Introduction

There is interest in understanding the drivers of the terrestrial carbon (C) budget on the global scale because of the effects of the increase of carbon dioxide (CO2) concentration in the atmosphere on global warming. Soil represents the major sink for C in terrestrial ecosystems, and the rate at which the soil C accumulates depends on the balance between C inputs from vegetation (directly through root exudates or indirectly through root turnover, leaf litter, etc.) and C losses due to CO2 efflux from the soil surface as a consequence of both autotrophic and heterotrophic respiration. In contrast to forest and grassland systems, agricultural systems are often regarded as a net source of atmospheric C (Schulze et al. 2010) for both direct causes (such as soil tillage and high partitioning of assimilated C to harvest products) and indirect causes (such as consequences of land use change and indirect CO2 emissions associated with land management). Soil respiration (Rs) represents a significant fraction of the carbon leaving terrestrial ecosystems, as a result of both root-derived (Ra) and microbial (Rh) respiration. In spite of a relatively extensive literature about respiration of aboveground organs, little attention has been devoted to the study of root respiration in orchards.

Various environmental conditions have been found to affect seasonal variability of Rs, of which soil temperature and moisture are often the most important (Tedeschi et al. 2006; Cook and Orchard 2008): both factors affect the rate at which soil organic matter (SOM) is decomposed by microbes, root growth and root metabolism. SOM quality and soil N availability may also control Rs by affecting the microbial growth and activity, while soil N may affect tree and root metabolism through an indirect effect on the partitioning of photosynthates and a direct effect on root respiration.

A range of methods have been used to partition the soil surface CO2 efflux into autotrophic (Ra) and heterotrophic (Rh) components (e.g., Hanson et al. 2000; Kuzyakov 2006; Subke et al. 2006), each one having specific applications. Trenching and girdling are probably the most common methods used, but their results are difficult to interpret due to the significant disturbance they introduce into the system (Kuzyakov 2006). Isotopic approaches making use of 13 C natural abundance discrimination (Rochette and Flanagan 1997; Fu and Cheng 2002; Millard et al. 2008) cause no disturbance, relying on the different isotopic composition of C3 and C4 plants due to their different photosynthetic pathways. In the absence of a C3:C4 comparison, this method has only been used once (Millard et al. 2010). An alternative approach to trenching or girdling is the root regression technique (Kuzyakov 2006; Rodeghiero and Cescatti 2006). This relies upon differing root densities in the soil, allowing soil surface efflux rate to be correlated with root density. A simple estimation of heterotrophic respiration can be obtained analytically as the y-intercept of the linear regression between soil-surface CO2 efflux and root biomass. Extrapolation back to a zero root density should, in theory, give an estimate of Rh, which can be achieved without the disturbance caused by most other methods. Drip-irrigated apple orchards potentially provide an excellent system in which to use this approach, as root density varies a great deal over short distances (Sokalska et al. 2009).

The aim of this experiment was to identify the role of environmental parameters and root density on soil respiration rates in an apple orchard. In doing so, we have exploited variability in soil and tree conditions by measuring soil respiration in different periods of the year, positions and orchard plots with a different history of soil nutrient supply. The study also aimed to apply the root regression technique to partition Rs into Ra and Rh.

Materials and methods

Experimental conditions

The experiment was carried out in an apple orchard located at the Experimental Station of the Faculty of Agriculture of the University of Bologna in Northern Italy (44°33’N, 11°21’E; 32 m a.s.l.). The orchard soil was a silty clay loam (18% sand, 50% silt, 32% clay), with a pH of 7.3, organic matter content of 2.3% and with 1.2% organic C and 0.13% total N. The wilting point and the water field capacity were 0.18 and 0.41 m3 m−3, respectively. Trees of cv. Gala grafted on M9 rootstocks were planted in winter 1996–1997 at 3.8 × 1.0 m spacing. The plot consisted of 30 rows, divided into blocks of six rows, each of them receiving a different nutrient supply regime since planting. For the measurements 8 tree rows were selected in 4 blocks: four rows (high fertilizer, HF) had been fertilized since 1997 with an annual application of 80 kg N/ha, 25 kg P/ha and 100 kg K/ha, while the remaining four rows (low fertilizer, LF) received half this amount of fertilizer. Nutrients were supplied through fertigation, split into several applications each year. The localized irrigation system used drippers along the tree rows (2.2 l/h) spaced every 40 cm. Soil underneath trees was kept weed-free using non-residual herbicides.

Soil respiration measurements

Measurements of Rs were taken on five separate occasions: 5 December 2006, 16 May, 7 and 27 August 2007 and 1 April 2008. Fruit harvest was carried out between 7 and 27 August 2007. Measurements were made with an EGM 4 gas exchange system (PP Systems, Amesbury, MA, USA), equipped with a soil respiration chamber (SRC-1) and a soil temperature probe (STP-1); soil water content was measured with a Theta Probe ML2x (Delta-t Devices). Soil temperature was measured at 10 cm depth, while volumetric soil water content was measured at 0–10 cm depth. In addition, over the whole experimental period, soil temperature was constantly recorded by a weather station installed in the orchard (Addcon Telemetry). One week before each measurement, plastic collars (11 cm diameter × 6 cm high) were placed into the soil (at 2 cm depth) between two adjacent trees, randomly selected in each tree row, along a line perpendicular to the tree row (Fig. 1). The central collar was placed at 30 cm from the nearest trunk along the row and the other 4 were placed at 30 and 60 cm from it, in opposite directions and perpendicular to the direction of the tree row. Leaf litter, which was only present in small amounts in December, was removed from the soil before placing the collars. The instantaneous measurement of Rs from each collar was obtained at four different times of the day, with each cycle of measurements starting at: 05:00, 10:00, 15:00 and 20:00 hours. Each cycle lasted about 2 h, during which Rs was measured on each individual collar (giving a total of 40 measurements). These instantaneous measurements of Rs from each collar were used to obtain the estimate of the daily average soil respiration. Similarly, instantaneous measurements of soil temperature and soil water content were integrated to obtain daily average values.

Fig. 1
figure 1

Diagrammatic representation of the location of the collars where soil respiration was measured (not to scale)

Full size image

Soil collection and analysis

The day after each Rs measurement cycle, except for 7 August, one soil core of 40 cm depth and 53 mm diameter was collected using a soil column cylinder auger (Ejkelkamp, Netherlands) from the center of each collar (40 soil cores each period). After being removed, the core was divided into three parts according to depth: 0–5 cm, 5–20 cm, and 20–40 cm. One sample of soil of approximately 40–45 g (wet weight) was collected from each layer where no roots were visually present, for subsequent organic C, total N and soil water content determinations. The samples were immediately weighed and dried at 35°C in a ventilated oven for 24 h; a sub-sample for C and N analyses was collected and weighed. A second soil sample was weighed and fully dried at 105°C for 24 h to obtain soil water content data. Organic soil C and total soil N concentrations were determined by an elemental analyzer (elemental analyzer EA1110; Carlo Erba Instruments). Soil NH4 and NO3 concentrations were determined only on samples collected in May and August 2007 from central collars (position 0; Fig. 1) and were obtained through extraction in KCL (2 M) and analyzed by an autoanalyzer (AxFlow AA3; Bran + Luebbe, Norderstedt, Germany). Average values of mineral nitrogen concentrations and soil water content were calculated for the 0–40 cm soil profile underneath the collars. To quantify root density, the remaining soil from each core was separated from roots. The first two layers of soil (0–5 cm and 5–20 cm) were pooled together but kept separated from the 200–400 mm soil layer. Each sample was immersed in a Na2CO3 solution (2 g l−1) to facilitate deflocculation, shaken for 12 h and then sieved under running water. Roots were carefully collected by tweezers and divided into three diameter classes: <2 mm (fine roots), 2–5 mm (medium roots) and >5 mm (coarse roots). Roots were dried in a ventilated oven at 65°C until constant weight was reached, and their weight was determined. Samples of roots collected under the collars present along the tree rows were ground and analyzed for C and N concentration.

Treatment of data and statistics

Seasonal data of soil CO2 flux, total soil organic C and N, root density, soil temperature and moisture were analyzed by two-way analysis of variance (ANOVA) with nutrient supply and collar position as factors. Data of fine root N concentration, and NO3 and NH4 soil concentration obtained from collars in the central position (position 0) were subjected to one-way analysis of variance with nutrient supply as the only factor. Instantaneous measurement data of CO2-flux were analyzed by multiple regression analysis using temperature and soil water content as independent variables.

Daily average Rs data (as dependent variable) were subjected to multiple regression analysis using temperature, soil water content, average soil profile water content, fine, medium and coarse root density and soil N and C concentration as independent variables.

We analyzed the scatterplots and the correlation between Rs and all the above variables to identify the potential regressors to include in the regression analysis. Then, we performed a linear multivariate regression analysis using the stepwise regression method to determine the best fit linear model with the minimum set of independent variables. As scatterplots showed a non-linear behavior for some of the above variables, we performed a non-linear multiple regression. Finally, as some of the variables (like temperature) were influenced by the season, we also performed a multivariate linear and non-linear regression by season to understand whether the influence of the regressors on Rs applied across a year or was confined to a single season.

Results

Soil chemical and physical characteristics

Average soil temperatures differed in the four periods of measurements and ranged from 9°C in December 2006 to 28°C in August 2007 (Fig. 2a). Average temperature in May 2007 and April 2008 were 20 and 14°C, respectively. Collar position had no effect upon soil temperature in December, while in May and August 2007, collars located 30 and 60 cm south showed an average daily temperature more than 2°C higher than the others.

Fig. 2
figure 2

Effect of collar position on the variability of a surface soil temperature, b surface soil water content, c fine root density and d, e) soil respiration. Values are the mean for each position (n = 32, only fine root density n = 8; bars are standard errors of the means). Data of August (e) refer to sampling before (pre) and after (post) harvesting of fruits. For collar positions, see Figure 1

Full size image

Soil water content (Fig. 2b) varied in the four periods of measurement and ranged from 17% (May 2007) to 33% (December 2006). Soil under collars differing for positions differed in soil water content, but this effect was not constant over the whole period: in May and August 2007, when trees were drip-irrigated, collars in position 0 showed higher soil water content values, while lowest values corresponded to collars located to the south. In December 2006 and April 2008, soil water content was uniform among soil positions.

Total soil C and N concentrations were higher in the top 0–5 cm soil layer (14.1 g kg−1 and 1.6 g kg−1, respectively) than in the 5–40 cm layer (9.8 g kg−1 and 1.1 g kg−1, respectively). Both soil C and N concentration were unaffected by either fertilization regime or soil position (data not shown). Soil mineral N was unaffected by the fertilization levels and increased from May (8.3 mg kg−1) to August (16.4 mg kg−1).

Measurements of Rs

Values of Rs were lowest in December 2006 and highest in August 2007 (Fig. 2d, e). The fertilization regime only affected Rs in December 2006 in positions 0 and 60 S, in which soil from HF collars had a 30–36% greater CO2 efflux than LF collars (data not shown). Soil respiration differed according to the position of the collars: in December 2006 and April 2008, the highest efflux was recorded from collars in position 0 (Fig. 2d), while in May and August 2007, the highest CO2 efflux was measured on collars located 30 cm S (Fig. 2d, e). In general, soil positions with the lowest Rs were 30 cm N and 60 cm N. Diurnal variation of Rs occurred only in April 2008, when Rs significantly increased from 10:00 to 15:00 hours (Fig. 3a), as a likely consequence of a temperature increase from 11.0 to 16.5°C (Fig. 3b). Interestling, Rs decreased again from 15:00 to 20:00 hours, even if soil temperature did not vary between these two times of the day.

Fig. 3
figure 3

Daily pattern of soil respiration (a) and surface soil temperature (b); values are mean of all collar positions for each season (n = 40, except for Apr. 2008 where n = 20; bars are standard errors of the means)

Full size image

Root density and root N concentration

Fertilization regime did not affect the density of roots of any size-classes (data not shown). There were significant differences in fine root density among the four measurement periods: highest values (Fig. 2c) were recorded in December 2006, while root density was similar in May and August 2007 and the April 2008 samplings (Fig. 2c). Fine root density changed according to collar position, being highest in the soil along the tree row (position 0) and in that below the collars in 30 cm S (May 2007 and April 2008). The lowest values of root density were recorded 60 cm from the center of the tree row (Fig. 2c). Fine root N concentration (Table 1) averaged 8.7 g kg−1 and was unaffected by the fertilization regime.

Table 1 Soil and root parameters when soil CO2 efflux was measured across the four measurement periods
Full size table

Effects of root and environmental characteristics on Rs

Measurements of Rs were positively and linearly correlated with instantaneous measurements of soil temperature and negatively and linearly correlated with soil water content (Table 2, eq. 1 and 2, respectively). Rs measurements were related to soil temperature (°C) and soil water content (% in volume) by a multiple regression analysis (Table 2, eq. 3). When non-linear regression analysis was performed, an exponential equation produced the best fitting to experimental data (Table 2, eq. 4).

Table 2 Relationships between soil respiration (g CO2 m−2 h−1) and measured parameters
Full size table

The variations of daily average soil CO2 efflux were studied in relation to environmental factors and root biomass. When multiple linear regression analysis was applied to the whole dataset of daily soil CO2-efflux, average soil temperature alone explained 26% (Table 2, eq. 5) of the variability of CO2 efflux data; considering fine root density in addition to average soil temperature improved the goodness of fit of the model (Table 2, eq. 6). Non-linear multiple regression analysis on the whole dataset again indicated soil temperature and fine root density as major determinants of daily average values of soil CO2 efflux (Table 2, eq. 9).

The best fit of daily average Rs versus soil temperature using non linear regression was obtained by an exponential model (R 2 = 0.71; Fig. 4b). Surface soil water content also related to Rs by a second order curve (Table 2, eq. 8; Fig. 4a). As surface soil temperature data were negatively correlated with surface soil water content data with r = −0.77 (P = 0.01, not reported in tables), the models to predict daily average Rs using the two variables together were discarded.

Fig. 4
figure 4

Soil respiration in relation to a surface soil water content, b surface soil temperature and c fine root density. Each point represents the mean value for each position (n = 32, except for April 2008 when n = 16)

Full size image

As surface soil temperature was the major environmental parameter varying among the four measurements periods, we tested whether the effect of temperature on Rs was confirmed within each period or if other factors explained more of the variability in the data. In December 2006, fine root density and soil water content explained 48% of the Rs value (Table 3). In May and August 2007, soil temperature was the only factor affecting Rs, while in April 2008 Rs depended upon soil water content (Table 3).

Table 3 Multiple linear regression analysis applied to daily average soil CO2 flux (g CO2 m−2 h−1) within each measurement period
Full size table

To assess the heterotrophic-derived component of soil respiration, Rs was related to the density of roots of different size-classes. Only fine root density was related to Rs and this relationship was only significant in December 2006 (Table 4), when the rate of Rh was estimated being 0.096 g CO2 m−2 h−1.

Table 4 Coefficients of determination, significance and equations for the linear regression analysis of data of daily average soil CO2 flux (y in g CO2 m−2 h−1) versus fine root density (x in kg m−3), in soil cores excavated at the center of the collar after respiration measurements
Full size table

Discussion

As soil respiration determines the major fluxes of CO2 in many terrestrial ecosystems (Law et al. 1999; Bolstad et al. 2004; Borken et al. 2006; Euskirchen et al. 2006), it is of great importance to understand to what extent Rs is affected by environmental drivers and the activity of the trees (Knorr et al. 2005). In our study, Rs was measured under a relatively high range of soil temperatures, soil water contents and an even greater range of root densities (Table 1). In contrast to other studies which have quantified Rs across a range of sites with different characteristics (e.g., Maekiranta et al. 2008; Martin et al. 2009), the variation in soil C and soil N concentration in the soil under our collars was relatively low (C.V. = 13 and 8%, respectively; Table 1).

Instantaneous Rs data showed a large temporal and spatial variation due to the interactive effects of time of the day, the period of the year and the position of the collars in the soil, ranging from 0.06 g CO2 m−2 h−1 in December 2006 to 1.49 g CO2 m−2 h−1 in the south-exposed soil areas in August 2007 (data not shown). The fact that Rs was generally slower in the north exposed rings was probably due to their lower fine root density, cooler soil temperature and higher soil water content (Fig. 2a–c). Our minimum values fall in the range reported by Davidson et al. (1998) for temperate mixed hardwood forest and by Law et al. (1999) for a ponderosa pine plantation. For apple trees, the only available data refer to the studies by Blanke (1996, 1998) who measured soil respiration in the weed-free strip underneath apple trees and concluded that maximum soil respiration can reach values as high as 3.8 g CO2 m−2 h−1, which, he concluded, exceed values of leaf photosynthesis in the corresponding periods.

The regression analysis applied to the whole dataset revealed that soil temperature was the environmental parameter accounting for most of the variance in Rs, with instantaneous measurements less well correlated to soil temperature than were daily average values. This effect was likely explained by diurnal variation of soil temperatures, resulting in similar temperatures being recorded at different periods of the year at different times of the day. Temperature affects almost all aspects of respiration processes (Luo and Zhou 2006) being either root-dependent or caused by the activity of heterotrophic soil organisms. Exponential models better explain the variation of soil respiration to soil temperature. As shown in Fig. 4a, b, Q10 of 2.4 for soil respiration was found in our experiment, which is slightly higher than that reported in a pine plantation under a similar range of temperature (Q10 = 1.8), lower than the value of 3.1 reported by Boone et al. (1998), but equal to the estimated global median value reported by Raich and Schlesinger (1992). To estimate yearly soil CO2 flux, we used its exponential relationship with soil temperature obtained by the weather station, which from 2006 to 2008 averaged 17.6°C. Simulation indicated a total flux of 2.5 kg CO2 m−2 year−1, a value slightly higher than that reported by Liguori et al. (2009) in a citrus orchard soil. Scaling up our data to the orchard soil area underneath the weed-free strips in the centre of tree rows, where most apple roots resides, suggests that some 3.2 t C ha−1 were lost by soil respiration annually; this value is lower than that of aboveground primary productivity of apple trees measured in the same experimental orchard in its sixth year after planting (Scandellari et al. 2010).

As a predictor, fine root density significantly improved the goodness of the prediction, when associated to temperature in non-linear models, but did not improve the linear model (Table 2). The other parameters measured (medium- and large-sized root density, soil nitrogen concentration and the average of soil water content over the 0–40 cm profile) had no significant effect on Rs.

Soil water content negatively correlated to instantaneous values of Rs (Table 2, eq. 2) and, when associated to soil temperature, it increased the ability to predict soil respiration under multiple linear regression analysis compared to soil temperature alone. Xu and Qi (2001) have shown that the relationship between Rs and soil temperature is affected by soil water content, being significantly lower under soil water limitation. In our experiment, water limitations rarely occurred because the orchard was irrigated (Table 1). When we analyzed the factors affecting soil respiration within each single period of the year (Table 3), we found that soil temperature represented a significant predictor in May and August 2007. In contrast, in December 2006, fine roots and soil water content, together, explained almost half of the variability of soil respiration, while in April 2008, only soil water content related to soil respiration. In December and April, soil temperature was relatively cool and showed limited variations among collars (Fig. 2a). The inverse correlation between surface soil temperature and surface soil water content using the whole dataset might suggest that the negative effect of soil water content on Rs was largely due to the positive effect of temperature. As the effects of soil water content were significant in periods when soil temperature showed a limited variation across the soil positions (Table 3; Fig. 2a), the data suggest that soil water content effects were independent from soil temperature. Soil respiration often increases with soil water content from values at wilting point until values of approximately 50–60% of water-filled pore space are reached (Bryla et al. 2001), after which Rs decreases if the soil becomes wetter, due to a limitation in soil aeration and a depression of microbial activity (Linn and Doran 1984; Mielnick and Dugas 2000). Data in Fig. 4a are in line with those reported by Mielnick and Dugas (2000) who found a negative trend of Rs at soil water content data above 25% of volumetric soil water content. Low Rs values in dry soils are likely associated with reduced root growth (Espeleta and Eissenstat 1998) and reduced ion-uptake (Eissenstat et al. 1999).

On an annual scale, the inability to use fine root density to predict Rs is likely because fine root metabolism varies during the season, possibly as a function of their N concentration. Bouma et al. (2001) have shown that fine roots reach their maximum respiration rates in their early developmental phases, which soon declines. The periods of more intense production of new fine roots have been studied using minirhizotrons, but even in the same location data are not very consistent, showing either spring or summer peaks (Psarras et al. 2000; Eissenstat et al. 2006). The highest fine root density, at least from positions 60 cm S and 0, were measured at the end of autumn (Fig. 2c), suggesting a cumulative root production over the season.

The history of nutrient supply to the orchard soil had a small effect on soil respiration only in December 2006. Previous studies carried out in the same experimental orchard indicated that nutrient supply stimulated tree growth and enhanced tree N status when the trees were young (Rombolà et al. 2000). However, when the trees reached the adult phase, different nutrient supplies only slightly affected leaf N concentration (averaging 2.5 and 2.3% in HF and LF trees, respectively) and had no effect on leaf P and K, fruit yield and vegetative growth (Malaguti et al. 2006). Total soil N might not really reflect the N availability for root uptake, but it is also possible that several years of localized application of two levels of N did not markedly modify either total soil N or soil mineral nitrogen. The literature on the effect of soil and root N availability on Rs draws different scenarios, depending on the time scale of the observations. In the short period after N addition, an initial increase of Rs, due to a priming effect on the microbial component and a consequent enhanced litter decomposition can be expected (Söderström et al. 1983; Fog 1988; Berg and Matzner 1997; Bowden et al. 2004), while increased long-term N availability tends to cause a decline in Rs due to both, the increase of recalcitrant soil organic matter compounds and the reduction of microbial activity (Fog 1988; Berg and Matzner 1997; Bowden et al. 2004; Burton et al. 2004). In a recent paper, Migliavacca et al. (2010) suggested that on the whole an increased total N deposition in forests as well as in any anthropogenic managed ecosystem tends to reduce soil respiration. This tendency has also been observed by others (Bowden et al. 2004) and can in part be explained by a reduction in mycorrhizal colonization and fungal development under nutrient-rich soil conditions (Bryla and Eissenstat 2005). In trees, the long-term effect of soil N availability usually leads to a reduction in C allocation to the belowground plant organs in favor of aboveground organs (Vogt et al. 1990; Haynes and Gower 1995), but not necessarily to a reduction of total amount of photosynthates allocated to the roots. Roots with higher N availability usually show higher specific respiration rates (Burton et al. 1996, 2002; Ryan et al. 1996).

Assessing the relative contribution of the heterotrophic and autotrophic components of soil respiration is an important goal in ecological studies as each one can differentially respond to soil climate, nutrient availability and management practices (Maekiranta et al. 2008). With the exception of December 2006 and, to a less extent, of April 2008, daily average soil temperature varied among soil collars (Fig. 2a). Considering that temperature has a major effect on root activity and its respiration, it is not surprising that we could only detect a significant linear relationship between fine root density and Rs in December, when soil temperature values were relatively constant. Several other studies have used a regression between Rs and fine root biomass to estimate Rh, by calculating the y-intercept. In a ponderosa pine plantation, Xu et al. (2001) estimated that Rh contributed 53% of Rs during the growing season; while in tropical deciduous forest, Behera et al. (1990) calculated that Rh contributed 49% of Rs. In another study, Rodeghiero and Cescatti (2006) estimated that Rh accounted for 42–84% of annual Rs in a range of evergreen forests. When we applied the root regression technique to our December 2006 data (Fig. 4c) to assess the contribution of Rh, we found that it contributed to some 65% of total Rs, the remaining 35% being attributed to Ra. Root regression has been critically evaluated as an approach to measure Rh. Kuzyakov (2006) pointed out that, compared with other techniques, high numbers of replicates are needed; there can be overestimation of root-derived CO2 when big and old roots are present, and that frequently there is a poor correlation between root amount and total CO2 with Rh estimates being biased by considerable extrapolation of the regression line. A clear limitation of the present study was our inability to estimate Rh when the trees were actively growing, when the overall contribution of Rh to Rs would likely have been smaller.

The increase of Rs values in August 2007 in the post-harvest period, as compared to the pre-harvest period, is likely associated to fruit removal (Fig. 2e), as surface soil temperature and surface soil water content were respectively lower and higher in post-harvest compared to pre-harvest (Fig. 2a, b). Fruits are a major sink for carbohydrates (Wardlaw 1990), and their removal results in a shift of carbohydrate allocation to the root system. This has been already reported in a study by Glenn and Welker (1993), who observed new fine root production immediately after fruit removal in mature peach trees. This suggests that the increase in CO2 efflux after fruit removal is to assign to the autotrophic, rather than the heterotrophic, component of Rs.

In conclusion, our study demonstrated that, on an annual scale, surface soil temperature was the most important environmental factor affecting respiration of the orchard soil. Surface soil water content showed an adverse effect on soil respiration and controlled a great part of the variability of soil respiration measured in December and April. The history of soil fertilization, soil C and N concentration, as well as root N and soil mineral N had negligible or no effect on soil respiration. Fine root density, but not medium- and large-sized root density, contributed to explain part of the yearly variability of soil respiration, and proved to be a good predictor of soil respiration in December; in this period only, autotrophic component of soil respiration could be estimated by the root regression technique and represented roughly 35% of total soil respiration.