Introduction

Land-use change (LUC) accounts for about 20% of human-induced carbon dioxide (CO2) emissions (Tubiello et al. 2015), contributing ~ 1.15 Pg of carbon (C) per year (yr) into the atmosphere over recent decades (Houghton et al. 2012). Globally, LUC has been dominated by deforestation and grassland conversion to crop or pastoral land, which commonly led to a C debt through vegetation removal and increase on soil organic carbon (SOC) via mineralization (Smith et al. 2016). Afforestation, in turn, may induce a positive C balance by increasing the C sink strength and residence time of sequestered C (Powlson et al. 2011; Minasny et al. 2017). Commercial forest plantations are a paramount business in South America (IBÁ 2019). The recent expansion of eucalypt production on native grasslands of the Pampa Biome, notably in the southernmost of Brazil, and the little quantitative information on its impact on C cycling have raised ecological concerns about this LUC (Oliveira et al. 2017). Previous studies have reported increased soil CO2 effluxes after conversion of the Pampa Biome into row-crop agriculture (Pillar et al. 2012), prompting the need for investigating similar conversions of native grassland to plantation forest.

Short-rotation eucalypt plantations create an ephemeral C sink through rapid C accumulation in their biomass (Bouillet et al. 2013). Conversely, the large production of aboveground litter (Gatto et al. 2010) and root allocation to deep soil layers (Laclau et al. 2013) might yield SOC accumulation, and thus promote long-term C sequestration through afforestation. A high soil C input, however, can be counterbalanced by increases in soil CO2 efflux. Soil CO2 efflux is the most common proxy used to measure soil respiration (Rs), i.e., the flux of microbially (Rh) and plant-respired (Ra) CO2 from the soil to the atmosphere (Maier et al. 2011). Thus, to foresee the feedback of planted forests to human-induced CO2 emissions it is imperative to consider and better understand Rs and its components. Despite the vast area occupied by eucalypt plantations in Brazil (IBÁ 2019), few studies directly address Rs (e.g. Binkley et al. 2006; Nouvellon et al. 2012b; Ryan et al. 2010). Most of the available data report Rs measurements at discrete time periods, mainly at the end of rotation. Although site preparation can be a major driver of the soil CO2 efflux dynamics, the early-impact of eucalypt afforestation on soil CO2 efflux has rarely been investigated (La Scala et al. 2006; Fialho et al. 2018; Teixeira et al. 2020). The majority of eucalypt stands in Brazil are row-cropped established following minimum cultivation practices, in which soil preparation (e.g., subsoiling, ripping, ridge-tillage) and fertilizations are restricted to within the planting row (W) (Gonçalves et al. 2013). These practices might create distinct functional zones within a stand (W and between-row (B)) that likely influence patterns of Rs and its components (Liebig et al. 1995), but it has been neglected in most forest studies.

Abiotic factors such as soil temperature and moisture are strong drivers of Rs seasonality (Davidson and Janssens 2006; Yuste et al. 2007; Hu et al. 2018). Rs is commonly reported to increase exponentially with soil temperature (Raich and Tufekciogul 2000; Yu et al. 2017), and to be linearly correlated with soil water content (Cook and Orchard 2008; Moyano et al. 2012). However, Rs components (Rh and Ra) respond differently to changes in soil temperature and moisture according to geographic location, vegetation type and developmental stage, and other soil properties (Tang and Baldocchi 2005; Heinemeyer et al. 2012; Hinko-Najera et al. 2015). Thus, environment-induced changes on Rs might not be as clearly predicted. For instance, in two studies under temperate climate, soil temperature was a good predictor of Rh, but not of Ra in a eucalypt site (Hinko-Najera et al. 2015), while it could explain both Ra and Rh in grassland sites (Heinemeyer et al. 2012). In the same studies, there was no correlation between Rh and soil moisture in the eucalypt site (Hinko-Najera et al. 2015), but Rh was inversely proportional to soil moisture in the grassland site (Heinemeyer et al. 2012). Additionally, Rs variability may be attributed to variations in the consumption and transport of CO2 from and within the soil profile to the atmosphere (Cook et al. 2007; Cook and Orchard 2008; Maier et al. 2012), requiring better knowledge of subsurface controls on Rs. Soil preparation practices can alter soil physical properties, which is expected to influence subsurface CO2 dynamics (Blagodatsky and Smith 2012); however, to our knowledge, this has been rarely documented in short-rotation forests (but see Teixeira et al. 2020).

Afforestation of grasslands alters belowground C allocation (Anderson-Teixeira et al. 2013), soil microbial community structure (Lupatini et al. 2013), and substrate availability (Don et al. 2011); consequently, afforestation may potentially alter the soil-to-atmosphere C flux by altering both Rh and Ra. The flux of readily oxidizable C and SOC mineralization drive Rh, while gross primary productivity and C allocation regulate the majority of Ra (Sey et al. 2010; Hopkins et al. 2013). A more dynamic and rapid influence on Rs components is expected to occur in young forest plantations, since afforestation will likely create distinct soil environmental conditions from those occurring in grassland soils through changes in above- and belowground litter input and turnover. Large belowground C allocation, enhanced Ra, and faster C cycling have been reported to result in greater Rs in grassland soils than in tropical or temperate forests (Raich and Tufekciogul 2000; Kellman et al. 2007). The partitioning of Rs following LUC can help to understand major Rs drivers and improve our ability to predict future changes in plant-soil C cycling, such as whether the afforested soils will potentially become net sources or sinks for atmospheric C, and define management practices to increase the sustainability of the new land use.

Here, we evaluate soil CO2 concentration [CO2], efflux, and partitioning for ~ 3 yr following the conversion of a grassland to a commercial eucalypt plantation in southern Brazil. We hypothesized that (1) eucalypt afforestation would result in progressively lower soil CO2 efflux following soil preparation; and (2) minimum cultivation practices result in two distinct soil functional zones (W and B) that differ regarding soil CO2 efflux rate and [CO2] dynamics over forest development. To test these hypotheses, we combined isotopic and root-exclusion approaches to distinguish Rs components and study how eucalypt development might change Ra, Rh, and Rs dynamics in a recently afforested site.

Materials and methods

Site description

This study was conducted in two nearby areas located at São Gabriel, Rio Grande do Sul, Brazil: a grassland (30°29′ S; 54°32′ W) and a eucalypt site (30°26′ S; 54°31′ W). The grassland site represents a native vegetation from the Pampa biome, which holds a diversity of soils and a complexity of plants with the prevalence of C4-grass species (Overbeck et al. 2007) and spans ~ 700,000 km2 in South America within Brazil, Argentina, Uruguay, and Paraguay (Di Giacomo and Krapovickas 2005). The Brazilian Pampa covers 63% of the Rio Grande do Sul state and has experienced the expansion of eucalypt plantations over the last decade. Currently, eucalypt plantations cover ~ 200,000 ha of the Brazilian Pampa (C.H. Prass, FEPAM, personal communication). The eucalypt site is a commercial plantation representative from the sites where the eucalypt expansion is occurring, in which standard silviculture practices have been applied (see details below). Several studies have been recently performed in adjacent stands aiming at understanding the interactions of climate, spacing, and genetics on eucalypt production (Binkley et al. 2017), as well as developing the best fertilization regime for the region (Ferreira et al. 2018).

The sites are in a humid subtropical climate with no defined dry season and hot summers (Cfa, Köppen’s classification). Monthly average temperature and precipitation ranged from 13.1 to 26.9 °C and 22 to 255 mm, respectively, during experimental period (November 2012–July 2015; Fig. 1). The soil in both sites was classified as an Oxyaquic Hapludalf (Soil Survey Staff 2014) with silty clay texture and low degree of soil weathering (Table 1). Markedly in the eucalypt site, the soil showed distinct red and gray mottles in deeper layers (below 60 cm depth), indicating some fluctuation in the water table depth over the year. Initial SOC and total nitrogen (TN) concentrations at 1 m depth were higher in eucalypt than in grassland site, while no pronounced differences on C/N ratio, soil porosity or soil density between sites were observed (Table 1).

Fig. 1
figure 1

Total monthly rainfall (left y-axis, bars) and air temperature (right y-axis, lines) over the experimental period. *indicates eucalypt seedling planting and black bars indicate CO2 sampling points

Full size image
Table 1 Soil characteristics of grassland and eucalypt sites at the beginning of the experiment (0.25-yr after eucalypt afforestation)
Full size table

The native grasslands used in our study had historically been grazed; however, grazing was stopped during our study and the only anthropogenic disturbance was occasional grass mowing to allow CO2 sampling (see Sect. 2.3). In the afforested site, grasses were eliminated through a single glyphosate application. Then, 2 Mg ha−1 of lime was broadcasted before soil preparation. Soil was prepared following minimum cultivation practices by subsoiling (up to 40 cm depth) and soil displacement only within the planting row (i.e., W), creating a ridge of ~ 20 × 80 cm (height × width). Simultaneously, mineral-phosphate (200 kg ha−1) was incorporated into the subsoiled rows using a tube attached behind the subsoiler shank. Eucalyptus dunnii Maiden L. seedlings were planted in August-2012 (end of winter season), spaced 3.3 m between-row (B) and 2.2 m along the W position. Soil was amended with 06:30:06 NPK fertilizer (150 g plant−1) at planting, and with 12 kg ha−1 of N as side-dress fertilization, 1-yr after planting.

Experimental design and soil air samplers

In each site, the experiment was set up as a randomized block design with four replicates three months after eucalypt planting, i.e., in the very early stage of the LUC. In the eucalypt stand, each block consisted of seven rows with 20 trees each totaling ~ 1000 m2. In the grassland site, an area of ~ 2500 m2 was fenced and equally split into four blocks.

Soil CO2 efflux was measured using two different approaches that captured bulk soil Rs and Rh, and thereby allowed Ra to be mathematically determined. We used chambers constructed from polyvinyl chloride (PVC) tubes with 40 cm diameter and 15 cm length buried 10 cm into the soil to assess Rs in both eucalypt and grassland sites. A longer PVC chamber (1.30 m length, buried 1.20 m into the soil) with the same diameter (40 cm) was used to preclude root development, and thus distinguish Rh from Ra. To install the Rh chamber, the soil was vertically excavated to a depth of 1.20 m around a 40 cm diameter section, resulting in an exposed intact vertical soil section. Then, the Rh chamber was sunk into the exposed soil and the excavated soil was disposed around the Rh chamber. In the eucalypt site, we installed two Rs and two Rh chambers in each block to account for each zonal tillage position (W and B). In the grassland site, one Rs chamber was installed per block, totaling four chambers. All chambers were installed around one month prior to the first measurement to reduce the effects of soil disturbance over the “settling period” (Li and Kelliher 2005). All chambers were kept in place throughout the experiment (2.91 yr).

We constructed a [CO2] tube to sample the [CO2] to 1 m soil depth (Fig. 2). The [CO2] profile tube was constructed from PVC (6 cm diameter × 1.10 m length), which was divided into five different sections (to measure the [CO2] of each specific soil layer) using nylon rods of 6 cm diameter × 2 cm length. Sections contained perforations (1 cm diameter) in their tube wall to balance the [CO2] inside the tube and the surrounding soil. Each nylon rod contained a specific 1 mm hose to allow the CO2 sampling from that section specifically. Similarly to Rs and Rh chambers, [CO2] tubes were installed in both W and B positions in the eucalypt site around 1 month before initial measurements and maintained in place throughout the experimental period. To install the [CO2] tubes, we excavated the soil with a stainless-steel core of 6 cm diameter × 1 m length. Afterward, the [CO2] tubes were vertically inserted into the excavated section, while the upper 10 cm section of the tube was left accessible above the soil surface. Therefore, the five tube sections matched with the respective soil depths: 0–10, 10–20, 20–40, 40–60, and 60–100 cm. The [CO2] tube remained capped between samplings to prevent it from filling with rainwater.

Fig. 2
figure 2

[CO2] profile tube used to sample the soil CO2 concentration at five soil depths: 0–10, 10–20, 20–40, 40–60, and 60–100 cm depth. The black circles represent perforations in the tube wall to allow air movement between internal and external atmosphere

Full size image

Soil CO2 sampling

Soil CO2 measurements were carried out across the first three years of stand development, which were divided into eight measurements (0.25, 0.50, 1.00, 1.25, 1.50, 1.91, 2.33, 2.91 yr after afforestation) distributed across different seasons (Fig. 1). All measurement campaigns occurred simultaneously during the daytime in the eucalypt and grassland sites. The litter overlying the soil inside the chambers in the eucalypt site was not removed for CO2 sampling. On the grassland site, the vegetation inside the chamber was clipped to allow capping. To determine CO2 efflux, the chambers were capped for 40 min and the air in the Rh and Rs chambers headspace was sampled at 0, 10, 20, and 40 min after capping. The internal air was accessed through a rubber septum inserted on the top of each lid. For the [CO2] tube, we did not use a CO2 increment time sequence, i.e., the air within each hose (each layer) was sampled only once after being homogenized with each section atmosphere by repeated syringe flushing. We used 50 mL syringes connected with a three-way stop for all samplings, which were then placed in cooler-containers and transported to the laboratory. The concentration [CO2] and isotopic C signature (δ13C) of CO2 were quantified in a cavity ring-down spectrometer (CRDS G2131-i Analyzer, Picarro, Sunnyvale, USA). Soil moisture and temperature at 10 cm soil depth were determined using probe sensors (EC-5, Decagon Devices, Pullman, USA) simultaneously with air sampling.

Soil sampling

Soil samples were taken from the eucalypt site at 0.25, 1.50, and 2.91 yr of afforestation. In each block, five soil samples were collected within each zonal tillage position (W and B) and depth (0–10, 10–20, 20–40, 40–60, and 60–100 cm), and composited for a single sample per position and depth. Samples were air-dried and ground (< 2 mm) to prepare for determination of δ13C-SOC using an elemental analyzer interfaced with an isotope ratio mass spectrometer (EA-IRMS ANCA-GSL 20–20 system, Sercon, Crewe, UK).

Soil respiration calculations

Soil CO2 efflux, a proxy for Rs (Maier et al. 2011), and Rh were calculated using the formula:

$$R_{s,h} = \frac{{\left( {\frac{dC}{dt}} \right) \times P \times V \times M}}{R \times T \times A} .$$
(1)

where dC/dt is the rate of CO2 change during 40 min, estimated by the slope from the linear regression between soil CO2 concentration and sampling time; P is the atmospheric pressure inside the chamber; V is the chamber volume after capping; R is the ideal gas constant; T is the air temperature (K) during air sampling; M is the molecular weight of CO2; and A is the chamber area.

The Rs in the afforested site was measured in both W and B positions. To compare the Rs in the eucalypt site with that in the grassland, we averaged Rs at W and B positions by considering their proportional area within the stand, i.e., 35 and 65% respectively.

We used Miller and Tans (2003) approach to calculate the δ13C value of Rs (δ13CO2). Briefly, the δ13CO2 value was multiplied by the respective CO2 concentration at each sampling time (0, 10, 20, and 40 min) and plotted against the CO2 concentration. The slope of the linear regression was considered equivalent to the δ13CO2.

The contribution of eucalypt-derived CO2 to soil CO2 efflux (C3-derived CO2) over the experiment was calculated as:

$$C_{3} \,derived\, CO_{2} = {\raise0.7ex\hbox{${\left( {\delta^{13} CO_{2x} - \delta^{13} CO_{2i} } \right)}$} \!\mathord{\left/ {\vphantom {{\left( {\delta^{13} CO_{2x} - \delta^{13} CO_{2i} } \right)} {\delta^{13} CO_{2EUC} - \delta^{13} CO_{2i} }}}\right.\kern-0pt} \!\lower0.7ex\hbox{${\delta^{13} CO_{2EUC} - \delta^{13} CO_{2i} }$}}$$
(2)

where δ13CO2 is the isotopic C signature of soil CO2 efflux in each sampling time (x) and at the beginning of the experiment (i), and δ13CO2EUC is the δ13CO2 from eucalypt roots respiration (− 32‰) which was obtained through root incubation.

The relationship between Rs and soil temperature or moisture was performed by nonlinear (Eq. 3) and linear regressions (Eq. 4), respectively, as follows:

$$Rs =\upalpha \times e^{\beta T}$$
(3)
$$Rs = a \times m + b$$
(4)

where T is the air temperature (°C), m is the soil water content (m3 m−3) and α, β, a, and b are the estimated parameters. The Q10 of Rs was calculated as the factor by which Rs is multiplied when temperature increases by 10 °C (Davidson and Janssens 2006):

$$Q_{10} = e^{10 \times \beta }$$
(5)

Data analysis

Data were checked for normality (Shapiro–Wilk test) and power-transformed to meet normality assumptions when needed. The influence of land use and time on the soil CO2 efflux rate and δ13CO2 was tested using a two-way repeated measures ANOVA. Within the eucalypt stand, a two-way repeated measures ANOVA (position × sampling time) evaluated the effects of zonal tillage position and time after afforestation on soil CO2 efflux and Rh, while a three-way repeated measures ANOVA was used to compare [CO2] and δ13[CO2] across the zonal tillage position, time after afforestation, and soil depth. When interactions were significant (p < 0.05), means were compared using Fisher’s LSD post hoc test (p < 0.05). All ANOVA’s were carried out using SISVAR software (Ferreira 2011). We evaluated the influence of δ13C-SOC on δ13[CO2] by comparing the slopes of lines relating the two variables at each soil depth using the Statgraphics Centurion XVI software package (StatPoint Technologies, Inc., Virginia, USA).

Results

Soil CO2 efflux in the grassland and eucalypt sites

Soil CO2 efflux exhibited pronounced variation under both land uses, but a larger seasonal variation was observed in the grassland site (Fig. 3a). The CO2 efflux rate in the grassland ranged from 0.21 to 1.34 g CO2 m−2 h−1 during our observation period (Fig. 3a). The highest CO2 efflux rate in the grassland soil occurred during the first six months of our experimental study, while lower CO2 efflux rates were observed during winter seasons. In the eucalypt site, the average soil CO2 efflux was 0.37 ± 0.20 g CO2 m−2 h−1 (Fig. 3a). The highest CO2 efflux rate (0.76 ± 0.07 g CO2 m−2 h−1) coincided with the beginning of the experiment and was 2.4 times higher than the soil CO2 efflux after 2.91 yr of afforestation. Soil CO2 efflux rates were lower in the eucalypt than in the grassland throughout the experiment, except at 1.00 yr and 2.91 yr after afforestation, when soil CO2 efflux was similar in both sites (Fig. 3a).

Fig. 3
figure 3

Soil CO2 efflux rate (a) and its 13C isotopic signature (δ13CO2; b) in the grassland site and in the eucalypt plantation; and the contribution of eucalypt-derived CO2 (C3-derived CO2) to soil CO2 efflux over 2.91 yr of afforestation (c). Errors bars represent standard errors (n = 4). n.s. means no significant (p > 0.05) difference between grassland and the eucalypt afforested site at that specific period. p-values for the two-way repeated measures ANOVA are presented in Online Resource 1

Full size image

The δ13CO2 exhibited lower variation in the grassland site, averaging − 15.6 ± 1.6‰ throughout the experiment (Fig. 3b). In the eucalypt site, δ13CO2 exhibited a downward trend, with extremes being observed at the beginning of the experiment (− 14.1‰) and after 2.91 yr of afforestation (− 29.4‰). Peaks on δ13CO2 were observed at 1.50 and 2.33 yr of afforestation, which coincided with lower rainfall period and higher temperature (Fig. 1). The δ13CO2 in the eucalypt site was generally lower (more negative) than in the grassland, except at 0.50 yr, when no difference between the land uses was observed.

There was an increasing contribution of C3-derived CO2 in soil CO2 efflux through time in the eucalypt stand (Fig. 3c), except at 1.50 and 2.33 yr of afforestation when C3-derived CO2 reduced. The C3-derived CO2 ranged from 3% at the beginning of the experiment to 85% after 2.91 yr of afforestation.

Soil CO2 efflux in the grassland was inversely proportional to soil moisture (p < 0.05; Fig. 4a). Fitting an exponential function (Eq. 3) for the soil CO2 efflux rate in the grassland against soil temperature resulted in a Q10= 1.49 (Fig. 4b). In the eucalypt site, neither soil moisture nor soil temperature could be modeled through linear or exponential functions, respectively, to explain soil CO2 efflux variability (p > 0.05; Fig. 4a, b).

Fig. 4
figure 4

Relationship between soil CO2 efflux rate (Rs) in the grassland and in the eucalypt site with soil moisture (m, a) and soil temperature (T, b) measured during 2.91 yr of afforestation. Regressions were fitted with the 36 paired observations, but values are represented (symbols) by the average (n = 4) of each sampling time. Regression line is shown only for grassland since equation coefficients were not significant for explaining Rs in the eucalypt site

Full size image

Influence of silvicultural practices on soil CO2 efflux, Rs partitioning, and [CO2] to 1 m soil depth in the eucalypt site

Averaged across the entire observation period, soil CO2 efflux was 20% higher in W (0.42 g CO2 m−2 h−1) than in B (0.34 g CO2 m−2 h−1) (Fig. 5a). Greater soil CO2 efflux in W than in B occurred at 1.25, 1.91 and 2.91 yr of afforestation. The soil CO2 efflux at 0.25 yr was 3.6- and 1.3-fold greater than that after 2.91 yr of afforestation in B and W, respectively. Distinct pattern of Rh over the experimental period was observed between B and W (Fig. 5b). High Rh proportion in W occurred at 1.50 and 2.33 yr of afforestation. Conversely, lower Rh proportions in W were observed at 1.25 yr after afforestation and in the winter seasons (1.91 and 2.91 yr).

Fig. 5
figure 5

Soil CO2 efflux rate in the eucalypt site measured within planting row (W) and between-row (B) positions (a), and the proportion of heterotrophic respiration (Rh) to the soil CO2 efflux in both positions (b). Values are mean ± standard error. n.s. means no significant difference in the soil CO2 efflux between zonal tillage positions. p-values for the three-way repeated measures ANOVA are presented in Online Resource 2

Full size image

Zonal tillage management influenced the [CO2] in the soil profile and the δ13[CO2] (Fig. 6; Online Resource 2). Overall, higher [CO2] occurred in the soil profile of B than W, except after 2.91 yr of afforestation, in which [CO2] was similar in B and W (5102 μmol mol−1) (Fig. 6a, b). Differences in [CO2] between B and W were more pronounced below 10 cm depth. In both positions, [CO2] increased with soil depth, but differences among depths were larger in B than W, and less pronounced through time in both positions (Fig. 6a, b). The greatest [CO2] in W was measured after 0.25 yr of afforestation (11,743, 13,636, 18,557, 20,183 and 17,996 μmol mol−1 in 0–10, 10–20, 20–40, 40–60 and 60–100 cm soil layers, respectively, Fig. 6a). The [CO2] concentration in all soil layers of W at the end of the experiment (2065, 4081, 4665, 6566 and 9444 μmol mol−1 in 0–10, 10–20, 20–40, 40–60 and 60–100 cm soil layers, respectively) differed from all other sampling times, but 1.50 and 2.33 yr.

Fig. 6
figure 6

Soil CO2 concentration (a, b) and isotopic signature of CO2 (δ13[CO2]; c, d) in the soil profile (0–100 cm depth) as influenced by zonal tillage management [within planting row and between-row] over 2.91 yr of afforestation. Errors bars represent standard errors. p-value bars represent the least significant difference (p < 0.05) of the three-way ANOVA. p-values for the three-way repeated measures ANOVA are presented in Online Resource 2

Full size image

The [CO2] at the 0–10 and 10–20 cm soil depths in B followed similar patterns: compared with the initial [CO2] (3014 and 5554 μmol mol−1 in 0–10 and 10–20 cm, respectively), there was a significant increase (~ 3.5-fold) until 1.00 yr of afforestation and then a subsequent [CO2] decrease until 2.91 yr of afforestation (2303 μmol mol−1 and 3794 μmol mol−1 at the 0–10 and 10–20 cm depths, respectively; Fig. 6b). However, [CO2] in both depths at 2.91 yr did not differ from that at 0.25 yr of afforestation. In deeper soil layers, [CO2] increments were observed until 1.25 yr (19,241, 23,772 and 27,729 μmol mol−1 at the 20–40, 40–60 and 60–100 cm depth), and then [CO2] decreased until the end of the experiment. For deeper depths, the [CO2] at 2.91 yr of afforestation (4346, 5378 and 8375 μmol mol−1 at the 20–40, 40–60 and 60–100 cm depth) was lower than those at the beginning.

Overall, a more negative δ13[CO2] occurred in W than in B, except at 0.25 and 1.91 yr (Fig. 6c, d). Differences in δ13[CO2] between W and B were greater below 20 cm depth. The δ13[CO2] decreased with depth mainly in W (Fig. 6c), and it became more negative as afforestation aged in both soil zones. Although, the decreasing δ13[CO2] trend was not observed at 1.50 and 2.33 yr, i.e. at this sampling times the δ13[CO2] become less negative (higher) than the previous measurement.

Coefficient of determination (R2) of δ13C-SOC × δ13[CO2] regressions ranged from 0.01 to 0.55 (Fig. 7). Higher R2 was observed in W than B across all soil layers but 40–60 cm. There was a positive correlation between δ13[CO2] and δ13C-SOC in all soil layers of W (Fig. 7a). δ13C-SOC had a greater explanation power on δ13[CO2] at 10–20 and 20–40 cm depths of W (R2 = 0.55 for both depths), in which higher slopes of the linear regressions were observed. A positive correlation between δ13[CO2] and δ13C-SOC was also observed in B, except at 0–10 and 20–40 cm depth; the highest slope was observed at 10–20 cm depth (Fig. 7b). Regression slopes were higher in W than in B across all soil layers.

Fig. 7
figure 7

Isotope C ratio (δ13C) relationship between the soil CO2 concentration (δ13[CO2]) and the SOC (δ13C-SOC) within 1-m depth of each soil zone over 2.91 yr after afforestation. Significance of regression slope is represented after each soil depth by: °p < 0.10; *p < 0.05; **p < 0.01

Full size image

Discussion

Our study provides quantitative information about the potential feedback of soil CO2 emissions following the afforestation of subtropical grassland with eucalypt. Soil CO2 efflux in the afforested site was lower than in the grassland. The eucalypt-derived CO2 contribution (C3-derived CO2) to soil CO2 efflux increased through eucalypt stand development, while the [CO2] in the soil profile and the isotopic C signature of CO2, either δ13CO2 or δ13[CO2], decreased over time. Site preparation created two distinct soil C cycling functional zones within the eucalypt stand, in which [CO2] across the soil profile was overall greater in B, while soil CO2 efflux rate was higher in W. We observed that Rh exerted a lower control on the magnitude of Rs at the end than at the beginning of the experiment, but the contribution of Rh to soil CO2 efflux varied spatially and temporally, i.e., between tillage zones and through forest development.

Land use and its influence on soil CO2 efflux

The soil CO2 efflux in the grassland site averaged 0.85 g CO2 m−2 h−1 (Fig. 3a), which is in a higher range than the rates reported for either a non-degraded (0.27 g CO2 m−2 h−1) or degraded (0.10 g CO2 m−2 h−1) temperate grasslands of South Africa (Abdalla et al. 2018), and also higher than the global average of tropical grasslands (0.25 g CO2 m−2 h−1) (Wang and Fang 2009), but quite similar to a managed grassland under temperate conditions (0.82 g CO2 m−2 h−1) (Heinemeyer et al. 2012). The Rs is a complex process regulated by the interactions of several biotic (e.g., vegetation type, net primary productivity) and abiotic drivers (e.g., clay and SOC content, climatic conditions) resulting in a highly variable Rs global database (Balogh et al. 2011; Moyano et al. 2012; Zimmermann et al. 2015). The Pampa biome has unique edaphoclimatic conditions, such as the combination of relatively fertile soils with high mean annual temperature (~ 18 °C) and total precipitation (~ 1500 mm), that should support high gross primary production, and thus increased Rs when compared to other grassland sites (Yang et al. 2008; Wang and Fang 2009). Additionally, the intensity of grassland management could also impact Rs. For instance, grazing results in larger belowground C allocation, higher soil microbial biomass and belowground turnover rate than ungrazed areas (López-Mársico et al. 2015; Wilson et al. 2018), which is expected to promote Rs. In our study, the influence of grazing might have reflected in the higher soil CO2 efflux rates observed in the first two samplings which were 22.5% greater than those measured in the same seasons afterwards (e.g., 1.25, 1.50 and 2.33 yr), since grazing was withheld after our observations began.

The global soil CO2 efflux in grassland is reported higher (~ 20%; Raich and Tufekciogul 2000) compared to forests growing under similar conditions, with differences increasing in temperate regions (+41%; Kellman et al. 2007). Greater availability of photosynthate allocated belowground, increased root respiration and/or microbial respiration may contribute to an elevated soil CO2 rate in grasslands (Raich and Tufekciogul 2000; Kellman et al. 2007). In our study, the soil CO2 efflux in the eucalypt site averaged 56% lower than in the grassland. In afforested sites, a large fraction of root-respiration, i.e. Ra, can remain within the tree system rather than diffusing into the soil (Bloemen et al. 2016). The CO2 moving internally from root-to-shoot-to-atmosphere via xylem varies during eucalypt tree growth (Cerasoli et al. 2009), and in a 4-yr-old eucalypt tree it can divert about one-fourth of the root respiration (Grossiord et al. 2012). Aubrey and Teskey (2009) observed that the CO2 moving internally via xylem in forest plantations can be even greater, reaching an equivalent amount of the CO2 diffusing directly from the soil to the atmosphere. Therefore, to fully account Rs in forested landscapes, the internal CO2 flux should have been considered. If we consider that 25% of Ra was transported via xylem (Grossiord et al. 2012), Rs in the eucalypt site would average 0.40 g CO2 m−2 h−1 over the experiment, which is less than 50% of grassland Rs rate. Moreover, if the soil CO2 efflux in the afforested site represented only half of the total Rs in the eucalypt plantations (Aubrey and Teskey 2009), Rs would still average 35% less than grassland Rs rates. The eucalypt and grassland sites were relatively nearby, under the same soil class, and shared similar land use history, but the afforested site had higher SOC at the start of the experiment (Table 1). While these differences may undermine appropriate comparisons, the lower soil CO2 efflux even with higher SOC in the eucalypt site demonstrates that the conversion from grassland to eucalypt plantations in southern Brazil is not likely to contribute to increased soil CO2 efflux rates.

The influence of soil temperature and soil moisture on CO2 efflux were evident in the grassland but not in the eucalypt site. The absence of a consistent relationship of Rs in forest systems with temperature or soil moisture has also been observed in other studies (Livesley et al. 2009; Wei et al. 2010). Specifically, due to rapid eucalypt growth, phenological variations could overshadow and decouple Rs from temperature and moisture response patterns (Curiel Yuste et al. 2004; Davidson et al. 2006; Machmuller et al. 2018). Additionally, we posit that the LUC effects on the initial availability and quality of the fresh C input to soil microbial community (Lupatini et al. 2013; Veen et al. 2015), and the reduced Ra rates when root biomass was still incipient, might also have helped decouple Rs from temperature or moisture changes.

Soil CO2 dynamics in a eucalypt afforested site

During the first few years of afforestation, alterations in SOC equilibrium induced by the increase in C outflows with soil preparation (Fialho et al. 2018), and the decrease in C inflows due to the lower litter input, might change soil CO2 efflux dynamics (Nouvellon et al. 2008, 2012a). We thus extensively compiled studies that have evaluated Rs in situ (100% of these studies used the soil CO2 efflux to assess Rs) across eucalypt stands in Brazil to obtain a broader perspective of this new land use (Table 2). The average soil CO2 efflux observed over 2.91 yr of afforestation (0.37 g CO2 m−2 h−1) is within the range reported for multiple age short-rotation eucalypt plantations in the country (0.30–0.55 g CO2 m−2 h−1), but lower when longer (27–29-yr) rotations are evaluated (D’Andréa et al. 2011; Vicentini et al. 2019; Table 2). The highest soil CO2 efflux observed in our study (0.76 g CO2 m−2 h−1) could be considered slightly above the nationwide average, but it was measured at the earliest stage of LUC (0.25 yr). Only one of the 13 compiled studies evaluated early stand development (Teixeira et al. 2020—Table 2) like our study, and the numbers are surprisingly similar considering the differences in geographic locations and environmental conditions. At this stage, the forest is expected to experience higher daily soil temperatures due to reduced canopy and greater soil exposure, which might promote microbial activity (Li et al. 2017) and soil CO2 efflux (Gomes et al. 2016). Moreover, tillage operations at planting can enhance soil CO2 efflux (Fialho et al. 2018). Despite the broad range of environmental conditions, biome domain, soil type, number of eucalypt rotations, or stand development stage, the overall agreement of the Rs range across eucalypt plantations in Brazil is intriguing. Likewise, the Rs range reported across eucalypt plantations in Brazil agrees with those in eucalypt stands worldwide (Giardina and Ryan 2002; Livesley et al. 2009; Fest et al. 2009; Grossiord et al. 2012). Collectively, these findings suggest a strong vegetation-type feedback to Rs (Metcalfe et al. 2011).

Table 2 Reported mean soil respiration of eucalypt plantations in diverse regions of Brazil
Full size table

The isotopic signature of Rs in the eucalypt site become more 13C-depleted (more negative) over time (Fig. 3b), reflecting an increase in C3-derived CO2 (Fig. 3c) either from decomposition of eucalypt-derived C (Rh) or roots respiration (Ra), and the diminished contribution of grass biomass (C4-C) through time. Interestingly, C3-derived CO2 showed a sharp decrease during warmer and drier periods of afforestation (1.50 and 2.33 yr), which resulted in less negative δ13CO2 (Fig. 3b) and δ13[CO2] down the soil profile (Fig. 6c, d). Likewise, during the same periods there was a reduction in soil CO2 efflux and Ra/Rh ratio of the W position (Fig. 5). Collectively, these findings indicate a likely reduction in tree photosynthetic activity under drought conditions (Nouvellon et al. 2008; Ruehr and Buchmann 2010). When the amount of eucalypt-derived organic matter is still small, an incremental stimulation of SOC decomposition under higher soil temperature (Thurgood et al. 2014) is likely to result in less 13C-depleted CO2, which may explain differences on δ13CO2 and δ13[CO2] between 1.50 yr and 2.33 yr of afforestation.

Zonal tillage management created spatial heterogeneity of soil CO2 efflux and [CO2] in the soil profile of the eucalypt stand. The soil CO2 efflux in W was ~ 20% greater than that in B (Fig. 5a). Conversely, B had 25% more [CO2] within 1-m depth and large [CO2] differences between soil depths when compared with W (Fig. 6a, b). Using row-only soil preparation might have impacted the soil CO2 efflux and [CO2] as a consequence of changing soil physical properties related with gas transportation (Liebig et al. 1995; Blagodatsky and Smith 2012) and modifying biological processes, such as the SOC decomposition and root density and exploration (Williams et al. 2017) in W. Indeed, W had lower soil bulk density (1.06 vs. 1.21 for the 0–100 cm soil layer; see Online Resource 3), and higher fine-root biomass and more homogeneous fine-root distribution (see Ferreira et al. 2018). On one hand, the incremental C3 contribution to SOC as afforestation progressed led to more negative δ13[CO2] in W, mainly in the depths 10–20 and 20–40 cm (Fig. 7a), which concentrate higher fine root biomass within 1-m soil depth (Ferreira et al. 2018; Santos et al. 2020). Such evidence indicates that fine-root activity may be a primary driver of [CO2] in W position. On the other hand, the absent influence of δ13C-SOC on δ13[CO2] at 0–10 and 20–40 cm of B, despite the more 13C-depleted [CO2] in these depths over time, suggests that the C input from eucalypt tree is being rapidly respired and not transferred to soil organic matter pools. Thus, we posit that zonal tillage management in eucalypt plantations imposes distinct drivers of Ra and Rh through stand development and should be considered in studies evaluating C cycling dynamics.

Differences in the soil CO2 efflux between W and B (observed at 1.25, 1.91 and 2.91 yr; Fig. 5a) were driven by a decreased Rh in the W position (Fig. 5b). The Rh contribution to Rs ranged from 14 to ~ 100% (Fig. 5b). Although a 100% Rh is very unlikely during active stand growth, this value indicates that Ra was negligible during some periods. It could be also a result of an artificially increased Rh because of moister soil from inhibited lateral water flow or root water uptake or a possible increase of dead root biomass within Rh chambers (Subke et al. 2006; Shrestha et al. 2016; Han et al. 2018). Nevertheless, we did not observe a clear seasonal trend in the Ra/Rh quotient, which could help distinguish these processes and possible artifacts. For instance, a greater proportion of Rh was observed at 1.00 yr, differing from the pattern observed in other winter sampling times (1.91 and 2.91 yr), which, although, could be another artifact created by the side-dress fertilization performed specifically during that period. Drastic seasonal variability of Rh had been also reported in other forest ecosystems (Saiz et al. 2006; Brown and Markewitz 2018). Thus, Ra/Rh dynamics might be a result of several co-varying factors, such as temporal variations in root production, exploration and turnover, microbial-C processing, and CO2 transport within the soil profile, that were possibly different between W and B (Liebig et al. 1995; Williams et al. 2016, 2017).

Conclusions

Our findings provide quantitative information to account for the soil CO2 efflux caused by LUC in countries that eucalypt plantations are a growing economic activity. We demonstrate that eucalypt afforestation of grasslands is not likely to contribute to greenhouse gas emissions through increased soil CO2 effluxes. Our data suggest that minimum cultivation practices, which are commonly adopted in eucalypt plantations in Brazil, create two distinguished functional soil zones within the stand (W and B), which vary as afforestation aged regarding Rs rates and sources (Ra and Rh), and [CO2] in the soil profile. Therefore, soil management practices in production forestry can influence soil C dynamics and leverage the role of planted forests to mitigate CO2 emissions, and should be accounted for in C cycling models. Observations across entire and multiple rotations are still needed to capture equilibriums in C fluxes and pools in short-rotation production forestry and infer the feedback of eucalypt planted forests to climate change by addressing it properly through management practices.