• In a forest grove at Korup dominated by the ectomycorrhizal species Microberlinia bisulcata , an experiment tested whether phosphorus (P) was a limiting nutrient.
• P-fertilization of seven subplots 1995–97 was compared with seven controls. It led to large increases in soil P concentrations. Trees were measured in 1995 and 2000. M. bisulcata and four other species were transplanted into the treatments, and a wild cohort of M. bisulcata seedlings was followed in both. Leaf litter fall from trees and seedlings were analysed for nutrients.
• Growth of trees was not affected by added P. Transplanted seedlings survived better in the controls than added-P subplots: they did not grow better with added-P. M. bisulcata wildlings survived slightly better in the added-P subplots in yr 1 but not later. Litter fall and transplanted survivors had much higher concentrations of P (not N) in the added-P than control subplots.
• Under current conditions, it appears that P does not limit growth of trees or hinder seedling establishment, especially of M. bisculcata , in these low-P grove soils.
Determining the factors that control and possibly limit the growth and dynamics of tropical rain forests allows an understanding of their structure and ecosystem processes, and contributes to explanations of their tree species composition. Primary forests on nutrient-poor soils are expected to involve adaptations of trees to low rates of elemental supply. The limiting factors can sometimes be first suggested from correlations between soil chemistry and floristic composition across sites, but to show them as causes requires an experimental approach. In this paper the results of a field test of phosphorus limitation at a site, Korup, in a central African rain forest are presented and discussed in the context of a long-term study of forest dynamics and nutrient cycles.
The large ectomycorrhizal tree species, Microberlinia bisulcata A. Chev., Tetraberlinia bifoliolata (Harms) Haumann and T. korupensis Wieringa (= T. moreliana Aubrév.), in the Amherstieae/Caesalpiniaceae, form groves on the low-elevation (40 m asl), relatively flat, sandy and phosphorus and potassium-poor soils in the southern part of the Korup National Park (Gartlan et al., 1986; Newbery et al., 1988; Newbery & Gartlan, 1996; Moyersoen et al., 1998a). At higher elevations (80–120 m asl) further north, soils are more organic, less sandy and much richer in nutrients, and here the ectomycorrhizal species do not occur but are replaced by other large-treed species (Gartlan et al., 1986; Newbery & Gartlan, 1996). Whilst the association between ectomycorrhizal-dominated forest and low soil phosphorus (P) and potassium (K) appears strong, there is as yet no experimental evidence to show that in fact P, or perhaps even K, is a limiting controlling factor that determines forest type, species composition and ecosystem characteristics. Phosphorus would perhaps, from the general literature, be the most likely candidate given the postulated role of ectomycorrhizal symbiosis in improved P acquisition (Smith & Read, 1997, and see Moyersoen et al., 1998b for Korup).
Tree groves at Korup appear to have modified their soils as processes of integration and aggregation developed over time. Concentrations of P in the surface layers of the soil within the main grove (on ‘transect P’) in Korup are 1.6-fold those in the surrounding nongrove forest (Newbery et al., 1997). A combination of fine root-mat and mycorrhizas, together with a ‘fast cycling’ of P, has led to a finely balanced but probably quite unstable ecosystem (Newbery et al., 1997; Chuyong et al., 2000, 2002). This increased P is presumably largely derived from mineral reserves deeper in the soil profile. The ‘grove effect’ is relatively small when compared with the difference between soil types across the Park (2–3 and 12–17 µg g−1 available P on the southern sandy soils vs richer brown earths 8 km north, respectively; Gartlan et al., 1986).
This main grove on ‘transect P’ is characterized by a stand of very large M. bisulcata trees (≥ 50, often ≥ 100, cm in stem diameter) which dominate and emerge from the main canopy. The two Tetraberlinia species also contribute appreciably to the total basal area but they are less abundant and do not reach the maximum 1.7 m diameter that M. bisulcata does. At this site M. bisulcata has a severe deficit in its recruitment, which suggests a transitional phase in the long term cyclic dynamics of this forest type (Newbery & Gartlan, 1996; Newbery et al., 1998). For the two Tetraberlinia species, regeneration is considerably better with much higher chances of replacement than M. bisulcata if current conditions were to persist (Newbery & Gartlan, 1996). The presently strongly negatively skewed frequency distribution of sizes of M. bisulcata may have been, it is postulated, the result of large disturbances two to three centuries ago (Newbery & Gartlan, 1996).
Despite its copious seed production in most years and ensuing large seedling cohorts on the forest floor (Newbery et al., 1998), M. bisulcata has extremely few individuals surviving to 1 cm dbh (diameter at breast height) (c. 5–10 yr old) and the forest is practically devoid of saplings and small pole-sized trees. Younger, largely juvenile trees (< 50 cm stem diameter) exist at very low densities (X. van der Burgt, D. M. Newbery, M. A. Moravie, unpubl. data). One or more factors possibly act together to inhibit recruitment of the species but as yet this complex has not been elucidated. Low light (PAR) within the closed and shaded understorey has an important negative effect on the survival of the relatively small-seeded M. bisulcata (Green & Newbery, 2001a). Green & Newbery (2001b) suggested that M. bisulcata seedlings that were repeatedly defoliated by herbivory may have become relatively P-limited, but they did not test this with addition of P in the forest. In noncompetitive nursery conditions M. bisulcata seedlings showed no growth response to the addition of P (roughly a doubling of available concentrations in the soil), although P was concentrated in the leaves (Green & Newbery, 2001a). Indications of P-limitation were found under forest, but not nursery, conditions.
Large-scale manipulation of the forest within the Park cannot be undertaken for obvious conservation reasons. In 1995, however, a new road from Ndian to Isangele annexed the Park's SE corner and within this area lies a small grove of M. bisulcata and T. moreliana trees. The annexe was degazetted soon after and this allowed the P-fertilization trial at Isangele Road (that reported here) to take place.
Three related questions were asked from this experiment: Does the addition of P fertilizer lead to generally increased tree growth and nutrients in litter fall; enable transplanted seedlings of M. bisulcata, and four other species with which it could be compared, to establish better; and result in higher survivorship of wild cohorts of M. bisulcata seedlings?
Methods and Materials
A 250-m × 350-m plot was demarcated in 1995, and divided into 35 50-m × 50-m subplots. Each of the seven rows of five subplots formed a block. Two subplots within each block were selected at random. One was allocated as a control and the other for treatment.
Enumeration of the trees
All trees ≥ 50 cm dbh in the 8.75-ha were number-tagged, mapped and identified. Stem diameter was measured with a relascope, usually above the buttresses (20 July 1995). In the 14 experimental subplots, all trees 10- < 50 cm dbh were similarly tagged, mapped, tape-measured, and identified (25 July−24 August 1995). In a 30-m × 30-m area central to each of these 14 subplots, five of the nine 10-m × 10-m subsubplots were taken at random and in them stems 1- < 10 cm dbh tagged and measured, but only identified when they were of the six main ectomycorrhizal caesalp species (18 September 1995–30 April 1996).
Triple-superphosphate fertilizer (ADER, Douala; 46% P2O5) was applied evenly across the seven added-P subplots on 7 November 1995 at the rate of 50 kg ha−1 (Fig. 1). Then on 2 March and 4 October 1996, on 28–29 May and 21 November 1997, fertilizer was again applied at the rate of 25 kg ha−1 (a total of 150 kg ha−1 over 2 yr).
Soil samples were taken on the following dates: 4–7 November 1995; 28 February−2 March, 19–20 September and 19–22 December 1996; 25–27 February, 1–3 June 1997 and 8–9 January 1998. The first two samples were just prior to the first two applications, the third 2 wk before the next one, the fourth and fifth were between applications, and the last two several weeks after the last two applications (Fig. 1).
Using a 50-mm-diameter opening steel corer, soil was taken at 18 random locations within the nine central 10-m × 10-m subsubplots of each experimental subplot. Surface litter was brushed aside before coring. The profile was cut into three layers: organic-litter (O-horizon), 0–5 cm depth (including the surface fine root layer), and 15–20 cm (i.e. within the mineral soil). Samples from the same layers were bulked within subplots and homogenized. The slightly wet soils (these lightly air-dried when too wet) were sieved through a 2-mm mesh and subsampled to give c. 200-g amounts. These were then transported to Ekona station and stored in a refrigerator at 2–4°C until analysis.
To assess down-slope lateral movement of fertilizer, five randomly positioned cores were taken in 5-m × 50-m-strips adjacent to each of 12 of the control and fertilized subplots on 4–7 November 1995, 18–20 December 1996 and 1–5 July 1997.
Four measures of soil P were made. These were (1) inorganic P or Olsen's P (Pi), by extracting 1.0 g of soil for 16 h in 30 ml of 0.5 M NaHCO3 (pH 8.5); (2) organic P (Po), as the difference between total P of the Olsen's extract and Pi, obtained from digestion of 5 ml of extract (1) with 0.2 ml H2SO4 and 5 mg K/Cu/Se catalyst at 360°C for 3 h. (3) Bray's no. 2 P (Pb), a common agricultural assay for acid tropical soils, by extraction of 2 g soil for 1 min in 14 ml 0.1 N HCl – 0.03 N NH4F. (iv) Total P (Pt), from 0.25 g air-dried soil in 3 ml concentrated H2SO4, treated as in (2). Concentrations were corrected to oven-dried weights (105°C). Phosphorus was determined on an autoanalyzer (Technicon Instrument Corp., Tarrytown, NY, USA)) using the molybdenum-blue colorimetric technique.
Between 24 October and 5 November 1995, 2-m × 2-m quadrats were established, initially two at random, within each of the control and added-P subplots. Twelve of the 28 had M. bisulcata seedlings from the 1995 mast fruiting. Six quadrats were located in the control subplots, and six in the added-P subplots, which gave the same intensity of sampling in both treatments. (Four subplots had two quadrats each, and four had one quadrat each.) Seedlings were mapped and labelled (n = 815 in total).
Seedlings were first censused on 27 November 1995 and thereafter 28 times almost monthly until 5 June 1998, recording survival and heights of survivors. Any seedlings that had been obviously cut at ground level (presumably by mammals) were noted. New seedlings establishing from the next masting event in 1997 were tagged on 29 October 1997 as a second cohort (i.e. from recording occasion number 24). Their demography was followed in a similar manner to cohort 1, but only on four further occasions.
In each of the control and added-P subplots, two further 2-m × 2-m quadrats were located at random and demarcated. Seven seedlings each of M. bisulcata, T. bifoliolata and T. moreliana, as the ectomycorrhizal tree species, and Oubanguia alata Bak. f. and Strephonema pseudocola A. Chev. as the nonectomycorrhizal ones, were planted in a partial 7 × 7 Latin square arrangement. The first four species were raised from seed at a neighbouring nursery at Mana Bridge. Seedlings grew for 3.5 months under 12% PAR (Green & Newbery, 2001b) and were planted out between 23 and 27 November 1995. Wildings (germinants that year) of S. pseudocola were collected in Korup near the main transect-P plot on 17 December 1995, and transplanted to Isangele Road the next day. In the parallel nursery experiments of Green & Newbery (2001a), potted plants of the ectomycorrhizal species became well infected with ectomycorrhizas. It is highly likely that transplants from the same stock of raised seedlings were also infected.
Measurements on all species started on 20 December 1995. On 5 June 1998, after 27 (respectively 26) almost-monthly spaced recordings for survivorship and height, the above-ground parts of the survivors were harvested. Shoot mass (SHM) was separated into leaf (LM) and stem (SM) mass; and stem diameter was recorded at the base. Leaf area was measured with a leaf-area meter (Delta-T, Cambridge, UK). Plant material was dried at 70°C in an oven and weighed. From these primary variables, stem mass (SM : SHM), leaf area (LA : SHM) and leaf weight (LM : SHM) ratios were derived.
For the three ectomycorrhizal species, leaves from different seedlings were bulked for the same quadrat. The extent of bulking depended on the amount of material but typically a sample was derived from one to six individuals to give roughly similar amounts. Often there were two or three such bulked samples per quadrat. Leaf material was milled in a steel ball-mill, and subsamples wet-ashed in a H2SO4/H2O2/Se-catalyst mixture. Nitrogen was analyzed by the modified Bertholet reaction (dialysis) and P by the molybdenum blue reaction, both colorimetrically determined with an auto-analyzer (Skalar Analytical B.V., Breda, Netherlands). Using the bulked dry masses of leaves for each analysed sample, the weighted mean concentrations of N and P were found for seedlings within each quadrat.
Photosynthetic active radiation (PAR, 400–700 nm) was recorded over each demographic and each experimental quadrat using sensors (model SKP215, Skye Instruments, Powys, UK), following the protocol and methods in Green & Newbery (2001b). PAR was recorded late in the morning, close to midday and in the early afternoon, and the average PAR found for each quadrat. PAR was recorded on 24–25 February 1998. The external sensor was positioned on the Isangele Road c. 300 m from the western edge of the plot.
Ten galvanized 1.5-mm mesh litter traps (40 cm × 40 cm in area, 30 cm deep; as in Chuyong et al., 2000) were placed at random within the central 20-m × 20-m area of each of the 14 experimental subplots on 18–21 March 1998. Litter was collected for two fortnightly intervals (on 3 and 16 April 1998), and bulked in the field over replicate traps per subplot. It was then separated into the four fractions: leaves, small wood, reproductive parts, and ‘other material’, following Chuyong et al. (2000). After oven-drying and weighing, the leaf fractions were analysed for total P and N as above. At the time of the setting up of the traps and after the last litter collection (17 April) the mass of litter on the forest floor (in the four fractions) was estimated by sampling 10 randomly placed 40-cm × 40-cm-quadrats per experimental subplot.
Tree mortality and growth
All trees that were enumerated in 1995 were re-censused between 14 and 27 March 2000 and the incidence of mortality was recorded (Zimmermann, 2000). The gbh (or dbh for small stems) was measured on the survivors of those trees 1- < 50 cm dbh in 1995). Increments on larger trees, which mostly required the use of a relascope, were too inaccurate for reliable growth estimates. Absolute (agr) and relative (rgr) growth rates were calculated for the intermediate trees as (dbh2 − dbh1)/t and [ln(dbh2) − ln(dbh1)]/t, respectively.
Among the trees ≥ 50 cm dbh the plot was dominated by M. bisulcata, with some trees of T. moreliana but none of T. bifoliolata (Appendix Table 1). Percentage basal area ectomycorrhizal (Newbery et al., 1988) was 39.6%. Among the intermediate-sized trees (10- < 50 cm dbh) there were no M. bisulcata, and ectomycorrhizal trees amounted to only 3.38% of the basal area (Appendix Table 2).
The 4820 trees 1- < 10 cm dbh enumerated in 0.7 ha consisted of 192 taxa, many of which could not be identified to a species and some of which were of uncertain species grouped within genera. These 1- < 10-cm trees formed a total basal area of 6.08 m2 ha−1. The dominant Cola semecarpophylla K. Schum. had 1090 stems (22.6%) and formed 1.52 m2 ha−1 (24.9%). Oubanguia alata had 99 stems (0.21 m2 ha−1), Tetraberlinia moreliana 41 (0.05 m2 ha−1), but M. bisulcata none.
The double-log model of size frequencies (Newbery & Gartlan, 1996) was fitted to the combined data sets of small, intermediate and large trees, and was a good fit (≥ 1 cm dbh; in 10-cm classes with midpoints 5, 15, … 145 cm; first class set equivalent to trees 1- < 10 cm dbh: ln(ln(n + 1.5)) = 2.32–0.181 dbh, r2 = 0.931, F = 190, d.f. = 1,13; P < < 0.001). This confirms that the plot's stand structure is basically similar to that of the four Korup transects to the north (Newbery & Gartlan, 1996).
Addition of P significantly increased the concentrations of P of most fractions in the litter and two soil layers in the added P compared with control subplots (Table 1). The most pronounced increases were for Pi and Pb in the organic matter per litter layer (2.7 and 8.7-fold, respectively) and for Pi and Pb in the 0–5 cm layer (3.1 and 8.1-fold) (Figs 2 and 3). Po and Pt increased relatively less. The control values from t2 −t7 remained similar to those at t1, except for Po in the 0–5-cm layer, which decreased during the course of the experiment. The 15–20-cm layer showed less change than the two layers above it.
Table 1. Concentrations of soil phosphorus averaged over the second to seventh sampling dates (i.e. after the first application of P) for control and added-P subplots ( n = 42 * ) at Isangele Road, Korup, for four P fractions in three layers, and for the first date averaged for the treatments
Phosphorus concentration (µg g−1)
Added-P (t 2 – t 7 )
F (d.f. = 1,6)
Control (t2– t7)
n = 35 for 15–20-cm layer (t 3 missing).
Bray's no. 2 (Pb)
The comparison of control subplots with down-slope strips to the added-P subplots, for times t4 and t6, showed marginally small increases for only two variables (Pi 0–5 cm, P = 0.035; Pb litter, P = 0.031; 1.25 and 1.88 fold increases). The other variables also tended (nonsignificantly) towards higher concentrations in the strips indicating some small degree of lateral transfer of added P. Trends with time for 15–20 cm were similar to, but weaker than, those at 0–5 cm (Figs 2 and 3).
Mean PAR for demographic quadrats was 1.053% of above canopy PAR (range of 0.46–3.70, n = 14; mean based on ln-transformed data, 95% confidence limits = 0.77–1.44) and the mean R : FR ratio 0.53 (SE = 0.0246, range of 0.30–0.70). In the transplant quadrats mean PAR was 0.909% (range of 0.41–2.99, n = 28; limits 0.781–1.058) and mean R : FR was 0.51 (SE = 0.02, range of 0.35–0.81).
Survival of seedlings of M. bisulcata in cohort 1 (1995) declined more slowly over the first 9 months in the added-P subplots than in the control subplots, but from c. 1 yr onwards it was very similar (Fig. 4a). Data from those quadrats within the four blocks with paired control and added-P subplots were used: this involved a total of 560 seedlings (control, 297; added-P, 263), cut seedlings excluded. At times 4–11 (Fig. 4a) counts of survivors differed significantly between control and added-P (χ2 − deviance = 5.48–9.48, d.f. = 1, P < 0.02; nC = 4, n+P = 4; GLM with binomial error) but were not significant (P > 0.2) at all other times (except at time 2, P < 0.05). The rate of decline in survival was steepest for the control subplots at times 2–4 (start of the dry season 1995–96) and steepest for the added-P subplots at times 11–13 (end of the wet season 1996).
The start of the seedling survival curve was more completely observed for cohort 2 (1997) as the new emergents were recorded 1 month earlier than in 1995 (Fig. 4b). There were, however, no significant differences in survival between control and added-P subplots (χ2 − deviance = 0.47–1.36, d.f. = 1, P > 0.05). The 129 new recruits (control, 73; added-P, 56) declined to 51 in the first 2 months (39.5% survival), and to 41 in the next 2 months (31.8%). The rate of decline between times 3 and 5 for cohort 2 (Fig. 4b) appeared very similar to that between times 3–11 for cohort 1 (a sharp change from time 3 onwards, Fig. 4a), so the average survival by time 3 (3 January 1996) in cohort 1, estimated effectively after c. 2 months (Fig. 4), at 82.5% (560–462 seedlings) overestimates the survival by time 3 (9 January 1998) in cohort 2 (at 39.5%). By inference, the difference of 43% is the likely average loss of seedlings between the time of emergence and the time of first recording in cohort 1 (1995).
Eleven of the 12 quadrats with M. bisulcata had > 15 seedlings of this species recorded at the start, the one with eight seedlings being omitted. The proportions surviving at the 14th and 28th (last occasion of recording) were, however, not significantly related to log10(%PAR + 1) (GLM binomial model, χ2-dev. change = 0.60 and 0.01, n = 11, P > 0.40 both; PAR range 0.46–1.50%).
Seedlings of M. bisulcata, T. bifoliolata and O. alata survived significantly better in the control than the added-P subplots (P < 0.02, Table 2), whilst for T. moreliana there was no difference and for S. pseudocola a nonsignificant one although again better survival in the control than added-P subplots. On average, T. moreliana (0.95) and T. bifoliolata (0.86) survived better than M. bisulcata (0.53).
Table 2. Survivorship of transplanted seedlings of five tree species after 2.5 year in control and added-P subplots of an experimental plot at Isangele Road, Korup. Df for change in deviance = 1
Mean proportion surviving
The proportion of dead seedlings after 1 year declined significantly with increasing PAR (GLM with binomial error, PAR as log10(%PAR + 1); χ2-deviance change = 22.0, d.f. = 1, P < 0.001, n = 27 [with one outlying high PAR value removed]; deviance ratio of the model corrected for dispersion = 5.73, d.f. = 1,25; P < 0.05). Adding P was marginally positively significant (χ2-dev. change = 5.36, d.f. = 1, P < 0.05; deviance ratio of new model = 3.62, d.f. = 2,24, P < 0.05), and the interaction between PAR and P was very weak (dev. change = 0.07, d.f. = 1, P > 0.05). This indicates that the increase in survival with increasing PAR had a similar slope for control and added-P subplots, confirmed by separate GLM analyses for the two treatments.
There were no significant differences (P < 0.05) between control and added-P for any growth variable of the seedlings of the five species transplanted, and only one case at P = 0.05 (Table 3). A general tendency for better growth in the controls was evident. Just the means of the two treatments are reported in Table 3. Using mean quadrat PAR as a covariate only slightly improved the detection of differences in the analyses of variance. Therefore, adding P did not increase seedling survivorship or growth: it decreased it.
Table 3. Growth of transplanted seedlings of five species in control and added-P subplots of an experimental plot at Isangele Road, Korup. All seedlings were raised from seed, except S. pseudocola which were wildings. Values in the table are the means of the two treatments ( n = 14 subplots). There was only one case in which the difference between control and added-P was significant, at P = 0.05 (in 30/35 cases, F with 1,6 d.f.: P > 0.30). This and two other marginally significant differences are shown as footnotes. PAR as a covariate was significant at P ≤ 0.05 for those growth variables indicated
In these cases only C and + P differed marginally significantly, when the covariate PAR was used: (a) C 4.49, + P 3.52, P = 0.082; (b) C 2.39, + P 1.80, P = 0.066; (c) C 271, + P 202, P = 0.050.
Seedlings of M. bisulcata, T. bifoliolata and T. moreliana all had significantly higher concentrations of phosphorus in their leaves in the added-P subplots compared to the control ones (by 23, 25 and 26%, respectively; P < 0.03; Table 4). Concentrations of K were higher in added-P than control subplots for T. moreliana only. There were no significant differences between treatments for N and Mg (Table 4).
Table 4. Concentrations of phosphorus, potassium, nitrogen and magnesium (mg g −1 ) in the leaves of surviving seedlings transplanted into subplots with two treatments (control and added-P) in an experimental plot at Isangele Road, Korup
The litter of trees in the added-P subplots had 38% higher concentrations of P than the control subplots (0.866 vs 0.628 mg g−1; F = 16.7, d.f. = 1,6; P = 0.007). Nitrogen concentrations differed little (mean 18.52 mg g−1, F = 11.6, P = 0.133) and K concentrations were almost identical for control and added-P subplots (mean 4.03 mg g−1, F < 0.01, P = 0.95). Control and added-P subplots did not differ (0.2 > P < 0.8) in litter fall or litter mass for any of the fractions or their totals (Table 5).
Table 5. Mean litter mass on, and litter fall to, the forest floor within the experimental plot at Isangele Road, Korup, 18 March–17 April 1998
Litter mass (g m−2)
Litter fall (g m−2 month−1)
Tree mortality and growth
Mortality of small trees was significantly higher in the added-P than control subplots (Table 6a). This was partly due to a large tree fall in one of the added-P subplots. If this subplot is excluded the mortality for added-P remains higher than control but insignificantly so (P > 0.05). Intermediate-tree mortality showed no significant differences between treatments although it was again higher in added-P than control subplots. Species (averaged over all subplots) varied by 5–6 fold in their mortality rates (Appendix Tables 3 and 4).
Table 6. Mean subplot (a) annualized mortality rates (m a ), and (b) absolute (agr) and relative (rgr) growth rates, of trees in subplots of two treatments (control, added-P) in an experimental plot at Isangele Road (1995–2000), in two size classes: intermediate-sized (10- < 50 cm dbh) and small (1- < 10 cm dbh) trees of all species enumerated. Df for GLM χ 2 -deviance (in a), and F in ANOVA of treatment differences (in b): 1 and 1,6, respectively. For numbers of trees involved and criteria for valid growth rates see text
Dbh class (cm)
Excluding subplot with large treefall (a) m a = 0.962 (χ 2 -dev. = 1.08, P = 0.298) (b) m a = 0.828, χ 2 -dev. = 1.07, P = 0.300).
Corresponding values of F ( P ) when all trees were used: c, 0.74 (0.42); d, 0.04 (0.85); e, 0.04 (0.86); f, 0.03 (0.88).
Pooling all species and testing at the subplot level, intermediate trees (10- < 50 cm dbh) did not differ significantly between control and added-P subplots in either their agr or rgr (Table 6b). Small trees (1- < 10 cm dbh) also showed no significant differences. The trees were first measured in the middle of the wet season in 1995 but remeasured at the start of the wet season in 2000, so possibly some negative growth rates were due to stem shrinkage. However, in March the forest soils are normally sufficiently rewetted, as was the case in 2000 (with 126 mm rainfall), and any dry season shrinkage should have been mostly readjusted.
For these calculations, the total numbers of small and intermediate trees at the start of the interval were 4830 and 1265, respectively, and of these 233 and 50 correspondingly died. (Totals are slightly higher than those summarizing the floristic composition in 1995 due to minor revisions to the taxonomy in 2000.) Of the 4597 surviving small trees, 46 (1.0%) were omitted with unusually large negative differences in dbh between start and end of the period (< −3.1 mm), and 23 (0.5%) with unusually large positive differences (> 17.4 mm). This left 4528 ‘valid’ individuals for the agr and rgr calculations. Similarly, of the 1215 surviving intermediate trees, 36 (3.0%) with large negative differences (< −4.1 mm) and six (0.5%) with large positive differences (> 114.5 mm) were omitted to leave 1173 ‘valid’ individuals. Using all trees changed the statistics very little (P > 0.40; Table 6b). With one exception, the species composition of the excluded species (those with the highest relative densities) was close to that of all species.
Nineteen species had ≥ 10 intermediate-sized trees (in the 14 subplots together) of which two had fewer than five trees in one treatment (the minimum considered for statistical tests). Comparisons with the Mann–Whitney U-test (because the frequency distributions of agr and rgr were mostly strongly skewed and not readily transformable) for the 17 suitable species, with individuals pooled for either control or for added-P subplots, showed significant differences (P < 0.05) in only two cases (Appendix Table 3). (These results were also qualitatively matched by the outcome of t-tests on log-transformed data.) In one case, trees in the control subplots grew faster than those in the added-P subplots, and in the other it was the converse. For M. bisulcata and T. bifoliolata there were insufficient numbers of intermediate trees with which to compare treatments. The mean growth rates are given in Appendix Table 3.
Of the small trees (1- < 10 cm dbh), 15 taxa with ≥ 20 enumerated trees were reliably identified to the species level. In only three cases were there significant differences (P < 0.05) between control and added-P subplots (Appendix Table 4): two had faster growth in the control than added-P subplots, whilst one showed a weakly opposite result. Mean growth rates are recorded in Appendix Table 4. T. moreliana grew faster also in control than added-P subplots, but not significantly so. There were no small stems of M. bisulcata with which test the effect of added P.
Importance of phosphorus for groves in Korup
Except for the lack of T. bifoliolata, the grove composition at Isangele Road was similar to that on transect P in the Park and can be taken as a representative of that large area. T. bifoliolata could be found separated 100–200 m north of the Isangele Road plot, however. The addition of P over 2 yr was successful in raising soil P that was available to the trees. There were no visual signs of deficiencies of other elements in leaves. There were neither strong imbalances in N : P ratios, nor obvious side-effects in forest processes such as change in appearance of litter on the forest floor. Although the addition of P as fertilizer is a rather special and unnatural event, or even a disturbance to the ecosystem, the release of P would have been gradual over several months and years. A further point is that adding TSP fertilizer may have lowered the soil pH in an already low-pH system. The consequences of this are unknown.
Addition of P to the grove ecosystem resulted in increased concentrations of P in plant leaves but overall no increase in growth and survival of seedlings and trees. This simply means that the plants took up the extra P supplied to them. From this set of consistent results it may be concluded that P was not limiting growth and survival of tree seedlings, and in particular that of M. bisulcata, and it was not limiting the growth of small and medium-sized trees of the forest overall. The exception was the demographic study of M. bisulcata in which added P led to better survival up to c. 1 yr and then this effect was lost. Indeed the added-P treatments led to decreased survival of some transplanted species and a tendency towards lowered growth compared with the control. Unfortunately, there were no intermediate-sized M. bisulcata trees on which to specifically test the effect of added P.
What might limit and control survival and growth of seedlings could be quite different for trees. Here the argument that PAR might limit tree response to added P is less plausible because larger trees display their leaves higher in the canopy and are better illuminated. The evidence for P not limiting tree growth is stronger. Possibly P is a limiting factor when the groves are establishing after a disturbance (Newbery & Gartlan, 1996), and competition for P among saplings and small trees on these P-poor soils leads to a selection of the ectomycorrhizal species. Later in the established stands P might become less of a premium and other elements, also beneficially acquired and cycled by the then larger ectomycorrhizal trees, take on a more controlling role. Gartlan et al. (1986) demonstrated that the occurrence of ectomycorrhizal caesalps was associated with low K as well as low P soils in Korup and it may be that K now becomes a limiting factor to tree growth. The means of P cycling in mature stands dominated by the ectomycorrhizal caesalps was found to be highly efficient with a strong feedback mechanism, in that the concentration of P in leaf litter was high and decomposition was fast (Chuyong et al., 2000, 2002) plus the fact that a not insubstantial amount of P is inputted annually in incident rain (Chuyong, 1994, G. B. Chuyong, D. M. Newbery, N. C. Songwe; unpubl. data). As the forest grove establishes into a stand of mature trees the importance of leaching of K from the vegetation is likely to increase and, despite the presence of the fine-root and ectomycorrhizal mat (Newbery et al., 1997), control of K cycling on an also generally K-poor soil might suggest that K, not P, limits growth at later stages. Indications of K were only recently appreciated from throughfall and canopy leaching results (G. B. Chuyong, D. M. Newbery, N. C. Songwe; unpubl. data).
If all of the tree species in Korup were already adapted to low-P soils then perhaps they would not be expected to respond at all to added P (Tanner et al., 1998). This seems unlikely on two counts: first M. bisulcata and the other ectomycorrhizal caesalps have high tree growth rates (Zimmermann, 2000; DM Newbery et al. unpubl. data) which is not an expected characteristic of slow growing tolerant and conservative species, and second many of the tree species in Korup are widespread on a range of soil types in Cameroon suggesting no localized specialization to low nutrient soil per se, for example S. pseudocola in the present experiment, and Irvingia gabonensis (Green & Newbery, 2001a) did not respond to added P. It would seem unlikely that any species is adapted to a single factor, but to a suite of related factors, which includes inter alia limited water supply in the dry season, low external radiation and intense canopy leaching with the heavy rains in the wet season.
Comparability of the experimental site with the main grove at Korup
Soil phosphorus concentrations Soil P concentrations at Isangele Road (Table 1) may be compared with those from a much more detailed study on transect P (Newbery et al., 1997, Table 10 loc. cit.). Total P in the leaf litter layer at Isangele Road was considerably less than that on transect P (803 µg g−1, cf.). In the 0–5 cm soil layer Pi was higher but Po lower than at transect P (1.34 and 9.02 µg g−1, respectively). Pt was also higher than on transect P (210 µg g−1). For comparison, transect-P-values were calculated with the formula: [(0.8 × root layer concentration) + (4.2 × mineral layer concentration)]/5.0 since the root layer was estimated to be 0.8 cm deep on transect P (Newbery et al., 1997). On transect P the mineral soil was sampled below the root layer to a depth of 5 cm so these values are not fully comparable to the ones at Isangele Road for 10–15 cm. Nevertheless, total P was very similar for this layer between the two sites (192 µg g−1), with Pi and Po also, respectively, higher and lower at Isangele Road than transect P (1.1 and 7.2 µg g−1). The conclusion is that the soils at Isangele Road are broadly similar to those on transect P but with a tendency to having more Pi but less Po, which corresponds to the different leaf litter concentrations at the sites.
Nutrient concentrations of transplants The P concentration in the leaves of the transplanted M. bisulcata seedlings in the control plots at Isangele Road (1.72 mg g−1; Table 4) was 18% higher than the corresponding concentration in the transplants in 1995–96 on transect P estimated from linear regression equations of concentration on log10(%PAR + 1) at the same mean %PAR of 0.91 (1.46 mg g−1; mean of two trials, Green & Newbery, 2001b). For T. bifoliolata and T. moreliana the sites were similar (1.32 and 0.99 mg g−1, respectively, on transect P). Nitrogen concentrations in the three species were close (26.1, 21.5 and 19.2 mg g−1, respectively, on transect P).
In the first transplant experiment on transect P in 1989–90 (Newbery et al., 2000), M. bisulcata then had even higher leaf P concentrations (2.28 mg g−1), but again similar concentrations for T. bifoliolata and T. moreliana (1.27 and 0.91 mg g−1). In 1989–90 N concentrations tended to be lower than at Isangele Road (21.3, 18.8 and 15.2 mg g−1 for the three species, respectively). Taken together, these results from the various studies consistently show that the lower the average PAR the higher the P concentration in the leaves of M. bisulcata, but this was not so, or was much less so, for T. bifoliolata and T. moreliana. Nitrogen differences were smaller and less consistent in direction. In the nursery Green & Newbery (2001a) found no significant interaction between increasing PAR and ± added P, even when seedlings were given more light (up to 36% PAR) and growth was much increased.
Adding P to the soil in the present experiment simply enhanced this effect, by increasing leaf P concentrations further. In the nursery experiment of Green & Newbery (2001a), P concentrations in leaves of M. bisulcata at 1 yr were 1.63 and 1.91 mg g−1 in two trials with a PAR level of 1.3%. Furthermore, Chuyong et al. (2000) reported that mature canopy green leaves of M. bisulcata had P concentrations of 1.18 mg g−1 (Table 6, loc. cit.), lower than those of seedlings, but N concentrations at 22.3 mg g−1 were much less different.
Litter fall and its nutrient concentrations Leaf litter fall was rather higher (by 53%) at Isangele Road for the same months of March and April in 1991 and 1992 on transect P (43.7 g m2 −1 per month; Chuyong et al. (2000), five plots with high abundance of ectomycorrhizal trees; Fig. 1(b), loc. cit.). Both cases were nonmast fruiting years. Mean litter P concentration for the control plots was very similar to the value for the HEM (high ectomycorrhizal forest) bulked leaf litter concentration on transect P (0.68 mg g−1; Chuyong et al. (2000), Table 3, loc. cit.) but that of N was slightly higher (17.1 mg g−1).
Fertilization trials in rain forest in a broader context
Walker & Syers (1976 ) suggested that long-term pedogenesis leads to soils poor in P and to P-limitation, especially on highly weathered oxisols in the tropics. Young soils, such as those on volcanic lavas, have been shown to be largely N-limited in the early successional stages ( Vitousek et al., 1993 ), but they become more P-limited in later stages ( Herbert & Fownes, 1995 ).
Experiments in primary tropical forests on nutrient limitation have been conducted rarely, with only one reported study in primary lowland forest by Mirmanto et al. (1999) for Bornean Indonesia. The investigations of Tanner et al. (1990, 1992) in Jamaica concerned montane forest. Tanner et al. (1998) have, however, suggested that pantropically N might be the limiting nutrient to forest growth in montane areas and P the limiting nutrient in the lowlands. As yet there are rather too few observations to test this hypothesis: the present study does not anyway support it.
Possible comparisons between Korup and other old tropical primary forest sites are limited. The detailed studies in Hawaii, on young montane volcanic substrates, showing early woodland succession are not quite so relevant: Tanner et al. (1998) give a useful overview of fertilizer trials and tree growth and nutrients on these soils, and highlight large positive growth responses to added N. In Jamaican montane forest, Tanner et al. (1990) compared a plot with added N (150 kg ha−1 yr−1) and one with added P (50 kg ha−1 yr−1) with two controls, the fertilizer being applied annually between 1983 and 1986. Adding N did not increase foliar N concentrations in four common tree species but it decreased P concentration in the two species that had higher P concentrations at the start. Trunk growth rate (trees ≥ 25 cm gbh) was 2-fold higher in the added-N plot compared with the controls. Adding P did not raise leaf P concentrations in the two already P-rich species but it did double it in the other two P-poor species. Added P increased trunk growth by 50%. Nevertheless, the effects took 2–3 yr to begin to show. Tanner et al. (1990) concluded that the forest was probably generally N-limited but that some species were P-limited in growth. On this basis the 4.7-yr interval for tree growth measurements in the present study at Korup should have been adequate.
In Venezuelan montane forest, Tanner et al. (1992) applied N, P, N + P fertilization in comparison with a control. There were five replicate plots per treatment. Levels of N and P were 225 and 75 kg ha−1 yr−1, respectively, in yrs 1 and 2, returning to the Jamaican levels in yr 3 (1986–88). Growth was recorded for 104 trees, but only in the control and N + P treatment plots. After 28 months, trunk growth with added-‘N + P’ was c. 2-fold that in the control, and further remeasurements showed that this difference was held until 1992 (E. V. J. Tanner; pers. comm.). From this it is unclear whether either P or N singly were limiting growth, and even then significant differences took 3–4 yr to appear. Adding N, P or N + P did not affect the leaf litter N concentrations, but after 2.5 yr the added P and added N + P treatments led to slight insignificant increases in the concentration of litter P. Thus whilst added N + P was associated with better growth, changes in N and P leaf concentrations were not that obvious. No estimate of retranslocation was made. Tanner et al. (1992) suggest that the N and P may have gone elsewhere in the tree besides into leaves, but there does remain the possibility that much of it was held in the soil or was lost from it by leaching. No confirming tests of soil available concentrations were made (unlike the present study). The longer-term data showed that after 6 yr, whilst added-N did lead to more litter fall, the concentrations of N and P in this litter did not change. In the context of the present discussion it can be conjectured that N per se may not have led to increased tree growth but its indirect effect on the soil altered one or more other controlling factors that allowed a growth increase.
The one other lowland study to date (besides the present one) by Mirmanto et al. (1999) in lowland dipterocarp forest in central Borneo, Indonesia, used a randomized block design with N, P and N + P applied, with a control, to four plots within each of four blocks. The levels of N and P used between 1994 and 1997 were 225 and 75 kg ha−1 yr−1, respectively, as in Tanner et al. (1992). All three treatments led to a 30% increase in litter fall over the control, although these data were only recorded early in the trial (1994–95) and may have been a temporary ‘reaction response’. In this 1 yr, leaf P, but not N, leaf litter concentrations increased for all treatments compared with the control. Trunk growth did not increase significantly with fertilization. From this study it is not possible therefore to conclude whether N or P limited tree growth or leaf production (when litter fall is taken as a surrogate) in the longer term (cf. Tanner et al., 1990; Tanner et al., 1992), but the lack of effect is tentative support for the conclusion of the present work in Korup.
The interpretation of fertilization experiments in vegetation requires some caution. The idea is often that introducing a simple treatment will have a simply interpretable response on a 1 : 1 cause-effect basis. Forests, however, are complex, long-lived and highly interconnected ecosystems in which short-term adjustment responses and time lags are to be expected. Adding large quantities of fertilizer to such a presumably near-equilibrium system (in terms of nutrient cycling) is tantamount to a major disturbance. Furthermore, adding a single nutrient can have a multitude of side-effects which lead, perhaps in the short term, to contrary, ameliorated or even opposite effects than expected.
In conclusion, P supply does not appear to limit the establishment and growth of seedlings, nor the growth of trees, of the Korup tree species as tested under current conditions. Phosphorus-limitation is not the reason why the present seedling-sapling recruitment of M. bisulcata remains so poor at this site. Whether P is important at an earlier successional stage of the proposed cyclic-mosaic process, that is after a strong disturbance event, is not yet known. Different limiting nutrients may control tree growth at different stages.
This research was financially supported by the European Commission (DG XII-contract TS3-CT93-0233) based first at the University of Stirling and then at the University of Bern from 1996. We thank the Institute of Agricultural Research for Development (IRAD) of Cameroon for facilitation of our Programme, and the Ministries of Forests and Environment and of Scientific and Technical Research for permission to work in Korup. We are grateful to E. Abeto for the field assistance in Korup, especially for the recording of the seedling quadrats; M. Zimmermann for technical assistance with the leaf nutrient analyses in Bern, and the director of the Ekona Centre of IRAD, Dr S. Zok, for making available facilities for soil analysis. Two anonymous reviewers are thanked for their comments.
Table 1. Density (no of trees in plot) and basal area abundance of trees 50 cm dbh of species with ≥ 5 individuals in the 8.75-ha Isangele Road plot, Korup
No. of trees
Basal area (m2 ha−1)
Includes five trees undetermined. Anis, Anisophyllaceae, Caes, Caesalpiniaceae; Comb, Combretaceae; Euph, Euphorbiaceae; Gutt, Guttiferae; Irvi, Irvingiaceae; Myri, Myristicaceae; Rubi, Rubiaceae; Ruta, Rutaceae; Scyt, Scytopetalaceae; Verb, Verbenaceae.
Table 2. Density (no of trees in plot) and basal area abundance of trees 10– <50 cm dbh of species with ≥ 20 individuals in 3.50 ha (14 subplots) of the 8.75-ha Isangele Road plot, Korup. Also shown are other ectomycorrhizal species with ≥ 5 individuals
Table 3. Mean per-tree absolute (agr) and relative (rgr) growth rates of intermediate-sized (10– < 50 cm dbh) trees averaged over subplots of the two treatments (control, added-P) of the experimental plot at Isangele Road (1995–2000), with mean annualized mortality rates for species with > 40 trees. Significant differences between treatments are indicated by superscripted small letters referring to the foot of the Table
Table 4. Mean per-tree absolute (agr) and relative (rgr) growth rates of small (1– < 10 cm-dbh) trees averaged over subplots of the two treatments (control, added-P) of the experimental plot at Isangele Road (1995–2000), with mean annualized mortality rates for species with > 40 trees. Significant differences between treatments are indicated by superscripted small letters referring to the foot of the Table