Tree Diversity in Tropical Rain Forests: A Validation of the Intermediate Disturbance Hypothesis Jean-François Molino and Daniel Sabatier

An exact measure of the impact of disturbance on tree species diversity could theoretically be achieved only through a complete monitoring of all trees, from seedlings to adults, during decades after disturbance. However, even on the oldest and most heavily studied permanent sample plots (such as BCI), only trees above 1 cm dbh are monitored on the whole area, and only for the past 10 to 20 years. On the other hand, although an increasing number of studies document germination and growth rates of various species in different light environments, they are too limited in space and time scales as well as in the number of monitored taxons, to allow extrapolation to the whole tree community.

It is thus necessary to admit somewhat arbitrary limits for the portion of the tree community to be studied. While choosing these limits, we kept in mind three considerations:

(i) if the minimum dbh size is too high, the studied dataset would mainly consist of trees already present before the disturbance event (not enough time allowed for new trees to be recruited in this dbh size class). Thus changes in species composition and diversity would only reflect the disappearance of trees killed by disturbance;

(ii) if the maximum dbh size is too low, the dataset would not include trees that have been most influenced by disturbance in their germination and establishment (at the time of census, they would already have been recruited in the higher dbh size class). In that case, although still perceptible, the effects of disturbance could be partly concealed, thus more difficult to emphasize.

(iii) tree density decreases steeply from lower to higher dbh size classes: therefore, environmental conditions are much more constant for a given number of individuals of small dbh than for the same number of big trees, because the latter are scattered on a larger area. Consequently, it should be easier to separate disturbance from other sources of variation (e.g. topography, soils, etc.) for lower dbh size classes.

Artificial disturbances in the 9 treated Paracou plots resulted in a fall of the mean stem density (stems with dbh 10 cm) from 620 stems/ha before 1986 to 473 stems/ha in 1989 (a 24% decrease) (1). After this 1989 minimum, mean stem density increased each year, but in 1997 it was still 7% inferior to the 1986 value (577 vs. 620 stems/ha) (1). Thus the composition and diversity of the above 10 cm dbh size class in 1996-1997 were still more marked by the mechanistic effects of disturbance (trees killed in 1987-1989) than by ecological ones (changes in light environments that influenced new recruitments after 1987). On the other hand, this means also that most trees that have germinated after the main disturbance event are still in lower dbh size classes. We think justified, according to points (i) and (ii), to restrict our dataset to trees with dbh below the conventional 10-cm limit.

The lower dbh limit was chosen in order to fit with our working capacities.

Supplemental Table 1. Tree species richness at Paracou and Piste de St Elie, as compared with BCI. N = number of stems; S = number of species. See Paracou Map (Supplemental fig. 1) for the positions of the theoretical 50-ha plots. Because Paracou-PSE theoretical plots were only partially censused, their greater species richness, as compared with BCI's fully censused plot, is highly significative.
50 ha areas	censused area		dbh (cm)	N	S
	ha	%
Paracou
transects 1, 2 & 3	1.5	3	2	5273	436
			2 and < 4	2475	344
transects 4, 5 & 6	1.5	3	2	5035	411
			2 and < 4	2215	324
transects 9, 10 & 11	1.5	3	2	5199	408
			2 and < 4	2410	337
PSE
transect 5B10B	1	2	2	2916	369
			2 and < 4	1326	257
BCI
whole plot a	50	100	1	244000	303
all gaps b	6.1	12.2	1 and < 4	22834	229
a. data from (2); b. data from (3); c. not given in (2) nor (3), but P at least equal to the value for all gaps.

Supplemental table 2. List of species and morphospecies identified during the census of all trees with dbh 2 cm in the 6.04-ha study area in French Guyana [5 ha at Paracou and 1.04 ha at Piste de St. Elie (PSE), see Supplemental fig. 1]. Vouchers for all taxa are deposited at Cayenne Herbarium (CAY). Family circumscriptions follow (4). H: Heliophilic guild; P: Pioneer; NP: Non-pioneer. SB: Pioneer species of the seed bank; The pioneer and non-pioneer heliophilic guilds were circumscribed through compilation of earlier independent studies. Except for one case (a species listed as pioneer by one author, and as shade-loving by another), we retained in the pioneer guild all Paracou-PSE species that were listed as pioneer in at least one of these earlier studies. The heliophilic guild was formed by adding to this pioneer guild the species described as sun-loving. Ref.: source references for H/SB classifications; misidentifications and taxonomical changes since publication of these studies were taken into account, but are not explicited here (5).

Supplemental Figure 1. Map of Paracou experimental site, showing camp, main trails and the twelve 9-ha (300 m × 300 m) experimental plots. Colours refer to the type of treatment applied in 1986-1988. Green = control plots (P1, P6 & P11). Blue = Treatment 1: commercial logging only, of trees of dbh > 50 cm, on average ca. 10 stems per ha (P2, P7 & P9). Yellow = Treatment 2: commercial logging + thinning (poison-girdling) of all non-commercial trees of more than 40 cm dbh (P3, P5 & P10). Pink = Treatment 3: commercial logging + logging for fuelwood of all non-commercial trees of dbh between 40 and 50 cm + thinning of trees of more than 50 cm dbh (P4, P8 & P12). Gray strips represent the ten 0.5-ha transects of the present study. Dashed rectangles show theoretical 50-ha (500 m × 1000 m) plots used in Supplemental table 1 for comparison with BCI plot.

Medium version | Full size version

Supplemental Figure 2. The heliophilic guild shows distinctive and homogeneous response to disturbance, as compared with the remaining tree community. In all figures, each black dot represents one of the 67 non-trail 20 m x 20 m terra firme quadrats of the Paracou study area. The relationship between %HS and Lost Basal Area (LBA) (see printed paper for definitions) is given for comparison (red dots). (A) A pool of 97 species (the same size as the heliophilic guild) was randomly chosen among the 449 non-heliophilic species. The percentage of stems that belong to this group (%RCNH) was calculated in each quadrat. The operation was repeated 40 times, in order to obtain a mean percentage of RCNH stems (%RCNHS) for each quadrat. Mean %RCNHS significantly decreased with increasing LBA (F_1,65 = 22.45, P<0.001); error bars are 99% confidence intervals; (B) Same as A, but with 40 groups of 97 randomly chosen species (RC) among the whole 546-species Paracou dataset. Here, we found no relationship between the mean percentage of RC stems (%RCS) and LBA on the same 67 quadrats (F_1,65 = 0.04, P = 0.8).

Medium version | Full size version

Supplemental Figure 3. Studies of seedlings germinating from forest soil that was experimentally exposed to full sunlight resulted in the characterization of a subset of species that are undoubtedly and markedly pioneer, and constitute the core of the pioneer guild (19). They have typically small, long dormant seeds that are disseminated over relatively large areas by bats or birds (15), and that germinate exclusively in newly opened large gaps (7, 13, 14, 15, 19). Using such independent studies (7-9, 13-15), 40 soil seed bank species (Psb) have been identified among the 61 pioneer species of our Paracou dataset, representing 54.2 % of all heliophilic stems. (A) As well as for the whole pioneer guild, the percentage of soil seed bank pioneer stems (%PsbS) increased with LBA (F_1,65 = 17.92, P<0.001). (B) Meanwhile, the percentage of non-seed-bank heliophilic stems (%HnsbS) is also positively influenced by LBA (F_1,65 = 12.60, P<0.001).

Medium version | Full size version

Supplemental Figure 4. Species richness [E(S₄₀), see printed text] of the 2 to 10 cm dbh tree communities in the 99 Paracou 400 m^² terra firme quadrats, as a function of the percentages of seed bank pioneer stems (%Psb) (see Supplemental fig.3) or other, non seed bank heliophilic stems (%HnsbS) in the same dbh class. (A) E(S₄₀) as a function of %PsbS (F_2,96 = 22.67, P<0.001). (B) E(S₄₀) as a function of % HnsbS (F_2,96 = 23.89 P<0.001); quadrats in red are those having %PsbS > 30 (highly disturbed); if they are excluded, r^² = 0.4372, F_2,82 = 31.85, P<0.001. Both relationships are hump-backed as in fig. 2A of the printed text; the left ascending part of the latter curve (slightly disturbed quadrats) is better explained by %HnsbS, whereas the right descending part (highly disturbed quadrats) is better explained by %Psb. The combination of %Psb and %HnsbS in a single estimator (i.e., %HS) allows the best partitioning of all quadrats as regards the diversity-disturbance relationship.

Medium version | Full size version

Supplemental Figure 5. The relationships between species richness [E(S₄₀), see text] of the 2 to 10 cm dbh tree communities in the 99 Paracou 400 m² terra firme quadrats and the percentages of stems that belong to randomly chosen groups of species (see Supplemental fig. 2 for definitions) in the same dbh class, emphasize the reliability of our definitions of the heliophilic and pioneer guilds. (A) E(S₄₀) as a function of the mean percentage of stems that belong to 97 randomly chosen non-heliophilic species (%NHRCS); error bars are 99% confidence intervals. (B) Same as A (horizontal scale has been changed for better understanding), but quadrats partitioned in the same 5 classes as in Fig. 2C of the printed text (q.v.); Class 1: F_1,16 = 18.44, P < 0.001; Class 2: F_1,15 = 5.67, P<0.05; Class 3: F_1,15 = 2.65, N.S.; Class 4: F_1,13 = 0.07, N. S.; Class 5 F_1,30 = 34.65, P<0.001; note that quadrats are ordered as in Fig. 2 of the printed text, but symmetrically. (C) Black dots: E(S₄₀) as a function of the mean percentage of stems that belong to 97 species randomly chosen among the 546 species (%RCS); error bars are 99% confidence intervals; red dots: the relationship between E(S₄₀) and %HS (same as in fig. 2B of the printed text, given for comparison); the marked condensing of all %RCS values is another confirmation of our choice of heliophilic species.

Medium version | Full size version

Supplemental Figure 6. Species richness [E(S₄₀), see printed text] of the 2 to 10 cm dbh tree communities in the 67 non-trail 400 m^² terra firme quadrats of Paracou, as a function of the percentage of heliophilic stems (%HS) in the same dbh class (F_2,64 = 26.34, P<0.001). Because soil seed bank pioneers could indicate not only canopy light-gap disturbances, but also soil disturbances, and because such soil disturbances are more important on skid trails than elsewhere, we tested for the relationship between %HS and species richness on non-trail quadrats alone (this figure). For the same purpose, we verified that the percentage of stems in the 59 heliophilic, non-seed-bank species (%HnsbS) showed similar relationships with disturbance (Supplemental fig. 3B) and diversity (Supplemental fig. 4B).

Medium version | Full size version

REFERENCES AND NOTES

1. S. Gourlet-Fleury, "Dispositif sylvicole expérimental de Paracou. Aperçus sur l'état et la dynamique des peuplements 13 ans après installation" (CIRAD-Forêt - Silvolab-Guyane, 2000).
2. R. Condit et al., J. Ecol. 84, 549 (1996).
3. S. P. Hubbell et al., Science 283, 554 (1999).
4. A. Cronquist, An Integrated System of Classification of Flowering Plants (Columbia Univ. Press, New York, 1981).
5. For example, Rollinia exsucca is cited in (8, 15) under the synonym R. resinosa Benth; Hydrochorea corymbosa (Rich.) Barneby & Grimes in (6) is actually Abarema jupunba (pers. observ.). The relevant taxonomical literature is too abundant to be cited here, but interested readers could refer to monographs published by the New York Botanical Garden for the Flora Neotropica project, to the Flora of the Guianas series [A. R. A. Görts-van Rijn, Ed., Utrecht University, Utrecht, Netherlands], and to (20).
6. V. Favrichon, Rev. Ecol. (Terre Vie) 49, 379 (1994).
7. M.-F. Prévost, Turrialba 31, 121 (1981).
8. M.-F. Prévost, Rev. Ecol. (Terre Vie) 38, 121 (1983).
9. M.-F. Prévost, unpublished data.
10. D. Toriola, thesis, Paris VI University (1997).
11. D. Toriola, J. Trop. Ecol. 14, 373 (1998).
12. D. Toriola, P. Chareyre, A. Buttler, J. Trop. Ecol. 14, 323 (1998).
13. H. de Foresta, thesis, Montpellier University (1981).
14. H. de Foresta, M.-F. Prévost, Biotropica 18, 279 (1986).
15. H. de Foresta, P. Charles-Dominique, Ch. Erard, M.-F. Prévost, Rev. Ecol. (Terre Vie) 39, 369 (1984).
16. Larpin, thesis, Paris VI University (1988).
17. B. Riéra, thesis, Toulouse University (1983).
18. J. P. Schultz, in The Vegetation of Suriname, I. A. de Hulster, J. Lanjouw, Eds. (Van Eedenfonds, Amsterdam, 1960), vol. 2.
19. D.-Y. Alexandre, Dynamique de la régénération naturelle en forêt dense de Côte d'Ivoire (ORSTOM, Paris, 1989).
20. J. Boggan et al., Checklist of the plants of the Guianas (Guyana, Surinam, French Guiana), ed. 2 (Smithsonian Institution, Washington, DC, 1997).