Soil CO2 efflux in an old-growth southern conifer forest (Agathis australis) – magnitude, components and controls

Total soil CO2 efflux and its component fluxes, autotrophic and heterotrophic respiration, were measured in a native forest in northern Aotearoa–New Zealand. The forest is dominated by Agathis australis (kauri) and is on an acidic, clay rich soil. Soil CO2 efflux, volumetric soil water content and soil temperature were measured bi-weekly to monthly at 72 sampling points over 18 months. Trenching and regression analysis was used to partition total soil CO2 efflux into heterotrophic and autotrophic respiration. The effect of tree structure was investigated by calculating an index of local contribution (Ic, based on tree size and distance to the measurement location) followed by correlation analysis between Ic and total soil CO2 efflux, root biomass, litterfall and soil characteristics. The measured mean total soil CO2 efflux was 3.47 μmol m−2 s−1. Autotrophic respiration accounted for 25 % (trenching) or 28 % (regression analysis) of total soil CO2 efflux. Using uniand bivariate models showed that soil temperature was a poor predictor of the temporal variation in total soil CO2 efflux (< 20 %). In contrast, a stronger temperature sensitivity was found for heterotrophic respiration (around 47 %). We found significant positive relationships between kauri tree size (Ic) and total soil CO2 efflux, root biomass and mineral soil CN ratio within 5–6 m of the sampling points. Using multiple regression analysis revealed that 97 % of the spatial variability in total soil CO2 efflux in this kauri-dominated stand was explained by root biomass and soil temperature. Our findings suggest that biotic factors such as tree structure should be investigated in soil carbon related studies.


Introduction
Soil CO 2 efflux (soil respiration) is the largest CO 2 flux from terrestrial ecosystems into the atmosphere (Raich and Potter, 1995;Janssens et al., 2001a;Bond-Lamberty and Thomson, 2010a).Quantifying the magnitude of soil CO 2 efflux and examining the spatial and temporal heterogeneity is critical in characterising the carbon (C) dynamics in terrestrial ecosystems (Schlesinger and Andrews, 2000;Trumbore, 2006;Smith and Fang, 2010) as even a small change in soil CO 2 efflux could have a strong impact on atmospheric CO 2 concentration (Andrews et al., 1999;Rustad et al., 2000).Advancing the understanding of soil CO 2 efflux and its driving factors is also important to predict the effects of land-use conversion and climate change on the net C sink of the terrestrial biosphere (Giardina et al., 2014).
Soil CO 2 efflux varies widely in space and time according to changes in various abiotic and biotic factors.Soil temperature is often the main abiotic factor explaining the temporal variability of soil CO 2 efflux (Raich and Schlesinger, 1992;Jassal et al., 2005;Bond-Lamberty and Thomson, 2010b).Many studies show a positive correlation between soil temperature and soil CO 2 efflux (Reich and Schlesinger, 1992;Lloyd and Taylor, 1994;Rustad et al., 2000).This relationship is often expressed as a Q 10 function (relative increase in soil CO 2 efflux rate per 10 • C difference) (van't Hoff, 1898) or modified Arrhenius function (Lloyd and Taylor, 1994).However, other abiotic factors have been found to influence Published by Copernicus Publications on behalf of the European Geosciences Union.
the temporal and spatial variation in soil CO 2 efflux.For example, several studies have shown a parabolic relationship between soil water content and soil CO 2 efflux with the highest soil CO 2 efflux occurring at an intermediate soil water content (Davidson et al., 1998(Davidson et al., , 2000;;Schwendenmann et al., 2003).Other soil factors driving the spatial variability in soil CO 2 efflux in forest ecosystems include the quality and quantity of soil organic matter (Rayment and Jarvis, 2000;Epron et al., 2004) and microbial biomass (Xu and Qi, 2001).
Biotic factors that influence rates of soil CO 2 efflux include plant and microbial components.Vegetation type and structure are important determinants of soil CO 2 efflux because they influence the quantity and quality of litter and root biomass supplied to the soil and they also mediate the soil microclimate (Fang et al., 1998;Raich and Tufekcioglu, 2000;Metcalfe et al., 2007Metcalfe et al., , 2011)).For example, litter addition experiments have shown that increasing litterfall enhances soil CO 2 efflux (Sulzman et al., 2005;Sayer et al., 2011).A few studies have investigated the effect of stand structure and tree size on soil CO 2 efflux in temperate (Longdoz et al., 2000;Søe and Buchmann, 2005;Ngao et al., 2012) and tropical forests (Ohashi et al., 2008;Katayama et al., 2009;Brechet et al., 2011).Those findings demonstrate that the spatial distribution of emergent trees strongly affects the root distribution and litterfall, partly explaining the spatial variation of soil CO 2 efflux (Katayama et al., 2009;Bréchet et al., 2011).Some studies show that soil CO 2 efflux at the base of emergent trees is significantly higher than soil CO 2 efflux at greater distances from the trees (Katayama et al., 2009;Ohashi et al., 2008).
Soil CO 2 efflux is the result of CO 2 production by heterotrophic and autotrophic respiration and gas transport (Fang and Moncrieff, 1999;Maier et al., 2011;Maier and Schack-Kirchner, 2014).Heterotrophic respiration mainly originates from microbes decomposing plant detritus and soil organic matter while autotrophic (= root/rhizosphere) respiration comes from plant roots, mycorrhizal fungi and the rhizosphere (Hanson et al., 2000;Bond-Lamberty et al., 2011).The relative contribution of autotrophic respiration to total soil CO 2 efflux varies widely (10-90 %) depending on the type of ecosystem studied (Hanson et al., 2000;Subke et al., 2006;Bond-Lamberty et al., 2011).Various methods (i.e.trenching, regression analysis, isotopic methods) have been developed to separate heterotrophic and autotrophic respiration under both laboratory and field conditions and are described in the review papers by Hanson et al. (2000), Kuzyakov (2006) and Bond-Lamberty et al. (2011).Separating total soil CO 2 efflux into autotrophic and heterotrophic sources is important to more accurately predict C fluxes under changing environmental conditions as heterotrophic and autotrophic respiration respond differently to abiotic and biotic factors (Boone et al., 1998;Davidson et al., 2006;Brüggemann et al., 2011).For example, heterotrophic respiration was found to be more susceptible to seasonal drought in a Pinus contorta forest (Scott-Denton et al., 2006).Other studies showed that autotrophic respiration is more temperature-sensitive than heterotrophic respiration and total soil CO 2 efflux (Boone et al., 1998;Högberg, 2010).
Soil CO 2 efflux has been measured in a wide range of mature and old-growth forests across the globe (Schwendenmann et al., 2003;Epron et al., 2004;Sulzman et al., 2005;Adachi et al., 2006;Bahn et al., 2010;Bond-Lamberty and Thompson, 2014).Southern conifer forests are an exception to this (but see Urrutia-Jalabert, 2015).They include kauri (Agathis australis D. Don Lindl.ex Loudon, Araucariaceae) forests in Aotearoa-New Zealand.Old-growth kauri forests are considered to be one of the most C-dense forests worldwide (Keith et al., 2009) with up to 670 Mg C ha −1 in living woody biomass (Silvester and Orchard, 1999).Kauri is endemic to northern New Zealand (north of latitude 38 • S) (Ecroyd, 1982) and is the largest and longest living tree species in the country.Kauri has significant effects on the soil environment (Whitlock, 1985;Verkaik et al., 2007) and plant community composition (Wyse et al., 2014).Phenolic compounds in kauri leaf litter (Verkaik et al., 2006) and low pH values (around 4) (Silvester, 2000;Wyse and Burns, 2013) partly explain the slow decomposition rates of kauri litter (Enright and Ogden, 1987) which result in thick organic layers in undisturbed kauri stands (Silvester and Orchard, 1999).
Organic layers (= forest floor composed of leaves, twigs and bark in various stages of decomposition above the soil surface) are important C reservoirs (Gaudinski et al., 2000) and can be a considerable source of CO 2 .Organic layers can also contain a large amount of roots which may result in increased soil CO 2 efflux (Cavagnaro et al., 2012).Mature kauri trees have an extensive network of fine roots which extends from the lateral roots into the interface between organic layer and the mineral soil (Bergin and Steward, 2004;Steward and Beveridge, 2010).A recent study also showed that roots and root nodules of kauri harbour arbuscular mycorrhizal fungi (Padamsee et al., 2016).Roots colonized by mycorrhizal fungi have been found to release more CO 2 than non-mycorrhizal roots (Valentine and Kleinert, 2007;Nottingham et al., 2010).
It remains unknown, however, how much soil CO 2 is released from these C-rich southern conifer forests and which factors are driving the temporal and spatial variability in soil CO 2 efflux.It has been shown that kauri has a significant influence on soil properties but the influence of kauri tree distribution on soil carbon related ecosystem processes is untested.Quantifying the magnitude of soil C loss and identifying the controls of this significant C flux are essential for the assessment of the C balance of these C-rich and longlived forest stands.
The aim of this study was to determine the magnitude, components and the driving factors of soil CO 2 efflux in an old-growth southern conifer forest.The specific objectives of our study were (i) to quantify total soil CO 2 efflux, (ii) to test the effect of collar insertion depth on soil CO 2 efflux, (iii) partition total soil CO 2 efflux into autotrophic and heterotrophic respiration, (iv) to identify the factors controlling the temporal variation of total soil CO 2 efflux and its component fluxes, and (v) to test the effect of kauri tree size and distribution on total soil CO 2 efflux and soil properties.We used direct (trenching) and indirect (regression technique) approaches to partition total soil CO 2 efflux into the autotrophic and heterotrophic components.Given that oldgrowth kauri forests are often characterised by thick organic layers, deep collars were deployed to assess the effect of insertion depth on total soil CO 2 efflux and to quantify the proportion of autotrophic and heterotrophic respiration in this layer.

Study site
The study was conducted in the University of Auckland Huapai reserve.The reserve is a 15 ha forest remnant surrounded by farmland (Thomas and Ogden, 1983) and is located approximately 25 km west of central Auckland on the northern fringe of the Waitakere Ranges (36 • 47.7 S, 174 • 29.5 E).Within the long-term research plot (50 × 40 m), the diameter at breast height (DBH) of all trees ≥ 2.5 cm was measured, the species were identified and their location mapped (Wunder et al., 2010) (Fig. 1).The plot is dominated by kauri (770 stems ha −1 ) with a basal area of 75 m 2 ha −1 , equating to approximately 80 % of the stand basal area (Wunder et al., 2010).Kauri tree size distribution differs within the plot.Four emergent kauri trees (up to 180 cm in DBH, ∼ 300 year old) are found on the upper slope of the plot.At the lower slope tree fall and removal of five large kauri trees in the 1950s created gaps which are now dominated by a cohort of younger kauri trees.Silver ferns (Cyathea dealbata) are also highly abundant (785 stems ha −1 ) (Wunder et al., 2010).Less-numerous species are a mixture of podocarps and broadleaved species, including Phyllocladus trichomanoides, Myrsine australis, Coprosma arborea and Geniostoma ligustrifolium.
L. Schwendenmann and C. Macinnis-Ng: Soil CO 2 efflux in an old-growth southern conifer forest Total annual rainfall, measured from 2011 to 2013 at a weather station located in the vicinity of the reserve, is approximately 1200 mm with 70 % occurring during austral winter (June-August).Annual mean temperature is 14 • C (Macinnis-Ng and Schwendenmann, 2015).The soils are derived from andesitic tuffs and are classified as Orthic Granular Soils (New Zealand soil classification;Hewitt, 1992) or Humults (US soil classification;Soil Survey Staff, 2014).The clayey soil is fairly sticky when wet, and hard and fragile when dry (Thomas and Ogden, 1983).The thickness of the organic layer varies between 5 and 15 cm and consists mainly of partly decomposed kauri leaves and twigs.

Experimental setup
The long-term research plot was subdivided into six equal quadrats.Within each quadrant two soil CO 2 efflux sampling points (in total 12) were randomly located (Fig. 1).For each sampling point we measured the distance to the closest tree with a DBH ≥ 2.5 cm.At each of these 12 sampling points, a cluster of measurements was made.There was one surface measurement and three inserted measurements as described below.
Total soil CO 2 efflux was measured on the surface of the forest floor by gently pressing a polyvinyl chloride (PVC) ring attached to the soil respiration chamber (see below for details) down on the forest floor during measurements to avoid cutting fine roots.The sampling points were marked with flags and kept free of vegetation.Total soil CO 2 efflux was measured over 18 months from August 2012 to January 2014 at each location.These sampling points were named Plot_Surface.
To measure the effect of collar insertion depth and to quantify the proportion of autotrophic and heterotrophic respiration to total soil CO 2 efflux in the organic layer, a cluster of three "deep" PVC collars (10 cm in diameter, 20 cm in height) was inserted next to each sampling point for surface soil CO 2 efflux measurements.Three collars per cluster were spaced evenly around the circumference of a circle 2 m in diameter, with small adjustments in the spacing to avoid large roots.Each collar was driven right through the organic layer and 1-2 cm into the mineral soil layer to cut off the roots growing in the organic layer.In order to prevent CO 2 uptake, any vegetation inside the collars was regularly removed.The thickness of the organic layer at each grid point was measured using a ruler outside each collar.The deep collars were inserted in November 2011 and left in place over the measurement period.Efflux was measured from August 2012 (9 months after insertion) to January 2014.Here after, these sampling points are known as Plot_Inserted.
We used the trenching approach to separate heterotrophic and autotrophic respiration in the organic layer plus mineral soil to 30 cm depth.To avoid disturbing the long-term research plot the trenching experiment was set-up directly adjacent (upslope) to the research plot.In July 2012, six 2 × 2 m plots were trenched to 30 cm depth based on a preliminary study showing that the majority of fine roots (over 80 %) are located in the organic layer and top 30 cm of the mineral soil.The trenches were double-lined with a water permeable polypropylene fabric and backfilled.During trenching, trampling and disturbance inside the 2 × 2 m plots were avoided as far as possible.Two types of measurements were conducted.First, total soil CO 2 efflux was measured at two sampling points outside each trenched plot (Outside_Trench_Surface, n = 12) in the same way as the Plot_Surface samples were measured (see above).Second, two collars were randomly placed inside the trenched plots (Trench_Inserted).The collars were inserted 1-2 cm into the mineral soil layer (deep collars) as described above.Soil CO 2 efflux was measured bi-weekly to monthly from August 2012 until December 2013.

Soil CO 2 efflux measurements
Soil CO 2 efflux was measured with a portable infrared gas analyser (EGM-4, PP Systems, Amesbury, MA, USA) equipped with a soil respiration chamber (SRC-1, PP Systems, Amesbury, MA, USA).The CO 2 concentration was measured every 5 sec over 90-120 sec between 09:00 and 02:00 LT (local time) and the change in CO 2 concentration over time was recorded.Diurnal soil CO 2 efflux measurements conducted in January 2013 showed that soil CO 2 efflux rates between 09:00 and 02:00 LT were comparable as there was no significant diurnal trend (data not shown).
Soil CO 2 efflux (µmol m −2 s −1 ) was calculated as follows (Eq.1): where CO 2 / t is the change in CO 2 concentration over time (t), calculated as the slope of the linear regression (µmol mol −1 s −1 = ppm s −1 ); P is the atmospheric pressure (Pa), V is the volume of the chamber including collar (m 3 ), R is the universal gas constant, 8.314 m 3 Pa K −1 mol −1 ), T is the temperature (K), and A is the surface area of ground covered by each chamber (0.007854 m 2 ).
Soil temperature (Soil temperature probe, 10 cm probe, Novel Ways Ltd, Hamilton, New Zealand) and volumetric soil water content (Hydrosense II, 12 cm probe, Campbell Scientific Inc., Logan, UT, USA) were measured at the same time next to each of the collars.

Litterfall, root and soil characteristics
Litterfall (including leaves, twigs, fruits, flowers, cone scales, etc.) was collected from 12 litter traps (pop-up planters, 63 cm in diameter) located next to each soil CO 2 efflux cluster within the long-term research plot (Fig. 1).Litterfall was collected bi-weekly from January 2012 to January 2014, dried at 80 • C to constant mass, sorted and weighed (Macinnis-Ng and Schwendenmann, 2015).Organic layer and mineral soil samples (0-10 cm depths) were taken next to each collar with a core sampler in November 2011 (research plot) and July 2012 (trenched locations).Samples were ground and analysed for total C and N concentration using an elemental analyser (TruSpec, LECO Corporation, St. Joseph, Michigan, USA).Soil (LECO Lot 1016, 1007) and leaf (NIST SRM 1515 -Apple Leaves) standards were used for calibration.The coefficient of variation was 0.5 % for C and 1 % N for plant material (45 % C, 2.3 % N) and 1 % for C and N for soil (2-12 % C, 0.2-1 % N). 10 % of samples were replicated and results were within the range of variation given for the standards.
Organic layer and mineral soil samples (0-15, 15-30 cm) were collected for soil analysis and root biomass estimation adjacent to six clusters within the plot (Fig. 1) and the trenched plots.Organic layer samples were collected from 20 cm × 20 cm quadrats.Mineral soil samples were taken using a 15 cm diameter steel cylinder.Samples were dried at 60 • C (forest floor) and 40 • C (mineral soil).Mineral soil samples were sieved at 2 mm.pH was measured in a 1 : 2.5 soil-water suspension (SensION 3 pH meter, HACH, Love-land, CO, USA).The organic layer samples were wetted and fine roots were manually picked with tweezers.The flotation method was used to separate roots from the clay rich mineral soil.Roots were dried at 60 • C to constant mass and weighed by size class (fine roots: < 2 mm, and small (coarse) roots: 2-20 mm).Litterfall, root and soil data are summarized in Table 1.

Data analysis
Efflux, soil temperature and volumetric soil water content measured at the three individual deep collars per cluster (Plot_Inserted) and the two samples of Out-side_Trench_Surface and Trench_Inserted were averaged before statistical analysis.Normality of the data distribution was examined using a Kolmogorov-Smirnov test.
Descriptive statistics (minimum, maximum, mean and median values, standard deviation, standard error, coefficient of variation) were used to describe soil CO 2 efflux, soil temperature and volumetric soil water content.Differences between total soil CO 2 efflux between treatments (Plot_Surface vs. Plot_Inserted; Outside_Trench_Surface vs. Trench_Inserted) and seasons were tested using a mixed model where deep collar insertion and trenching were considered as a fixed effect and sampling dates as a random effect.
To estimate annual total soil CO 2 efflux, a linear temperature function (see below) and continuous half-hourly soil temperature measurements (plot centre, 10 cm depth, 107 temperature probe, Campbell Scientific, Logan, UT, USA) were used.
Two methods (trenching and regression analysis) were used for partitioning of total soil CO 2 efflux.In the trenching approach, the Plot_Inserted and Trench_Inserted sampling points represents heterotrophic respiration in the organic layer and organic layer plus mineral soil to 30 cm depth, respectively.Measurements from the soil surface (Plot_Surface and Outside_Trench_Surface) represent total soil CO 2 efflux.Autotrophic respiration in the organic layer and organic layer plus mineral soil to 30 cm depth was calculated as the difference between total soil CO 2 efflux and the efflux measured from the Plot_Inserted and Trench_Inserted sampling points, respectively.Heterotrophic respiration from Plot_Inserted and Trench_Inserted sampling points was not corrected for decomposing root-derived CO 2 efflux.For the regression-analysis approach (organic layer plus mineral soil to 30 cm depth), the heterotrophic respiration was derived analytically as the y-intercept of the linear regression between root biomass (independent variable) and total soil CO 2 efflux (dependent variable) (Kucera and Kirkham, 1971;Kuzyakov, 2006).Autotrophic respiration was then estimated by subtracting the heterotrophic respiration from total soil CO 2 efflux.
Univariate and bivariate models were used to investigate the relationship between total soil CO 2 efflux, heterotrophic and autotrophic respiration and the abiotic factors soil temperature and volumetric soil water content.Data from within the research plot and trench sampling points were combined.The temperature response of soil CO 2 efflux was tested using a linear, exponential (Q 10 , van't Hoff, 1898) and modified Arrhenius function (Lloyd and Taylor, 1994).Linear and quadratic functions were used to assess the soil water dependence of soil CO 2 efflux.The combined effect of soil temperature and soil water content on soil CO 2 efflux was tested using a polynomial function.Coefficient of determination (R 2 ), standard error of estimate (SEE), and Akaike Information Criterion (AIC) were used to evaluate model performance.The analysis was conducted using Sigma Plot (Version 13, Systat Software Inc., Chicago, IL, USA).
The influence of kauri tree size and distribution on total soil CO 2 efflux, litterfall, root biomass and soil properties was tested using an index of local contribution (I c ).As described in Bréchet et al. (2011), the I c index is a function of (1) the trunk cross section area (S, m 2 ) and (2) the distance (d, m) of kauri trees from the sampling points.The following functions were tested: uniform, I c = S; linear, where a is a coefficient of form and r is a fitted radius of influence (r, in m) (Bréchet et al., 2011).It was assumed that all kauri trees had the same radius of influence (r, i.e. the distance above which their contribution would become negligible).The coefficient of determination was used to assess the strength of the relationships between litterfall, root biomass or soil CO 2 efflux and the sum of the I c .
The spatial variability in soil CO 2 efflux was quantified at the plot scale using the coefficient of variation.Multiple regression analysis was used to assess the spatial controls (soil temperature, soil moisture, organic layer thickness, soil C and N, root biomass) of total soil CO 2 efflux.
Statistical analyses were performed using SPSS v. 22 (IBM SPSS Statistics, IBM Corporation, Chicago, IL, USA).The local contribution analysis (I c ) was conducted using R (v3.1.0,R Core Team, 2014).Significance for all statistical analyses was accepted at p < 0.05.Data are available in the figshare data repository (Schwendenmann and Macinnis-Ng, 2016).

Results
3.1 Seasonal variations in soil CO 2 efflux, soil temperature and volumetric soil water content and the effect of deep collar insertion and trenching During the study period, soil temperature and volumetric soil water content (SWC) varied with season (Fig. 2b and c).Soil temperatures peaked at about 17 • C during austral summer and early autumn (January-March) while minimum temperatures of around 11 • C (Fig. 2b) were measured in late winterearly spring (July-September).Annual mean soil temperature was 14.2 ± 0.1 • C (Table 1).The SWC was highest during late winter-early spring (July-September) with values of 55 % and soil was driest during late summer-early autumn (March-May) with around 25 % (Fig. 2c).Annual average was 43.9 ± 0.9 % (Table 1).Across the study period, an average of 1.9 ± 0.1 kg m −2 litter fell at the sampling points and the organic layer was 8.8 ± 0.9 cm thick (Table 1).Other information on soil and vegetation characteristics is summarised in Table 1.Total soil CO 2 efflux rates (Plot_Surface) measured at 12 sampling points within the research plot varied from 0.7 to 9.9 µmol CO 2 m −2 s −1 during the 18-month study period.Total soil CO 2 efflux was positively skewed with the mean larger than the median (Table 2).The mean total soil CO 2 efflux (± SE), averaged over the 12 plot sampling points and all sampling dates, was 3.6 ± 0.1 µmol CO 2 m −2 s −1 .Higher efflux rates were measured during austral summer and early autumn (December-March, 2.7-4.7 µmol CO 2 m −2 s −1 ) compared to winter (June-August, 1.8-3.9µmol CO 2 m −2 s −1 ) (Fig. 2a).However, differences among seasons were not significant (p > 0.05).In contrast, soil temperature differed significantly between summer (16.5 • C) and winter (11.8 • C).We also detected significant seasonal differences in SWC with drier conditions during summer (mean SWC = 31 %) compared to winter (mean SWC = 47 %).Deep collar insertion (Plot_Inserted) had a significant effect on total soil CO 2 efflux but no effect on soil temperature and SWC was found (Table 2).Soil CO 2 efflux from inserted collars (3.0 ± 0.1 µmol CO 2 m −2 s −1 ) was 17 % lower compared to total soil CO 2 efflux (3.6 ± 0.1 µmol CO 2 m −2 s −1 ) (Table 2).The overall temporal pattern of soil CO 2 efflux was similar between inserted and surface collars (Fig. 2a).However, soil CO 2 efflux from surface collars varied considerably during the dry summer in 2013 (Fig. 2a and c).Higher soil CO 2 efflux from surface collars in April 2013 coincided with rain events after a long dry period with high litter input (see Macinnis-Ng and Schwendenmann (2015) for details).
Total soil CO 2 efflux measured outside the trenched plots ranged from 0.6 to 6.9 µmol CO 2 m −2 s −1 with a mean of  2).Volumetric soil water content in the trenched plots was significantly higher (56.8 %) compared to the untrenched sampling points (44 %).In contrast, soil temperature was not significantly affected by trenching (Table 2).

Contribution of autotrophic respiration to total soil CO 2 efflux
Mean autotrophic respiration derived from the trenching approach was 0.8 ± 0.1 µmol CO 2 m −2 s −1 .The contribution of autotrophic respiration to total soil CO 2 efflux (to 30 cm depth) was 25 %.Excluding the roots from the organic layer through deep collar insertion showed that roots in the organic layer contribute around 17 % to total soil CO 2 efflux.The proportion of autotrophic respiration to total soil CO 2 efflux tended to be lower during summer-early autumn (December-April) compared to winter (June-August).However, differences were not statistically significant due to high variability in autotrophic respiration, especially during summer (data not shown).Mean total soil CO 2 efflux (Plot_Surface plus Out-side_Trench_Surface; n = 18, mean = 3.47 µmol CO 2 m −2 s −1 ; SE = 0.20 µmol CO 2 m −2 s −1 ) was positively correlated with total (organic layer plus mineral soil to 30 cm depth) root biomass (R 2 = 0.394, p = 0.042, intercept = 2.49 µmol CO 2 m −2 s −1 ) (Fig. 3).Using the regression approach produced an autotrophic respiration estimate of 0.98 µmol CO 2 m −2 s −1 .The proportion of autotrophic respiration to total soil CO 2 efflux derived from the root biomass regression approach was 28 %.

Effect of soil temperature and volumetric soil water
content on the temporal variability in total soil CO 2 efflux, heterotrophic and autotrophic respiration Independent of the model used, soil temperature explained less than 20 % of the temporal variation in total soil CO 2 efflux (Fig. 4a, Table 3).The Q 10 values for total soil CO 2 efflux was 1.6 (Table 3).A slightly stronger soil temperature response was found for heterotrophic respiration (Fig. 4b, Table 3) with a Q 10 value of 2.2 (Table 3).However, all temperature models for heterotrophic respiration had higher AIC values compared to total soil CO 2 efflux (Table 3) which suggests a poorer performance.No significant relationship was found between soil temperature and autotrophic respiration (Fig. 4c, Table 3).Neither a linear nor a quadratic function resulted in a significant relationship between SWC and total soil CO 2 efflux (Fig. 4d, Table 3).Heterotrophic respiration decreased significantly with increasing SWC (Fig. 4e, Table 3).In contrast a weak, but significant quadratic relationship was found between SWC and autotrophic respiration (Table 3).Bivariate polynomial functions did not result in higher R 2 or better AIC values compared to univariate models (Table 3).

Spatial variation in total soil CO 2 efflux and environmental factors
The spatial variability of total soil CO 2 efflux between the 12 sampling points in the research plot was relatively high, with a coefficient of variation (CV) of 43 % (Table 2).We found a good relationship between the tree local contribution index (I c ) and total soil CO 2 efflux.The relationship was strongest (coefficient of determination, R 2 = 0.342, p = 0.030, linear model) within a radius of 5 m (Fig. 5a  and b).
The spatial variation in total root biomass (organic layer plus mineral soil to 30 cm depth, 0.9 to 8 kg m −2 ) was very high (CV > 95 %, Table 1).Similar to total soil CO 2 efflux, a radius of 5 m provided also the best correlation between root biomass and I c .The coefficient of determination was R 2 = 0.985 (p = 0.021, univariate model, Fig. 5c and d).
Compared to root biomass and soil CO 2 efflux the spatial variation in litterfall (total amount over the 18-month period, 1.1-2.2kg m −2 , Table 1) was small (CV = 20 %, Table 1).We did not find any significant correlations between litterfall and I c (data not shown).
Using multiple regression analysis revealed that most of the spatial variability in total soil CO 2 efflux within the plot could be explained by soil temperature and root biomass (R 2 = 0.977, Adjusted R 2 = 0.953, F = 41.972,p = 0.023).3.
Our findings suggest that soil CO 2 efflux in a conifer dominated forest can be as high or even exceed the efflux rates from broadleaf forests.This is in contrast to previous studies which found that soil CO 2 efflux in conifer forests are lower than those in broadleaf forests (Raich and Tufekcioglu, 2000;Curiel Yuste et al., 2005a).However, these former studies were limited to temperate study sites and based on direct comparisons of sites where forest type was the principal variable differing among pairs.Mean annual soil temperature has been shown to be a good predictor of large-scale variation in total soil CO 2 efflux in non-water limited systems independent of vegetation types and biome (Bahn et al., 2010).With a mean annual temperature of 14 • C, this study site was relatively warm compared to sites along the Pacific coast of North America partly explaining the high soil CO 2 efflux rates in this kauri dominated forest.
The amount of litterfall has also been associated with differences in soil CO 2 efflux at the scales of biomes (Davidson et al., 2002;Reichstein et al., 2003;Oishi et al., 2013).Annual C input via litterfall in this kauri dominated forest was 410 and 760 g C m −2 in 2012 and 2013, respectively (Macinnis-Ng and Schwendenmann, 2015).This litter C flux is substantially higher than those values from conifer and mixed conifer-hardwood forests in the Northern Hemisphere (50-400 g C m −2 yr −1 ; mean: 164 ± 14 g C m −2 yr −1 ; n = 43; Bond-Lamberty and Tompson, 2014;Holland et al., 2015).Kauri litterfall is within the range of values (110-700 g C m −2 yr −1 ; mean: 345 ± 30 g C m −2 yr −1 ; n = 22) reported for old-growth tropical forests (Chave et al., 2010; Table 3. Relationships of total soil CO 2 efflux, heterotrophic and autotrophic respiration with soil temperature (T) and volumetric soil water content (W) using uniand bivariate functions.Holland et al., 2015;Bond-Lamberty and Tompson, 2014).High litter input, together with high annual temperature, can be another major factor explaining the comparatively high soil CO 2 efflux rate in this southern conifer forest.This is somewhat surprising as one would assume that organic matter mineralisation and thus soil CO 2 efflux is reduced given the slow decomposition rate of kauri litter.In four kauri forests ranging from pole to mature forests mean residence times between 9 and 78 years were estimated for 8 to 46 cm thick organic layers (Silvester and Orchard, 1999).According to Silvester and Orchard (1999), sites with higher litterfall were accompanied by faster breakdown and no relationship was found between litterfall and the depth of the organic layer.The organic layer in our study sites was only 5 to 15 cm thick.Possible reasons for a lack of litter accumulation and build-up of a thick organic layer are the following: removal and disturbance of the organic layer as a consequence of tree fall and removal of five large kauri trees in the 1950s (Thomas and Ogden, 1983) and topography.The topography of the study site (moderate to steep slope) likely explains the negative correlation between organic layer thickness and elevation (r = −0.539,p = 0.021).Erosive removal of the organic layer and mineral soil on steep slopes and deposition downslope have been shown to affect soil characteristics and C cycling (Vitousek et al., 2003;Yoo et al., 2005;Stacy et al., 2015).For example, in a temperate forest in Japan (Nakane et al., 1984) and a tropical seasonal forest in Thailand (Takahashi et al., 2011) soil CO 2 efflux decreased with increasing slope.However, we did not find any significant correlation between elevation and total soil CO 2 efflux, root biomass, and soil water content suggesting that forest structure (see Sect. 4.2) may have had a stronger effect on soil characteristics than topography.
While mean annual soil temperature partly explains the overall high mean soil CO 2 efflux measured in this forest, soil temperature was not a very good predictor of the temporal variation in total soil CO 2 efflux.Independent of the regression model used, soil temperature explained a small proportion (< 20 %, Fig. 4a, Table 3) of the seasonal variation in total soil CO 2 efflux.In temperate forest ecosystems in the Northern Hemisphere (Ngao et al., 2012;Bond-Lamberty and Tompson, 2014) soil temperature often explains more than 50 % of the temporal variability in total soil CO 2 efflux.It is important to note that the soil temperature range in this kauri forest was narrow (around 7 • C) compared to other temperate forests with a larger seasonal soil temperature amplitude (> 10 • C, Paul et al., 2004).Thus, a seasonal temperature effect may not have been visible in this kauri forest.The Q 10 value (1.6, Table 3) was at the lower end of the range reported for mixed and evergreen forests (Q 10_10−20 • C ; 0.5-5.6;Bond-Lamberty and Tompson, 2014).However, low Q 10 values have also been reported for other conifer forests, especially at sites characterized by mild winters (Borken et al., 2002;Curiel Yuste et al., 2005b;Sulzman et al., 2005).Low Q 10 values in evergreen forests have been explained by the lack of a distinct seasonality in photosynthesis and substrate supply (Curiel Yuste et al., 2005b).
No significant relationship was found between SWC and total soil CO 2 efflux (Fig. 4d, Table 3).However, total soil CO 2 efflux tended to decline with increasing SWC.Excess SWC may negatively affect CO 2 efflux rates by reducing soil aeration and thus CO 2 diffusivity (Janssens and Pilegaard, 2003).Further, low levels of oxygen as a result of high SWC decreases activity of plant roots (Adachi et al., 2006) and the heterotrophic decomposition of soil organic matter (Linn and Doran, 1984).This may be particularly relevant in the clayey soils under study.

Forest structure and the spatial variation in soil CO 2 efflux
The spatial variability (CV = 43 %) of total soil CO 2 efflux in this study is slightly higher compared to other studies with similar numbers of measurements and/or plot size (32-39 %; Epron et al., 2006;Kosugi et al., 2007;Bréchet et al., 2011).The higher spatial variation might be related to differences in tree size and distribution across the plot.The stand is clearly dominated by kauri trees in all size classes (Fig. 1).However, kauri occurs in clusters around the four largest and emergent kauri individuals whose neighbourhood is generally characterised by relatively few trees (see lower centre of Fig. 1).
The influence of forest structure (here: kauri tree distribution and tree size, I c ) on total soil CO 2 efflux is confirmed by the significant relationships between I c and total soil CO 2 efflux, root biomass and mineral soil C : N ratio.Previous studies have shown that kauri has significant effects on soil processes such as pH and nitrogen cycling (Silvester, 2000;Jongkind et al., 2007;Verkaik et al., 2007;Wyse et al., 2014).This is the first study showing that kauri exerts a substantial influence on soil C related processes.Our results also corroborate a study by Katayama et al. (2009) suggesting that the spatial arrangement of emergent trees in a tropical forest is an important factor for generating spatial variation of soil CO 2 efflux.Studies in European beech forests also show that the combination of root, soil and stand structure help to understand the mechanisms underlying soil CO 2 efflux and that forest structure has some influence on the spatial variability of soil CO 2 efflux (Søe and Buchmann, 2005;Ngao et al., 2012).The relationship between total soil CO 2 efflux and forest structure was strongest within a radius of 5 m (Fig. 5a and b).In a tropical forest, the strongest correlation between soil CO 2 efflux and forest structural parameters was within 6 m from the sampling points (Katayama et al., 2009).A radius of 5 m also provided the best correlation between root biomass and I c .As measurements of the lateral root extension are not available for kauri, it remains unknown if this distance equals the maximum lateral extension of fine roots from the trunk or represents the distance where fine root density is highest.Based on observations, large lateral roots of mature kauri trees often extend beyond the width of the crown and an extensive network of fine roots extends from the lateral roots into the interface between organic layer and the mineral soil (Bergin and Steward, 2004).The radial fine root spread in mature Northern Hemisphere conifer stands varies considerably (6-20 m) depending on site characteristics and stand structure (Stone and Kalisz, 1991).
In contrast to other studies (e.g.Bréchet et al., 2011;Katayama et al., 2009), we did not find a significant correlation between litterfall and forest structure.Tree size and architecture have been reported to affect the pattern of litterfall distribution on the forest floor (Ferrari and Sugita 1996;Staelens et al., 2004;Zalamea et al., 2012).However, de-L.Schwendenmann and C. Macinnis-Ng: Soil CO 2 efflux in an old-growth southern conifer forest spite a 3-fold difference in tree size across the plot we did not see a significant effect of tree size on total litterfall.This is also reflected in a small within-plot variation in litterfall (CV = 21 %, Table 1).This is confirmed by a litterfall study in four remnant kauri forests where a small variation in litterfall (CV = 17-26 %) was found across a wide range of litter trap positions (Silvester and Orchard, 1999).
Spatial variability in total soil CO 2 efflux was largely attributed to soil temperature and the amount of fine root biomass and associated rhizosphere, with 97 % of the variation explained.This implies a relationship with tree productivity which is in agreement with findings from other conifer forests (Janssens et al., 2001b;Luo and Zhou, 2006).Although roots accounted for less than 30 % of total CO 2 efflux, recent research has shown that both recent photosynthate and fine root turnover can be important sources of C for forest soil CO 2 efflux (Epron et al., 2011;Warren et al., 2012) as discussed below.

Components of total soil CO 2 efflux
Collar insertion through the organic layer into the mineral soil resulted in a 17 % reduction in soil CO 2 efflux.Similar reductions were found in other ecosystems and demonstrates that collar insertion by only a few centimetres cuts off fine roots (Heinemeyer et al., 2011) and contributions by ectomycorrhizal fungal mats (Phillips et al., 2012) reducing total soil respiration.Thus, collar insertion can cause underestimation of total soil CO 2 efflux.This may be a particular problem in ecosystems where a large amount of roots and mycorrhiza are found in the organic layer and at the interface between the organic layer and an organic rich mineral soil as in this kauri forest.
The partitioning of total soil CO 2 efflux into its main components: heterotrophic respiration (oxidation of soil organic matter) and autotrophic respiration (root and associated mycorrhiza respiration) remains technically challenging.Differences in the proportion of autotrophic or heterotrophic respiration to total soil CO 2 efflux might vary not only among species and ecosystems but also with the method used for partitioning total soil CO 2 efflux (Kuzyakov, 2006;Subke et al., 2006;Millard et al., 2010).Cutting roots through inserting deep collars and trenching increases the dead root biomass (Heinemeyer et al., 2011).As we did not correct our estimates of soil CO 2 efflux for decomposing rootderived CO 2 fluxes the heterotrophic respiration may have been slightly overestimated (Hanson et al., 2000;Kuzyakov, 2006;Ngao et al., 2012).However, both techniques used in this study, trenching and regression-analysis, showed similar results.The proportion of autotrophic respiration in this kauri was between 25 % (trenching) and 28 % (regression analysis) of total soil surface CO 2 efflux.The contribution of autotrophic respiration to total soil CO 2 efflux can account for as little as 10 % to more than 90 % worldwide (Hanson et al., 2000) but values of 45-50 % are typical (Subke et al., 2006).Our estimate is at the lower end of values observed for Northern Hemisphere conifer and tropical broadleaf forests (30-70 %, Epron et al., 2001;Högberg et al., 2001;Bond-Lamberty and Tompson, 2014;Taylor et al., 2015).This suggests that root and/or rhizosphere activity in this forest is comparatively low.However, a similar proportion of autotrophic respiration (23 %) was estimated for a New Zealand old-growth beech forest (Tate et al., 1993) and an old-growth Douglas-fir site in the Cascades, Oregon (23 %) (Sulzman et al., 2005).Another factor accounting for the differences in values is the depth of trenching (Hansen et al., 2000;Kuzyakov, 2006;Bond-Lamberty et al., 2011).The contribution of autotrophic respiration may have been underestimated as we only trenched to 30 cm depth.It is recommended to trench to a depth beyond the main rooting zone (Subke et al., 2006) and in some studies the trenched plots are dug down to the solid bedrock (Díaz-Pinés et al., 2010).
Total soil CO 2 efflux is not only directly affected by the amount of autotrophic respiration but also by the supply of C through root turnover and root exudates.The decomposition of root debris has been shown to increase microbial activity and thus heterotrophic respiration (Göttlicher et al., 2006).Despite a low root and/or rhizosphere activity the total soil CO 2 efflux in a mycorrhizally associated Douglasfir forest was dominated by belowground contributions due to the large pool of rhizospheric litter with a relatively high turnover rate (Sulzman et al., 2005).In addition, root exudates containing carbohydrates, sugars and amino acids supply energy for the decomposition of soil C ("priming") (Högberg et al., 2001).Further, a recent study showed that a common root exudate, oxalic acid, promotes soil C loss by releasing organic compounds from mineral-protected aggregates.This indirect mechanism has been found to result in higher C losses compared to simply increasing the supply of energetically more favourable substrates (Keiluweit et al., 2015).
Root activity may also affect physical soil conditions.In some studies, SWC and fine root biomass were negatively correlated (Coomes and Grubb, 2000;Ammer and Wagner, 2002).High uptake of water by kauri fine roots concentrated in the organic layer may lead to lower SWC and slightly higher soil temperatures (Verkaik et al., 2007;Verkaik and Braakhekke, 2007).The drier conditions at the base of trees might be an indicator of good soil aeration that enhances the diffusivity of soil CO 2 into the air (De Jong and Schappert, 1972;Tang et al., 2003).
The soil temperature response was stronger for heterotrophic respiration (Fig. 4b, Table 3).This is in line with other studies conducted in temperate mixed forests (Kirschbaum, 1995;Boone et al., 1998) and suggests a higher sensitivity of heterotrophic respiration to temperature.Below 50 % SWC autotrophic respiration increased with increasing water content (Fig. 4f).A positive correlation between soil water content and autotrophic respiration have also been reported for temperate and tropical forests (Zang et al., 2014;Brunner et al., 2015;Doughty et al., 2015).This is in contrast SOIL, 2, 403-419, 2016 www.soil-journal.net/2/403/2016/ to other studies which reported that dry conditions enhanced the growth of fine roots in the surface soil resulting in higher proportions of autotrophic respiration (Bhupinderpal-Singh et al., 2003;Noguchi et al., 2007).

Conclusion
Our study has two significant findings for southern conifer forests.Firstly, this is the only study quantifying the amount of soil CO 2 efflux in an old-growth kauri forest.Our findings suggest that the loss of soil CO 2 (1324 ± 121 g C m −2 yr −1 ) from this forest type is considerable.Although the contribution of autotrophic respiration is comparatively low (< 30 %), root biomass explained a high proportion of the spatial variation in soil CO 2 efflux.This suggests that the total soil CO 2 efflux in this forest is not only directly affected by the amount of autotrophic respiration but also by the supply of C through roots and mycorrhiza.Any modification in root/rhizosphere will most likely result in long-term modifications of the soil CO 2 efflux.This is of relevance given that many kauri forests are threatened by Phytophthora agathidicida (Weir et al., 2015) which infects the roots and can lead to tree death (Than et al., 2013).Secondly, this study is the first to confirm that kauri not only exerts a strong control on soil pH and nitrogen cycling but also on soil carbon related processes.The effect of kauri tree size and distribution on total soil CO 2 efflux demonstrates the need to include biotic parameters for better prediction of the spatial variability in soil CO 2 efflux.

Data availability
The underlying research data set is available in the figshare data repository (doi:10.17608/k6.auckland.3505796).

Figure 1 .
Figure 1.Overview of the research plot showing the position of all trees ≥ 2.5 cm diameter (circle size scaled to diameter -Factor 5) and sampling points for soil CO 2 efflux measurements (Plot_Surface, Plot_Inserted), litter and soil sampling.The trench plots are located adjacent (upslope) of the research plot at around 92 m a.s.l.

Figure 2 .
Figure 2. Soil CO 2 efflux (a), soil temperature (b) and volumetric soil water content (c) measured in the research plot from August 2012 to January 2014.Values show mean ± standard error of the Plot_Surface (n = 12) and Plot_Inserted (deep collar, n = 12) sampling points, respectively.Volumetric soil water content was not measured in March 2013 due to equipment failure.Fortnightly rainfall (c) was measured in a paddock in the vicinity of the research plot.

Figure 3 .
Figure3.Regression of total root biomass (organic layer plus mineral soil to 30 cm depth) vs. total soil CO 2 efflux.

Figure 4 .
Figure 4. Upper panels: relationship between soil temperature and total soil CO 2 efflux (a), heterotrophic respiration (b) and autotrophic respiration (c).Lower panels: relationship between soil volumetric water content and total soil CO 2 efflux (d), heterotrophic respiration (e) and autotrophic respiration (f).Regression lines are only displayed for significant linear relationships.The results for other uni-and bivariate functions are shown in Table3.

Figure 5 .
Figure 5. Relationships between the sum of local contribution indices of surrounding trees within the fitted radius of influence and total soil CO 2 efflux (a, b), root biomass (c, d) and mineral soil CN ratio (e, f).The arrows in (a), (c) and (e) indicate the best coefficients of variation (highest R 2 value) with models shown in (b), (d) and (f).M1 = univariate model (I c = S); M2 = linear model (I c = S × (1 − (d/r))) where S = trunk cross section area (S, m 2 ), d = distance between the trees and the measurement point (d, m), r = fitted radius of influence (r, m).

Table 1 .
Descriptive statistics for litter, root, and soil characteristics.Samples were taken in the vicinity of the surface soil CO 2 efflux sampling points (n = 12, except for root biomass, n = 10).

Table 2 .
Descriptive statistics of soil CO 2 efflux, soil temperature and volumetric soil water content measured within the research plot (Plot) and adjacent to it(Trench).Measurements were conducted between August 2012 and January 2014.Different letters (a, b for plot; c, d for trench) for a given variable indicate a significant difference between total soil CO Schwendenmann and C. Macinnis-Ng: Soil CO 2 efflux in an old-growth southern conifer forest 3.1 ± 0.1 µmol CO 2 m −2 s −1 (Outside_Trench_Surface, Table 2).Soil CO 2 efflux from Trench_Inserted collars was significantly lower (25 %) compared to total soil CO 2 efflux (Table www.soil-journal.net/2/403/2016/SOIL, 2, 403-419, 2016 L.