Tropical tree rings reveal preferential survival of fast-growing juveniles and increased juvenile growth rates over time

Authors

  • Danaë M. A. Rozendaal,

    1. Section of Ecology and Biodiversity, Institute of Environmental Biology, Faculty of Science, Utrecht University, PO Box 80084, 3508 TB, Utrecht, The Netherlands
    2. Programa Manejo de Bosques de la Amazonía Boliviana (PROMAB) – Universidad Autónoma de Beni (UAB), PO Box 107, Riberalta, Bolivia
    3. Instituto Boliviano de Investigación Forestal (IBIF), PO Box 6204, Santa Cruz, Bolivia
    Search for more papers by this author
  • Roel J. W. Brienen,

    1. Section of Ecology and Biodiversity, Institute of Environmental Biology, Faculty of Science, Utrecht University, PO Box 80084, 3508 TB, Utrecht, The Netherlands
    2. Programa Manejo de Bosques de la Amazonía Boliviana (PROMAB) – Universidad Autónoma de Beni (UAB), PO Box 107, Riberalta, Bolivia
    3. School of Geography, University of Leeds, Woodhouse Lane, Leeds LS2 9JT, UK
    Search for more papers by this author
  • Claudia C. Soliz-Gamboa,

    1. Section of Ecology and Biodiversity, Institute of Environmental Biology, Faculty of Science, Utrecht University, PO Box 80084, 3508 TB, Utrecht, The Netherlands
    2. Programa Manejo de Bosques de la Amazonía Boliviana (PROMAB) – Universidad Autónoma de Beni (UAB), PO Box 107, Riberalta, Bolivia
    Search for more papers by this author
  • Pieter A. Zuidema

    1. Section of Ecology and Biodiversity, Institute of Environmental Biology, Faculty of Science, Utrecht University, PO Box 80084, 3508 TB, Utrecht, The Netherlands
    2. Programa Manejo de Bosques de la Amazonía Boliviana (PROMAB) – Universidad Autónoma de Beni (UAB), PO Box 107, Riberalta, Bolivia
    Search for more papers by this author

Author for correspondence:
Danaë M. A. Rozendaal
Tel: +31 30 253 6876
Email: d.m.a.rozendaal@uu.nl

Summary

  • Long-term juvenile growth patterns of tropical trees were studied to test two hypotheses: fast-growing juvenile trees have a higher chance of reaching the canopy (‘juvenile selection effect’); and tree growth has increased over time (‘historical growth increase’).
  • Tree-ring analysis was applied to test these hypotheses for five tree species from three moist forest sites in Bolivia, using samples from 459 individuals. Basal area increment was calculated from ring widths, for trees < 30 cm in diameter.
  • For three out of five species, a juvenile selection effect was found in rings formed by small juveniles. Thus, extant adult trees in these species have had higher juvenile growth rates than extant juvenile trees. By contrast, rings formed by somewhat larger juveniles in four species showed the opposite pattern: a historical growth increase. For most size classes of > 10 cm diameter none of the patterns was found.
  • Fast juvenile growth may be essential to enable tropical trees to reach the forest canopy, especially for small juvenile trees in the dark forest understorey. The historical growth increase requires cautious interpretation, but may be partially attributable to CO2 fertilization.

Introduction

Most juvenile trees in the understorey of tropical forests are strongly light-limited, resulting in very low diameter growth rates (Chazdon & Fetcher, 1984; Clark & Clark, 1999). Slow-growing juveniles experience higher mortality rates (Wyckoff & Clark, 2002) and remain longer in the understorey where risk of damage from falling debris is high (Clark & Clark, 1991). They may therefore have a much lower chance of reaching the canopy, compared with fast growers (Enright & Hartshorn, 1981; Swaine et al., 1987; Baker, 2003). If fast growers preferentially reach the canopy, this implies that extant canopy trees will on average have grown more rapidly as juveniles compared with extant juveniles (Fig. 1a); or, when expressed against time, that juvenile growth rates realized in the distant past (i.e. in extant adults) will have been higher than those attained in the recent past (i.e. in extant juveniles; Fig. 1b). In a temperate forest, Landis & Peart (2005) indeed found higher juvenile growth rates for extant canopy trees compared with extant juveniles using the same approach.

Figure 1.

 Graphical representation of the hypotheses. (a, b) ‘Juvenile selection effect’ and (c) ‘historical growth increase’.

Another factor that may influence juvenile tree growth in tropical forests is climatic change (Clark, 2007; Lloyd & Farquhar, 2008; Malhi et al., 2008). There is strong evidence that the growth rates of tropical forest trees have increased over the last few decades (Laurance et al., 2004b; Phillips et al., 2008), probably as a result of increased atmospheric CO2 concentration (cf. Lloyd & Farquhar, 2008). Thus, for juvenile trees in tropical forests, one would expect that growth rates in the distant past (i.e. in extant canopy trees) would have been lower than those in the recent past (i.e. in extant juveniles); the ‘historical growth increase’ (Fig. 1c).

Clearly, the ‘juvenile selection effect’ and a ‘historical growth increase’ have opposing consequences for temporal patterns in juvenile growth (Fig. 1). It is unclear which of these factors has a stronger effect. This may depend on the size of the juvenile trees under consideration. For instance, one could expect the juvenile selection effect to be stronger for juvenile trees in the forest understorey, as these typically experience higher mortality risk (e.g. Lieberman et al., 1985; Condit et al., 1995b). On the other hand, a historical growth increase may also be greater in smaller trees, as CO2 fertilization is expected to have a stronger effect on light-limited trees (Lloyd & Farquhar, 2008). Artificial CO2 enrichment indeed increased seedling growth in a forest in Panama (Würth et al., 1998), although this was not compared with the responses of larger individuals. It is therefore likely that both the juvenile selection effect and a historical growth increase influence growth patterns of juvenile trees, although their relative magnitudes remain unclear.

Here, we present the results of a tree-ring study on five tree species from the Bolivian Amazon in which we tested whether temporal patterns in juvenile growth rates are consistent with the juvenile selection effect and/or a historical growth increase in various diameter classes. Tree rings provide a very effective tool for this purpose, as historical growth rates over the last few centuries can be obtained, a period much longer than that for which plot data are available (Condit, 1995; Clark, 2007). Specifically, we tested the hypotheses, as depicted in Fig. 1, that juvenile growth rates increase with extant tree size (‘juvenile selection effect’) and that juvenile growth rates increase over time (‘historical growth increase’). To our knowledge, this is the first study to test the juvenile selection hypothesis and evaluate centennial-scale growth changes for tropical forest trees.

Materials and Methods

Study sites and species

Fieldwork was carried out in three lowland semi-deciduous moist forests in Bolivia: Los Indios, La Chonta and Purísima, which are 200–600 km apart. Los Indios (10°26′S, 65°33′W) and Purísima (11°24′S, 68°43′W) are situated in the department of Pando, La Chonta (15°47′S, 62°55′W) in the department of Santa Cruz. Mean annual precipitation is similar for Purísima and Los Indios, at c. 1700 mm yr−1, and somewhat lower for La Chonta, at 1580 mm yr−1; La Chonta is located in the transitional region from moist to dry forest. All sites experience a dry season (< 100 mm per month) from May to September.

Forests in the three locations have canopy heights of 25–35 m. Stem densities vary somewhat across sites: 367 ha−1 (of stems of diameter at breast height (dbh) > 10 cm) in La Chonta (Peña-Claros et al., 2008), 423 ha−1 in Los Indios (Toledo et al., 2008), and c. 544 ha−1 in northern Bolivia in forest similar to that at Purísima (R. J. W. Brienen, unpublished data).

Our study included five tree species (Supporting Information Table S1): Cedrela odorata L. (from Purísima), Cedrelinga catenaeformis (Ducke) Ducke, Clarisia racemosa Ruíz & Pavón, Peltogyne cf. heterophylla M. F. Silva (all three from Los Indios) and Pseudolmedia laevis (Ruíz & Pavón) J. F. Macbride (from La Chonta). All species tolerate shade, but to different extents (cf. Brienen & Zuidema, 2006). Cedrela odorata and C. catenaeformis are light-demanding, while C. racemosa, P. cf. heterophylla and P. laevis are shade-tolerant species. The annual character of the juvenile and adult rings was proven using C14-bomb peak dating for C. catenaeformis, C. racemosa and P. cf. heterophylla (C. C. Soliz-Gamboa, unpublished results), and rainfall correlations for C. odorata (Brienen & Zuidema, 2005) and P. laevis (see Note S1 and Fig. S1).

Sample collection and ring-width measurements

Selective logging had taken place during 2002 in Purísima, and during 2006 in both Los Indios and La Chonta. Fieldwork in Purísima and Los Indios was carried out just after selective logging, and for La Chonta just before. For each species, we randomly selected 60–69 trees of < 50 cm dbh and 29–53 trees of > 50 cm dbh (not possible for P. laevis; because of small stature only three trees of > 50 cm dbh were included). We selected individuals of < 50 cm dbh such that they were evenly distributed among diameter categories. In order to reduce spatial autocorrelation in growth rates, selected individuals needed to be at least 20 m apart. Damaged juveniles (< 10 cm dbh) were not included. In C. catenaeformis we checked for hollowness by pre-coring trees, and preferentially selected nonhollow trees to allow measurement of juvenile rings.

Discs or increment cores were obtained from the selected individuals. In the case of large trees (> 50 cm dbh), we always collected discs from stumps of felled trees. For small trees, increment cores were obtained for C. odorata (2–3 per tree), and discs for all other species. Sampling height varied from 0.5 to 1.5 m in C. odorata and was kept constant in all other species, at either 0.5 or 1 m.

A digital picture of each disc was taken, to calculate fresh disc area using pixel-counting software (SigmaScan Pro 5.0, Systat Software Inc., Chicago, Illinois, USA). Discs were air-dried and sanded with progressively finer sandpaper to a grit of 1000. Rings were marked in three or four radii (one to three for C. odorata) – selected to correspond to total disc area – using a stereomicroscope (×6.3–40). Every tenth ring was connected over the whole disc to control for errors in ring identification. Ring widths were measured perpendicular to the ring boundaries using a LINTAB 5 (Rinntech, Heidelberg, Germany) measurement device. For C. catenaeformis, C. racemosa and P. cf. heterophylla the last ring (2006, growth period 2006–2007) was left out for trees of < 50 cm dbh, as this ring was formed after selective logging. For all species we succeeded in cross-dating part of the samples. Cross-dating proved difficult or impossible for the juvenile phase as a consequence of nonclimatic influences on growth, such as suppression and release of growth caused by canopy dynamics (cf. Brienen & Zuidema, 2006). This means that juvenile growth may have been slightly overestimated as a result of missing rings for slow-growing trees. Missing rings may have led to a slight underestimation of both the juvenile selection effect and the historical growth increase. Ring widths of all radii of the discs were averaged, and corrected for desiccation and irregular growth using the mean radius of the fresh disc area. For the increment cores of C. odorata, average ring width was corrected with the dbh of the tree. Ring width was then converted to basal area growth, as this provides a better proxy for biomass growth.

Data analysis

We calculated the juvenile growth rate for each individual as the median basal area growth rate within a narrow range of diameters (i.e. diameter classes 0–1, 1–2, 2–3, 3–4, 4–6, 6–8, 8–10, 10–15, 15–20 and 20–30 cm) for all study species, except for the faster growing C. odorata and C. catenaeformis. For these species we used wider diameter classes at smaller sizes: 0–5 and 5–10 cm. By using different diameter classes for these species, we maintained a similar number of rings per diameter class for all species. The use of a median growth rate per diameter class allowed comparison of growth rates among trees at the same diameter, and thus at the same ontogenetic stage. For each individual and each diameter class, we also calculated the mid-point of the ages of all rings, which we termed year before present (YBP). The calendar year for the ‘present year’ in the calculation of YBP was 2002 for C. odorata and 2006 for all other species.

We then related median growth rates (diameter and basal area) for a given diameter class to the extant diameter of all the trees using Pearson’s correlation. For instance, for the diameter class of 0–1 cm, we correlated the median growth rates of all individuals with their current, extant size (as in Fig. 1a). A positive correlation between these variables indicates that juvenile growth rates of extant adults have been higher than those of extant juveniles, consistent with the ‘juvenile selection effect’.

Similarly, we related median growth rates to YBP in order to determine whether growth rates of similarly sized trees have changed over time. Relations between growth rate and YBP are expected to be comparable to those for growth rate and extant diameter, as age and size are not independent over the studied extant size range. Temporal autocorrelation in growth and persistent growth differences among trees may cause the relation of growth and extant dbh to deviate from that between growth and YBP, but the effect is probably small for juvenile growth (Brienen et al., 2006). We checked whether correlations for diameter growth differed to those for basal area growth: this was not the case for 76 out of 80 correlations performed. All statistical analyses were conducted in spss 16.0 (SPSS Inc., Chicago, Illinois, USA).

Results

Ages and juvenile growth rates

Growth rates varied considerably across species, with C. odorata and C. catenaeformis showing much faster juvenile growth than the other species (Fig. 2). On average, juveniles of C. catenaeformis and C. odorata reached 10 cm diameter in 26 and 24 yr, respectively, a period more than three times shorter than that for the other species, which took 77–80 yr (Fig. 2).

Figure 2.

 Growth trajectories to 10 cm diameter for five tropical tree species. Grey lines indicate trees > 10 cm diameter at breast height (dbh), and black lines indicate trees < 10 cm dbh. Hollow trees were not included, as for those trees exact age could not be determined. (a) Cedrela odorata (86 trees), (b) Cedrelinga catenaeformis (63 trees), (c) Clarisia racemosa (95 trees), (d) Peltogyne cf. heterophylla (85 trees), and (e) Pseudolmedia laevis (60 trees).

Juvenile growth and extant diameter

In C. odorata, extant trees of > 10 cm dbh showed a much wider range in growth trajectories than trees of < 10 cm dbh (Fig. 2a). Cedrelinga catenaeformis juveniles of 1–10 cm dbh showed slower initial growth than extant trees of > 10 cm dbh of the same species (i.e. less steep growth trajectories; Fig. 2b). For C. racemosa, P. cf. heterophylla and P. laevis this pattern was less marked (Fig. 2c–e). In four out of five species we found significant correlations between juvenile growth and extant diameter (Figs 3,4,7). Positive as well as negative correlations were observed. For C. catenaeformis, C. racemosa and P. cf. heterophylla, positive correlations between juvenile growth and extant diameter were found mostly in the smallest diameter classes, whereas negative correlations were observed at larger sizes for C. racemosa and P. laevis. For all diameter classes > 6 cm, we only found significant correlations for C. catenaeformis in the diameter classes 5–10 and 20–30 cm (Fig. 7). For C. racemosa and P. cf. heterophylla a positive correlation between growth and extant diameter was found for the 0–1 cm diameter class (Fig. 3c,d), showing that extant large trees had faster juvenile growth than extant juveniles. Similarly, juvenile growth was positively correlated with extant diameter in C. catenaeformis juveniles (0–5, 5–10, 15–20 and 20–30 cm diameter; Figs 3b,4b,7b).

Figure 3.

 Relationship between basal area growth in the 0–1 cm diameter class (0–5 cm for Cedrela odorata and Cedrelinga catenaeformis) and extant diameter for five tropical tree species. Pearson’s correlation coefficient is indicated. *, < 0.05; **, < 0.01; ***, < 0.001; ns, not significant.

Figure 4.

 Relationship between basal area growth in the 2–3 cm diameter class (5–10 cm for Cedrela odorata and Cedrelinga catenaeformis) and extant diameter for five tropical tree species. Pearson’s correlation coefficient is indicated. *, < 0.05; **, < 0.01; ***, < 0.001; ns, not significant.

Negative correlations between juvenile growth and extant diameter were apparent in three of the five study species: in C. racemosa (2–4 cm diameter), P. cf. heterophylla (1–2 cm diameter) and P. laevis (2–6 cm diameter) juvenile growth rates decreased significantly with extant diameter (Figs 4c,e,7). Thus, in those species extant juveniles grew more rapidly than extant adult trees had done as juveniles.

Juvenile growth and time before present

As expected, the relationships between juvenile growth rate and YBP (Figs 5,6) were similar to those for growth and extant diameter (Figs 3,4), although for the former we found slightly more significant correlations (Fig. 7). Positive correlations between juvenile growth and YBP were found for the growth of small juveniles of C. racemosa and P. cf. heterophylla (0–1 cm dbh; Fig. 5c,d) and C. catenaeformis (0–5 and 5–10 cm dbh; Fig. 5b). This is in accordance with the finding presented above that larger extant trees grew more rapidly at the juvenile stage than extant juveniles in these diameter classes.

Figure 5.

 Relationship between basal area growth in the 0–1 cm diameter class (0–5 cm for Cedrela odorata and Cedrelinga catenaeformis) and year before present (YBP) for five tropical tree species. YBP is the median value per tree. Pearson’s correlation coefficient is given. *, < 0.05; **, < 0.01; ***, < 0.001; ns, not significant.

Figure 6.

 Relationship between basal area growth in the 2–3 cm diameter class (5–10 cm for Cedrela odorata and Cedrelinga catenaeformis) and year before present (YBP) for five tropical tree species. YBP is the median value per tree. Pearson’s correlation coefficient is given. *, < 0.05; **, < 0.01; ***, < 0.001; ns, not significant.

Figure 7.

 Pearson’s correlation coefficients among basal area growth, extant diameter at breast height (dbh, circles) and year before present (YBP, squares) in diameter classes to 30 cm. Nonsignificant correlations are indicated with open symbols, and filled symbols are used for significant correlations (< 0.05). YBP is expressed as number of years before present (median for all rings considered per diameter class). A positive correlation indicates a juvenile selection effect, whereas a negative correlation indicates a historical growth increase. The minimum and maximum sample sizes are indicated in the lower part of the graphs.

For juveniles of intermediate size, we found negative correlations between juvenile growth and YBP in four species (Fig. 7), suggesting increased juvenile growth rates over time. Such correlations were found for juveniles of 10–15 cm diameter in C. odorata, 2–4 cm diameter in C. racemosa, 1–3 and 6–8 cm diameter in P. cf. heterophylla and 2 to 8 cm diameter for P. laevis (Figs 5,6). Thus, except for one diameter class in C. odorata, no negative correlations between juvenile growth and YBP were observed for trees of > 8 cm diameter (Fig. 7).

Discussion

Juvenile selection effect

Many researchers have hypothesized that fast-growing juvenile trees in tropical forests have a higher chance of reaching the canopy; the ‘juvenile selection effect’ (e.g. Enright & Hartshorn, 1981; Swaine et al., 1987; Terborgh et al., 1997; Baker, 2003; Brienen & Zuidema, 2006). To our knowledge, our study is the first to actually test this hypothesis for tropical tree species through a direct comparison between current and historical juvenile growth rates. For three out of five study species we found that extant canopy trees have had faster growth as small juveniles compared with extant juveniles, consistent with this juvenile selection effect. Our results suggest that fast growth until a diameter of 1 cm is attained (or 30 cm in C. catenaeformis) increases the chance of reaching the canopy. These findings are in accordance with those of Landis & Peart (2005), who showed higher juvenile growth in extant adult trees compared with extant juveniles in three temperate forest species until a diameter of 5 cm was attained.

Our results probably underestimate the strength of the juvenile selection effect, as all trees included in our sample had already survived to 1 cm diameter. Therefore, the effects of slow growth on survival to 1 cm diameter were not taken into account here. As slow growth leads to low survival (Terborgh et al., 1997; Wyckoff & Clark, 2002), the consequences of slow growth in the 0–1 cm diameter class are probably larger than we determined here. However, finding a juvenile selection effect in trees that have already survived to 1 cm diameter indicates that fast growth until that size is attained has implications for survival to larger diameters. Such enhanced survival of initially fast-growing saplings may be the result of continued favourable growing conditions or a larger leaf area and larger carbohydrate reserves developed by these individuals early in life (cf. Landis & Peart, 2005).

The juvenile selection effect is probably driven by differences in light availability among trees, but we cannot exclude other factors that may have caused a growth decline over time. However, in the case of a climate-related growth decline, we would expect a growth decrease for all species over more diameter classes. It is possible that successful trees have established in high light conditions. This agrees with the results of Brienen & Zuidema (2006), who found fast early growth for Amburana cearensis. Similarly, Baker & Bunyavejchewin (2006) found that almost 40% of the canopy individuals of shade-tolerant Neolitsea obtusifolia had recruited in gaps, while this was > 90% for the shade-intolerant species Melia azederach. Thus, juvenile trees that establish in a canopy gap may have a greater chance of reaching the canopy.

The preferential survival of fast-growing juveniles to large size implies that fast-growers have a higher chance of reaching reproductive size. Thus, fast-growing juvenile trees probably make a large contribution to population growth and fitness (Zuidema et al., 2009). This means that selection pressure would act to increase the capacity for fast growth in juvenile trees. However, the majority of saplings in the understorey of tropical rainforests grow in the shade (Chazdon & Fetcher, 1984; Clark & Clark, 1999) and trajectories to the canopy can be long (Clark & Clark, 2001; Brienen & Zuidema, 2006), putting strong selection pressure on traits that increase survival in shade. Thus, one would expect the juvenile selection effect to be absent or smaller in shade-tolerant species compared with light-demanding species. Landis & Peart (2005) found evidence for juvenile selection also in shade-tolerant species, but it was less strong than in shade-intolerant species. We found that the juvenile selection effect was weaker in two of the three shade-tolerant species, compared with C. catenaeformis, a light-demanding species (Brienen & Zuidema, 2006). Also, the effect in the two shade-tolerant species was restricted to much younger life stages (i.e. up to 1 cm diameter) compared with C. catenaeformis, which needs high light conditions up to a diameter of 30 cm. These differences probably indicate differences in shade tolerance among species. However, we did not find a juvenile selection effect in the light-demanding species C. odorata. This is at odds with the findings of Terborgh et al. (1997), which showed that slow-growing C. odorata juveniles had lower chances of survival. This may partly be attributable to the fact that we did not include trees from 1 to 5 cm dbh for this species, or to less consistency in the height at which samples were taken.

Our results have implications for modelling tropical tree growth. In the presence of a juvenile selection effect, the average growth rate of extant juveniles is unlikely to be the correct measure to use in growth models or projections (e.g. Lieberman & Lieberman, 1985). Some researchers have already recognized this problem, and apply only above-average growth rates in tree growth projections (e.g. Condit et al., 1995a; Terborgh et al., 1997; Laurance et al., 2004a) or include the link between growth and light conditions (Metcalf et al., 2009). Although our sample of tree species was limited, our findings suggest that using above-average growth rates is generally to be recommended. However, the degree to which growth rates need to be adjusted depends on the shade tolerance of the species. Baker (2003) found that for shade-tolerant species mean growth rates may give relatively accurate estimations of tree age, whereas the use of mean growth rates for shade-intolerant species will greatly underestimate long-term growth rates and hence overestimate tree age. To accurately predict long-term growth rates and model growth dynamics, species-specific data on the relation between growth and mortality are needed (e.g. SORTIE; Kobe et al., 1995; Pacala et al., 1996). As a short-cut to such modelling techniques, a comparison of (long-term) tree-ring data and increment data for extant trees may provide valuable insights into the degree to which adjustments of growth rates are needed for different species.

Increasing juvenile growth rates over time?

Juvenile growth rates increased over time in four of the five species in certain size classes (Fig. 7). In temperate forests, similar increases in growth over time have been found (Rolland et al., 1998; Soulé & Knapp, 2006; Voelker et al., 2006; Wang et al., 2006; Johnson & Abrams, 2009; but see Landis & Peart, 2005). A cautious interpretation of our results and those of others is required, as sampling biases may result in spurious correlations between growth and time. Imagine that individual trees differ in long-term growth rates, such that slow-growing trees remain slow-growers, and fast-growers keep growing rapidly (Brienen et al., 2006). Combined with some degree of size dependence in the mortality probability at larger diameters, this will lead to a shorter lifespan of persistently fast-growers compared with slow-growers (i.e. subgroups with different mortality in the population; cf. Sheil & May, 1996). In temperate forests, slow-growing trees had indeed a longer lifespan than fast-growing trees of the same species (Black et al., 2008; Bigler & Veblen, 2009). Thus, when a population of large trees is sampled at a specific moment in time, the subpopulation of trees recruited in the distant past contains a smaller proportion of fast-growing trees than the subpopulation of trees recruited in the more recent past. This may lead to an apparent increase in growth rates in the recent past, even when growth rates did not increase (sampling bias I). A second sampling bias may occur when trees are sampled from some size threshold onwards. Then, slow-growing trees will not have reached this threshold at the moment of sampling, which could mean that more recently the sample may be biased towards fast-growing individuals (sampling bias II). Sampling bias I, II, or their combination may yield spurious positive correlations between growth and time.

Sampling bias I may have affected our results, although we do not have evidence for size-dependent mortality in the study species. There may also have been some impact of sampling bias II, as we sampled a relatively high number of large trees. The sampling threshold in our study was 1 cm dbh (for C. odorata, 5 cm dbh), which is rather low. However, we did not sample in proportion to the population structure, and thus trees in larger size classes have been overrepresented in our sample. If we assume that equal proportions of slow- and fast-growers establish at every time step, fast-growers will reach larger size classes at a younger age than slow-growers. Even when evaluating growth at small tree sizes, we might include more fast-growers than slow-growers and have a bias towards fast, recent growth.

We believe that our findings of historical growth increases are probably not the result of the above-mentioned sampling biases, although we cannot rule out the possibility that there has been some impact. If such sampling biases occurred, this should have led to historical growth increases also for the larger diameter categories, but these were not found. It is clear that the sampling biases described may significantly affect historical growth patterns. A full understanding of their magnitude requires a combination of empirical and modelling studies.

There are various explanations for the observed historical growth increase. Long-term changes in precipitation could have affected tree growth, but this is not likely as precipitation did not increase over the last few decades in our study region (Malhi & Wright, 2004). Another explanation is the occurrence of recent large-scale disturbances leading to increased growth over recent time periods. We did not find large, synchronous increases in growth rates that would indicate large-scale disturbances, nor do we have indications that such disturbances have taken place in our study areas over the last two centuries. A final – and often proposed – cause of historical increase in juvenile growth is the rise in atmospheric CO2. The observed historical growth increase may be consistent with growth stimulation by elevated atmospheric CO2. Remarkably, the historical growth increase in our study was confined to very small trees. This is consistent with the greater growth increase for young trees found in temperate forests (Voelker et al., 2006; Wang et al., 2006), although in those studies only non-light-limited trees were included. We found hardly any indication of a historical growth increase for larger trees. Only for C. odorata were increasing growth rates over time found for trees from 10 to 15 cm diameter, and for none of the species did we observe a growth increase over time in the larger size classes (Fig. 7).

Concluding remarks

The juvenile selection effect and historical growth increase probably interact. This could be the reason why evidence was not found for either of the two effects in certain diameter classes and species, where they may have cancelled out. Thus, not finding a historical growth increase could be the result of a strong juvenile selection effect. Similarly, this interaction may cause an underestimation of the historical growth increase in certain diameter classes.

The strength of the two effects probably varies among species. The juvenile selection effect is likely to be strongest in shade-intolerant species, while a historical growth increase may be stronger in shade-tolerant species. These species-specific patterns may have contributed to the switch from a juvenile selection of small (0–1 cm diameter) trees to a historical growth increase in slightly larger trees which we observed in two shade-tolerant species. We do not know whether a historical growth increase is associated with shade-tolerant species (cf. Kerstiens, 2001) or whether it is just harder to detect in shade-intolerant species that show strong selection of fast-growing juveniles.

Given the importance of changes in tropical forest biomass for atmospheric CO2 concentrations (Malhi et al., 2008), there is a clear need to understand the drivers of long-term growth patterns of tropical trees. Our study shows that tree-ring analysis is a promising tool to evaluate temporal changes in growth of tropical forest trees, at decadal to centennial scales. When combined with measurements on stable isotopes (e.g. McCarroll & Loader, 2004; Hietz et al., 2005) or analysis of recruitment patterns (e.g. Baker et al., 2005), tree-ring analysis may help to unravel the causes of such temporal growth changes.

Acknowledgements

We are grateful to Mart Vlam for help with the ring measurements. We thank Nazareno Martínez, Miguel Cuadiay, Adhemar Saucedo, Don Eugenio Mercado, Jeroen Wiegeraad, Edwin Rodríguez, Jan Rodenburg, Adhemar Cassanova Arias and many others for help with the fieldwork. We are grateful to the staff and personnel of PROMAB-UAB, IBIF and the Universidad Autónoma Gabriel Rene Moreno for logistic support. Logging companies ‘La Chonta Woods’ and ‘Maderera Boliviana Etienne S.A.’ are acknowledged for permission to work in their concessions and for logistic support. Niels Anten, Heinjo During, Marinus Werger and three anonymous reviewers provided constructive comments on earlier versions of the manuscript. DMAR was supported by grant W 01.53.2004.047 from the Netherlands Foundation for the Advancement of Tropical Research (WOTRO).

Ancillary