 3 • C) during the dormant season [27].A lower dormant season R S in temperate forests (below 0.3 µmol m −2 s −1 ) of North China [18] and in a boreal black spruce forest in Canada [25] may be due to the lower soil temperature, which reached −10 • C at 2 cm depth [25] or 5 cm depth [18] in harsh winter.These results indicated that winter R S was mainly controlled by soil temperature as at the Maoershan site (Figures 2 and 3).
We found that Q S10 and Q H10 in the non-growing season were between 11.13 and 692.29, but dropped to between 2.34 and 3.06 in the growing season.Numerous studies have demonstrated exceptionally high Q 10 for the beneath-snow respiratory flux in cold-winter ecosystems.For instance, the Q S10 and Q H10 in the dormant season (all above 4) were higher than those in the growing season (often <2) in two forests in the Loess Plateau of China [29].Although our Q s10 of 692.29 might be unreasonable from a biotic view, a much higher value (1.25 × 10 6 ) was reported in a subalpine conifer forest [23].It has been argued that the high temperature sensitivity could be attributable to the combined effects of exponential growth of snow molds, and the exponential response of their respiration rate to small changes in temperature [57].
Our Q S10 was higher than Q H10 in both non-growing season.There was a debate on which is higher for Q S10 and Q H10 in the literature [58,59].Lower temperature sensitivity for R S versus R H was reported in a mixed beech forest in Switzerland during the dormant period [27], in a boreal black spruce plantation in Canada [60], and in an oak chronosequence in China [13].However, we found that Q S10 was more often higher than Q H10 in the literature [12,14,25,29,61,62].The winter higher Q S10 compared with Q H10 indicated a higher temperature sensitivity of roots than microbes.However, the apparent Q 10 during the growing season may reflect the effect of plant phenological patterns [48,63,64].

Non-Growing Season and Annual Soil Respiration
Our mean annual R S (869 g C m −2 yr −1 ) was similar to that of the previous studies in broadleaved deciduous forests at the same site (781-813 g C m −2 yr −1 ) [38], as well as the mean of global temperate forests (829 ± 337 g C m −2 yr −1 ) [3].Our mean accumulated R S for non-growing seasons (94 g C m −2 , Table 3) was also well in the range of the reported values in temperate forests   54,55,[65][66][67].Lower R S values during the non-growing seasons between 2009 and 2010 than those between 2010and 2011 were probably due to the corresponding lower soil temperature (1.6 versus −0.5 for T 5 ; Figure 1a) [66].We argue that the interannual variations in soil temperature and the corresponding R S were mainly due to the difference in D S before January, rather than the maximum D S (Figure 1b).Our non-growing season contribution to annual R S (10.8%) was at the low end of those in literature (6%-23%) [18,24,44,54,66,[68][69][70].However, the contribution of non-growing season R H to annual R H (14.5%) was over triple that of R R (4.5%).Using the non-growing season R S contribution to annual R S (10.8%), to scale growing season measurements, to the annual scale would introduce substantial biases in the annual R H (−33.70 g C m −2 yr −1 or −6.2%) and R R (15.70 g C m −2 yr −1 or 4.9%).

Contribution of Rhizospheric and Heterotrophic Respiration
Our annual R H contribution to R S (63.3%) was slightly higher than the global means of the temperate deciduous forests: 54% by Subke et al. [35] and ~50% by a recent meta-analysis [14].Our higher R H contribution might be due to the relative higher SOC, as indicated by a positive correlation of R H with the SOC at the same site [28].However, our estimates of R H contribution were much higher than those in the broadleaved deciduous forests at the same site (23%-38%) [28].The discrepancy may be primarily due to different treatment of root decomposition effect.
Heterotrophic contribution varied with the seasons, which might reflect different controlling of R R and R H .The R H dominated R S (84.8%) in the non-growing season, but its contribution declined in the growing season (60.7%).Similar patterns were widely observed in other northern or subalpine forests [9, 12,25,27,29,71].For example, the contribution of R H to R S was lowest in the growing season (54% in an oak forest and 40% in a black locust plantation in the Loess Plateau, China), and it increased up to 88% and 94% during the dormant season [29].The largest contribution of R H to R S was 70% during the dormant period, and as low as 41% during the summer in a mountain mixed temperate forest [27].In a subalpine ponderosa pine plantation in the United States, the R H /R S ratio varied from 44% during the growing season to 84% during the non-growing season [71].We found that using the R H contribution during the growing season to estimate the non-growing R H and R R would underestimate R H but overestimate R R by 22.51 g C m −2 yr −1 , or 7%.The colder soil in the former winter had a higher R H contribution versus the latter (89.1% versus 81.3%), although the lower soil temperature suppressed the R S and its components (Table 3).Reduced winter R S was widely reported in snow removal experiments [24,54], and between years in natural conditions [66], but the partitioning of R S is rarely reported.The reduced total respiration but increased R H contribution indicates a higher temperature sensitivity of roots versus microbes.This is also supported by the higher Q S10 versus Q H10 (Figure 3).Therefore, a single value of R H contribution from growing season measurements should not be applied throughout an annual cycle [9].

Effects of Plot Trenching on Soil Respiration
The decomposition of dead roots in trenched subplots is one of the most important uncertainties in R S partitioning.Estimates of this effect in the literature have ranged from 2% to 54% [28, 30,35,42,72], with the effect being the most pronounced within the first year after trenching, and declining to minimal levels after two years [26,73] or more [74,75].The estimate of error introduced by the additional decay of trenched roots for the second year following trenching indicates that R H might be overestimated by as much as 20% in boreal forests [34] and 14% in temperate forests [64].The root decomposition accounted for 23%-35% of the annual R S in a previous study at our site [28], which might be overcorrected due to (1) overestimation of coarse root biomass using allometric equations, which did not account for the declination of root biomass with increasing distance from the stump; and (2) ignoring the carbon input of fine root mortality in natural conditions [76,77].These uncertainties in dead root decomposition and fine root mortality cautions future studies involving the trenching method [49,73,76,77].
To minimize the effect of root decomposition, we used the measurements in the third and fourth years after trenching, as done in some previous studies (e.g., [61,78]).However, measurements conducted many years after trenching might underestimate the R H /R S ratio, due to a lack of below-ground litter input in trenched subplots compared with control plots [35].Furthermore, eliminating the priming effect of root inputs in trenched plots may decrease decomposition [79], and thus underestimate R H . Therefore, we did not exclude the root decomposition in the trenched plots.Accurate estimates of R H by trenching experiment need detailed carbon budgets in trenched plots [77].
Differences in soil moisture and temperature between control and trenched plots may be another potential artifact of the trenching experiment [28,49].The increase in T 5 by trenching treatment was due to the removal of understory vegetations, which enhances the light reception of the forest floor.However, the trenching effect on T 5 reversed in the non-growing season, which might be attributed to the shallower snowpack or the disturbance introduced when measuring CO 2 efflux.The increase of SWC 10 (3% or 0.05 m 3 m −3 ) by trenching might be caused by different soil drainage conditions and water uptake by trees between plots.A previous study reported that the effect of soil moisture was 16% in a pine forest in Sweden [73].Nevertheless, R H and its contribution to R S , based on our plot-specific T 5 and SWC 10 data and the plot-treatment-specific respiration models, may not be largely influenced by these environmental changes from trenching, because the regression models and the driving variables had already accounted for those changes.

Conclusions
Soil rhizospheric and heterotrophic respiration in a temperate deciduous forest was partitioned by a trenching experiment throughout two years.The non-growing seasons' Q S10 and Q H10 were both much higher than those in the growing season.Mean non-growing season R S , R H , and R R for the two years was 94, 79, and 14 g C m −2 , respectively, which contributed 10.8%, 14.5%, and 4.5% to the corresponding annual fluxes (869, 547, and 321 g C m −2 yr −1 , respectively).The contributions of R H to R S for the non-growing season, growing season, and annual period were 81.3%, 60.7%, and 63.3%, respectively.Using the fixed non-growing season contribution of R S to scale growing season measurements to the annual scale would introduce significant biases on annual R H (−34 g C m −2 yr −1 or −6%) and R R (16 g C m −2 yr −1 or 5%), and using the R H contribution measured in the growing season to partitioning R S in the non-growing season would overestimate R R by 23 g C m −2 yr −1 (7% of annual flux).We concluded that it was important to take non-growing season measurements on the partitioning of R S components in temperate forests.

Figure 1 .
Figure 1.Seasonal dynamics of (a) daily air temperature at 16 m height (Ta), daily soil temperature at 5 cm depth (T5), and daily photosynthetically active radiation (PAR); (b) daily soil volumetric water content at 0-10 cm depth (SWC10), snow depth (DS), and (c) soil CO2 efflux (RS) and CO2 efflux from trenched plots (RH) from November 2009 to October 2011.The error bars are standard deviations of the three plots.The vertical dash lines in the figure indicate the end of the year, and the horizontal dash line is zero temperature.

Figure 1 .
Figure 1.Seasonal dynamics of (a) daily air temperature at 16 m height (T a ), daily soil temperature at 5 cm depth (T 5 ), and daily photosynthetically active radiation (PAR); (b) daily soil volumetric water content at 0-10 cm depth (SWC 10 ), snow depth (D S ), and (c) soil CO 2 efflux (R S ) and CO 2 efflux from trenched plots (R H ) from November 2009 to October 2011.The error bars are standard deviations of the three plots.The vertical dash lines in the figure indicate the end of the year, and the horizontal dash line is zero temperature.

Figure 2 .
Figure 2. Relationships between soil temperature at 5 cm depth (T5) and (a,c) soil CO2 effluxes (RS) and (b,d) CO2 efflux from trenched plots (RH) during the non-growing season.The RS0 and RH0 are normalized respiration rate at 0 °C for RS and RH, respectively, and QS10, QH10 are temperature sensitivities for RS and RH, respectively.All p-values of the regression models were <0.001.

Figure 3 .
Figure 3.Relationships between soil temperature at 5 cm depth (T 5 ) and (a,c) soil CO 2 effluxes (R S ) and (b,d) CO 2 efflux from trenched plots (R H ) during the growing season.The R S0 and R H0 are normalized respiration rate at 0 • C for R S and R H , respectively, and Q S10 , Q H10 are temperature sensitivities for R S and R H , respectively.All p-vales of the regression models were <0.001.

Figure 4 .
Figure 4. Contributions of the non-growing season to annual (climatic year) respiration for soil (RS), heterotrophic (RH), and rhizospheric (RR) respirations (a); the contribution of RH to RS for the nongrowing season (NGS; from November to April), the growing season (GS; from May to October) and annual fluxes (b).The error bars are standard deviations of the three plots.

Figure 4 .
Figure 4. Contributions of the non-growing season to annual (climatic year) respiration for soil (R S ), heterotrophic (R H ), and rhizospheric (R R ) respirations (a); the contribution of R H to R S for the non-growing season (NGS; from November to April), the growing season (GS; from May to October) and annual fluxes (b).The error bars are standard deviations of the three plots.