Introduction

Soil is the largest organic carbon pool in terrestrial ecosystems (Lal 2004). Global soil respiration (R S), which has been estimated to be 98 ± 12 Pg C yr−1 in 2008 (Bond-Lamberty and Thomson 2010), is an important source of atmospheric CO2 (Raich and Schlesinger 1992; Friedlingstein et al. 2006). Soil respiration contains many components that are difficult to distinguish (Kuzyakov 2006). However, to obtain a better understanding of C-cycling on regional and global scales, previous studies have developed methods to quantify the components of R S (Hanson et al. 2000; Kuzyakov 2006; Subke et al. 2006). Soil respiration is usually divided into autotrophic respiration (R A) and heterotrophic respiration (R H) (Hanson et al. 2000; Subke et al. 2006). In forest ecosystems, trenching (or girdling) has been widely used to partition R A and R H (Hogberg et al. 2001; Subke et al. 2006; Li et al. 2010; Sayer and Tanner 2010), and the contribution of R H was 45–70 % of R S, which declined with increasing annual R S (Subke et al. 2006) . Since litter decomposition commonly makes a significant contribution to R S, litter removal has been used to quantify the contribution of aboveground litter decomposition (R AL) to R S (C AL) in forest ecosystems (Bowden et al. 1993; Sulzman et al. 2005; Sayer 2006; Wang et al. 2013). Combining the above, the decomposition of soil organic matter (SOM) can thus be calculated as follows: R SOM = R S–R A–R AL, thereby enabling estimates of the contribution of SOM decomposition to R S (C SOM) (Rey et al. 2002; Sulzman et al. 2005; Chang et al. 2008).

However, the above equation may underestimate the contribution of SOM to R S. For example, Rey et al. (2002) reported lower estimated values of C SOM compared with measured values based on experiments combining trenching with litter removal. This discrepancy is due to the priming effect caused by both rhizosphere activity (Fu and Cheng 2002; Kuzyakov 2002) and litter decomposition (Park et al. 2002; Kalbitz et al. 2007). Subke et al. (2004) indicated that since both litter decomposition and rhizosphere activity can promote SOM decomposition, they may positively interact. Therefore, to correctly partition the components of R S, it is important to quantify this interaction. However, few studies have investigated this interaction in forest ecosystems (see Subke et al. 2004).

Using experimental data collected from a subtropical montane cloud forest in SW China, this study tested the hypothesis that litter decomposition and rhizosphere activity has a positive interaction effect on soil respiration. In this forest, SOM mainly comes from roots and aboveground litter. Since roots release both low- and high-weight substances, such as sugars, amino acids, enzymes, and mucilage (Nguyen 2003), and litter decomposition percolates dissolved organic carbon into the mineral soil (Kalbitz et al. 2000), the trenching and litter removal treatments prevented labile input. Temperature-quality hypothesis suggests that old, low-quality SOM causes higher temperature sensitivity (Q 10) due to the higher activation energy required for the decomposition of low-quality SOM (Bosatta and Agren 1999); previous studies have supported this hypothesis (Davidson and Janssens 2006; Conant et al. 2008; Wetterstedt et al. 2010; Suseela et al. 2013). Therefore, we also hypothesized negative effects of litter decomposition and rhizosphere activity on the temperature sensitivity of R S.

Materials and methods

Site description

This experiment was conducted at the Ailaoshan Station for Subtropical Forest Ecosystem Studies (24°32′N, 101°01′E; 2,480 m above sea level) of the Chinese Ecological Research Network, which is located in Jingdong County, Yunnan Province. Over the past 10 years (2002–2011), the annual mean air temperature was 11.3 °C, with a minimum monthly mean temperature of 5.7 °C in January and a maximum monthly mean temperature of 15.6 °C in July. The average annual rainfall was 1,778 mm, with 86.0 % falling in the rainy season (May–October) (Fig. 1). The forest is influenced by the southwest monsoon and is exposed to frequent and intense wind and mist events throughout the year. The forest is described as a subtropical cloud forest, given the abundant moisture and persistent cloud cover (Song et al. 2012; Zhang et al. 2012). The dominant tree species in the forest are Vaccinium duclouxii, Lithocarpus chintungensis, and Schima noronhae, along with Sinarundinaria nitida in the shrub layer, and the litterfall is 864 g m−2 year−1. The soils are Alfisols, which have a pH value of 4.5, soil organic carbon 304 g kg−1, and total nitrogen 18 g kg−1 in the humus horizon (Chan et al. 2006).

Fig. 1
figure 1

Seasonal variations in air temperature and rainfall averaged from 2002 to 2011 at the Ailaoshan Station for Subtropical Forest Ecosystem Studies

Full size image

Experiment design

Three plots (10 × 10 m) were selected in the forest, and four subplots were established in each plot: control (CK), litter removal (NL), trenching (NR) and trenching with litter removal (NRNL). Cover structures (i.e., a bamboo framework covered with 1-mm nylon mesh, 1 × 1 m) were established in the NL subplots at a height of 1.2 m above the ground to prevent new litter dropping. Visible litter in the subplots was removed at the beginning of this experiment. In the NR subplots, PVC pipes (diameter 630 mm, height 500 mm) were used for trenching. A circular trench (width about 300 mm) was dug to 500 mm, to form a cylinder of soil contained by the PVC pipe; soil was backfilled by its original layers with topsoil over subsoil (see Fig. S1). NRNL plots were established by combining the NL and NR treatments. In the center of each subplot, one PVC connector was permanently inserted into the soil to a depth of 20 mm at the beginning of the experiment, and a PVC top-closed pipe (diameter 200 mm, height 200 mm) was mounted on the connector, constituting a respiration chamber when we measured soil CO2 efflux.

Data collection

The experimental setup was finished on 15 January 2010 and measurements began on 7 February 2010, continuing for two years. Soil CO2 efflux was measured monthly using a gas analyzer (LI-840; Li-cor, Lincoln, NE, USA) between 9:00 and 11:00 (Beijing Time) to avoid diurnal fluctuations. Soil temperature (T, °C) was measured at 50 mm depth with a digital thermometer (6310; Spectrum, Illinois, USA) and soil water content (W, %) was measured by time domain reflectometry (MP-KIT; Beijing Channel, Beijing, China).

Soil CO2 efflux (R) was calculated as follows:

$$ R=\frac{M}{V_0}\cdot \frac{P}{P_0}\cdot \frac{T_0}{T\mathrm{a}}\cdot H\cdot \frac{ dc}{ dt} $$
(1)

where R is the soil CO2 efflux (μmol m−2 s−1); M is the CO2 molar mass; V 0, P 0 and T 0 are constants (22.4 L · mol−1, 1013.25 hPa, and 273.15 K, respectively); Ta is air temperature (K); H is the height of the respiration chamber (m); and dc/dt is the slope of CO2 concentration variation with time over the measurement period.

Calculations

As soil temperature and soil water content affected soil respiration in forest ecosystems, previous studies have developed two-factor regression models to reflect the relationship of soil respiration with soil temperature and soil water content (Xu and Qi 2001; Qi et al. 2002). At the present site, soil temperature and soil water content showed similar seasonal variations (Fig. 2). Considering their interaction effect on soil CO2 effluxes, their product was taken as variable. Thus, a two-factor regression model was developed to reveal the relationship of soil CO2 efflux with soil temperature and soil water content, and given high regression coefficients of determination (R2 values) ranged from 0.76 to 0.93 (Fig. S2). The two-factor regression model equation was as follow:

Fig. 2
figure 2

Seasonal variability in soil temperature and soil water content of each treatment (mean + SE). Open circles represent control (CK), black circles represent litter removal (NL), open squares represent trenching (NR), and triangles represent trenching together with litter removal (NRNL)

Full size image
$$ R=\mathrm{a}\cdot T\cdot W+\mathrm{b} $$
(2)

where a and b are constants estimated from regression model (details see Fig. S2), T is soil temperature (°C), and W is soil water content (%).

R S, R NL, R NR, and R NRNL represent the soil CO2 effluxes of CK, NL, NR, and NRNL, respectively. We divided soil respiration into four components: basic SOM respiration (R SOM), litter respiration (R L), root respiration (R R), and a component representing the interaction between litter decomposition and rhizosphere activity (R INT) (Fig. 3). The two-factor regression model showed that both T and W had positive effects on soil CO2 efflux (Fig. S2), while the treatments affected the soil microclimate (Sayer 2006; Sayer and Tanner 2010), especially W (Fig. 2). Therefore, to eliminate the biases due to soil microclimate change, effluxes should be compared under the same environmental conditions. Accordingly, the mean soil temperature and soil water content measured in control subplots were used to correct the values of R S, R NL, R NR, and R NRNL by Eq. (2). The same T and W variables, but different parameters (a and b) for each subplot, were used for this correction (Fig. S2). The correction did not change R S and R NL, but it reduced R NR and R NRNL (Fig. S3).

Fig. 3
figure 3

Soil respiration divided into four components: basic SOM decomposition (R SOM), litter respiration (R L), root respiration (R R), and interaction (R INT). R A is autotrophic respiration including R R and R INT, R AL is aboveground litter decomposition including R L and R INT, R H is heterotrophic respiration including R SOM and R L, and R NL is soil CO2 efflux after litter removal including R SOM and R R

Full size image

We calculated the components as follows (details see Fig. 3):

$$ \mathrm{Soil}\ \mathrm{total}\ \mathrm{respiration}:{R}_{\mathrm{S}}={R}_{\mathrm{S}\mathrm{OM}}+{R}_{\mathrm{L}}+{R}_{\mathrm{R}}+{R}_{\mathrm{INT}}\underset{x\to \infty }{ \lim } $$
(3)
$$ \mathrm{Basic}\ \mathrm{SOM}\ \mathrm{respiration}:{R}_{\mathrm{SOM}}={R}_{\mathrm{NRNL}} $$
(4)
$$ \mathrm{Litter}\ \mathrm{respiration}:{R}_{\mathrm{L}}={R}_{\mathrm{NR}}-{R}_{\mathrm{NR}\mathrm{NL}} $$
(5)
$$ \mathrm{Root}\ \mathrm{respiration}:{R}_{\mathrm{R}}={R}_{\mathrm{NL}}-{R}_{\mathrm{NRNL}} $$
(6)
$$ \mathrm{Interaction}:{R}_{\mathrm{INT}}=\left({R}_{\mathrm{S}}+{R}_{\mathrm{NR}\mathrm{NL}}\right)-\left({R}_{\mathrm{NR}}+{R}_{\mathrm{NL}}\right) $$
(7)

Temperature sensitivity (Q 10) was calculated from R T+10/R T; R T was a regression model of efflux with soil temperature in field experiments. For example, an exponential equation (Q 10 = eb*T) was used to estimated Q 10 previously, which was derived from a one-factor regression model (R = R 0*eb*T). However, one-factor regression model could not well reveal the relationship of efflux with soil temperature. Therefore, as mentioned above, two-factor regression models were developed for better understanding of the relationship. Q 10 was not a constant; it changed with the variations of soil temperature and soil water (Davidson and Janssens 2006).

Base on the two-factor regression model (Eq. 2), we could get R T and R T+10.

$$ {R}_{\mathrm{T}}=\mathrm{a}\cdot T\cdot W+\mathrm{b} $$
(8)
$$ {R}_{\mathrm{T}+10}=\mathrm{a}\cdot \left(T+10\right)\cdot W+\mathrm{b}=\mathrm{a}\cdot T\cdot W+\mathrm{b}+10\mathrm{a}\cdot W $$
(9)

where W in R T was the same to R T+10, because R T+10 meant efflux rate after increasing 10 °C soil temperature, and thereby soil water content should keep the same compared to R T. Therefore, we developed Q 10 equation as follow:

$$ {Q}_{10}=\frac{ RT+10}{ RT}=1+\frac{10\mathrm{a}\cdot W}{\mathrm{a}\cdot T\cdot W+\mathrm{b}} $$
(10)

where a and b are as estimated by Eq. (2) for each subplot (Fig. S2). Eq (10) revealed the relationship of Q 10 with soil temperature and soil water content as reviewed by Davidson and Janssens (2006); Q 10 should be compared under the same environmental conditions for different treatments. Therefore, the mean soil temperature and soil water content measured in control subplots were used to calculate Q 10 values of all subplots for each measurement time.

Statistical analysis

All the data were subjected to tests for normality and homoscedasticity before ANOVA analyses. Univariate general linear modeling (GLM) was used to test the interaction effect of litter and root treatments. All differences were tested for statistical significance at the 95 % level. Statistical analyses were performed with SPSS 13.0 (SPSS Inc., Chicago, IL, USA).

Results

Influence of interaction between litter and rhizosphere on soil respiration

The Univariate GLM analyses showed that both rhizosphere activity and litter decomposition had significant effects on R S, and there were significant interactions between them (Table 1A).

Table 1 F and p values of univariate GLM analyses on soil CO2 efflux (A) and temperature sensitivity (B)
Full size table

R INT exhibited seasonal variation, increasing rapidly from the dry season to the rainy season (Fig. 4a). The average value in dry season (0.82 ± 0.28 μmol · m−2 · s−1) was significantly different from that in rainy season (2.55 ± 0.50 μmol · m−2 · s−1) (Fig. 4b).

Fig. 4
figure 4

Seasonal variability in soil respiration components a and their contributions to soil respiration c: open circles represent R SOM or C SOM, upward-facing triangles represent R INT or C INT, downward-facing triangles represent R R or C R, and squares represent R L or C L. Comparisons among components in dry season average, rainy season average, and annual mean b, d: black vertical bar represents R SOM or C SOM, dark gray represents R INT or C INT, gray represents R R or C R, and white represents R L or C L; all data passed the normality and homoscedasticity tests before one-way ANOVA analysis; different letters for the same measurement time indicate significant differences (as indicated by LSD’s post-hoc test); *means significant difference between dry and rainy seasons (independent t testes). Data are mean + SE (n = 3)

Full size image

The contribution of R INT to R S (CINT) showed a similar pattern to its flux; i.e., higher contributions in the rainy season (Fig. 4c). However, C INT in dry season (27 ± 8 %) did not significantly differ to that in rainy season (31 ± 6 %). R INT contributed 30 ± 7 % to R S as the annual mean (Fig. 4d).

Single-factor regressions of R INT as a linear function of T only, and as a linear function of W only, showed that T and W explained 64 and 85 % of the variation in R INT, respectively. Linear regression of T with C INT and logarithmic regression of W with C INT showed that T and W explained 41 and 93 % of the variation in C INT, respectively (Fig. 5).

Fig. 5
figure 5

a, b Linear regression of R INT onto soil temperature and soil water content, c Linear regression of C INT onto soil temperature, d Logarithmic regression of C INT onto soil water content

Full size image

Comparisons to other components

R SOM showed similar patterns to those for R INT; i.e., higher fluxes in the rainy season. R L showed minor seasonal variations whereas R R was largely unchanged throughout the year (Fig. 4a). There was significant difference for R SOM between dry season and rainy season, however, no significant differences were observed for R R and R L (Fig. 4b).

The contribution of R SOM to R S (C SOM) showed a similar pattern to that of C INT, whereas the contribution of R R to R S (C R) showed the opposite pattern, with large contributions during the dry season. The contribution of R L to R S (C L) showed no seasonal variations. Values of C SOM, C R, and C L were respectively 40 ± 4 %, 24 ± 2 %, and 9 ± 3 % in the dry season, and 49 ± 2 %, 11 ± 2 %, and 9 ± 4 % in the rainy season, among which only C R had significant difference between dry season and rain season (Fig. 4d).

In dry season, R INT only had significant difference with R L, while in rainy season and annual time, it had significant differences with not only R L but also R SOM and R R (Fig. 4b). The pattern also applied for C INT that was significant different from C L in dry season, while with C L,C SOM and C R in both rainy season and annual time (Fig. 4d).

Temperature sensitivity

Q 10 showed similar seasonal patterns in CK and NL, with both lowest and highest values in dry season. The lowest values (1.35 ± 0.06 and 1.24 ± 0.03, respectively) appeared in March 2010 while the highest values (2.19 ± 0.07 and 2.06 ± 0.03, respectively) in January 2012. Similar seasonal patterns were also observed in NR and NRNL, with lowest values (1.62 ± 0.01 and 1.63 ± 0.01, respectively) in July 2011, and highest values (2.61 ± 0.15 and 2.71 ± 0.14, respectively) in January 2011 (Fig. 6a). Q 10 values of CK, NL, NR, and NRNL were significantly different between dry season and rainy season (Fig. 6b).

Fig. 6
figure 6

Seasonal variation of Q 10 values of the four treatments, calculated using Eq. (10) for each measurement time point a (mean + SE, n = 3). Comparisons among treatments in dry season average, rainy season average and annual mean b (mean + SE, n = 3): all data passed the normality and homoscedasticity tests before one-way ANOVA analysis; different letters for the same measurement time indicate significant differences (as indicated by LSD’s post-hoc test); *means significant difference between dry and rainy seasons (independent t testes, p < 0.05)

Full size image

Univariate GLM analyses showed no significant effect of interaction between litter decomposition and rhizosphere activity on Q 10 values, but a significant effect of rhizosphere activity on Q 10 (Table 1B). One way ANOVA analyses showed Q 10 values were not different between CK and NL (p > 0.05), but they were significantly higher in NR and NRNL than in CK, in dry season, rainy season, and annual time (Fig. 6b).

Discussion

As shown previously, intact rhizosphere activity is directly linked to litter decomposition (Subke et al. 2004; Subke et al. 2011). In the present study, the interaction between rhizosphere activity and litter decomposition had a significant impact on soil respiration (Table 1A). In terms of annual mean, R INT accounted for 30 % of R S (Fig. 4b). Clearly, our work indicates that R INT plays an important role in determining the dynamics of R S. The mechanism linking litter and rhizosphere is possibly microbial activity (Li et al. 2004; Subke et al. 2004; Feng et al. 2009). Litter and roots provide easily decomposable carbon for microbial growth, which leads to a positive feedback on SOM decomposition (Dighton et al. 1987; Chapela et al. 2001). In addition, we found that R INT and C INT were controlled by soil moisture (Fig. 5), and the lower values we observed in the dry season are perhaps due to competition for soil moisture between roots and rhizosphere microbes (Kuzyakov 2002).

As it shown in Fig. 3, R INT was in the realm of autotrophic respiration if soil respiration was divided into autotrophic and heterotrophic respiration, and was also in the realm of aboveground litter decomposition (R AL) if that was divided into aboveground and belowground respiration. Conclusively, the contribution of SOM to R S which was estimated from above two partition methods as mentioned in introduction would be underestimated.

R INT may include two components: SOM decomposition primed by both litter and rhizosphere activities, and litter decomposition primed by rhizosphere activity alone. Previous studies have shown that rhizosphere activity (Fu and Cheng 2002; Kuzyakov 2002) and litter decomposition (Park et al. 2002; Kalbitz et al. 2007) can cause priming effects on SOM decomposition. Over short timescales (in our case, 2 years) and in small areas (in our case, 1 m2), litter removal treatments cannot be expected to affect forest function (Sayer 2006), and nor can they be expected to influence R R, because R R is driven primarily by photosynthesis (Kuzyakov and Cheng 2001; Wang et al. 2010). However, rhizosphere activity can promote litter decomposition (Subke et al. 2004; Kuzyakov et al. 2007; Subke et al. 2011).

The lack of significant differences in Q 10 values between the CK and NL treatments, combined with the significant differences between the CK and NR (and NRNL) treatments (Fig. 6b), indicate that roots have a strong effect on the Q 10 of soil respiration; this inference was supported by the results of Univariate GLM (Table 1B). The results obtained for the NR treatment are consistent with temperature-quality hypothesis, as shown previously (Fierer et al. 2005; Knorr et al. 2005; Davidson and Janssens 2006; Conant et al. 2008; Hartley and Ineson 2008; Suseela et al. 2013), although some contrary results have also been reported (Giardina and Ryan 2000; Reichstein et al. 2000). Other studies have reported similar temperature dependences for different kinds of SOM (Cox et al. 2000; Fang et al. 2005; Jones et al. 2005), similar to our results for NL. The contrasting results we obtained for NR and NL may be explained in terms of labile carbon. The NL treatment allowed root rhizosphere activity, thereby permitting labile carbon as input into the soil. Although the NR treatment allowed litter input, little labile carbon made its way into the soil because trenching prevented the process of litter decomposition. This explanation is further supported by the fact that we observed no significant interaction between the litter and roots treatments (Table 1B). The influence of interactions between litter and rhizosphere activities on soil respiration have received little attention to date (Subke et al. 2004; Subke et al. 2011), and our study not only demonstrates the importance of such interactions, but also the need for further research to quantify these types of interactions in different ecosystems over longer timescales.

Conclusion

We found a significant positive effect of the interaction between litter decomposition and rhizosphere activity on soil respiration. The annual mean average values for C INT, C SOM, C R, and C L were 30 %, 46 %, 15 %, and 9 %, respectively. R INT and C INT showed seasonal variations driven by soil water content. Our finding of an interaction effect on soil respiration (R INT) suggests a way to identify the contribution of SOM to soil respiration and thereby enables an improved understanding of the carbon cycle on the regional scale. Litter decomposition had no effect on the temperature sensitivity of soil respiration, whereas rhizosphere activity had a strong effect. Together, they had no significant interaction effect on temperature sensitivity. Our results provided solid results to prove the temperature-quality hypothesis and showed a viable way for further studies to gain more insights on global carbon cycle.