Plant Ecology

, Volume 213, Issue 7, pp 1081–1092

Tree growth and death in a tropical gallery forest in Brazil: understanding the relationships among size, growth, and survivorship for understory and canopy dominant species

Authors

    • Departamento de BiologiaUniversidade Federal de Lavras
  • Robin Chazdon
    • Department of Ecology and Evolutionary BiologyUniversity of Connecticut
  • Bruno S. Corrêa
    • Centro Federal de Educação Tecnológica de Minas GeraisCampus IX–Nepomuceno
Article

DOI: 10.1007/s11258-012-0067-8

Cite this article as:
van den Berg, E., Chazdon, R. & Corrêa, B.S. Plant Ecol (2012) 213: 1081. doi:10.1007/s11258-012-0067-8

Abstract

We studied the relationships between size, growth, and survival for two functional groups, the persistent canopy and understory dominant tree species in a tropical gallery forest in Southeastern Brazil. In 28 plots of 10 × 30 m we tagged, identified, and measured the diameter of all trees with diameter at ground level ≥5 cm in 1993/1994, 1998, and 2004. We choose the three dominant canopy species (Protium spruceanum, Copaifera langsdorffii, and Pera glabrata) and two dominant understory species (Ixora brevifolia and Trichilia emarginata) for the comparisons. We assessed the relationship between previous growth rates and mortality, variation in growth and mortality rates among size classes, and temporal correlations in growth rates. Trees (whole community) with null or negative growth had a higher mortality, although this effect was not significant for individual species. Growth patterns were consistent along consecutive periods of evaluation for canopy species, but not for understory species. Canopy species had higher and more variable growth rates than understory species, which we attribute to greater access to light at the canopy level and also to a wider range of light conditions experienced during growth. Canopy species and one understory species, T. emarginata, showed accelerated growth as they became larger. Mortality rates were higher for the smallest trees for the community overall and for P. spruceanum.

Keywords

Functional groupForest stratificationForest dynamicsTropical forestTree growth

Introduction

Understanding the relationships among growth, mortality, and size for different functional groups of trees is a key step toward understanding tropical forest dynamics and structure (Chazdon et al. 2010; Rozendaal et al. 2010; Rüger et al. 2011; Terborgh et al. 1997; Uriarte et al. 2010). Several papers have evaluated these relationships at the level of tree communities and populations. Most of these approaches are related to detailed inventories conducted in the last decades within a network of large permanent plots in the tropics (see CTFS in Condit et al. 1995, 2000, 2005, 2006), but others resulted from other permanent plots distributed across the tropics (e.g. Fashing et al. 2004; Guimarães et al. 2008; Higuchi et al. 2008; Laurance et al. 2009; Taylor et al. 2008). Although some of these papers addressed relationships among growth, mortality, and size at the species-level (e.g., Baker and Bunyavejchewin 2006; Clark and Clark 2001; Condit et al. 1993; Lewis et al. 2004; Rozendaal et al. 2010; Rüger et al. 2011), approaching those aspects using functional groups is an exciting perspective that needs to be better explored because it changes the focus from taxonomic entities to ecologically meaningful groups (Bohlman and O’Brien 2006; Feeley et al. 2007; Poorter et al. 2008).

More recently, quantitative traits and their responses in terms of growth, recruitment, and mortality have been used to classify species into functional groups. Wood density, maximum size, position in the forest stratification, and leaf traits have been strongly emphasized (Bohlman and O’Brien 2006; Chazdon et al. 2010; Gourlet-Fleury et al. 2005; Hérault et al. 2011; King et al. 2006; Köhler et al. 2000; Poorter et al. 2008; Wright et al. 2007, 2010). The populations’ dynamics in terms of growth, recruitment and mortality patterns in clearly defined environmental conditions and/or successional stages are particularly important indicators of the species’ functional roles in a community (Baker and Bunyavejchewin 2006; Chazdon et al. 2010; Clark and Clark 2001; Rüger et al. 2011).

Approaching forest dynamics in a robust and quantitative way is strongly limited by the number of sampled individuals and by the time period of evaluation (Poorter et al. 2008). Most tree species in diversified tropical forests present low densities, which restricts population-level studies to all but the most abundant species (Condit et al. 2006; Rüger et al. 2011). On the other hand, the most abundant species are believed to have a central role in the community ecology relationships (Balvanera et al. 2005; Sato 2009), and long term data about their dynamics should provide fundamental information about the functioning of the community. Our aim is to develop a deep understanding of functional differences between persistent canopy and understory species (see Methods for definitions) in tropical forest and to elucidate causal links between tree growth and mortality. We focused on five dominant species (three canopy ones and two understory ones) in an undisturbed gallery forest in Brazil based on three consecutive inventories, spanning more than 10 years.

Our objectives were to compare growth and mortality of tree species that occupy the understory forest stratum but that differ in the stratum and environmental conditions where adult individuals thrive. We compared canopy dominant species to understory dominant species to evaluate if the probability of mortality is related to growth rates for the whole community and for the studied species. We also analyzed at the species-level if individual trees maintain their growth rates patterns across measurement periods and compared variation in growth and mortality among size classes. Finally, we evaluated variation in species growth across the main environment gradient in the forest, from the streamside to the border habitat.

We hypothesized that trees experienced negative or null growth during the first period would be more prone to die during the second period and, conversely, that trees with positive growth during the first period would have reduced mortality rates during the second period. For the survivors, we assumed that the growth pattern for the trees (positive or negative or null) persists along the consecutive periods. We also postulated that canopy species grow faster than understory species and that canopy species grow faster when they are larger, with their crowns closer to the canopy. On the other hand, understory species are predicted to grow monotonically with size. Canopy species, because they are more dependent on light opportunities, are predicted to show higher growth variability than understory species. We expected that mortality rates would be higher in the smaller size classes of canopy species, but would be independent of size in understory species. We further postulated that species growth patterns would change across the main environment gradient in the area, from the streamside to the habitats closer to the forest border with the surrounding grassland, following decreases in soil moisture and increases in light availability.

Methods

Study area

We studied a 7.55 ha area of riparian forest, characterized as a gallery forest (IBGE 1993), located in the municipality of Itutinga (21°21′ S and 44°36′W, altitude of 920 m), Minas Gerais State, Brazil. It follows a narrow stream tributary to the Rio Grande and has sharp boundaries with montane grassland (Fig. 1). During the wet season, no parts of the forest experience flooding or have their soil saturated by water table elevation (van den Berg and Oliveira Filho 1999). The adjacent montane grassland corresponds to the same kind of vegetation that covers most of the region. Most of the studied forest has apparently not suffered extensive disturbance over the past years, showing no signs of selective logging.
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-012-0067-8/MediaObjects/11258_2012_67_Fig1_HTML.gif
Fig. 1

Map of the studied gallery forest (21°21′ S and 44°36′W, municipality of Itutinga, state of Minas Gerais, Brazil). The contour lines (vertical distances in meters from the stream level) are shown close to the map’s frames

The climate was classified as Cwb of Köppen, with wet summers and dry winters. Data from the Meteorological Station of Lavras (21°13′40″S, 44°57″50″W, 918 m of altitude) obtained between 1960 and 1992 show a mean annual temperature of 19.6 °C, with monthly means varying between 16.0 °C in June, and 21.8 °C in February. Annual mean rainfall of 1517.0 mm is concentrated (93 % of total) between October and March; the monthly mean rainfall varies from 19.2 mm (July) to 293.3 mm (January) (van den Berg and Oliveira Filho 1999).

The soils of the area were classified as Dystric Cambisol, on steeper slopes (sample blocks A, B, and D in Fig. 1) and Plinthic Ferralsol, in flatter areas (block C) (FAO-UNESCO 1988). The predominant parent material is the mica-slate.

Sample design

Between February 1993 and May 1994, we inventoried the tree community by laying out 28 plots of 10 × 30 m displayed along the topographic gradient from the portion close to the water course (Streamside) to the portion close to the boundaries of the forest with the grassland (Border), sampling also the intermediary area (Middle) (van den Berg and Oliveira Filho 1999). The rectangular plots were laid out within blocks with their longer side perpendicular to the topographic gradient to reduce intra-plot environmental heterogeneity (Causton 1988). Across this gradient, soil moisture decreases and light levels increase (van den Berg and Santos 2004).

Every stem with DGL (diameter at ground level) ≥5 cm was recorded, following the same criteria of other forest surveys carried out in the Upper Rio Grande Region (Pereira et al. 2006). The identification of trees was based on a former floristic survey (van den Berg and Oliveira Filho 2000). Their circumference at ground level was measured (precision of 0.5 cm).

The inventory was repeated in 1998 (average of 4.42 year from the first one) and, again, in 2004 (6.33 year). As we lacked precise dates for the measurements by plot, we used the average period between successive inventories in growth rate calculations. Although this probably increased the error in the growth estimates, we assumed that errors were randomly distributed, since plots on the main environmental gradient (streamside, middle, and border) were measured during each inventory.

The decision for the rather unusual measurement height (circumference at ground level) was done before 1992 by the initial research team that initiated the first inventories in the region. We were obligated to repeat this measurement during the next inventories to allow comparisons. Because large trees with buttresses are very rare in these gallery forests, we did not have problems using the ground level measurements. In the few cases of low buttresses, we measured the trees just above the buttresses.

Based on Importance Value data from van den Berg and Oliveira Filho (1999), we selected the three most important canopy species and the two most important understory species to analyze their behavior in terms of growth, mortality, and size classes. The canopy species were Protium spruceanum (Benth.) Engler (Burseraceae), Pera glabrata (Schtt.) Bailllon (Euphorbiaceae), and Copaifera langsdorffii Desf. (Fabaceae-Caesalpinioideae). The selected understory species were Ixora brevifolia Benth. Arg. (Rubiaceae) and Trichilia emarginata (Turcz.) C.DC. (Meliaceae). We defined canopy species as those whose adult individuals reach the canopy and have their crowns directly exposed to solar irradiance. We defined understory species as those whose larger and mature individuals remain below the forest canopy. We considered both canopy and understory species as persistent species because one can find them abundantly in the smaller size classes evaluated (van den Berg and Oliveira Filho 1999) and they show no clear decrease in terms of number along the evaluated period (data not published).

We calculated the absolute increment growth rates by dividing the difference in diameter between two consecutive periods by the time in years between the two measurements.

Data analysis

We choose to perform all analyzes related to mortality simply by comparing the observed and expected distribution of dead trees by classes (size or previous growth) using the χ2 test. The possibility of analyzing explicit mortality rates by classes was impaired by the low number of dead trees, since we would need to use replications (plots) to have variance measurements. To analyze how growth rates affected the mortality of trees, we created growth categories using the first interval data for the whole community. We first ranked the trees from the ones with most negative to the ones with most positive growth. Using this order, we divided the trees with negative growth into two groups with equal number of trees. We did the same with the ones with positive growth, creating a total of five growth categories: (1) strong negative growth (82 trees), (2) weak negative growth (83 trees), (3) null growth (373 trees), (4) weak positive growth (694 trees), and (5) strong positive growth (694 trees). Using χ2 test, we compared the observed number of dead trees in the second interval in each category with the expected number calculated assuming that mortality was independent of previous growth. For each species, we used the χ2 test to determine if the mortality (proportion of dead trees) during the second period was higher for plants with null or negative than for ones with positive growth in the first period.

We analyzed the persistence of growth pattern from the first period to the second using non-parametric Spearman’s correlation test. The growth values did not conform to a normal distribution (Shapiro–Wilk test), therefore, we could not use Pearson’s correlation test. Significant positive correlation (we did not find any negative correlation) indicated the growth pattern in the first period tended to persist in the second. To check if canopy species grow faster than understory ones, we compared the focal species growth using a Kruskall–Wallis test followed by a Gao’s non-parametric multiple test (Gao et al. 2008). We used the non-parametric approach because the data were not normally distributed (Shapiro–Wilk test).

To analyze mortality in relation to size classes for the entire community and for each studied species, we divided the trees into three diameter size classes with equal ranges. As expected, there was a strong decline in the number of trees from the first to the third class, but we opted for that approach considering that mortality should vary with size and access to light (Coomes and Allen 2007a; Coomes et al. 2011). We tested whether overall community and species-specific mortality patterns among the size classes were different from a random pattern using a χ2 test that compared the observed proportion of dead trees by diameter classes with the expected proportion supposing a random mortality. Because, for some species, the numbers of dead trees were too low for the two larger size classes, we also performed the χ2 test combining these last two classes.

We compared the growth variability among species using box plots of species growth rates. We analyzed if growth was related to size by correlating initial DGL (first inventory) with growth rate for the whole period (10.35 year), using Spearman’s correlation.

We carried out all statistical analysis in the R program version 2.13.1. We performed the Gao’s test in the Mutoos package (R program). The graphics were carried out in R.

Results

For the tree community as a whole, mortality during the second period was dependent on previous growth rates (Fig. 2). Individuals with negative (strong or weak) growth or null growth during the first period had higher mortality than expected and trees with strong positive growth had less mortality than expected. Only individuals with weak positive growth had mortality similar to what would be expected for a random pattern. However, mortality was not dependent on growth rates for any of five studied species (Table 1). The community as a whole showed higher mortality in the smallest diameter class (5–25 cm) than in the larger size classes. Larger size classes had fewer dead trees than expected (Table 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs11258-012-0067-8/MediaObjects/11258_2012_67_Fig2_HTML.gif
Fig. 2

Comparison between the observed and the expected number of dead trees by growth classes for the whole community. Growth classes (cm year−1): −1.2243 ≤ class 1 < −0.0720, −0.0720 ≤ class 2 < 0, 3 = 0, 4 = 0 < class 4 ≤ 0.2160, 0.2160 < class 5 ≤ 3.7448

Table 1

Comparison between observed and expected proportion of dead trees for individuals that had null or negative growth during the first evaluated period

Species names

Observed proportion of dead trees

Expected proportion of dead trees

χ2

pa

Protium spruceanum

0.1087

0.0522

1.9386

0.1638

Pera glabrata

0.1250

0.0827

0.3000

0.5839

Copaifera langsdorffii

0.0556

0.0341

0.2443

0.6211

Ixora brevifolia

0.0323

0.0531

0.0137

0.9069

Trichilia emarginata

0.0000

0.0141

aχ2 = Chi square test, p = significance level

Table 2

Comparison among diameter (DGL, diameter at ground level) classes in terms of proportion of dead trees, for the community and for each species

Diameter classesa, b

Community

Protium spruceanum

Pera glabrata

Copaifera langsdorffii

Ixora brevifolia

Trichilia emarginata

Total number in the 1st inventory

1926

230

133

176

113

71

1

156 (147.6)c

12 (8.3)

9 (7.3)

6 (4.5)

4 (3.9)

1 (0.8)

2

6 (12.7)

0 (2.8)

2 (2.9)

0 (1.3)

2 (1.9)

0 (0.1)

3

0 (1.7)

0 (0.9)

0 (0.8)

0 (0.2)

0 (0.2)

0 (0.1)

χ2

6.2166

5.539

1.6472

2.1338

0.2328

0.3193

p

0.0447

0.0627

0.4389

0.3441

0.8901

0.8524

1

156 (147.6)

12 (8.3)

9 (7.3)

6 (4.5)

4 (3.9)

1 (0.8)

2 + 3

6 (14.4)

0 (3.7)

2 (3.7)

0 (1.5)

2 (2.1)

0 (0.2)

χ2

5.1771

4.1262

0.6609

0.9696

0.0000

0.0000

p

0.0229

0.0422

0.4161

0.3248

1.0000

1.0000

aDiameter classes (cm) vary with the category: Community: 1 = 5 < 25, 2 = 25 < 45, 3 = 45 < 65; Protium spruceanum: 1 = 5 < 17, 2 = 17 < 29, 3 = 29 < 41; Pera glabrata: 1 = 5 < 18, 2 = 18 < 31, 3 = 31 < 44; Copaifera langsdorffii: 1 = 5 < 24, 2 = 24 < 43, 3 = 43 < 62; Ixora brevifolia: 1 = 5 < 11.5, 2 = 11.5 < 18, 3 = 18 < 24.5; Trichilia emarginata: 1 = 5 < 9.5, 2 = 9.5 < 14, 3 = 14 < 18.5

2: Chi square test, p: significance level

cThe values are number of observed and expected (in parenthesis, calculated based on random mortality) number of dead trees for diameter classes

All three canopy species (P. spruceanum, P. glabrata, and C. langsdorffii) showed significant positive correlation between growth at the first and the second periods. On the other hand, the two understory species (I. brevifolia and T. emarginata) did not show any significant correlation between time periods (Table 3; Fig. 3b as example). This pattern persisted when species where pooled together by functional groups (canopy and understory species) and when we excluded individuals of canopy species that were larger than the largest individual of understory species so that functional groups’ behavior could be compared in the same scale of environment variation (Table 3; Fig. 3a).
Table 3

Values of Spearman’s correlation (rho) between DGL growth at first and second interval (DGL growth 1–2 × 2–3) and between DGL at first inventory and DGL growth across the whole period (DGL × DGL growth 1–3) for group of species (Canopy and Understory) and for species

Species and group of species

 

DGL growth 1–2 × 2–3 (cm year−1)

DGL (cm) × DGL growth 1–3

Habitat

rho (all trees)a

rho (DGL ≤ 26.11 cm)a

rho (all trees)a

rho (DGL ≤ 26.11 cm)a

Canopy species

Whole area

0.3647***

0.3854***

0.2686***

0.2651***

Border

0.4847***

0.5017***

0.2777**

0.1928*

Middle

0.2493***

0.3334***

0.3662***

0.4242***

Streamside

0.3384***

0.2718*

0.1768*

0.0163

Understory species

Whole area

0.0816

 

0.0375

 

Border

0.0279

 

−0.0233

 

Middle

0.0844

 

−0.0185

 

Streamside

0.1589

 

0.3630**

 

Copaifera langsdorffii

Whole area

0.4438***

0.4622***

0.3085***

0.3312***

Border

0.6400***

0.5583***

0.4498***

0.2916*

Middle

0.2682*

0.3933**

0.3233**

0.3982**

Streamside

0.4043*

0.2353

0.1755

0.2493

Protium spruceanum

Whole area

0.3099***

0.3458***

0.3081***

0.2384***

Border

0.3601*

0.4455**

0.0798

0.1085

Middle

0.3168***

0.3317***

0.5796***

0.4820***

Streamside

0.1546

0.1621

0.0774

0.0775

Pera glabrata

Whole area

0.2302*

0.2532**

0.2451**

0.2429*

Border

0.3366**

0.3818**

0.2762*

0.2088

Middle

0.1248

0.1568

0.3264*

0.3280*

Streamside

−0.3659

−0.3541

−0.3958

−0.2607

Ixora brevifolia

Whole area

0.1279

 

−0.0443

 

Border

0.1691

 

0.0912

 

Middle

0.2096

 

−0.0932

 

Streamside

0.1059

 

0.2718

 

Trichilia emarginata

Whole area

−0.0334

 

0.3189**

 

Border

−0.8660

 

−0.5000

 

Middle

−0.1815

 

0.1586

 

Streamside

0.2308

 

0.5438**

 

The results are present for the whole area and for topographic gradient habitats (Border, Middle and Streamside). For Canopy species rho values are also present for trees with DGL ≤26.11 cm, which corresponds to the maximum DGL found for Understory species in the area

aSignificance levels: p ≤ 0.05: *, p ≤ 0.01: **, p ≤ 0.001: ***

https://static-content.springer.com/image/art%3A10.1007%2Fs11258-012-0067-8/MediaObjects/11258_2012_67_Fig3_HTML.gif
Fig. 3

Relationships between first period (DGL growth 1–2) and second period (DGL growth 2–3) of growth for the Canopy and Understory trees (with species pooled together by their respective groups) (a), for Copaifera langsdorffi and Trichilia emarginata (b), and relationships between initial DGL and whole period of growth (DGL growth 1–3) for the Canopy and Understory species pooled together (c) and for Copaifera landsdorffii and Trichilia emarginata (d). See Table 4 for Spearman’s correlation values and significance

When we analyzed the data by different habitats we found that, for canopy species, the number of significant positive correlation between growth at the first and second period increased from streamside to border. C. langsdorffii was the only canopy species that showed significant correlation between the two periods in the streamside, but only when all the individuals were analyzed together. P. glabrata only showed significant correlation in the border (Table 3). Both understory species lacked significant correlations between growth rates in first and second periods when we analyzed their data by habitat (Table 3).

The growth rates varied among the species (Kruskal–Wallis χ2 = 68.6204, p = 4.44E − 14), with canopy trees growing faster than understory species (Table 4; see also Fig. 4). Within the canopy group, P. spruceanum had higher growth rates than the other two species. Growth rates did not vary significantly between the two understory species. Canopy species clearly had higher growth variability than the understory species (Fig. 4).
Table 4

Comparison among species growth rates, using the non-parametric multiple test of Gao after verification of significant differences among the species by Kruskall–Walis test

 

Protium spruceanum

Pera glabrata

Copaifera langsdorffii

Ixora brevifolia

Trichilia emarginata

Comparison summarya, b

a

b

b

c

c

Protium spruceanum

Gao = 2.7669

Gao = 4.4902

Gao = −7.9173

Gao = −7.1609

p = 0.0121

p = 3.8545e–05

p = 2.9601e–13

p = 7.6844e–11

Pera glabrata

 

Gao = 1.4588

Gao = −4.4646

Gao = −3.9563

p = 0.1457

p = 5.0572e–05

p = 3.3045e–04

Copaifera landsdorffii

  

Gao = −3.0755

Gao = −2.6061

p = 6.9645e–03

p = 1.9786e–02

Ixora brevifolia

   

Gao = 0.3443

p = 0.7310

aSpecies followed by different letters had different growth rates. The letters order corresponds to the decreasing growth order

bGao: Gao’s test value, p: significance level

https://static-content.springer.com/image/art%3A10.1007%2Fs11258-012-0067-8/MediaObjects/11258_2012_67_Fig4_HTML.gif
Fig. 4

Dispersion of growth for the five studied species. Species code: first two letters correspond to functional group (Cn canopy, Un understory); last two letters correspond to the initials of species name (Cl = Copaifera langsdorffii, Pg = Pera glabrata, Ps = Protium spruceanum, Ib = Ixora brevifolia and Te = Trichilia emarginata)

The significant positive correlation between DGL and growth during the whole period (10.5 year) found for the canopy species was stronger in habitats farther away from the creek and closer to the forest border. (Table 3; see also Fig. 3c). In the streamside habitat, the temporal correlation was weak (all individuals together) or non-significant (individuals with DGL ≤26.11). At the species-level, C. langsdorffii showed a significant positive correlation between DGL and growth for border and middle habitats (see also Fig. 3d). P. spruceanum only showed a significant correlation in the middle habitat. P. glabrata presented a significant temporal correlation for middle and border habitats, but, in the last case, only for all individuals together (Table 3). The understory species, T. emarginata and I. brevifolia did not show significant correlation between DGL and growth, except for T. emarginata in the streamside (Table 3; see also Fig. 3d). The smaller number of sampled individuals for understory species may have decreased the change of finding significant correlations for the comparisons by habitats. Probably this effect was strongest in the border, where the understory species are rarer.

Among species, P. spruceanum showed a result similar to the overall community, with the mortality concentrated in the first diameter class (5–17 cm) and absent in the larger classes. No other species showed differential mortality among size classes (Table 2).

Discussion

Most of our initial hypotheses were confirmed. Overall, trees with null or negative growth had higher mortality, although this effect was not significant for individual species. Growth patterns tend to persist over consecutive periods of evaluation for canopy species. Trees with higher or lower growth during the first period tended to maintain higher or lower rates, respectively, in the second one. This behavior varied across the topographic gradient, and was stronger in the areas farther way from the creek and closer to the forest border. In contrast, growth patterns between consecutive periods were not correlated for understory species. Canopy species had higher and more variable growth rates than understory species, which can be attributed to greater access to light at the canopy level and to a wider range of light condition experienced during the growth. Our data confirmed the idea that canopy species show accelerated growth as they became larger and their crowns probably experience better light conditions. This pattern was more accentuated in the border and middle habitats, probably because of higher light levels. The hypothesis that understory species would grow monotonically with increasing size was only confirmed for I. brevifolia. T. emarginata did experience higher growth rates in larger sizes (mainly in the streamside habitat), indicating that even understory species can be favored by better light conditions.

Our data support the hypothesis that mortality rates should be higher for the smallest trees, for the community as a whole. These results concur with other studies that small individuals have higher rates of mortality in forests (Coomes and Allen 2007a, b; Wyckoff and Clark 2002).

The canopy trees in our study grew faster and had higher growth variability than the understory species. Canopy species attain higher light conditions than understory species, enabling higher growth rates. Growth and mortality of tree populations are strongly related to each other and depend on species traits and environmental conditions. Light availability increases markedly from the forest floor to the canopy, and competition for this resource is considered a strong driver for observed patterns of growth or mortality (Coomes and Allen 2007a, b; Coomes et al. 2011). Also, canopy gaps (from falling of trees to herbivory attach, going through loss of branches) temporarily generate light increase that provide opportunities for increased growth (Baker and Bunyavejchewin 2006; Coomes and Allen 2007a). On the other hand, trees with different strategies, as canopy and understory species, must experience and explore light opportunities in different ways (King et al. 2006). Furthermore, differences in functional traits (e.g., wood density and leaf traits) and allometric relationships in terms of biomass allocation to height and diameter growth or crown expansion also affect rates of tree growth (see Bohlman and O’Brien 2006; Poorter et al. 2008; Uriarte et al. 2010), since canopy species tend to grow faster than understory ones even within similar size classes (King et al. 2006). At least, for wood density, based on the published information, there are no clear differences among the species. Wood density varied from 0.60 to 0.70 g cm3 for C. langsdorffii, 0.65 to 0.69 g cm3 for P. glabrata. The unique value found for Protium spruceanum was 0.56 g cm3, but possible with a large variation. We did not find any data for I. brevifolia or for T. emarginata, but, for Ixora spp., the values go from 0.69 to 0.96 g cm3 and for Trichilia spp. from 0.34 to 0.90 g cm3 (Zanne et al. 2009).

Canopy trees experience a greater range of light conditions than understory species, and should present accelerated growth rates as they became larger (Brienen et al. 2006; Clark and Clark 2001; Coomes and Allen 2007a; Coomes et al. 2011; King et al. 2006; Poorter et al. 2008; Rozendaal et al. 2010; Rüger et al. 2011; Wyckoff and Clark 2002). Their growth rates must be more variable because of the more extensive light gradient experienced by them (Hérault et al. 2011). Confirming this, all canopy species we analyzed showed increasing growth rates with size as well as higher variation on the growth rates when they were compared to understory species. This behavior was accentuated closer to forest border, probably because of the higher light intensity there (van den Berg and Santos 2004). Surprisingly, T. emarginata, an understory species, showed higher growth with larger diameter, indicating that, understory species also benefit from the probable higher light conditions achieved at larger sizes. Baker and Bunyavejchewin (2006) showed that higher light conditions benefited all species, independent of their functional traits. On the other hand, Rüger et al. (2011) found no relation between growth and light or size variation in the shade-tolerant understory species Faramea occidentalis. This behavior was very similar to our observations for the understory species I. brevifolia, which is, also in the Rubiaceae.

A persistent condition of negative or null growth must increase the likelihood of tree mortality, particularly for species that are poorly adapted to low light conditions in the forest understory (Brienen et al. 2006; Rozendaal et al. 2010; Terborgh et al. 1997). We confirmed this trend for whole community, as mortality in the second period was strongly associated with negative or null growth in the first period. The process that culminates with the death of a tree begins with a previous period of low or negative growth. Kobe et al. (1995) and Wyckoff and Clark (2002) observed that mortality is frequently preceded by the reduction of growth and the growth rates are good predictors of mortality. Possibly because of the small number of dead trees for each species analyzed, we did not find evidence for this pattern at the species-level (see Poorter et al. 2008; Uriarte et al. 2010).

The consistency of diameter growth rates between consecutive periods for canopy species may reflect the persistence of environmental conditions that limit or promote growth (Brienen et al. 2006). Compared to understory species, canopy ones have a broader size variation and must experience light in a more constant way while they grow larger and the overtop suppression tends to lessen. On the other hand, understory species while thriving in the lower forests stratum, must experience more dynamic and transitory light conditions and that can lead to the absence of relationships for growth rates between both evaluated periods. In this sense, the absence of growth persistence from one period to another would be more related to environment conditions than to species traits. If that was true, smaller individuals of canopy species also should not present persistent growth patterns between consecutive periods. But even when we restrict these comparisons for the canopy species to the same size range of understory species, the correlation between growth rates in the consecutive periods was maintained. This result suggests that the persistent growth patterns for canopy species and the lack of persistent growth patterns for understory species along consecutive periods are related to species’ functional traits (Thomas 1996). Apparently, canopy species, even when they are small, had a larger variation in terms of growth rates than understory species and this variation is persistent at the individual level over time. On the other hand, understory species tend to have low and less variable growth rates. This means that canopy species probably respond more strongly and persistently to light opportunities than understory species. To test these ideas would require explicit investigations of the relationships between growth and variation in environment conditions, mainly light, at the individual tree level for the two different functional groups (sensu Clark and Clark 1992). Also, further investigation in terms of physiology, leaf morphology and anatomy, and allometry could point out possible different species strategies for exploring their environments (Hérault et al. 2011). Both approaches are beyond the data we had but point out some directions for further research.

Although we chose the most abundant species registered in the study area to test our hypotheses, our analysis failed to detect effects of growth or size on mortality at the species-level, mainly because of the small number of dead trees registered. Clearly, for evaluating mortality pattern and its relation to population structure and growth much larger population sizes are needed. Therefore, testing these hypotheses for the same functional groups in larger permanent plots would be an exciting possibility. Also, it would be desirable to extend these analyses to other similar gallery forests and other species to corroborate or refute the observed patterns for canopy and understory species. Another interesting approach would be to compare behavior of these species in forests in different successional stages. In this case, the present area can be used as a control, since it is largely undisturbed by human activities.

Acknowledgments

Eduardo van den Berg was supported by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (BEX 6641/10-5). Robin Chazdon was supported by the National Science Foundation of the US Grant DEB-0639393 and DEB-1147434.

Copyright information

© Springer Science+Business Media B.V. 2012