• Regeneration of the dominant ectomycorrhizal tree Microberlinia bisulcata in groves in Korup, Central Africa, is very poor. The hypothesis was tested that this species is more shade intolerant than other co-occurring species.
• In two 1-yr trials, each with M. bisulcata and four other species at a nursery close to Korup, growth was measured under five PAR levels, with ± added P and ± watering in the dry season. In parallel experiments the effects of PAR with two R : FR ratios were investigated.
• Increasing PAR had a consistent effect on the rates of increase in plant mass and on changes in the other variables. Doubling soil P, watering and halving the R : FR ratio had almost no effect. However, across species, mass at low PAR and relative growth rate related positively and negatively, respectively, to seed mass.
• One contributing factor for the poor recruitment of M. bisulcata is therefore its low survival and slow growth at low PAR, due to its small seed size. The two codominant ectomycorrhizal grove species of Tetraberlinia, with larger seeds, were less affected by low PAR.
Where a tree species has strong spatial aggregation at the adult stage yet fails to regenerate in situ, major factors are presumably operating to create and maintain the patches and then to restrict recruitment. At a scale of several decades it might be concluded that the species is becoming locally extinct, but at the scale of centuries to millennia such an inconstancy could be part of longer-term processes, ones that are more stable when larger spatial scales and time–space interactions are considered (Newbery & Gartlan, 1996). Dominance-decay cycles are not uncommon in nature (Watt, 1947), and where stochastic extreme events drive vegetation processes, host-predator or host-pathogen cycles might have a reinforcing role. Contemporary evidence of failure to regenerate provides an insight into the general factors that may be acting at more moderate intensities for a majority of forest tree species whose present population imbalances are less easily observable.
Aubréville (1938) put forward the idea that in African forests there were mosaics of species cycling from location to location, as local dominants were replaced by other species and degenerating populations re-establishing elsewhere. This idea fits into the broader framework of ‘pattern and process’ proposed by Watt (1947). It implies that a species that is temporarily successful at a site makes local conditions for its own replacement poor (Swaine & Hall, 1988, Newbery, Songwe & Chuyong, 1998). What biological factors would lead to and control this shifting mosaic? Among the rainforest regions this phenomenon is predominantly found in Africa (Richards, 1996) suggesting that dominance-decay cycles are partly due to continental biogeography (Letouzey, 1968; Schnell, 1976). Where dominance-decay is strong it provides good opportunities to test hypotheses. To date, no general explanation has been put forward to explain the Aubréville phenomenon.
The Atlantic coastal forest (la forêt biafréenne) of Central Africa (Aubréville, 1968; Letouzey, 1968, 1985; White, 1983) is a spatially and temporally complex forest type characterized by patches in which species of the Caesalpiniaceae (‘caesalps’) dominate. These patches are of varying composition and size (typically 1–10 km across), and they show varying degrees of replacement. Some have strong regeneration, tending towards sustained monodominance, whilst others have weaker replacement where dominant species give way to other codominant caesalps (Letouzey, 1968). The Aubréville phenomenon is variable in intensity and to some degree caesalps, as a guild, form self-replacing mosaics amongst regionally more widespread species. The caesalps concerned are all ectomycorrhizal (largely of the tribe Amherstieae; Newbery et al., 1988; Alexander, 1989). This situation is more complex than monodominance, shown for example by Gilbertiodendron dewevrei in the Congo and Zaire (Connell & Lowman, 1989; Hart et al., 1989; Hart, 1995), because these forests are codominated by a guild of species (Newbery et al., 1998).
In Korup National Park such codominating patch dynamics are represented by Microberlinia bisulcata A. Chev., Tetraberlinia bifoliolata (Harms) Haumann and Tetraberlinia moreliana Aubrév. [= T. korupensis Wieringa spec. nov.] (Gartlan et al., 1986; Newbery et al., 1988; Newbery & Gartlan, 1996; Wieringa, 1999), forming well-defined ‘groves’. Only one other species has been found to be relatively more abundant inside the grove than outside of it in Korup, Calancoba glauca, which is not ectomycorrhizal (Newbery et al., 1997). Although the site has a very high rainfall (c. 5.2 m yr−1) it is strongly seasonal with December to February typically having < 250 mm; (Newbery et al., 1998), and some species in the forest are deciduous. M. bisulcata, a huge emergent tree, is the principal grove species in terms of basal area abundance but it is almost totally failing to regenerate beneath the parent trees (Newbery et al., 1998). In its diameter frequency distribution, M. bisulcata is not so unusual in this forest type as c. 47% of tree species at Korup, and neighbouring Douala-Edea (Gartlan et al., 1986; Newbery et al., 1986; Newbery & Gartlan, 1996), lack small trees. M. bisulata is distinctive in the southern part of Korup because it is restricted to groves, and with no regeneration outside of them (Newbery & Gartlan, 1996).
The dominance-decay for M. bisulcata is therefore a more restricted form of a general phenomenon envisaged by Aubréville (1938). What is, however, paradoxical is that this species is strongly ectomycorrhizal (Newbery et al., 1988) and its contribution to total basal area and overall phosphorus dynamics is considerable, by increasing surface soil organic P to 1.6-fold (Newbery et al., 1997). With respect to nutrient acquisition and cycling it is perhaps the best adapted species to this very P-poor site (Newbery et al., 1997), yet it has the worst regeneration of the main canopy and emergent trees. The other two less abundant codominant caesalps (T. bifoliolata and T. moreliana) form a quasi-successional sequence behind M. bisulcata. They have smaller trees as mature adults, but show better regeneration as saplings and poles (Newbery et al., 1998). This forest subtype appears to have its own special internal vegetation dynamics.
There are two important questions. (1) Does M. bisulcata have a particularly low shade-tolerance? The generally low-light environment at the forest floor within groves may be the reason for its very low survivorship. (2) Is M. bisulcata more light-responsive than other forest species? Newbery & Gartlan (1996) have suggested that an extreme past event, in the form of a set of droughts, may have led to the establishment of the groves. Higher understorey light conditions might have selected M. bisulcata and the two Tetraberlinia spp. In this paper, growth experiments are reported for M. bisulcata and eight other tree species under nursery conditions. The aim was to assess their species’ relative responses to increasing light quantity, change in light quality, soil P level and dry-season water availability, with regard to whether they formed groves or not and their ectomycorrhizal status. In Green & Newbery (2001), the same species were studied as forest transplants.
Methods and Materials
Two types of experiment were conducted on an open site close to the Mana River footbridge entrance to Korup National Park (5°1′ N 8°52′ E). A 0.3-ha area of 4-yr-old oil palms was cleared in December 1995. The site had previously supported mature palms and been fertilized with ground rock phosphate until 1982. This site was selected because it had the same soil type as that in the main forest study site within the Park (on transect P) and, lying 8 km west, it was the closest possible place outside of the forest at which to establish a nursery. The seasonal patterns in rainfall, temperature and radiation are shown in Fig. 1 for the years of the reported experimental trials.
Photosynthetically active radiation (PAR) at the nursery site was measured using a quantum sensor and data logger (SKP 215, SDL 2850; Skye Instruments, Powys, UK) mounted on the roof of a guard post. PAR measurements every 30 s were integrated over 30-min periods, for most of the days over which the trials ran. Sporadic battery failure unfortunately resulted in missing values (33% of days) and these had to be estimated with data from a Gunn-Bellani radiation integrator (Baird and Tatlock, Essex, UK) sited at the Bulu meteorological station (5 km to the SW). The radiation integrator was calibrated with the quantum sensor readings using all nonmissing days (PAR = 0.91 + 3.42(RAD) −0.079(RAD2); r2 = 0.69, P < 0.001, n = 711; PAR in mol m−2 d−1, RAD in ml water d−1 evaporated).
Species and pretreatment conditions
One experiment investigated the effects of light quantity, P and water supply, and the other, running concurrently, investigated the effects of light quantity and light quality. Each experiment comprised two 1-yr trials. Each trial used three ectomycorrhizal and two nonectomycorrhizal species, with one species common to both trials. The species in trial 1 were M. bisulcata, with Anthonotha fragrans (Bak.f.) Exell and Hillcoat, Berlinia bracteosa Benth., Irvingia gabonensis (O’Rorke) Baill. and Vitex ferruginea K. Schum. & Thonn. In trial 2 they were M. bisulcata, with Oubanguia alata Bak.f., Strephonema pseudocola A. Chev., T. bifoliolata and T. moreliana (Table 1).
Table 1. Species selected for the seedling experiments in two trials from those occurring in the 82.5-ha study plot on transect P, Korup National Park, Cameroon; ranked according to their abundance in the forest, showing germination type, whether they formed groves, and mean ± SE (n) seed mass. The number and basal area (BA, m2 ha−1) for large trees ≥ 50 cm diameter at breast height (dbh), and basal area for all trees ≥ 10 cm d.b.h., in the 82.5 ha plot (DM Newbery and colleagues, unpublished data). Species’ mycorrhizal associations are indicated (from Newbery et al., 1988). Species are ranked according to BA of ≥ 50-cm trees
1 Abbr. to families: Caes, Caesalpiniaceae; Comb, Combretaceae; Irvi, Irvingiaceae; Scyt, Scytopetalaceae; Verb, Verbenaceae. 2 EM = ectomycorrhizal; nonEM = nonectomycorrhizal. 3 Types: 1, epigeal/phanerocotylar/foliaceous; 2, epigeal/phanerocotylar/fleshy; 3, hypogeal/phanerocotylar/fleshy; 4, hypogeal/cryptocotylar; after Hladik & Miquel (1990) and Garwood (1996). 4Mass of endosperm or cotyledons. Estimated mass for V. ferruginea; SEs not available for three species. 5These species are known to have arbuscular mycorrhizas (Newbery et al., 1988).
Species were chosen from those abundant in a large permanent plot in Korup on Transect P (Newbery et al., 1998, DM Newbery & colleagues, unpublished data). These species constituted nine of the 13 most abundant, in terms of basal area (trees ≥ 50 cm diameter at breast height (dbh)), within the study plot (Table 1). Basal area abundances for trees ≥ 10 cm d.b.h. highlight the importance of the abundant tree O. alata in the subcanopy. Seeds were collected from six to 20 adult trees of each species within the study plot (except for V. ferruginea), and they were well mixed before sowing. The selected species fruited in different years, at different times of year and seed of most species could not be stored due to a lack of dormancy. Fruit production therefore determined which species could be used in each trial. No attempt was made to start all species from a common date. Although periods of growth differed slightly among species (Table 2), especially for V. ferruginea, all species spent exactly 1 yr in each trial and thus they all experienced a full year of seasonal change. Their nursery growth and development was concurrent with that of wild seedling cohorts in the forest and matched the courses of the comparative transplant trials (Green & Newbery, 2001).
Table 2. Starting date for each species in the Mana nursery experiments in two trials. All species were harvested after 1 yr. Total PAR and rainfall are those received by the seedlings over 365-day periods, and the number of days from the starting date to the beginning of the dry season. The dry season was defined as the period for which 30-d running rainfall totals were < 100 mm. The 1994/95 dry season started on 13.12.94 and the 1995/96 dry season on 11.1.96
Total PAR (kmol m−2)
Number of days to dry season
PAR/R : FR experiment:
For five species, seeds were collected from several individuals across the study site, oven-dried and a random subsample taken (Table 1). Individual seed mass, that is the mass of the endosperm or cotyledon (exocarp removed), was found. For V. ferruginea, which has a large woody nut, the endosperm is tiny and was recorded simply as an estimated mean of ‘< 0.01 g’, the limit of the balance. For three other species (data supplied later by G. B. Chuyong) the mean was found from single weighings of bulked samples.
Seeds were germinated on raised beds of moist soil and at the age of 3 months were transplanted, this being the start of the trial (Table 2). In the first trial the seedlings allocated to a particular PAR treatment were pretreated at that PAR level. In the second trial all seedlings were pretreated at 12% of incident PAR in a single large shade house.
PAR did not affect germination of the species in trial 1, except for V. ferruginea, which did not germinate or grow adequately at the lower PAR levels. Very young wild seedlings (< 1 month old) of this species were collected from a forest gap in the Park. These wildings were transplanted into each of the PAR pretreatment conditions and raised for a further 2.5 months (aged then c. 3 month). At the lowest PAR level (0.22%), survival of these wildings was nevertheless poor and only one replicate seedling per block could be used at this level: none survived to harvest.
Seedlings were grown in 23-cm-diameter polyethylene pots (volume 8.3 l) filled with soil from the nursery site. This soil was dug to a depth of 20 cm and mixed well. Mycorrhizal inoculation was achieved by adding soil from the root layer (0–2 cm) under parent trees of each species collected in the transect-P plot, and a handful applied as a top dressing to both germination beds and pots. The presence of ectomycorrhizas was assessed on a random sample of young (< 3-month) seedlings of M. bisulcata, T. bifoliolata and T. moreliana from the germination beds. All seedlings were found to be well inoculated (B. Moyersoen, personal communication). Later in the trials fruiting bodies of an ectomycorrhizal fungus of the genus Inocybe (P. Roberts, pers. comm.) were found growing in pots of all the ectomycorrhizal species in this study (at 12 and 36% PAR).
Experiment investigating the effects of PAR, phosphorus and water
The design was a split-plot arrangement with five PAR levels (the ‘plot’ unit) replicated (as houses) over four randomised blocks. Within each plot was a factorial arrangement of water and P treatments, and species (as ‘within-plot’ units). Water and P treatments each had two levels (control and ‘added’). There were four replicate seedlings of each of the five species investigated and thus a total of 320 seedlings of each species and 1600 in each trial.
Five PAR levels were chosen on the basis of the results of a preliminary survey in the forest study plot (Green & Newbery, 2001). The lowest level (0.22%) was below that commonly encountered in full shade in the forest, the next equivalent to that under full shade (complete canopy, 1.3%) and the highest (36%) typical of that found in the centre of small tree-fall gaps. PAR levels were achieved by construction of wooden-framed shade houses (3.6 m × 2.4 m × 1.5 m high) covered with layers of fine, black, high-density woven polyethylene netting of different standard mesh sizes (Tildenet, Bristol, UK) on both the top and sides. Shade houses were randomly arranged within each of four blocks and spaced 3.5 m apart. The distance between parallel blocks was 6 m.
The final PAR treatments (levels 1–5) were 0.22, 1.3, 6.1, 12 and 36% PAR transmission (Table 3). The mean daily total photosynthetic photon flux densities (PPFD) within the shade houses over the two trials were, correspondingly, 0.054, 0.32, 1.49, 3.0 and 8.8 mol m−2 d−1. Mean temperatures and relative humidities within the shade houses in the middle of the day, relative to external values of 32.1°C and 78% were +1.1, +0.8, +0.7, +0.9, +0.5°C, and −3.6, −2.5, −2.3, −1.0, −1.0%, for PAR levels 1–5, respectively. That is temperature increased slightly, and relative humidity decreased slightly with increasing shading.
Table 3. The mean percentage PAR and R : FR ratio in light treatments in Mana nursery experiments. SEs are shown for the PAR/P/water experiment: those in the PAR/R : FR experiment are the average SEs across the trials (n = 4 shade houses)
R : FR
Mean R : FR for trial 1 was 0.61 and for trial 2 was 0.89 (two blocks removed due to detioration of filter material). Levels of the R : FR factor: H, high; L, low.
The added-P treatment was an initial top dressing of 1.6 g per pot ground rock phosphate (GRP, 13% P, equivalent to 208 mg P per pot), followed by three quarterly applications of 0.4 g GRP per pot (52 mg P per pot). This level was based on a common tropical agriculture application rate of 50 kg P ha−1, or 384 kg ha−1 GRP (Tanner et al., 1990; Tanner et al., 1992), with the aim of raising the available P to a level equivalent to that in richer Korup soils (Gartlan et al., 1986). To check the efficacy of the P applications, a soil sample from each pot with M. bisulcata in trial 1 was taken (a slice of the whole pot’s depth at the time of harvesting) and the four replicates per shade house (PAR levels 2–5 only) bulked for the two water treatments. The air-dried material was later analysed for labile inorganic P in Bern using Olsen’s reagent.
Under the added-water treatment, pots each received 1.5 l of water weekly in the dry season whilst the control pots were not watered. The positions of the pots within shade houses was re-randomised at monthly intervals to remove positional effects.
Experiment investigating the effects of PAR and red/far-red ratio
The effects of light quantity (PAR) and quality, in terms of the red to far-red (R : FR) ratio, were investigated with a two-factor randomised block design. Five levels of PAR were used, which differed slightly from those in the PAR/P/water experiment above (Table 3) but covered the same range. Two R : FR ratios were used, one similar to ambient light (R : FR = 1.1) and one with enhanced far-red light (R : FR = 0.6). The lower R : FR ratio was typical of the lower end of the range of measurements in the field (Green & Newbery, 2001). Light quantity × quality treatments were achieved by constructing small shade houses (1.8 m × 1.2 m × 1.2 m high), 10 per block and replicated over four blocks (rows). Houses were spaced 2.5 m apart within each block, and the blocks were 7 m apart.
The low R : FR treatments was set using a layer of green filter (No. 122, Fern Green; LEE Filters, Andover, UK), combined with layers of black, polyethylene netting of different densities (as used in the PAR/P/water experiment). The high R : FR treatment was set using a layer of clear plastic (neutral) filter (No. 130; LEE Filters) and a higher density of netting to achieve the same PAR transmission and equalize microclimatic effects with only a very slight effect on light transmission. Shade and filter materials covered both the top and sides of the houses. Holes were cut in the sheets before construction in a pattern of rectangular slots (25 cm × 5 cm; and which removed 20% of the area) to permit adequate ventilation and entry of rain. Thus, 20% of light in the shade houses was unfiltered. Light levels were measured as in the PAR/P/water experiment (Table 3).
There were four replicate plants of each species within each shade-house and therefore 160 plants of each species and 800 plants in total in each trial. All seedlings received pretreatment growth conditions of 3 month under neutral shade (high R : FR). All plants were watered weekly in the dry season, receiving 1.5 l of water pot−1 (i.e. equivalent to the added water treatment in the PAR/P/water experiment). No P was applied. The analysis of variance model differed from that of the other experiment since PAR level and R : FR treatment were independently replicated in each block in a two-factor crossed design. There were no within-house (plot level) treatments; species were nested within houses, however.
PAR and R : FR levels inside shade houses were measured using three quantum sensors, and a 660/730 (R : FR) sensor (SKR 110, Skye Instruments), simultaneously with an external PAR sensor. Internal PAR sensors were mounted on a horizontal rotating beam, on a camera tripod, at a height of 70 cm. A spirit level counterbalanced the sensors and ensured that the sensors were horizontal. The radius of the beam was 1.2 m and the sensors were positioned such that they described arcs which were in the middle of concentric annuli of equal area. (Distances of the sensors to the centre of the beam were 35, 84 and 109 cm). Spot readings were taken with the arm (stationary) at 16 points of the compass. Each house was measured in morning (09h00–11h00), mid-day (11h00–13h00) and afternoon (13h00–15h00) periods on different days, to take account of the angle of the sun. PAR treatments and blocks were taken in a random order.
Seedling height, number of leaves and deaths were recorded at monthly intervals. Seedlings lost due to damage by rats were excluded from the analysis. At harvest, seedlings were separated into leaf, stem, coarse root (> 2 mm diameter) and fine root (≤ 2 mm). The woody rachis of the mesophyllous pinnate species (A. fragrans, B. bracteosa and V. ferruginea) was put into a separate category and not included with leaves when calculating specific leaf area, leaf weight ratio and nutrient concentrations. In the microphyllous pinnate species (M. bisulcata and T. moreliana) the rachis was thinner and included with leaflet samples as it was considered equivalent to the green midrib of mesophyllous species. Root systems were separated from the soil by washing all of the soil in pots through a mesh of 5 mm. Roots were carefully sorted by size using a 2-mm diameter rod as a gauge. Leaf area was measured manually in trial 1 but with a leaf area meter (Delta-T area measurement system, Cambridge, UK) in trial 2. For the manual method, a sample of 20 leaves was taken in a systematic way, evenly spaced on the seedling, and their lengths and breadths measured. The mean leaf area from the sample was multiplied by the number of leaves per seedling. For each species a calibration of length × breadth with leaf area was produced by measuring a sample of 100 leaves on squared paper. Lamina missing due to herbivory was estimated to the nearest 10% of leaf area. All plant parts were oven-dried to constant mass (80°C) and weighed to ±0.01 g. Mean relative growth rate between germination and harvest was calculated as follows, RGR = (ln(total mass) − ln(seed mass))/seedling age (mg g−1 d−1, Kitajima, 1994). The aim was not to estimate the instantaneous growth rate, but rather to examine the relative change in mass between seed and seedling at harvest.
Leaf samples were milled and subsamples returned to Bern for analysis of foliar N and P concentrations. Redried material was digested in Se/H2SO4/H2O2 acid mixture, N analysed by the modified Bertholet reaction (dialysis) and P with the molybdenum blue reaction (Allen et al., 1989), both colourimetrically determined with an auto-analyser (Skalar b.V., Breda, Netherlands: methods 503–324 for N and 155–316 for P).
In the PAR/P/water experiment, analysis of variance with the split-plot model was used to test for treatment effects on the primary variables of total mass (TM), height (HT) and leaf area (LA) and nutrient concentrations, and on the derived growth variables of specific leaf area (SLA = leaf area : leaf mass), leaf area ratio (LAR = leaf area : total mass), leaf weight ratio (LWR = leaf mass : total mass), fine root weight ratio (FRWR = fine root mass : total mass) and root-shoot ratio (RSR). TM, HT and LA were logarithmically (ln) transformed to correct for the heteroscedasticity of the variances in the data (especially across PAR levels). Species were first analysed individually to allow simpler examination of the treatment interactions. Then the five species in each trial were compared with a combined analyses of variance. In the PAR/R : FR experiment the analysis followed a two-factor (PAR × R : FR) randomized block design. Where mortality counts were involved general linear modelling (GLM) was applied.
Seedling mortality at low PAR levels
All species had high mortalities after 1 yr at 0.22% PAR in the PAR/P/water trials (Fig. 2a,b). The median number of mortalities (based on counts per shade house) was 3.5 and 1.0 out of 16 in trials 1 and 2, respectively, excluding V. ferruginea.
The differences in mortality between species in trial 1 were not significant (angular transformation, one-factor analysis of variance with blocking, F4,12 = 1.58, P > 0.05; confirmed by GLM with binomial error model; GENSTAT 5.32, Payne and the Genstat 5 Committee, 1993). The rank order of dying at 0.22% PAR in trial 1 (Fig. 2a) was M. bisulcata > A. fragrans > I. gabonensis > B. bracteosa. In trial 2, the species factor was significant (F4,12 = 3.86, P < 0.05; also confirmed by GLM). The mortality of S. pseudocola was significantly higher than that of T. moreliana, O. alata and T. bifoliolata, whilst the mortality of M. bisulcata was significantly higher than that of T. bifoliolata only (Fig. 2b). The data in Fig. 2(a,b) are based on the counts pooled across blocks (maximum n = 64), whilst the analysis of variance used block means: only for M. bisulcata in trial 2 was the between-block heterogeneity large enough to reduce the final mean survival from 0.25 (Fig. 2b) to 0.19.
At the 0.50% PAR level in the PAR/R : FR experiment, the mortalities were substantially lower for most species compared with those at 0.22% PAR in the other experiment. Analysis of variance revealed highly significant differences between species in trial 1 (F4,22 = 12.7, P < 0.001): the mortality of V. ferruginea and M. bisulcata was greater than that of the other three species (Fig. 2c). In trial 2 (Fig. 2d) there were no significant differences (F4,22 = 2.39, P < 0.1). V. ferruginea was the only species with moderately high mortality at 1.3% PAR (0.45), whilst all other species in trial 1 had mortalities < 0.20. There were no significant effects of R : FR ratio on seedling mortality at the lowest PAR level (0.50%) in the PAR/R : FR experiment.
Species generally had lower mortality at low PAR in trial 2 compared with trial 1. This effect is seen by comparing the reference species M. bisulcata in the two trials (Fig. 2). Seedlings are likely to have had greater carbon reserves when entering trial 2 as they were pretreated at 12% PAR for 3 months, and not at their treatment PAR level (0.22 or 0.50%). For M. bisulcata, this difference ensured the survival of the majority of seedlings at 0.50% PAR in trial 2 (mortality 0.07) compared with trial 1 (mortality 0.81).
In trial 1, at 100 d (the approximate time by which species showed maximal differences), at 0.22% PAR (interpolating from curves), M. bisulcata had a mortality of 0.83, I. gabonensis 0.37, B. bracteosa 0.36 and A. fragrans only 0.21. For 0.50% PAR at 100 d M. bisulcata had a mortality of 0.50 whilst the other species improved to 0.07–0.19. All species lay within the range 0–0.09 mortality at 1.1 or 1.3% PAR.
The species may be ranked in terms of intolerance to very low PAR levels as follows: V. ferruginea > S. pseudocola > M. bisulcata > other species. M. bisulcata was more intolerant of very low PAR than either of the other two grove-forming ectomycorrhizal species, T. bifoliolata and T. moreliana. The most critical range of percentage PAR which separated species’ survival was 0.2–0.5 at c. 3 months.
Seedling growth in relation to PAR
Increasing the level of PAR had a consistent, highly significant (P < 0.001) and positive effect on the total dry mass of all species (Fig. 3a,b; Table 4). Because of the large number of deaths in the lowest PAR level in both trials, this level was excluded from the analysis of growth. Height and leaf area showed similar trends to total mass but with a tendency to reach maximal values at 12–36% PAR. PAR also had a significant negative effect in almost all cases for the derived leaf growth variables (LAR, LWR, SLA; Table 4). LAR showed the largest differences with PAR, especially at the lowest levels (Fig. 3c,d); LWR decreased slightly with increasing PAR – except for V. ferruginea whose LWR was overall much lower than the other species and whose SLA declined the steepest with increasing PAR. The relative allocation to roots (shown by FRWR and RSR) increased slightly with increasing PAR for most species.
Table 4. Results of analyses of variance in the PAR/P/water expt 1 and in the PAR/R:FR expt 2, each with two trials. PAR had a significant effect on all variables for all species (indicated by ‘ALL’) except where shown (by e.g. ‘-Vf’). Other factors had no significant effect on any variable, for any species, except those shown (– indicates no species was significant)
For only one variable was there consistent statistical support across trials for a difference between ectomycorrhizal and nonectomycorrhizal species. At the highest PAR level (36%), the FRWR of the nonectomycorrhizal species was significantly higher than that of the ectomycorrhizal species. That is, in trial 1 the FRWRs of I. gabonensis (0.087) and V. ferruginea (0.083) were significantly higher (P < 0.001, Fisher’s LSD = 0.010) than those of B. bracteosa (0.068), A. fragrans (0.061) and M. bisulcata (0.057). In trial 2, the FRWR of O. alata (0.121) and S. pseudocola (0.110) were significantly higher (P < 0.001, LSD = 0.008) than those of T. bifoliolata (0.072), M. bisulcata (0.064) and T. moreliana (0.063). At lower light levels (≤ 12% PAR) this difference was not significant.
M. bisulcata acted as the cross-reference between trials 1 and 2 and demonstrated the effect of pretreating seedlings, although the difference in climatic conditions between years cannot be separated. It was smaller at PAR 2 (1.3%) in trial 1 (total mass 0.87 g), when it was pretreated at that treatment PAR level, than in trial 2 (1.56 g) when it was pretreated at 12% (PAR 4) (Fig. 3a,b). Correspondingly, at PAR level 5, it was smaller in trial 2 than trial 1 (65.8 and 31.9 g, respectively), when pretreated below the treatment PAR level (Fig. 3a,b). Plant height and leaf area followed a similar cross-over in response, LAR and LWR conversely so. SLA changed in parallel with increasing PAR, M. bisulcata being lower in trial 2 than trial 1, similarly parallel for FRWR but with trial 2 higher than trial 1. In trial 2 the RSR in PAR 2 was much higher than in trial 1 and similar to that in PAR 4 in both trials. RSR established during the pretreatment period at 12% PAR has changed little during the course of trial 2.
Increasing PAR had a consistent negative effect on the N and P concentrations of leaves of all species, significant (P < 0.05) in 18 out of 20 cases (Table 4). Species differences in mean leaf N and P concentration (Table 5) were relatively large (N ranged from 15.5 to 25.5 and for P from 0.97 to 1.91 mg g−1 (F4,237 = 439 and 283, respectively, P < 0.001 for N and P in trial 1; and F4,232 = 813 and 825, P < 0.001 correspondingly in trial 2). Species differences were often maintained across PAR levels, leading to parallel declines with PAR.
Table 5. Concentrations of phosphorus and nitrogen in the leaves of surviving seedlings of the nine species grown in the PAR/P/water experiment. Data are means of the two water treatments at 1.3 and 36% PAR. The main values are for the control (no added-P treatment) whilst those in parenthesis (see text calculations for Microberlinia bisulcata) are for the added-P treatment
The relative growth response to increased PAR was calculated as a ratio of the total mass at 36% PAR to that at 1.3% PAR. In trial 1, M. bisulcata (ratio = 75.5) and V. ferruginea (56.9) had relatively very large responses compared with the other species (< 18). In trial 2, the responses were generally lower with M. bisulcata (22.7), T. moreliana (28.7) and T. bifoliolata (20.8) the highest. The differential response between trials was most pronounced for M. bisulcata.
The performance of the nine species can be compared at PAR level 4 (12%). In both trials all species were pretreated and then grown at this PAR. Total mass (TM) need only be considered as the other primary growth variables were strongly correlated with it. The performance of M. bisulcata differed between the trials: in trial 1, ln(TM) was 3.330 whilst in trial 2 it was 2.935 (ratio of 1.13). This corresponds to the difference in PPFD recorded in the two trials, 9.38 and 8.57 kmol m−2, respectively (ratio of 1.09) (Green & Newbery, 2001).
The four species apart from M. bisulcata in trial 2 were aligned with those in trial 1 by adding the difference in ln(TM) for M. bisulcata between the two trials (= 0.395). This gave a final rank order of ln(TM) as follows: O. alata (3.278), V. ferruginea (3.297), M. bisulcata (trial 1 = trial 2) (3.330), T. bifoliolata (3.593), T. moreliana (3.807), A. fragrans (3.828), B. bracteosa (3.866), I. gabonensis (3.879) and S. pseudocola (4.079). Back-transformed total masses ranged correspondingly from 26.5 to 59.1 g.
Seed mass and PAR responses
There was no obvious and simple relationship between species’ seed mass and mortality at very low PAR in trial 1 (Fig. 4). The one species with very small seeds (V. ferruginea) had the highest mortality at 0.22% PAR, with M. bisulcata (0.64 g) the next highest. But the largest-seeded species, A. fragrans (22.7 g) also had high mortality, appearing as an outlier in Fig. 4. Lowest mortality occurred at relatively intermediate seed weights of c. 2–5 g (B. bracteosa and I. gabonensis). The difference in survival for M. bisulcata in the two trials was, however, substantial alongside the range in survival of the other species, but relative to M. bisulcata in trial 2, O. alata, T. bifoliolata and T. moreliana had much lower mortality with just modest increases in seed size up to 1.6 g. S. pseudocola was also an outlier (14.6 g) in Fig. 4. Whilst A. fragrans declined linearly with time (Fig. 2a), S. pseudocola declined much faster in the first 100 d (Fig. 2b), suggesting a different process.
After a yr in light level PAR 2 (1.3%) the total mass of surviving seedlings was highly dependent upon seed mass (r2 = 0.97, P < 0.001; Fig. 5a). At this low PAR level, it appears there was little net growth after utilization of the seed reserves. The strong correlation suggests that differences in the effects of pretreatment were smaller than species’ differences. The mean growth rate of the nine species was well predicted by seed mass (Fig. 5b, r2 = 0.89–0.94), such that large-seeded species had lower growth rates. The relationships are highly significant despite the difference in pretreatment PAR between the two experiments. Relative log-growth response was similarly significantly (P < 0.05) related to seed mass at PAR levels 3, 4 and 5 (Fig. 5c). The effect of pretreatment PAR was demonstrated by trial 1 species being above the regression line and trial 2 species below the regression line, that is the relative effect of higher pretreatment PAR is greater at 1.3% than at 6.1%.
Responses to phosphorus addition
The effect of adding P as a treatment was highly significant on soil P, with a mean concentration at the end of trial 1 of 6.4 µg g−1 in the control and 12.2 µg g−1 in added P treatment (F = 91.0, df = 1,36, P < 0.001). Over the PAR levels 2–5, the growth of M. bisulcata depleted soil P to a similar extent in both treatments (Fig. 6; control, 4.1 µg g−1; added P, 5.4 µg g−1). The effect of watering was marginally significant (F = 4.18, df = 1,36; P = 0.048) with the drier control soils having slightly higher concentrations than watered ones (9.92 and 8.67 µg g−1, respectively). All other interactions were not significant (P > 0.05).
Based on the pot dimensions and the bulk density of mineral soil in the forest (Newbery et al., 1997), the 20-cm deep pots were estimated to each hold 9.0 kg d. wt of soil. The difference in soil P concentration between PAR levels two and five implied that the difference in uptake of P (assuming no other losses from the pots) was 36.9 mg in the control and 48.6 mg in the added-P treatment. Using concentrations of leaf P for M. bisulcata (Table 5), together with leaf masses, the amount of P in control PAR-2 leaves was 0.98 mg and in PAR-5 leaves 20.80 mg, a difference of 19.8 mg. For added-P, the corresponding amounts were 0.84 and 28.75 mg, with a difference of 27.9 mg. These differences in uptake are lower than the soil P depletions because there was no account of the P taken up into stems and roots. They are, however, in the correct order of magnitude (50–60% into leaves). The pot P concentrations were, furthermore, similar to mean labile Pi concentrations of 11.3 µg g−1 measured in the field by Newbery et al. (1997), notwithstanding the former agricultural use of the nursery site. The results suggest that added P did become adequately available and was not entirely adsorbed by the soil.
There were few significant growth responses to the P-addition treatment in either trial (Table 4). In particular, P had no significant effect on the total mass, height or leaf area for any species. The significant interactions between P and PAR for M. bisulcata (height and leaf area) and V. ferruginea (leaf area) were due only to significant differences between P levels at low PAR (level 2, Fisher’s LSD test, P ≤ 0.05). No evidence was found in these nursery trials, for any of the species, that P was limiting under high PAR levels where rapid growth was occurring.
By contrast to the lack of a growth response to P addition, there were significant effects on the leaf P concentration in two of the species in trial 1 (M. bisulcata and I. gabonensis) and three of the species in trial 2 (O. alata, T. bifoliolata and S. pseudocola;Table 4). The added-P treatment resulted in higher leaf P concentrations in all these cases. The response of M. bisulcata (Table 5) was repeated in the second trial, although was just outside the 5% significance level (F = 3.82, df = 1,34; P = 0.059). The lack of a significant growth response and yet the increase in leaf P concentrations in these species (especially noteworthy for M. bisulcata) further indicates that P availability was not limiting growth. P-addition led to increased N concentrations in only one species (I. gabonensis). It appears that some species took up more P (but not N) than others but no pattern relating to mycorrhizal grouping was seen.
Responses to the watering treatment
Rainfall was less in trial 1 than 2 (5.03 and 6.23 m, respectively) and the number of days to the start of the dry season fewer (19 and 37). One way to define dry seasons in lowland rain forest is periods of 30 d with < 100 mm rainfall (Walsh, 1996). Using this definition, the dry season of trial 1 ran from 13 December 1994 to 6 March 1995 (84 d) and the dry season of trial 2 from 11 January 1996 to 29 February 1996 (50 d). The dry season in trial 1 was more intense than in trial 2: in trial 1 the 30-d running total reached a minimum of zero mm rainfall (for 4 d), while in the trial 2 dry season the minimum reached was 17.2 mm.
Significant effects of watering were largely limited to A. fragrans in terms of primary growth variables (added-water > control) and I. gabonensis for derived leaf growth variables (control > added-water) in trial 1. For V. ferruginea there was a significant interaction between PAR and water, which may have been due to this species (unlike the other eight) dropping its older leaves in unwatered high PAR treatments. There were no significant effects of water on primary variables in trial 2, and an isolated case for the derived variables (Table 4). There were also very few significant effects of the watering treatment on leaf nutrient concentrations (Table 4). For those species with significant differences, P and N concentrations were always lower in the added-water treatment than the control. Across all species in both trials and for all variables in Table 4 (100 cases) only six showed significant cases of the P × water interaction (mostly P < 0.05) and four cases of the PAR × P × water interaction (all P < 0.05). In summary, watering had almost no overall effect on growth.
Seedling growth in relation to R : FR
Given the large effect of PAR on seedling growth it was essential in the interpretation of the PAR/R : FR experiment that R : FR treatments did not differ significantly in PAR at each set level of PAR. This was tested by analysis of variance (and Fisher’s LSD test at P ≤ 0.05) using the mean values of PAR for each shade house measured during each trial. In trial 1, the high and low R : FR ratio treatments at PAR level 4 were found to have significantly different PAR values (8.8 and 6.5%, respectively), although the same shade houses were not significantly different when measured in trial 2 (6.7 and 7.2%, respectively). PAR level 4 was therefore excluded from further analyses. PAR level 1 was also excluded from the growth analysis (as in the parallel PAR/P/water experiment) because of the high mortality of seedlings. At PAR levels 2, 3 and 5, R : FR treatments were not significantly different in percentage PAR in either trial. The analyses of variance of growth thus included just PAR levels 2, 3 and 5. By the end of trial 2, the roof filter material in two low R : FR houses at PAR 5 had become torn, causing both PAR and R : FR to be raised (Table 3). These blocks were also omitted from analyses of trial 2.
Light quality had no effect on survivorship of M. bisulcata at 0.50% PAR in either trial 1 or 2 (Fisher’s exact test, P = 0.358 and 0.493, respectively). PAR had a significant effect on all growth variables, for all species, with only two exceptions out of the 110 cases (Table 4). Patterns of response were the same as in the PAR/P/water experiment. By contrast, the R : FR treatment had very few significant effects either on primary or derived growth variables (Table 4). In trial 1, total masses of M. bisulcata, A. fragrans and I. gabonensis were significantly lower (P < 0.05) under low compared with high R : FR across PAR levels (5.8 vs 7.6 g, 19.3 vs 22.9 g and 14.1 vs 17.0 g, respectively) but this was not significant in trial 2 for M. bisulcata. (This may have been because the difference in R : FR treatments at PAR level 5 was smaller than in trial 1, see Table 3). There were also no significant interactions between PAR and R : FR. Significant effects of R : FR on leaf area (low R : FR < high R : FR) were recorded only for M. bisulcata and A. fragrans. No significant cross-over effects, such that R : FR ratio caused high values at low PAR and low values at high PAR, or vice versa, were recorded for any species or variable.
There were also few significant effects of R : FR ratio on derived growth variables and significant results were poorly consistent across species (Table 4). A significant effect of R : FR on the derived architectural-allometric variable TM/HT was found only for I. gabonensis, such that TM/HT was lower in low R : FR than high R : FR both across all PAR levels and at PAR level 5. There were no significant R : FR (or interaction) effects on SLA. The few significant effects that were recorded for other derived variables showed no consistent patterns between the different species. In summary, there were very few significant effects of R : FR.
Comparing the total mass in the PAR/P/water experiments with the PAR/R : FR experiments (using the 12% PAR level of the former and interpolating at 12% PAR on the ln(TM) vs PAR graph of the latter), gave ln(TM) values of 3.33 and 3.13, respectively, for trial 1, and 2.94 and 2.86 for trial 2, differences of only 6 and 3%. This shows strong between-experiment corroboration.
The maximum levels of PAR to which plants were exposed (≤ 36%) were lower than full sunlight because attention was focused on the lowest levels representative of the forest at Korup, and increasing these up to what would be typical for small gaps. The range in PAR used experimentally was similar to that recorded in the Korup plot (Green & Newbery, 2001), although the lowest at 0.2% was less frequent. In trial 1, seeds were germinated and pretreated under the conditions they were later grown in (as would be the case for a germinant in the forest), but in trial 2 they experienced an intermediate level of 12% (as is common in silvicultural practice). The steps in PAR were chosen to mimic ecological processes at the forest floor when leaf area index increased or decreased; that is when light decreased below what was generally sustaining for growth, that is PAR 2–1 (1.3–0.2%) or when a plant in the shade at 1.3% was exposed to higher PAR (i.e. raised to levels 3, 4 and 5, viz. to 6, 12 and 36%, respectively). The ratios of total mass at 36% PAR to that at 1.3% PAR represented the fold-growth increases expected when plants under typical conditions of closed canopy become exposed to a gap above or near them.
The experiments were carried out in situ in Cameroon, which had several advantages: (1) The local soil plus fresh forest inoculum could be used without storage and long transport. (2) The plants grew under local, seasonally changing, climatic conditions (in terms of daylength, RH, temperature and rainfall) in the same way as establishing plants would in the forest. However, it was not feasible to investigate below-ground measures in detail (e.g. specific root length) or levels of mycorrhizal infection. It is clearly not possible from this work to say how PAR, R : FR, P and water affected mycorrhizal infection, or the rates of N, P and K uptake.
Since seed storage was impossible (most of the species having recalcitrant seed), the choice of species used in trials 1 and 2 was dependent upon which species’ seeds were available in the two years. Fortunately, the central species, M. bisulcata, which was mast fruiting in 1995 also provided enough seeds in 1994 for a comparison between experiments.
Unavoidable differences in starting date may have slightly confounded the differences in response between species. This effect was small and does not invalidate the analysis or conclusions on the following grounds: (1) In trial 1, the four species (V. ferruginea aside) received very similar total irradiance (Table 2) and rainfall (except for I. gabonensis for which it was lower, but then the water treatments were not significant factors for most species). In trial 2, total irradiance was almost identical for the five species (Table 2), especially for the key three large ectomycorrhizal species. (2) In trial 1, M. bisulcata started at dates between A. fragrans and B. bracteosa, whilst in trial 2 it started on the same dates as S. pseudocola and T. bifoliolata. In trial 1, where the five species are ranked 1–5 in terms of their total mass at 1.3% PAR, the ranks of the start dates were 4,2,1,3,5 and those of total irradiance 4,3,1,2,5. Trial 2 had three tied dates: in both trials the sample sizes were too small for statistical tests. (3) T. bifoliolata and T. moreliana had similar seed sizes (Table 1), started growth 7 wk apart, yet the total mass increased with percentage PAR almost identically in three of the four PAR levels and was close for the fourth (Fig. 3b). The pattern of radiation for October–December 1995 and the same months in 1996 was very similar (Fig. 1).
The change in pretreatment conditions from trials 1–2 was due to the practical constraints of timing and shade-house space. So whilst the two consecutive trials did not allow a full comparison of all species over all PAR levels – this being limited to the common 12% level for pretreatment and growing – it did permit a measure of the effect of pretreatment PAR conditions on M. bisulcata survival and growth. Given the size of pots, dimensions of the shade houses and the experimental layout, growth over 1 yr was the longest period possible before crowding occurred and nutrients began to be exhausted. The hypotheses tested focused on those early months of establishment: larger field trials would be needed for plants of 2 yr and older.
V. ferruginea is the long-lived opportunist of the nine species (Richards, 1996). It has a very small seed endosperm within a woody nut and it failed to germinate and grow adequately in the lowest PAR level. Whilst comparisons with other studies are difficult because total dry mass of seed was usually recorded, and not endosperm or cotyledon reserves, the other eight species were much larger-seeded by general tropical standards (Kelly & Purvis, 1993; Hammond & Brown, 1995; Kelly, 1995; Grubb, 1998). V. ferruginea had to be raised differently from the other species. There is a good case for leaving it out of several of the main comparisons.
In both trials there was little evidence of limitation to growth due to P or water to explain any differences in species’ requirements. Water supply was reduced in the control during the dry season but clearly not enough to affect growth in the years studied. Adding P doubled soil available concentrations but control P was not limiting to growth even at higher light levels. This was a surprising result because this control concentration was similar to the labile P fraction in the field (Newbery et al., 1997). Moyersoen et al. (1998) found, under UK glasshouse conditions, that the uptake of P by O. alata and T. moreliana under conditions of low soil inorganic P concentration depended on the extent of mycorrhizal colonization, although neither species increased its dry mass due to infection or raised soil P concentration. This partially supports the findings here.
At 1.3% PAR and above all species survived the experimental trials well. In the range 0.2–0.5%, however, PAR became so low as to cause mortality and there were large differences between species, the largest being at c. 100 d. Unfortunately, trial 1 and 2 species cannot be compared exactly for their survival at 0.22% because of the different pretreatment PAR conditions. Nevertheless, in trial 2, M. bisulcata had considerably worse survival (Fig. 2.1) than the other two codominant species, T. bifoliolata and T. moreliana, and when M. bisulcata was pretreated at lower PAR in trial 1 (0.22%), its survival fell even further. Thus it would be expected that T. bifoliolata and T. moreliana, had they been pretreated at 0.22% in expt 1, would too have had reduced survival, though not as low as M. bisulcata. On that basis the most likely order of deep shade intolerance is M. bisulcata > T. moreliana > T. bifoliolata. When PAR was raised to 0.50% this pattern disappeared: M. bisulcata still fared relatively poorly though equal to the other two species in trial 2.
The shape of the survival curve for M. bisulcata in Fig. 2(c) is similar to that in Fig. 2(b) suggesting that the 3-month pretreatment at 12% was equivalent to the difference between 0.22 and 0.47% PAR treatment conditions. Survival in critically low light levels was not related in a simple way to seed size (Fig. 4) since very large-seeded species, such as A. fragrans and S. pseudocola also survived poorly in 0.22% PAR. One possible reason is that A. fragrans and S. pseudocola (like V. ferruginea) have hypogeal cryptocotylar germination but M. bisulcata, T. bifoliolata and T. moreliana (as well as I. gabonensis and O. alata) are epigeal phanerocotylar. (B. bracteosa is also hypogeal but phanerocotylar, Table 1.) The two large-seeded cryptocotylar species would not have been able to use their cotyledons for photosynthesis.
The survival of M. bisulcata in trial 1 was much worse than the two nongrove forming ectomycorrhizal species (A. fragrans and B. bracteosa) at 0.22% PAR (Fig. 2a) and less than the nonectomycorrhizal I. gabonensis. In trial 2, M. bisulcata survived less well than T. bifoliolata, T. moreliana and O. alata but not as low as S. pseudocola. S. pseudocola, inter alia, would most probably have fared also worse than M. bisulcata at 0.22% had it been grown in trial 1.
All species increased in growth with increasing PAR in a predictable and consistent way. Several studies have contrasted the growth responses of seedlings of different tropical tree species to increasing quantity and quality of light (Chazdon et al., 1996; Veneklaas & Poorter, 1998) but very largely concentrated on comparing pioneer with main canopy forest species (Agyeman et al., 1999, for Ghana), or used minimum PAR levels well above the compensation point (Grime & Jeffrey, 1965: 2.8% for eastern USA; Osunkoya et al., 1994: 2.5% for Australia; Poorter, 1999: 3%, for Bolivia). The eight main canopy tree species in this study differed in their tolerance of deep shade. When grown under an ecologically relevant range of PAR all species showed similar log-linear increases in mass with increasing PAR in the range 1.3–12% full sunlight, with slightly less than a log-linear increase in mass above 12% (Fig. 3a,b).
Morphological changes to increasing PAR were also predictable. In increasing PAR seedlings had thicker leaves (lower SLA), allocated more biomass to roots than shoots (higher R : S) and had lower ratios of leaf area to seedling mass (lower LAR). In general species differences in these derived growth variables were maintained across PAR levels (e.g. Figure 3c,d for LAR). In shade, the adaptations increase the ratio of photosynthesis to respiration at the whole plant level (Kitajima, 1994; Poorter, 1999).
Ectomycorrhizal (n = 5) and nonectomycorrhizal (n = 4) species did not differ as groups in their degree of shade-tolerance or light-responsiveness. Just in the allocation to fine roots did the groups differ, ectomycorrhizal species having fewer fine roots than nonectomycorrhizal ones, which is not unexpected.
The eight species were generally not responsive to change in light quality as has been generally reported (Tinoco-Ojanguren & Pearcy, 1995; Chazdon et al., 1996, but see Lee et al., 1996 for some differentiation for dipterocarps in SE Asia), although V. ferruginea showed the strongest interactive response to PAR and R : FR ratio (Table 4) further indicating its light-demanding ecology.
Responses to light and seed size
Final mass at each PAR level was strongly dependent on seed size of the species. Larger seed reserves conferred greater starting capital in terms of energy (Foster & Janson, 1985; Leishman & Westoby, 1994; Kitajima, 1996a; Saverimuttu & Westoby, 1996), especially important in the lowest PAR levels. Large seedlings, however, had relatively low leaf area ratios (Fig. 3c,d) resulting in relatively low ratios of photosynthesis to respiration, particularly in low PAR (Kitajima, 1994; Veneklaas & Poorter, 1998). This fundamental allometric-constrained carbon balance largely explains the relationship between growth rate and seed size. Smaller-seeded species had higher growth rates than larger-seeded ones at the four PAR levels tested which were above very deep shade (0.22%, Fig. 5b). In addition, the relative growth response (ratio of mass attained in shade to that in a higher PAR) was also highly dependent on seed size (Fig. 5c). Smaller-seeded species responded proportionally more to increased PAR than larger-seeded ones.
Bringing together survivorship and growth rates, both large- and small-seeded species had high mortality when grown in very low light below the photosynthesis : respiration compensation point (at 0.22% PAR). For small-seeded species this was most likely because of low starting capital, and for large-seeded species possibly because of low LAR ratio and hence proportionally higher respiration costs relative to photosynthetic gain. Medium-sized species showed the highest survival. The optimum seed size would be expected to change, however, depending on how far (or for that matter how long) species are under the compensation point. Under higher PAR levels where positive growth was possible, smaller-seeded species always had the higher growth rates. These higher growth rates, however, were not necessarily sufficient for small-seeded species to catch larger-seeded ones, because the latter had started as larger seedlings. Thus, if ecological fitness is measured in terms of size attained after 1 yr, being small-seeded and having a faster growth rate will not ensure success.
Few studies have followed survival of tree seedlings below the compensation point. Leishman & Westoby (1994) reported a positive relationship between seed size and longevity across 23 Australian spp. grown at 1% light, but working with 11 phylogenetical pairs at 0.12% PAR, Saverimuttu & Westoby (1996) found a similar relationship for longevity of the cotyledon stage but none for the leaf stage. Factors such as rate of storage mobilization, cotyledon type and patterns of carbon allocation (Kitajima, 1996b) could be most critical in influencing survivorship in the 0.2–0.5 (1.0)% PAR range. Growth and survival under 1% PAR and above are more clearly determined by the net positive carbon balance in the shoot.
The case of M. bisulcata
M. bisulcata did not appear to be special in any particular way regarding its rate of response to PAR. It fitted the seed-size dependent pattern. The key comparative level of PAR across trials, and therefore for comparing species, was number 4 (12%). Ranked at this PAR, M. bisulcata was not different from either of the other two grove ectomycorrhizal species, or from the nongrove and the nonectomycorrhizal ones. However, its seeds were less than half those of T. bifoliolata or T. moreliana in mass, which led to small but significant differences in growth rate and response. M. bisulcata survived poorly at low PAR because its seed size was critically small and the threshold between seed capital and carbon gain was the determining factor. V. ferruginea aside, of the eight species, M. bisulcata had the smallest seed. Under such nursery controlled conditions it seems that simple carbon-light-growth mechanisms operated. The one complicating factor was cotyledon-type. Comparing M. bisulcata with its competing codominants species, T. bifoliolata and T. moreliana, its smaller seed size led to poorer survival under low light conditions. Under higher light, its smaller seed size resulted in a more efficient seedling architecture, and hence a higher growth rate than T. bifoliolata or T. moreliana. The higher growth rate was not sufficient, however, for M. bisulcata to grow as large as T. bifoliolata or T. moreliana after 1 yr of growth. Clearly whether differences in growth rate are maintained beyond the first year will be important in determining recruitment into larger size classes. But over 1 yr, M. bisulcata had poorer survival or was smaller than T. bifoliolata or T. moreliana under all light levels. Factors other than light must therefore have been operating for M. bisulcata to come to dominate in groves in Korup, and then subsequently fail to regenerate in the groves.
In conclusion, M. bisulcata is at a survival disadvantage under low light conditions, and under higher light conditions its growth rate is insufficient to catch the larger seeded codominants. Light responsiveness is probably not the likely answer alone as to why M. bisulcata dominates in groves at Korup. Whether its small-seeded growth characteristics have further negative consequences in the field is investigated in Green & Newbery (2001).
This research was financially supported by the European Commission (contract TS3-CT93–0233). We thank the Institute of Agronomic Research (IRAD) of Cameroon for facilitation of our Programme, and the Ministry of Parks and Ministry of Education, Science and Technology for permission to work in Korup. We are grateful for the assistance of E. Abeto, S. J. Ramage and S. Ngibile in the field. N. C. Songwe and G. B. Chuyong (IRAD) provided much support and valuable discussions. B. Moyersoen kindly inspected the seedlings for mycorrhizal status. We are grateful to the (EU/ODA/GTZ/WWF) Korup Project for logistical support, to N. Timti of Pamol Plantations (Cameroon) Ltd for access to land for the nursery and for meteorological data; and to M. Zimmermann for undertaking the nutrient analyses. We thank two anonymous reviewers for their constructive improvements to the paper.