Tree Physiology 32, 696–706 doi:10.1093/treephys/tps038

Research paper: Part of an invited issue on carbon allocation

Stand-level patterns of carbon fluxes and partitioning in a Eucalyptus grandis plantation across a gradient of productivity, in São Paulo State, Brazil Otávio C. Campoe1,7, José Luiz Stape2, Jean-Paul Laclau3,4, Claire Marsden3,5 and Yann Nouvellon3,6 1Departamento

de Ciências Florestais, Universidade de São Paulo, USP, ESALQ, Piracicaba 13418-260, São Paulo, Brazil; 2Department of Forestry and Environmental Resources, North Carolina State University, Raleigh, NC 27695, USA; 3CIRAD, UMR Eco&Sols, 34060 Montpellier, France; 4Departamento de Ecologia, Universidade de São Paulo, USP, IB, 05508-900 São Paulo, Brazil; 5SupAgro, UMR Eco&Sols, 34060 Montpellier, France; 6Departamento de Ciências Atmosféricas, Universidade de São Paulo, USP, IAG, 05508-900 São Paulo, Brazil; 7Corresponding author ([email protected]) Received November 9, 2011; accepted March 15, 2012; published online April 27, 2012; handling Editor Daniel Epron

Wood production represents a large but variable fraction of gross primary production (GPP) in highly productive Eucalyptus plantations. Assessing patterns of carbon (C) partitioning (C flux as a fraction of GPP) between above- and belowground components is essential to understand mechanisms driving the C budget of these plantations. Better knowledge of fluxes and partitioning to woody and non-woody tissues in response to site characteristics and resource availability could provide opportunities to increase forest productivity. Our study aimed at investigating how C allocation varied within one apparently homogeneous 90 ha stand of Eucalyptus grandis (W. Hill ex Maiden) in Southeastern Brazil. We assessed annual above­ ground net primary production (ANPP: stem, leaf, and branch production) and total belowground C flux (TBCF: the sum of root production and respiration and mycorrhizal production and respiration), GPP (computed as the sum of ANPP, TBCF and estimated aboveground respiration) on 12 plots representing the gradient of productivity found within the stand. The spatial heterogeneity of topography and associated soil attributes across the stand likely explained this fertility gradient. Component fluxes of GPP and C partitioning were found to vary among plots. Stem NPP ranged from 554 g C m−2 year−1 on the plot with lowest GPP to 923 g C m−2 year−1 on the plot with highest GPP. Total belowground carbon flux ranged from 497 to 1235 g C m−2 year−1 and showed no relationship with ANPP or GPP. Carbon partitioning to stem NPP increased from 0.19 to 0.23, showing a positive trend of increase with GPP (R2 = 0.29, P = 0.07). Variations in stem wood production across the gradient of productivity observed at our experimental site were a result of the variability in C partitioning to different forest system components. Keywords: carbon budget, carbon partitioning, Eucalyptus, gross and net primary productivity, soil CO2 efflux, total belowground carbon flux.

Introduction The Eucalyptus genus is the most planted hardwood in tropical regions, covering >20 million hectares and providing an increasing share of wood to the worldwide demand (FAO 2010). Eucalyptus plantations were introduced in Brazil a

c­ entury ago, and their area now reaches ~4.7 million hectares (ABRAF 2011). Since their introduction in Brazil, continuous efforts in research have improved management practices (site preparation, fertilization, and control of pests and competing vegetation) and forest genetics (seedlings and clone genotypes;

© The Author 2012. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]

Carbon partitioning across a gradient of productivity  697 Gonçalves et al. 2004, 2008), allowing mean wood productivity of Brazilian eucalypt plantations to increase from ~0.4 kg C m−2 year−1 in the 1960s to 1 kg C m−2 year−1 nowadays, with some commercial stands reaching >1.5 kg C m−2 year−1 (Ryan et al. 2010, Stape et al. 2010, Le Maire et al. 2011b). At the end of a 6-year rotation, aboveground C stocks of Eucalyptus grandis × Eucalyptus urophylla S.T. Blake clones typically reach 5.8 and 7.1 kg of C m−2 under rainfed and irrigated conditions, respectively (Ryan et al. 2010). Maintaining or increasing wood production in these intensively managed forests requires a better understanding of the ecophysiological processes controlling their growth. Resource manipulations in experiments established over a wide range of climatic regions have increased our knowledge regarding the effects of the environment on plantation carbon budgets and stem wood productivity (Giardina et al. 2003, Forrester et al. 2006, Stape et al. 2008, Laclau et al. 2009, 2010, Ryan et al. 2010). Empirical patterns of growth have been documented from several sites across Brazil, but intensive ecophysiological insights that can be used in processbased models are still lacking (Almeida et al. 2004, Miehle et al. 2009). In particular, the adjustment of C allocation to varying resource availability and soil conditions is still poorly understood and quantified, resulting in considerable uncertainty in process-based model predictions of wood production over large zones (Landsberg et al. 2003, Litton et al. 2007). Carbon fluxes and partitioning in forest ecosystems have been studied for decades, using methodologies proposed by Raich and Nadelhoffer (1989), Ryan (1991) and Giardina and Ryan (2002), in which GPP is decomposed into its measured or estimated component fluxes: aboveground net primary production (ANPP: sum of leaf, branch, bark, and stem production), total belowground carbon flux (TBCF, sum of carbon fluxes corresponding to the production and respiration of coarse and fine roots, root exudation and production of substrates used by mycorrhizae) and aboveground autotrophic respiration from trees (RA). Litton et al. (2007), analyzing published data representing a wide range of planted and natural forests from tropical to temperate climatic regions, showed a general pattern of more C partitioned to wood production and less to TBCF with increasing water and nutrient supply. Fertilizer additions did not influence TBCF in eucalypt plantations, but reduced the fraction of GPP partitioned belowground in experiments established in Hawaii (Ryan et al. 2004) and in Brazil (Epron et al. 2012). Similar trends were observed in Brazilian eucalypt plantations under rainfed and irrigated treatments (Ryan et al. 2010). In another tropical eucalypt plantation, irrigation reduced the fraction of GPP partitioned belowground, but it also increased both wood production and TBCF by 12% relative to a rainfed ­treatment (Stape et al. 2008).

Recent studies have found that intensively managed eucalypt plantations under enhanced nutrient and water conditions present higher stem wood production than less productive plantations, mostly as a result of increased GPP and C partitioning to stem wood production (du Toit 2008, Stape et al. 2008, Ryan et al. 2010). In these studies, the availability of water and nutrients was experimentally manipulated, but to our knowledge no study has examined the patterns of C fluxes and partitioning along local, natural gradients of fertility in eucalypt plantations. Within the same stand, natural gradients of productivity are commonly observed despite homogeneous management (plant material, soil preparation, fertilization, etc.), due to large stand sizes (Le Maire et al. 2011a, 2011b), and due to topography and spatial variability in soil texture, soil nutrient content and soil water availability (e.g., soil water retention capacity and water table depth). Such gradients may be a significant source of uncertainty in forest production modeling, where growth is generally considered to be homogeneous within a given stand. Therefore, our objective in this study was to characterize the C budget and C partitioning across a gradient of productivity within one commercial E. grandis plantation stand in Brazil (São Paulo State) in order to gain insight into the processes driving GPP, C partitioning and stem production. Aboveground net primary production and TBCF were measured over 1 year on 12 plots distributed across the gradient of productivity. We hypothesized that the same trends as those reported by largescale studies (e.g., Litton et al. 2007) or by studies where the availability of water and nutrients was experimentally manipulated (e.g., Ryan et al. 2010, Epron et al. 2012) would also be observed across our within-stand gradient of productivity, i.e., plots with higher stem wood production would be associated with higher GPP and higher C partitioning to stem wood production.

Materials and methods Study site The study site was located in the Southeast of Brazil, in São Paulo State, at 22°58′04″S, 48°43′40″ W, in a typical operational plantation of E. grandis belonging to the Duratex company. Over the last 15 years, the mean annual rainfall was 1391 mm, with 75% concentrated from October to March. Mean annual temperature was 19.2 °C, ranging from 13.3 °C in the coldest months (June to August) to 27.2 °C in the warmest months (December to February). Temperatures occasionally fall below 5 °C in every cold season. The mean annual air relative humidity was 77%, with minimum values during winter (close to 45%). During the study period, from October 2008 to September 2009, the mean annual temperature and relative humidity were similar to historical values, while annual

Tree Physiology Online at http://www.treephys.oxfordjournals.org

698  Campoe et al. Table 1. ​Main physical and chemical soil attributes at the experimental site. Depth

Clay

(cm)

(%)

Higher clay soil 0–5 35.2 5–15 34.4 15–50 39.0 50–100 41.5 100–200 42.2 200–300 43.5 Lower clay soil 0–5 17.4 5–15 16.2 15–50 18.2 50–100 20.2 100–200 22.9 200–300 23.4

Silt

Sand

pH

P

Total C

Total N

K

Na

H2O

(mg kg−1)

(%)

(%)

mmolc kg-1

Ca

Mg

H

Al

SB

CEC

BS (%)

7.8 9.7 10.0 9.8 12.1 12.5

57.1 55.9 51.0 48.7 45.7 44.0

4.9 4.7 4.7 4.8 5.3 5.5

2.9 1.8 1.4 1.1 1.0 0.9

1.32 0.73 0.70 0.51 0.34 0.22

0.07 0.04 0.04 0.03 0.02 0.01

0.9 0.5 0.7 0.3 0.1 0.2

0.4 0.2 0.2 0.2 0.2 0.2

7.0 1.8 1.0 1.0 1.0 1.0

2.9 1.8 1.4 1.1 1.0 0.9

38.5 25.5 22.0 16.3 10.8 8.5

6.0 6.5 5.5 2.3 0.0 0.0

11.2 4.3 3.3 2.6 2.2 2.3

55.7 36.3 30.8 21.1 13.0 10.8

20.1 11.7 10.8 12.4 17.1 20.9

5.5 1.3 2.1 1.4 3.6 4.0

77.0 82.5 79.7 78.4 73.6 72.6

4.7 4.3 4.5 4.5 4.5 5.0

3.2 2.0 1.5 1.1 1.1 0.9

1.61 0.54 0.40 0.38 0.27 0.18

0.07 0.03 0.02 0.02 0.01 0.01

0.6 0.2 0.1 0.1 0.1 0.1

0.4 0.2 0.1 0.1 0.1 0.1

6.7 1.8 1.0 1.0 1.0 1.0

6.3 1.5 1.0 1.0 1.0 1.0

48.3 26.0 14.3 13.3 8.3 4.3

6.7 8.8 5.8 5.0 2.5 1.0

14.0 3.7 2.2 2.2 2.2 2.2

69.0 38.4 22.2 20.5 12.9 7.4

20.3 9.5 10.0 10.8 16.8 29.3

Soil sampling was performed on plots 1–4 for higher clay soil and 8–11 for lower clay (Silva et al. 2011). Total C and total N were obtained with a Hydra 20–20 mass spectrometer coupled to an automatic N analyzer (ANCA-GSL; Sercon Co., Crewe, Cheshire, UK). P was determined by Mehlich-1 and colorimetry; K and Na were determined by Mehlich-1 and photometry; Ca and Mg were determined by KCl extraction and atomic absorption. SB: sum of bases; CEC: cation exchange capacity; BS: base saturation.

rainfall was 18% higher (1646 mm) than that of the previous years. The 90-ha experimental plantation encompasses a range of soil physical and chemical attributes (Table 1). Soils are very deep oxisols (Soil Survey Staff 2010) developed on cretaceous sandstone in the upper part of the study site (750 m above sea level) with low clay content (≈ 20%) and basaltic material at the lowest elevation (725 m above sea level) with high clay content (≈ 40%, Figure 1). Nitrogen and potassium contents were 31 and 125% higher in the clayey soil downslope than in the sandy soil upslope, respectively. Phosphorus contents, pH, and cation exchange capacity (CEC) values were similar for these two sampled locations (differences 10 mm in diameter) of 20 trees spanning the range of sectional areas were fully excavated using a backhoe tractor, close to plots 1–4 and 8–11. After excavating, coarse roots and stumps were manually collected and separated from the soil. Each component of each tree, above- or belowground, was weighed on site individually, and representative samples were collected and dried at 65 °C until constant weight for dry mass determination. Stumps and coarse roots were washed in the field prior to the weighing procedure to minimize mineral soil contamination. Dry biomass (Bi,j) of component j (stump, coarse root, stem, bark, branch, and foliage) of tree i was estimated using the model:

Bi , j = a j + b j(Di2Hi )cj + ε i , j (1)

where aj, bj and cj are estimated parameters, and εi,j denotes the residual error not explained by the model.

The equations were fitted for each component by sampling location based on site index (SI, mean height of the four dominant trees at the age of 6 years; Marsden et al. 2010). Plots with SI > 30 m (plots 1–5) and plots with SI ≤ 30 m (7–11) were grouped separately leading to ‘local-specific’ models, or all plots combined leading to ‘global models’. This method aimed to (i) identify the effects of site heterogeneity on the relationship between tree size and component dry biomass, and (ii) provide accurate estimations of tree biomass for each plot. Model fittings were performed using PROC NLP of SAS (SAS Institute, Cary, NC, USA) and maximum likelihood estimations (Saint-André et al. 2005). Model comparisons were based on the Akaike information criterion (AIC  = −2 ML + 2p) and Bayesian information criterion (BIC = −2 ML + p log(n)), where p is the number of parameters included in the model, ML are the maximum likelihood estimates and n is the number of observations. Models with the lowest AIC and BIC were selected. When differences in AIC and BIC were  1 cm in diameter) were estimated using the allometric equations described above. The estimation of the carbon flux associated with the decomposition of stumps from previous rotations (900 stumps ha−1 on average) was based on stump circumference surveys and biomass equations. Decomposition rates were estimated individually for each plot by an exponential decay function (with an average first-order decomposition coefficient of 0.2 year−1, based on Stape et al. 2008). Fine root density sampled down to a depth of 18 m in a chronosequence close to the study site showed that total fine root biomass changed little at the end of the stand rotation (Christina et al. 2011). Furthermore, fine root biomass is considerably lower than coarse root biomass in 6-year-old eucalypt plantations (Laclau et al. 2000, Ryan et al. 2004). Therefore changes in fine root biomass over our study period were neglected. Aboveground net primary production was estimated as the sum of litterfall and biomass increments of each aboveground tree compartment, computed with Eq. (1) using D and H measurements at the beginning and at the end of the study period. Aboveground autotrophic respiration (RA) was estimated from ANPP, assuming constant carbon use efficiency (CUE) of 0.53, like in Epron et al. (2012). This value was reported by Giardina et al. (2003) studying growth and maintenance respiration of stem, branches and foliage on an intensively managed Eucalyptus saligna Smith plantation in Hawaii. Eucalyptus saligna is a species closely related to E. grandis, and CUE

Tree Physiology Volume 32, 2012

v­ alues around 0.5 have been reported for numerous tree species and forest types (Maier et al. 2004, DeLucia et al. 2007, Litton et al. 2007, Stape et al. 2008). Gross primary production (GPP) was calculated as the sum of ANPP, TBCF and RA.

Data analysis Patterns of carbon fluxes and partitioning with increasing GPP over the experimental site gradient were evaluated using simple linear regressions among the 12 plots, performed on the statistical software R 2.9.1 (R Development Core Team 2009). The coefficients of determination (R2) and root mean square errors (RMSE) were used to assess the strength of the relationships. The probability level used to determine significance was P ≤ 0.05. The data used to generate the regressions were tested for normality and variance homoscedasticity.

Results Allometry The heterogeneity of the experimental site affected the relationships between tree dimensions (diameter and height) and DM of the components. Except for the bark component, which was accurately predicted with a global allometric model, the AIC criterion showed that the dry biomass of all the other living components was better predicted using local-specific models (Table 2).

Forest growth and carbon stocks Stand productivity differed strongly across the experimental site. Site index ranged from 28.0 m (plot 7 located upslope on soil with a low clay content) to 34.6 m (plot 1 located downslope on soil with a higher clay content). Stand basal area followed the same pattern, ranging from 23.0 m2 ha−1 on plot 7 to 31.9 m2 ha−1 on plot 1 (increase of 39%; Table 3). Plot ­elevation showed a strong negative relationship with both SI (R2 = 0.73, P =  0.0004) and basal area (R2 = 0.78, P = 0.0002). The forest carbon stock, considering tree components (above and belowground), forest floor (O horizon) and stumps from the previous rotation, but excluding soil organic matter, ranged from 85 to 121 Mg C ha−1 across the gradient of productivity, with a coefficient of variation of 9.2% (Table 3). On average, across all plots, coarse roots (>1 cm), stem wood, stem bark, branches, and leaves accounted for 10.7, 76.3, 6.4, 3.6, and 3.0%, respectively, of the total carbon stored in tree biomass at the end of the study. Plots with higher C stored aboveground also showed higher ANPP (R2 = 0.35, P = 0.05).

Carbon fluxes and partitioning A strong variability in carbon fluxes and partitioning was observed across the 12 studied plots (Table 4). Stand GPP

Carbon partitioning across a gradient of productivity  701 Table 2. ​Final weighted dry biomass (DB) models for each component of the forest plantation. Numbers in the sampling location column represent the destructive sampling campaigns combined to generate the final weighted dry biomass models. Significance level (P) of all models were  30 m (models from sampling on plots 1–5). Table 3. ​Biometric characteristics and carbon stock for the forest plantation components on the 12 plots. Plots are classified by ascending order of elevation. Plot # Plot elevation (m) Site index (m, 6 years) Basal area (m2 ha−1, 6 years) C stock (Mg ha−1)  ​ ​Forest  ​ ​ ​ ​Foliage  ​ ​ ​ ​Branch  ​ ​ ​ ​Bark  ​ ​ ​ ​Stem  ​ ​ ​ ​Coarse root (diameter  > 1 cm)  ​ ​Forest floor  ​ ​ ​ ​Leaf  ​ ​ ​ ​Branch  ​ ​ ​ ​Bark  ​ ​ ​ ​Fragmented fraction  ​ ​Stumps Total

1 726

2

3

4

5

6

7

8

9

727

730

731

740

746

10

11

12

747

755

755

756

756

756

34.6 31.9

34.0 28.1

33.5 30.8

33.7 28.3

33.1 28.6

32.5 28.4

28.0 23.0

29.3 23.8

29.2 23.7

29.6 24.3

29.0 24.0

31.3 23.0

2.9 3.5 7.1 84.3 10.1

2.4 2.9 6.0 70.8 8.4

2.7 3.3 6.7 80.0 9.6

2.5 3.1 6.1 73.5 8.8

2.4 2.8 6.2 73.1 8.7

2.3 2.7 6.0 70.4 8.4

2.6 3.2 4.6 55.8 9.6

2.8 3.5 4.9 59.7 10.3

2.9 3.6 5.0 60.9 10.5

2.9 3.7 5.0 61.2 10.6

2.8 3.5 4.9 59.8 10.3

1.8 2.2 4.7 55.1 6.6

1.3 4.7 0.6 2.9 3.4 120.7

1.1 3.7 0.4 2.6 6.5 104.9

1.4 4.5 0.4 3.0 3.6 115.2

1.1 3.6 0.5 2.6 3.3 105.1

1.2 3.9 0.4 2.9 — 101.7

1.2 5.3 0.6 2.4 6.3 105.7

1.4 4.5 0.9 2.5 5.8 90.8

1.2 4.6 0.5 2.1 7.6 97.3

1.4 4.1 0.6 1.8 8.1 98.9

1.3 4.4 0.4 1.8 8.1 99.3

1.4 4.7 0.7 2.1 8.8 99.0

1.6 5.1 0.7 2.2 5.3 85.3

ranged from 2971 g C m−2 year−1 on plot 2 to 4132 g C m−2 year−1 (+39%) on plot 3. These two plots were also the most contrasted for ANPP, which attained 1208 g C m−2 year−1 on plot 2 versus 1791 g C m−2 year−1 (+48%) on plot 3, and for stem wood NPP (554 g C m−2 year−1 on plot 2 versus 923 g C m−2 year−1 (+67%) on plot 3). Although SI and basal area were strongly correlated with plot elevation, GPP, ANPP and stem wood NPP were not. Stem wood NPP showed a significant positive linear relationship with estimated forest GPP (R2 = 0.82, Figure 2a); note, however, that stem wood NPP was one of the measured variables used to estimate GPP.

Total belowground carbon flux ranged from 497 g C m−2 year−1 (plot 4, low elevation and median GPP) to 1235 g C m−2 year−1 (plot 7, higher elevation and high GPP) and showed no significant relationship with GPP (R2 = 0.10, P = 0.33) or ANPP (R2 = 0.08, P = 0.36). A significant positive relationship was found between plot elevation and TBCF (R2 = 0.34, P = 0.05). The two main component fluxes of TBCF were soil CO2 efflux (FS) and litterfall (FA). FS ranged from 1053 to 1615 g C m−2 year−1 with no correlation with ANPP (R2 = 0.12, P = 0.28), while FA increased linearly from 434 to 693 g C m−2 year−1 with increasing ANPP (R2 = 0.57, P = 0.005). Mean coarse root

Tree Physiology Online at http://www.treephys.oxfordjournals.org

702  Campoe et al. Table 4. ​Carbon fluxes and partitioning for the 12 plots. Plots are classified by ascending order of elevation. Plot # Plot elevation (m) m−2

1

2

3

4

5

6

7

8

9

10

11

12

726

727

730

731

740

746

747

755

755

756

756

756

1536 309 254 150 823 760 1135 536 −48

1208 293 262 100 554 691 1111 500 −57

1791 333 367 169 923 754 1205 673 −62

1642 313 423 112 794 497 1053 693 −32

1586 334 262 136 855 785 1154 555 -

1340 308 250 95 687 1087 1558 484 −134

1528 319 269 106 834 1235 1615 461 −120

1329 302 201 107 719 833 1275 449 −163

1435 303 252 120 760 849 1326 479 −170

1405 293 232 104 777 992 1464 434 −183

1351 302 296 101 653 833 1384 548 −190

1222 325 175 89 633 1024 1456 442 −109

98

66

110

95

102

82

143

124

130

133

112

76

111 1,362

70 1,072

174 1,588

74 1,456

84 1,407

65 1,188

58 1,355

47 1,179

42 1,272

12 1,246

75 1,198

44 1,084

3,658

2,971

4,132

3,595

3,778

3,615

4,118

3,341

3,557

3,643

3,383

3,329

0.536

0.458

0.515

0.483

0.539

0.513

0.546

0.541

0.530

0.533

0.483

0.518

0.225 0.420 0.208 0.372

0.186 0.407 0.233 0.361

0.223 0.433 0.182 0.384

0.221 0.457 0.138 0.405

0.226 0.420 0.208 0.372

0.190 0.371 0.301 0.329

0.202 0.371 0.300 0.329

0.215 0.398 0.249 0.353

0.214 0.403 0.239 0.358

0.213 0.386 0.272 0.342

0.193 0.399 0.246 0.354

0.190 0.367 0.308 0.325

year−1)

Fluxes (g C ANPP  ​ ​Foliage  ​ ​Branch  ​ ​Stem bark  ​ ​Stem wood Total belowground C flux  ​ ​Soil CO2 efflux  ​ ​Litterfall  ​ ​Stump dry biomass ​ change  ​ ​Coarse roots change (diameter > 1 cm)  ​ ​Litter layer change Aboveground autotrophic respiration (RA) Gross primary production C partitioning (fraction)  ​ ​Stem wood NPP : ANPP  ​ ​Stem wood NPP  :  GPP  ​ ​ANPP  :  GPP  ​ ​TBCF  :  GPP  ​ ​RA : GPP

Figure 2. ​Linear increase in stem NPP (a) and C partitioning to stem NPP (b) as a function of increasing GPP. Numbers represent plots.

Tree Physiology Volume 32, 2012

NPP was 106 g C m−2 year−1, with a CV of 23% across the 12 plots. Stump decomposition over the studied year led to a CO2 efflux ranging from 32 to 190 g C m−2 year−1, depending on the plots. Large variations in stumps per hectare were found across the studied plantation, as a result of a spatial variability of land use before the establishment of the commercial eucalypt plantation (data not shown). Higher values of ANPP were observed on plots with lower stump C stocks (R2 = 0.38, P = 0.04). The increase in C stock in the forest floor over the study period ranged from 12 to 174 g C m−2 year−1, showing a significant positive linear relation with ANPP (R2 = 0.45, P = 0.02). Carbon partitioning to ecosystem components also varied within the stand (Table 4). Carbon partitioning to ANPP varied between 0.37 (plot 12) and 0.46 (plot 4). Carbon partitioning to stem production ranged from 0.19 to 0.23, presenting a trend of linear increase with GPP (Figure 2b). In contrast, the fraction of primary production partitioned belowground was not correlated with GPP, while C partitioning to leaf production showed a significant negative correlation with GPP (Figure 3, Table 5). Comparing the two most contrasting plots in terms of GPP (plot 3 versus plot 2), the 67% difference in stem wood NPP was a combined result of higher photosynthesis (GPP was 39% higher in plot 3 than in plot 2) and higher C flux

Carbon partitioning across a gradient of productivity  703 ­ artitioning to stem wood NPP (0.22 and 0.19 in plots 3 and 2, p respectively).

Discussion Tree allometry The variability of topography and soil physical and chemical attributes had a significant impact on tree dimension and mass, affecting the relationship between D2H and DM, requiring local-specific models for each aboveground component (except for bark) and coarse roots.

Root mean square errors of global biomass models for stem wood, branches, leaves and coarse roots were on average 36% higher, compared with local-specific models. Applying ‘global models’ instead of ‘specific models’, neglecting effects of variable nutrient and water availability, may lead to errors in biomass estimations (Saint-André et al. 2005, Sicard et al. 2006). Treatment-specific biomass equations were also considered more accurate than global equations for most of the tree compartments in Acacia–Eucalyptus mixed-species plantations (Laclau et al. 2008) and in Eucalyptus plantations fertilized with different levels of NaCl on K-deficient soils (Almeida et al. 2010). These results highlight the fact that studies using a single biomass model for different treatments can improve the accuracy of their results by fitting treatment-specific biomass models.

Carbon stocks, fluxes and partitioning

Figure 3. ​Linear increase in leaf NPP (a) and decrease in C partitioning to foliage NPP (b) as a function of increasing GPP. Numbers represent plots.

Considering that the studied eucalypt plantation is even aged and received identical silvicultural treatments (seed lot, soil preparation, weed control and fertilizations), the large range of observed carbon stocks suggests that the variability in soil physical and chemical attributes and water availability across the study site strongly influenced forest productivity. Aboveground net primary production observed for our seedorigin Eucalyptus plantation (average of 1448 g C m−2 year−1 and maximum of 1791 g C m−2 year−1) was high compared with tree plantation productivity around the globe (Giardina et al. 2003, Maier et al. 2004, Litton et al. 2007, du Toit 2008). Compared with other intensively managed Brazilian Eucalyptus plantations, our ANPP values matched those of other sites with similar age, stocking, and fertilization regimes across climatic regions and soil types (Laclau et al. 2008, Stape et al. 2008, Ryan et al. 2010, Epron et al. 2012, Nouvellon et al. 2012). Improved genetic materials and enhanced silvicultural practices in Brazilian eucalypt plantations have made it possible to reach the highest productivities in the world for commercial forest plantations (Gonçalves et al. 2004, Stape et al. 2010). Carbon fluxes were substantially influenced by the spatial heterogeneity of the studied commercial plantation. Even though chemical analyses did not show large differences in nutrient contents between the two soil types (Table 1), the considerable differences in soil texture were reflected in the

Table 5. ​Regressions of the linear relationships (Y = a + b × GPP) between GPP and C partitioning to main ecosystem components. C partitioning

a

b

R2

P

RMSE

Mean value

Foliage NPP : GPP Branch NPP : GPP Stem bark NPP : GPP Stem wood NPP : GPP ANPP : GPP TBCF : GPP RA : GPP

0.148 0.077 0.022 0.120 0.367 0.307 0.326

−0.000017 −0.000001 0.000003 0.000025 0.000010 −0.000019 0.000009

0.755