1Growth patterns and ages of tropical forest trees are strongly governed by temporal variation in light availability. Periods of high growth after canopy disturbances (releases) are necessary for successful canopy regeneration, but their importance cannot be studied without lifetime data. The recent detection of annual rings in tropical forest trees enables such analyses.
2We used tree ring analysis to study lifetime growth patterns and age variation in six Bolivian rain forest species. Our aims were to evaluate the magnitude and sources of age variation of canopy trees, to analyse the frequency of suppression and release events, and to analyse the relation between temporal growth changes and tree age.
3The average age of trees of 60 cm diameter differed threefold between species and by two- to threefold even within species. This variation was mainly explained by variation in passage time through the juvenile categories.
4We used strong relative growth changes to detect release and suppression events. On average, canopy trees experienced 0.8–1.4 releases, with a maximum of 4.
5We distinguished four canopy accession patterns by which trees have attained the canopy (growth without major growth changes, one release event, one suppression event, or several release and suppression events), with increasing time required to attain the canopy. The distribution of trees over categories of canopy accession is therefore closely related to the average age of canopy trees and its variation.
6There were clear differences among species in how trees attained the canopy and in the length of slow-growth periods they experienced, suggesting differences in shade tolerance and growth responses to gaps, which are indicative of life-history differences among non-pioneer tree species.
7Canopy attainment of tropical rain forest trees does not occur by steady growth, but rather by irregular patterns of growth spurts and stand-stills, probably mostly caused by temporal variation in light. Differences in these patterns may largely explain differences in the ages of large tropical rain forest trees.
Many trees in the understorey of tropical rain forests grow at very low rates because of limiting light availability (Chazdon et al. 1996), except when released by canopy disturbances (e.g. tree and branch falls) (Canham 1985; Canham et al. 1990; vanderMeer & Bongers 1996b). Many non-pioneer tree species in tropical forests, particularly those species that attain large stature (canopy and emergent trees), are thought to require such releases from understorey light levels to enable them to reach the canopy (Denslow 1980; Brokaw 1985). The size of gaps, the frequency of gap formation and the growth response of such species to high-light levels, therefore determine the age at which individuals reach the canopy. This age is both an important life-history characteristic, as it usually coincides with the onset of reproduction (Zuidema & Boot 2002), and essential for forest management, as it indicates time needed to replace a canopy tree that has been logged for timber.
The size and number of gaps required for trees to attain the canopy is thought to vary among species due to differences in response to high-light conditions and in shade tolerance (Hartshorn 1978; Chazdon et al. 1996; Canham et al. 1999). Thus, the average growth trajectory of trees towards the canopy probably differs between species (cf. Canham 1985; Poulson & Platt 1989). Information on these trajectories, including the presence and frequency of periods of high growth (releases) and low growth (suppressions), is essential to understand differences in life histories among species. In addition, variation among individuals within a species in the occurrence of releases and suppressions may have important consequences for the age structure of the population.
We used tree rings to reconstruct lifetime growth patterns of six rain forest tree species in order to address the following questions. (i) What is the magnitude of variation in age of large conspecific trees, and what causes this variation? Does this differ among species? (ii) What is the frequency of suppression and release events in growth trajectories of individual trees? Do these frequencies vary among species and size categories? (iii) To what extent do differences in individual growth trajectories influence the time required to reach the canopy?
The selected species are non-pioneers (sensuSwaine & Whitmore 1988), differing in shade tolerance, that grow in northern Bolivia, where the occurrence of a distinct dry season causes them to produce clear annual rings (Brienen & Zuidema 2005). We identified relative growth increases and decreases in ring trajectories, which may indicate releases and suppressions due to canopy dynamics (e.g. Nowacki & Abrams 1997), although other causes are also possible. We then classified trees according to their growth pattern into the canopy and compared these among species. Our study provides reliable information on ages of tropical rain forest trees and is one of the first to unravel the causes of intraspecific variation in tree age.
Materials and methods
study areas and species
The study areas are situated in the northern part of the Bolivian Amazon in the departments of Pando and Beni. The vegetation in both areas is tropical lowland moist forest with a canopy height of 25–35 m, average basal area (d.b.h. > 20 cm) of 15 m2 ha−1 and density of 103 trees hectare−1 (Superintendencia Forestal 1999). The total precipitation is 1760 mm (Cobija, Pando) and 1690 mm (Riberalta, Beni), with a distinct dry season from May until September, when there is less than 100 mm of rain per month.
We collected samples of Amburana cearensis and Cedrela odorata from the private property ‘Purisima’ (11°24′ S, 68°43′ W), 50 km south of the town of Cobija. This area consists of 850 ha of mainly undisturbed tropical moist forest, on undulating terrain.
Samples of Bertholletia excelsa, Cedrelinga catenaeformis, Tachigali vasquezii and Peltogyne cf. heterophylla were collected from several adjacent logging concessions and private areas (10°55′ S, 65°40′ W), approximately 40 km east of the town of Riberalta, again consisting of mainly undisturbed tropical moist forest.
All study species grew clear annual rings (Brienen & Zuidema 2005) and are hereafter referred to by genus only (Table 1 gives full details). None of the species is a ‘classical’ pioneer (sensuSwaine & Whitmore 1988), but species do differ in shade tolerance or successional status. Amburana, Bertholletia, Cedrela and Tachigali were all characterized as relatively light-demanding species, occurring in earlier stages of regeneration and being intermediate in their shade tolerance as a seedling (Poorter 1999; Pena-Claros 2003) and Cedrelinga has previously been characterized as shade tolerant (Poorter 1999). Peltogyne is a shade-tolerant species with a relatively high abundance of saplings in the dark understory (R. Brienen, personal observations). The six species clearly differ in diameter growth potential (Table 1). Amburana, Bertholletia and Peltogyne showed relatively low maximum growth rates, while Cedrela and Cedrelinga had high maxima and Tachigali showed a very high growth potential.
Table 1. Adult stature, growth potential, largest observed diameter (R. J. W. Brienen, unpublished data) and sample size for the six study species. Growth potential is the mean of the five highest annual growth rates observed in different trees (cf. Clark & Clark 1999)
Growth potential (cm year−1)
Largest observed diameter (cm)
Amburana cearensis (Allemao) A. C. Smith
Bertholletia excelsa H.B.K.
Cedrela odorata L.
Cedrelinga catenaeformis (Ducke) Ducke
Peltogyne cf. heterophylla M. F. Silva
Tachigali vasquezii J. J. Pipeloy
sample collection and ring measurements
Between October 2002 and September 2003 we took a cross-sectional wood disc or segment from trees of at least 50 cm (d.b.h.) that had either been felled for timber or had died recently of natural causes. Most samples (> 90%) were taken at a height between 0.4 and 1.5 m, in exceptional cases up to 3.4 m. We tried to take samples above buttresses but, especially in Cedrela and Cedrelinga, buttresses were often more than 3 m high and a relatively large proportion of discs contained buttresses. We used pictures of each complete disc to calculate its surface area (A, cm2) and average diameter (D, cm = SQRT(A/π) * 2).
After air-drying, we polished the discs mechanically with sandpaper up to grit 600. For ring counting and measuring of ring widths we chose two to four radii such that their average length corresponded to the calculated average diameter of the disc. For trees with buttresses, we counted and measured the rings both in buttresses and in parts in-between buttresses.
Tree-rings were marked and every 10th ring was cross-correlated between radii to identify errors in ring marking. The ring widths were measured to the nearest 0.001 mm using a computer-compatible tree-ring measuring system (Velmex Inc., Bloomfield, NY, USA) and a 40 × 10 stereomicroscope. Ring measurements were performed along each of the radii in a straight line, and generally perpendicular to growth boundaries.
Annual ring formation in adult categories of the six study species was confirmed by correlations between ring width and rainfall, by counting tree rings on discs of known ages or by radiocarbon dating (Brienen & Zuidema 2005). However, there is some uncertainty as to the extent to which rings are formed annually in the younger individuals; rings were not cross-dated in the juvenile categories and only for Cedrela has annual ring formation been proven (Dunisch et al. 2002). Tachigali showed no ring formation in the juvenile wood and was thus excluded for analysis of the juvenile categories. We expect ring formation in juveniles of the other species to be predominantly annual, but missing rings or false rings are a potential bias. Missing rings during periods of suppression is a known problem and could yield underestimations of ages and overestimates of early growth (Lorimer et al. 1999). In Peltogyne, some rings may have been missing and ages may therefore be underestimated in this species (Brienen & Zuidema 2005). We do not, however, expect these potential biases to influence the outcome of the results substantially.
calculation of age-diameter relationships and passage time
We calculated the annual diameter increments of each tree by averaging ring-widths along the different radii and multiplying this value by two. Subsequently, we checked the difference between this calculated diameter value and the actual diameter at sample height (obtained from the disc area) to correct for over- or under-estimation of growth rates due to irregular growth patterns around the trunk (buttresses and depressions). The size of the correction for each annual increment value was based on the maximum difference between the increments along the radii for that particular year. Thus, no correction was applied if the different radii showed equal increments in a particular year, while large corrections were applied for those years with relatively high differences between the increments along the radii (e.g. in buttresses vs. depressions). Corrected growth values were validated by comparing them with growth curves calculated from area measurements of 10-year increment periods. The sum of corrections was less than 10% for over 75% of the trees, and amounted to more than 25% for only a few (3%) trees.
For each tree we established age-diameter relationships for its complete lifetime. When rings were lacking in the tree centre, we estimated the distance to the pith and used the average number of rings of the other samples of that species to estimate the age of the first visible ring. In Cedrelinga this was done for a substantial number of samples due to the frequent occurrence of hollow trees. For this species the number of rings to 5 cm diameter was estimated for 20 out of 33 samples. In the other species relatively few samples relied on such estimates (two samples in Amburana, five samples in Bertholletia, six samples in Cedrela and none in Peltogyne). Note that the ages presented here are calculated from stem discs obtained at > 40 cm above the ground and do not include the time required to grow from seedling to sampling height. This means that both average age and its variation are underestimated. The time to reach the minimum sampling height is probably less than 2–3 years for five of the study species (Amburana, Bertholletia, Cedrela, Tachigali and Cedrelinga; Poorter, 1999).
For each size class of 10 cm width between 0 and 60 cm diameter, the median, minimum and maximum passage time was calculated as the number of years spent in a size class. A Kruskal–Wallis test with the Dunn-test was used to test for differences between species in passage times in each size class. In order to assess the contribution of each size class to the eventual variation in ages at 60 cm diameter, we analysed the effect of passage time in each class on age at 60 cm diameter by a multiple regression that included all size classes except for those causing a high colinearity.
analysis of temporal growth patterns
Lifetime growth patterns of individual trees of Amburana, Cedrela, Cedrelinga and Peltogyne were first analysed to test whether the frequency of growth changes differs between different size classes. We then analysed ‘canopy accession patterns’ to test whether and how species differ in the way they grow to the canopy (or to canopy size). For Bertholletia we had insufficient complete growth trajectories and for Tachigali we had no ring data for the smaller size classes.
As we had no information on the relationship between light environment and growth, we defined the criteria of suppression and release as proportional decreases or increases in growth, instead of using an absolute threshold (e.g. Canham 1985; Landis 1999). Such proportional criteria have been used in other studies and proven to be reliable for detecting canopy disturbances in temperate forest trees (Lorimer & Frelich 1989; Nowacki & Abrams 1997; Lusk & Smith 1998; Rentch et al. 2003). We used moving averages to remove long-term age-size relations and very short-term variation in growth rates caused by variation in weather (Nowacki & Abrams 1997). As canopy closure in tropical forest gaps takes, on average, 15 years (vanderMeer & Bongers 1996b), a window of 10 years is considered to be sufficiently short to capture release and suppression effects brought about by canopy opening and closing.
where %GCi = percentage of growth change for year i, M1 = the preceding 10-year mean diameter growth (including the year of change), and M2 = the subsequent 10-year mean diameter growth. Hence, if M1 is the mean over the period 1960–69, the mean for M2 was calculated over the years 1970–79.
We regarded a growth increase of more than 100% as a growth release, and a decrease of at least 50% as a growth suppression. The year of strongest growth change was regarded as the year of the release or suppression event and a growth release lasting for more than 5 years is regarded as sustained. These operational definitions are used, regardless of the actual cause of the growth change (e.g. canopy disturbance, physical damage, or other factors influencing tree growth).
For each tree, we calculated the frequency of growth releases and suppressions in each size class and tested for differences in the frequency of events between understorey (< 30 cm in diameter) and canopy trees (> 30 cm in diameter) using a Wilcoxon signed ranked test. Most trees in our study reached the canopy at around 30 cm in diameter and received full or nearly full light above this, while only few trees below this diameter received full light levels (R. J. W. Brienen, unpublished data; Zuidema & Boot 2002).
Four patterns of canopy accession were distinguished and each individual tree was assigned to one of these patterns.
1‘No sustained release’. Canopy accession without the occurrence of major growth changes. A single temporary growth release (< 5 years) is allowed, as light climates generally improve when trees grow higher into the canopy even in the absence of a canopy opening (Montgomery & Chazdon 2001). Note that this pattern also includes trees that have been in high-light conditions since the start of the ring data.
2‘One sustained release’. Canopy accession through at least one clear release event. This pattern is distinguished to identify trees showing a clear growth increase, which is most likely due to the opening of the canopy. Trees belong to this type if they show one release event, which is either preceded by growth suppression or which is sustained for more than 5 years.
3‘One suppression’; canopy accession takes place after the occurrence of one suppression event. Trees of this type usually have a high initial growth rate, followed by a strong growth decrease and no subsequent growth releases.
4‘Multiple releases and suppressions’; canopy accession takes place through multiple growth releases and suppressions. Any growth release must always be followed by a growth suppression and/or separated from a new growth release by more than 5 years. For trees of this type, successful growth into the canopy involved repeated growth shifts, most probably due to canopy dynamics.
For each of the four species, we calculated the proportion of trees belonging to each canopy accession pattern and the time to reach 30 cm diameter for each combination of species and pattern. Differences between species in the distribution of trees in each of the canopy accession patterns and in the average number of releases were tested by a Fisher exact test. We tested for differences in time to reach 30 cm diameter between species and patterns, using a two-way anova with a Bonferroni post-hoc test for contrasts between species and patterns.
Median and maximum of longest consecutive periods of low growth (i.e. < 2 mm year−1) were calculated for each species. The length of such periods can be interpreted as an indicator of the ability of species to survive periods of suppressed growth (Canham 1985; Orwig & Abrams 1994; Landis 1999). We are aware that the choice of the growth threshold influences the lengths of the periods, but we used the results only for comparisons between species and these did not change when using lower thresholds. In many studies different thresholds were applied for different species based on the occurrence of extant juveniles in different microsites and their growth rates at different light levels (Lorimer et al. 1988; Landis 1999). We could not do this for our species, due to lack of such data. The application of one threshold probably yields equal ranking of species, as shown by Landis (1999).
We observed large differences in growth trajectories both among and within species (Figs 1 and 2). Bertholletia trees may live to over 400 years old (maximum observed, 427 years), while Tachigali usually dies before the age of 60 (25 years are added to the ring data to correct for the time needed to reach 10 cm in diameter; Poorter et al. 2005b). Maximum ages observed for Amburana and Peltogyne were 243 and 254 years, and the oldest tree of Cedrela was 308 years. Although large trees were sampled for Cedrelinga, their ages did not exceed 123 years.
Species differed strongly in mean age-size relations (Fig. 1). Mean ages at 60 cm in diameter varied from 49 years (Tachigali) to more than 150 years (Bertholletia and Peltogyne). Around 15 cm in diameter, mean ages did not vary much among Amburana, Cedrela, Cedrelinga and Tachigali, but age differences increased at larger diameters. The average age-size relations of Bertholletia and Peltogyne were very similar, showing the highest average ages at any diameter.
The species also differed in the shape of individual growth patterns (Fig. 2); most trees of Amburana and Bertholletia showed rather straight trajectories, with little variation (i.e. constant growth rates over time). Many trees of Cedrela, Cedrelinga and Peltogyne showed sigmoid curves, with slow initial growth, gradually increasing at larger diameters and then slowing again towards the largest sizes.
The highest variation in growth trajectories was observed in Peltogyne and Cedrela (Table 2); the coefficient of variation (CV) of ages at 60 cm diameter for these species was more than twice as high as the values for Cedrelinga, while Amburana and Bertholletia showed intermediate values.
Table 2. Part correlation coefficients indicating the effect of passage time through size categories on age at 60 cm in diameter, obtained by multiple regression analysis. The last columns show the mean, minimum and maximum ages at reaching 60 cm diameter and the coefficient of variation (CV, %) and the sample size (n). Categories for which part correlations had a high colinearity were excluded from the analysis (shown as –). All regression models were highly significant (P < 0.001). Significance indications for each of the part correlation coefficients are: NS = not significant, **:P < 0.01,***:P < 0.001. Tachigali and the first classes of Bertholletia and Cedrelinga were not included in the analysis, because of low sample size (< 10; shown as x)
Size categories (cm diameter)
Age at 60 cm diameter (year)
For Bertholletia and Peltogyne the passage times and age calculations start at 1.5 cm diameter, as the rings in this trajectory were missing or invisible in most samples.
Ages expressed here for Tachigali are ages from 10 cm diameter. For comparison with other species add 25 years as indicated in parenthesis, which is the projected time to grow to 10 cm diameter (Poorter et al. 2005b).
passage time and its significance for age variation
In all study species, we found the highest median passage time in the smallest size class (0–10 cm diameter, Fig. 3). The subsequent decrease was more marked in Cedrela, Cedrelinga and Peltogyne than in Amburana, and in Bertholletia the passage times increased again after 40 cm in diameter. Ring data for diameters < 10 cm were lacking for Tachigali, but estimated passage time from repeated diameter measurements is relatively high (25 years; Poorter et al. 2005b) compared with that of subsequent classes.
Differences between the maximum and minimum passage time were large for all species, except for Tachigali. Generally, the largest differences were found in the first size class and decreased towards larger size classes. This pattern is mainly caused by decreasing maximum passage time, as the minimum passage time varied little among the size classes. In Amburana and Peltogyne, the highest maximum passage time was found in the second and third size classes, respectively. Of all species and size classes, the largest absolute difference in passage time was found in the first size class of Cedrela; although trees can grow 10 cm in diameter within 8 years, this can also take over 70 years.
The contribution of passage time through each size class to age at 60 cm diameter was assessed by part correlation coefficients (Table 2). For all species, the highest part correlation coefficients were found in the smallest size classes (0–20 cm diameter). In Amburana, Cedrela and Peltogyne correlation coefficients strongly decreased with size, but no clear pattern was found in Bertholletia and Cedrelinga. These results indicated that growth in the smallest size classes is most important in determining tree age at 60 cm.
temporal variation in growth patterns
Growth trajectories of individual trees are highly variable, both for individual trees over time and among individuals, as shown by the eight examples in Fig. 4 where arrows indicate suppressions and releases as defined by proportional growth changes. Strong growth changes in the trajectories are mostly classified as suppressions and releases. Some trees show no growth releases or suppressions, while others show one or more such events, and there are remarkably long periods of very low growth rates (e.g. Figure 4f, Cedrela, 50 years of very low growth rates before the first release event occurred).
The frequencies of growth releases and suppressions differed among size classes (Fig. 5). Generally, we observed high frequencies of growth releases in the smaller size classes and substantially lower frequencies or no events at all in trees > 40 cm diameter. In Amburana, Cedrela and Peltogyne, the highest frequency of growth releases is observed in the smallest class, with a clear decrease for larger size classes. The average frequency of growth releases in the smallest size class was approximately one out of the five or six trees every decade. The highest frequency of release events in Cedrelinga occurred between 10 and 20 cm in diameter, and was two to six times higher than the frequencies observed in other species in this size class. The frequency of suppression events did not show a clear pattern, and species showed both lower (e.g. Amburana) and higher frequencies (e.g. Peltogyne) in the larger size classes. Cedrelinga trees that reached the canopy experienced few or no suppressions.
The average frequency of release events for all species was significantly higher for understorey trees (< 30 cm in diameter) compared with canopy individuals (30–70 cm in diameter). The frequency of suppressions was only significantly higher for understorey trees in Cedrela (Wilcoxon sign rank test, P < 0.05).
canopy accession patterns
The proportions of trees assigned to each of the canopy accession pattern types differed significantly between species (Table 3, Fisher exact test, P < 0.05), except between Cedrela and Peltogyne. It is remarkable that most trees of Amburana (> 40%) grew into the canopy without sustained releases. In Cedrela, three canopy accession patterns (no sustained release, one sustained release, and multiple releases) were encountered in comparable proportions, while in Cedrelinga nearly 80% of the trees showed one sustained release. Nearly half the Peltogyne trees showed multiple releases, but others grew into the canopy without releases or through one sustained release. Thus, both between and within species there are differences in the way in which trees attain the canopy.
Table 3. Proportions of trees in each of four canopy accession patterns; mean time required to reach 30 cm diameter and average and maximum number of releases to reach 30 cm in diameter. The last two columns show the mean and maximum of longest consecutive periods of slow growth (< 2 mm year−1) before reaching 30 cm in diameter. The distribution over canopy accession patterns differed between all species (pairwise comparisons; Fisher exact tests, P < 0.05), except for the comparison between Cedrela and Peltogyne. Both species and patterns had a significant effect on the time to reach 30 cm in diameter (two-way anova, P < 0.001; Cedrelinga excluded due to missing data). Significant differences are indicated by different capital letters for species and different lower case letters for patterns (Bonferroni post-hoc tests, P < 0.05). Median period of slow growth (< 2 mm year−1) differed among species (Kruskal–Wallis tests, P < 0.05). Different letters indicate differences between species (P < 0.05) using Dunn tests
Canopy accession pattern
Percentage of trees
Time to reach 30 cm diameter (year)
Number of releases until 30 cm
Period with growth < 2 mm year−1 (year)
The time to reach 30 cm and the number of releases in Cedrelinga only cover the growth trajectory from 5 cm diameter to 30 cm.
The time required to reach the canopy (i.e. 30 cm diameter) differed significantly between species and between different patterns (two-way anova, Fspecies = 17.0, Fpatterns = 18.2, P < 0.001; Table 3). Peltogyne needed more time to reach 30 cm diameter than Cedrela and Amburana. Time to reach the canopy for Cedrelinga could not be compared directly with the other species as it was calculated from 5 cm in diameter onwards.
In Amburana, Cedrela and Peltogyne the average time required to grow into the canopy was longest for trees with multiple releases, intermediate for trees with sustained releases and shortest for trees without sustained releases.
The highest mean number of releases to grow into the canopy was found in Peltogyne and the lowest number in Amburana. No significant differences were found between species in the mean number of releases (Kruskal–Wallis tests). Among all trees the absolute maximum of observed releases was found in Cedrela and Peltogyne, with up to four releases before reaching 30 cm in diameter.
An analysis of growth rates revealed that Peltogyne had significantly longer consecutive periods of growth < 2 mm year−1 compared with the other species (Table 3). The maximum period of slow growth was long for Cedrela and Peltogyne (i.e. 47 and 41 years), while it was much shorter for Amburana and Cedrelinga.
Intraspecific variation in age-size relationships was very high. Among the six species, Cedrela showed the highest variation in age per size category. For this species understorey trees of 10 cm d.b.h. may be as old as adult trees of 60 cm in the canopy. Several studies have reported a similar magnitude of variation in age-size relations (cf. Enright & Hartshorn 1981; Villalba et al. 1985; Worbes et al. 2003), thus reaffirming that diameter is a poor indicator of tree age (Harper 1977; Sarukhan et al. 1984).
The variation in ages at larger diameters (60 cm) was mainly caused by variation in passage time of trees through small size classes (< 20 cm in diameter). Growth variation in these size categories is high due to large differences in light conditions among individuals (Chazdon & Fetcher 1984; Montgomery & Chazdon 2001). Thus, the age of large canopy trees is strongly determined by the time they require to become 30 cm in diameter.
Few estimates exist on passage time of juvenile tropical trees through the smallest size classes. Condit et al. (1993) present such data for a subsample of fast-growing species, making comparison with our results difficult. The only estimates appropriate for comparison with our study are projections based on short-term growth data for six non-pioneer, canopy and emergent species in Costa Rica (Clark & Clark 1992, 2001). The projected passage times to reach 30 cm in diameter (34–59 years) based on maximum growth rates are close to those observed by us, although median growth rates predict more than fourfold longer passage times (177–462 years). This difference could be partly explained by interspecific growth differences but the fundamental difference between our sample and that of Clark & Clark (1992, 2001) is probably more important: our study trees represent a subset of successful trees that have reached the canopy, while Clark and Clark's projections are based on all juvenile trees in the forest, including those that will not reach the canopy. Our subset of successful trees may have had above-average growth rates, as has recently been shown for subalpine forest trees (Landis & Peart 2005). Slow growing juveniles have a higher probability of mortality as they stay longer in the understorey (Swaine et al. 1987; Terborgh et al. 1997; Arets 2005) and juvenile mortality is higher for suppressed trees (Kobe et al. 1995; Wyckoff & Clark 2002). Such selection of fast-growing trees may explain the discrepancy in passage time, but direct comparisons of age values obtained for the same species with different methods are required to confirm this (Bormann & Berlyn 1981; Martinez-Ramos & Alvarez-Buylla 1998; Baker 2003). To the extent that the above explanation holds, our results show that the use of median or mean growth rates tends to overestimate tree ages and that tropical forest canopy trees are younger than hitherto assumed.
temporal growth patterns
We used relative growth changes to analyse temporal growth patterns of individual trees. This method has been successfully applied in the analysis of canopy disturbance in temperate forests (Lorimer & Frelich 1989; Orwig & Abrams 1995; Nowacki & Abrams 1997; Lusk & Smith 1998; Rentch et al. 2003). It had not been used for tropical forest trees before, although canopy dynamics similarly govern tree growth in both temperate and tropical forests (Canham et al. 1990). We defined suppressions and releases based on strong and lasting growth differences, making our estimates of the frequency of these events rather conservative (cf. Lorimer & Frelich 1989; Nowacki & Abrams 1997; Rentch et al. 2003). For release events, the frequencies probably exclude causes other than canopy openings, as a doubling in growth rate over 10 years is expected to be related to an increase in light level. For suppression events, both canopy closure and crown damage (branch loss; Paciorek et al. 2000) may be responsible.
For all study species, we consistently found that small trees experienced more releases than large trees per decade. This decreasing frequency may be caused by the larger temporal shifts in light levels for trees in the understorey and subcanopy (Chazdon & Fetcher 1984; Montgomery & Chazdon 2001), but also by the weaker responses of large trees to increased light levels (Lorimer & Frelich 1989). Suppression frequencies, on the other hand, did not decrease with tree size and were also observed for large canopy trees. Crown damage and perhaps liana infestation are the most likely causes for suppressions in large trees (vanderMeer & Bongers 1996a).
Among trees of the same species we observed a high diversity of growth trajectories through the juvenile categories, indicating that trees can reach the canopy in many ways. The classification into four canopy accession patterns allowed us to group trajectories. Trees with ‘no sustained release’ probably experienced rather stable or slightly changing light conditions while growing into the canopy and, as they did not experience suppressions, required the shortest time to reach the canopy. Growth into the canopy through ‘one sustained release’ is most likely brought about by an abrupt and strong increase in light level as a result of gap formation: these trees grew relatively fast, but suppression before the release led to them entering the canopy later than those with no release. Amburana was the only species with individuals showing ‘one suppression’, usually after high initial growth rates. These trees may have been located in a canopy gap early in their life (Lorimer & Frelich 1989), and reached the canopy at a relatively young age. ‘Multiple release’ patterns are an indication of repeated canopy openings, which are often separated from each other by depressed growth rates due to canopy closure (or crown damage) and thus take longest to reach the canopy.
Differences in canopy accession patterns among trees therefore correspond to differences in the age at which they reach the canopy: low ages for trees that grew gradually to the canopy and high ages for trees that showed several periods of slow and fast growth. The distribution of trees of a given species over these four categories of canopy accession is probably related to the life history of the species, but also determines the age structure of canopy trees. For instance, an even distribution of individuals over canopy accession patterns, such as in Cedrela and Peltogyne, leads to a twofold higher variation in age at a certain size compared with Cedrelinga and Amburana, which predominantly followed one or two canopy accession patterns.
Our six study species were all non-pioneers (sensuSwaine & Whitmore 1988), but they differed strongly in longevity, age-size relationship and canopy accession patterns. Tree longevities differed more than sevenfold. The short maximum age of Cedrelinga and Tachigali (123 and 60 years) sharply contrasts with that of Bertholletia (427 years, Zuidema & Boot 2002). Especially in Tachigali, maximum age is low, only 60 years. This monocarpic species has very high growth rates and becomes reproductive at a young age (Poorter et al. 2005b).
Among the six study species, mean ages at 60 cm diameter varied more than threefold. For Amburana, Cedrela, Cedrelinga and Tachigali trees of < 20 cm diameter had comparable size-age relations, but at larger diameter, ages differed strongly among species. This divergence was probably caused by differences in potential growth rate that became apparent at higher light availability (Clark & Clark 1999), or by differences in light availability among species (Poorter et al. 2005a). The differences between species in patterns of suppression and release suggest differences in life history, in spite of the fact that all study species were non-pioneers. These differences result from interspecific variation in shade tolerance (Kobe et al. 1995; Canham et al. 1999) and in response to increased light level (Denslow 1980) and differences in light level (Poorter et al. 2005a). For instance, the absence of suppressions and the short periods of slow growth rates in Cedrelinga are most probably attributable to low survivorship in the shade. Juveniles of this species that have experienced longer periods of suppression may have died and are thus not present in our sample. In contrast, Cedrela and Peltogyne showed long periods of low growth (> 47 and 41 years with < 2 mm year−1) suggesting a higher ability to survive periods of suppression.
Based on ring analysis alone, it is difficult to draw conclusions about the frequency and size of canopy openings that are required for successful regeneration into the canopy of our study species. Such figures could be obtained by combining our results with information on the proportion of juveniles that is present in various forest light environments and their growth rates in these sites. This has been successfully applied in temperate forests (cf. Canham 1985; Lorimer et al. 1988; Cao & Ohkubo 1999; Landis 1999).
Our results show clear differences between non-pioneer species in lifetime growth patterns. In particular, the frequency of releases and the ability to withstand long periods of suppression differ among species. Such differences are not detected using short-term growth data. This points out the relevance of lifetime growth data for the analysis of life histories of tropical forest trees.
We are very grateful to Adhemar Cassanova Arias, Merlijn Janssens, Henri Noordman, Jeanette Pacajes, Anneke Rijpkema, Jan Rodenburg, Vincent Vos and Oliver Yancke for their indispensable assistance with the ring measurements, and to the staff of PROMAB and the ‘field team of Purisima’ for their help with the fieldwork. We thank the Instituto de Geología y Medio Ambiente (IGEMA) from the Universidad Mayor de San Andres (UMSA) in La Paz and Jaime Argollo for the use of their measurement equipment. Niels Anten, René Boot, Heinjo During, Matthew Landis, Francis Putz and Marinus Werger are acknowledged for constructive comments on earlier versions of this paper. This research is part of the Programa de Manejo de Bosques de la Amazonía Boliviana (PROMAB) financed by grant BO 009703 from the Netherlands Development Assistence (DGIS).