Forest Ecology and Management 304 (2013) 322–332
Contents lists available at SciVerse ScienceDirect
Forest Ecology and Management
journal homepage: www.elsevier.com/locate/foreco
Silvicultural disturbance has little impact on tree species diversity in a Central African moist forest ⇑ S. Gourlet-Fleury a, , D. Beina b,c, A. Fayolle d,e, D.-Y. Ouédraogo a, F. Mortier a, F. Bénédet a, D. Closset-Kopp c, G. Decocq c a UPR BSEF, CIRAD, Campus International de Baillarguet, F-34398 Montpellier, France b Université de Bangui – Cerphameta, BP 1450 Bangui, Central African Republic c Université de Picardie Jules Verne, UR ‘‘Ecologie et Dynamique des Systèmes Anthropisés’’ (EDYSAN, FRE 3498 CNRS), 1 rue des Louvels, F-80037 Amiens, France d Ministère des Eaux, Forêts, Chasse et Pêche, avenue du 19 janvier, BP 3314 Bangui, Central African Republic e Agro Bio-Tech, Université de Liège, 2, Passage des Déportés, B-5030 Gembloux, Belgium article info abstract
Article history: Timber production is an important economic sector in most forested countries of Central Africa, where Received 29 January 2013 about 14 million hectares of lowland moist forests are now planned for management. This production Received in revised form 8 May 2013 is expected to be sustainable, but the actual impact of logging on biodiversity is still questioned. Accepted 10 May 2013 To answer this question, we used a unique long-term controlled experiment implemented more than 20 years ago in an old-growth semi-deciduous moist forest of the Central African Republic (CAR). We tested whether (i) anthropogenic disturbances associated with silvicultural operations had an effect on Keywords: the composition and diversity of tree communities, and (ii) there is a relationship between diversity Anthropogenic disturbance and disturbance intensity in those forests. Logging Thinning For this, we botanically identified all trees P 10 cm DBH in 28 1-ha plots where no treatment (con- Regeneration guild trols), logging and logging + thinning operations were implemented 24 years ago and created a strong Diversity indices gradient of disturbance. We investigated the relationships between five diversity indices and a distur- M’Baïki experiment bance index calculated for each 1-ha plot, for all species and separately for three regeneration guilds. We found a strong positive monotonic relationship between the intensity of disturbance and the per- centage of pioneer species in the tree communities, which proved to be equally detrimental, in terms of relative abundance, to the non-pioneer light-demanding and the shade-bearing species. Overall, disturbance appeared to have a weak monotonous negative effect on diversity, irrespective to the index considered. The diversity of shade-bearers slightly decreased along the disturbance gradient without significant decrease in species density; disturbance had no effect on non-pioneer light demand- ers, but a clear significant negative effect on the diversity of pioneers, with a significant decrease in spe- cies density. This negative effect was associated with the massive recruitment of the early-successional, fast-growing Musanga cecropioides R. Br. (Urticaceae), which rapidly preempted space and resources in the most disturbed plots. Despite this effect, disturbance did not significantly affect the local heterogene- ity in species distribution. These results suggest that the semi-deciduous moist forests of CAR are locally resilient to small-scale disturbances associated with silvicultural operations. This may be a consequence of the past anthropo- genic and/or climatic disturbances, which have been stronger and more long-lasting than elsewhere within the tropical forest biome, and would have removed the most vulnerable species. Because logging intensity in these forests is usually low, we do not expect any direct major impact on tree species diver- sity, at least after the first felling cycle. Ó 2013 Elsevier B.V. All rights reserved.
1. Introduction
Timber logging plays an important role in the economy of Cen- ⇑ Corresponding author. Tel.: +33 (0)4 67 59 38 83; fax: +33 (0)4 67 59 37 33. E-mail addresses: sylvie.gourlet-fl[email protected] (S. Gourlet-Fleury), d_beina@ tral Africa. The formal forestry sector often represents the main yahoo.fr (D. Beina), [email protected] (A. Fayolle), dakis.ouedraogo@gmail. sector able to generate direct and indirect employment as well as com (D.-Y. Ouédraogo), [email protected] (F. Bénédet), deborah.closset- funding for infrastructure in still rural economies (Bayol et al., [email protected] (D. Closset-Kopp), [email protected] 2012). In the formal sector, logging occurs within concessions (G. Decocq).
0378-1127/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.foreco.2013.05.021 S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 323 granted to private companies within the permanent forest estate. ments (logging, logging + thinning) were added to natural gap An increasing part of these concessions is submitted to a manage- dynamics to create a strong gradient of disturbance intensity. ment plan, made mandatory in all forest laws since the 1990s (Nasi The experimental design at M’Baïki is similar to that used by et al., 2012). This has resulted in over 14 million ha of forests cur- Molino and Sabatier (2001) in French Guiana, and the forest type rently managed, out of 44 million ha of lowland moist forests allo- compares to the ‘‘moist forests’’ of Bongers et al. (2009) in Ghana. cated to timber logging in long-term concessions (Bayol et al., Our first goal was to examine whether anthropogenic distur- 2012). bances associated with the silvicultural operations still affect the Management plans implemented in Central Africa are supposed tree species composition in terms of regeneration guilds. We to be sustainable and thus to ensure ‘‘the production of a continu- hypothesized an increased importance of the most light- ous flow of desired forest products and services without undue demanding, early successional tree species to the detriment of reduction of (the forest) inherent values’’ (ITTO, 2005). Whether the most shade-bearing, late successional species in response to timber can be sustainably produced has been a highly controver- the increased light levels beneath the canopy. Our second goal sial topic (e.g. Bawa and Seidler, 1998; Rice et al., 1997, 1998; was to make profit from the disturbance gradient to examine Zimmerman and Kormos, 2012, vs Fredericksen and Putz, 2003; whether there is a relationship between diversity and disturbance Pearce et al., 2003; Putz et al., 2012), a central question being the intensity in those forests, in order to identify possible unsustain- potential negative impact of logging on tree species diversity. able levels of logging and thinning. For this, we used a variety of Three recent meta-analyses did evidence an overall negative im- diversity indices accounting for species richness, species evenness pact of logging on tree species richness (Clark and Covey, 2012, and combinations thereof. We expected (i) some diversity indices Gibson et al., 2011, Putz et al., 2012) but this impact resulted from like species richness to peak at an intermediate position along a high variety of logging situations (in terms of logging regime – the gradient, as observed by Molino and Sabatier (2001) and, to a frequency and intensity, initial state of forests, interaction with lesser extent, by Bongers et al. (2009); (ii) different shapes of the other types of disturbance) and hid a variety of results: in some re- disturbance-diversity relationship among the regeneration guilds, viewed cases, disturbance did increase species richness (e.g. Bobo as previously reported by Bongers et al. (2009). More specifically, et al., 2006; Swaine and Agyeman, 2008; van Andel, 2001). we expected a monotonic increase of species diversity – due to Disturbances associated with logging do not necessarily have a both increased richness and evenness – with increasing distur- negative impact on diversity. In mature forests, logging create gaps bance in pioneer and light-demanding species, and a concomitant which, when large enough, can be colonized by tree species requir- monotonic decrease of diversity in shade-bearing tree species. ing more light than the species previously settled in the canopy. When a forest has long remained undisturbed and is dominated 2. Materials and methods by slow growing late successional species, large and medium-sized gaps created by logging and post-logging treatments might in- 2.1. Study site crease diversity (e.g. monodominant Cynometra forests in Uganda, Plumptre, 1996) while, conversely, small natural gaps in forests The M’Baïki Experimental Station (3°900N, 17°930E) is located in dominated by early and mid-successional species might have lim- the Lobaye Province, Central African Republic (CAR), 110 km south- ited effect, or tend to reduce diversity (e.g. secondary forests in west to Bangui. The climate is humid tropical with a 3 to 4-months Panama, Hubbell et al., 1999; Sheil and Burslem, 2003). According dry season (November/December–February, with less than 50 mm to the intermediate disturbance hypothesis (IDH), diversity is ex- precipitation). The average annual rainfall is 1738 mm (1982–2007 pected to be the greatest at intermediate intensity or frequency period), the mean temperature is 24.9 °C (1981–1989 period) of disturbance, as well as at intermediate time since the last (meteorological station of Boukoko, A. Ougou, pers. comm.). The disturbance, when there is a balance between late successional plots are located within a 10 km radius on a large plateau (500– shade-bearing competitors and early successional light-demand- 600 m a.s.l) and the topography is generally flat. The Precambrian ing colonizers (Connell, 1978; Mackey and Currie, 2001; Sheil geological substrate is covered with a mix of schists, sandstones and Burslem, 2003). The IDH has been supported by empirical and quartzites (Ceccato et al., 1992). The alteration of the parent studies in tropical forests (time since major disturbance in moist material led to deep, locally gravelly, red ferralitic soils, corre- forests of Uganda, Sheil, 2001, intensity effect in moist forests of sponding to acrisols in the WRB soil classification (IUSS Working French Guiana, Molino and Sabatier, 2001, intensity and frequency Group WRB, 2006). The vegetation is a semi-deciduous moist forest effects in wet to dry forests of Ghana, Bongers et al., 2009) but not of the Guineo–Congolian type (White, 1983). The canopy is domi- systematically (Hubbell et al., 1999; Sheil and Burslem, 2003). nated by species from the Malvaceae, Meliaceae, Ulmaceae, and Studies reporting the impacts of logging on tree species diver- Sapotaceae plant families (Boulvert, 1986). sity in tropical moist forests are scarce (Clark and Covey, 2012, Gib- son et al., 2011, Putz et al., 2012 and references therein) and they are even scarcer for Africa. Moreover, most of these studies used 2.2. Experimental design observational, rather than experimental designs. This may hamper unambiguous comparison of forest communities’ response to dis- Ten 4-ha permanent plots were established in 1982 in an old- turbance as the latter is usually not definitely dated and well- growth forest to monitor the effects of silvicultural treatments characterized, and may be due to multiple confounding factors. (Bedel et al., 1998). Each 4-ha plot is composed of four 1-ha sub- Given the importance of logging in Central Africa, and the unsolved plots, inside which all trees with a DBH P 10 cm were individually debate about its potentially negative impact on tree species diver- marked, geo-referenced, and botanically identified. The plots have sity, there is an urgent need for the results of long term experi- been annually monitored since 1982, with all standing ments in old growth forests of the region. trees P 10 cm DBH recorded. Between 1984 and 1985, seven 4- In this study, we used a unique data set from a long-term con- ha plots were selectively logged, while three 4-ha plots were kept trolled experiment, which was implemented more than 20 years untouched to act as controls. During logging operations, trees from ago – a near complete felling cycle – in a previously undisturbed 16 timber species with a DBH P 80 cm were harvested (2–7 stems semi-deciduous moist forest of the Central African Republic per hectare). Between 1986 and 1987, four of these seven logged (CAR). We examined patterns of tree species diversity in 28 1-ha plots were additionally thinned to increase light availability in plots of the M’Baïki Experimental Station where silvicultural treat- the understory and promote tree regeneration. Thinning consisted 324 S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332
Table 1 Structural and diversity characteristics of the 28 1-ha subplots. Subplot: identity of the 1-ha subplot, treatment: C = control, L = logged, LT = logged + thinned. N: number of trees P 10 cm DBH in 2009, Nl, Gl and Vl: number of trees P 10 cm DBH, cumulated basal area and standing volume lost between 1984 and 1987, pNl, pGl and pVl: percentage of the number of trees P 10 cm DBH, basal area and standing volume lost, relative to the number of living trees, basal area and standing volume of trees in 1984, %P: 100 (number of pioneer trees / total number of trees P 10 cm DBH in 2009), %NPLD and %SB: same as %P, for the number of non-pioneer light demanding trees and shade bearing trees respectively, S: species richness, r: species density (rarefaction index calculated for N = 504 trees), H0: Shannon entropy, Fisher’s a: Fisher’s diversity index, J0: Pielou’s index, z: Arrhenius dissimilarity index. The indices H0, Fisher’s a, J0 and z were the mean values obtained for 1000 repetitions of 504 randomly selected trees in each subplot (see text). Calculation details for the diversity indices are given in the footnote.
Subplot N Nl pNl(%) Gl (m2 pGl Vl (m3 pVl %P %NPLD %SB Sr H’ Fisher’s J’ z (Treatment) (ha 1) (ha 1) ha 1) (%) ha 1) (%) (N = 504) a 111 (L) 600 63 10.5 7.7 24.6 82.9 28.5 10.7 25.2 64.0 144 132.6 4.09 42.8 0.873 0.933 112 (L) 646 21 3.6 2.2 7.8 22.6 9.0 8.0 28.9 62.7 127 116.3 3.97 37.4 0.862 0.915 113 (L) 555 131 21.3 10.8 31.9 100.0 33.0 15.9 23.2 60.5 119 114.2 3.96 34.2 0.872 0.923 114 (L) 651 43 6.9 1.0 3.8 7.7 3.3 12.7 27.5 59.6 135 121.9 4.05 39.6 0.873 0.927 131 (C) 627 12 1.9 0.3 0.9 1.8 0.7 8.3 27.3 64.1 154 140.6 4.23 48.4 0.887 0.935 132 (C) 615 3 0.5 0.5 1.4 3.6 1.2 5.5 32.4 62.0 127 117.6 4.05 37.7 0.879 0.920 133 (C) 575 6 1.0 0.8 2.6 8.9 3.0 7.5 23.1 69.2 127 120.6 4.09 38.4 0.885 0.935 134 (C) 585 2 0.3 0.8 2.5 8.4 2.6 5.0 22.6 72.3 131 124.0 4.06 40.6 0.873 0.920 141 (L) 579 116 20.3 11.8 32.2 136.1 36.8 14.2 26.9 58.5 122 115.9 4.03 36.3 0.879 0.930 142 (L) 634 69 11.5 4.6 15.6 48.6 17.8 13.1 27.3 59.5 132 120.5 4.10 39.1 0.886 0.933 143 (L) 562 106 18.6 7.7 25.4 88.0 30.1 14.2 24.6 61.0 114 109.2 3.87 32.5 0.857 0.915 144 (L) 606 113 19.8 10.4 28.5 105.3 31.1 14.9 24.6 60.6 125 114.8 3.90 34.8 0.855 0.915 151 (LT) 720 41 6.5 5.6 21.7 58.8 25.1 10.0 30.6 59.4 133 118.6 4.11 40.4 0.883 0.928 152 (LT) 717 59 9.8 10.8 35.3 114.1 39.5 8.9 29.7 61.4 126 112.2 4.03 36.8 0.877 0.923 153 (LT) 721 89 14.2 15.8 44.9 172.5 50.8 10.7 33.0 56.3 135 117.0 3.91 38.0 0.847 0.910 154 (LT) 658 52 9.0 12.7 36.8 130.9 40.3 9.3 29.2 61.4 134 121.0 4.08 39.7 0.879 0.930 161 (C) 696 1 0.2 0.6 1.7 6.3 1.8 8.0 29.7 62.2 138 122.9 4.20 41.6 0.898 0.939 162 (C) 701 15 2.5 0.7 2.0 6.7 2.1 6.4 32.0 61.1 139 123.2 4.16 42.0 0.890 0.934 163 (C) 641 22 4.1 1.0 2.9 8.8 2.4 8.0 27.3 64.3 131 120.1 4.07 39.6 0.876 0.929 164 (C) 711 33 5.9 1.6 5.9 12.5 4.7 7.2 24.3 68.1 135 118.1 4.01 38.6 0.867 0.926 211 (L) 504 84 16.1 12.3 32.7 130.1 36.1 18.1 20.4 61.5 114 114.0 3.94 34.4 0.865 0.917 212 (L) 617 54 9.2 8.5 25.8 98.8 31.7 11.3 24.5 63.7 125 115.7 4.06 36.9 0.883 0.929 213 (L) 553 28 6.0 4.1 12.1 55.8 16.7 14.3 26.0 59.1 124 119.4 4.09 37.3 0.890 0.933 214 (L) 581 63 12.0 8.6 26.8 97.8 31.7 12.2 23.2 64.5 131 123.3 4.11 39.5 0.886 0.934 221 (LT) 613 108 19.5 12.5 38.0 131.6 42.0 17.5 17.6 64.6 131 121.2 4.10 38.9 0.886 0.935 222 (LT) 670 39 7.0 11.5 35.4 121.9 39.0 8.7 21.6 69.3 137 123.4 4.21 41.5 0.901 0.946 224 (LT) 651 53 8.8 13.5 35.9 128.2 36.7 9.5 24.0 66.1 137 123.2 4.14 40.5 0.889 0.935 244 (C) 547 4 0.8 0.3 1.0 2.8 1.0 9.3 25.4 65.1 125 120.5 4.03 36.9 0.876 0.930 P N x N S: total number of species in each subplot. r ¼ EðS Þ¼ 5 1 1 = . i.e. the expected number of species in a sample of n trees (S ) selected at random from a n i¼1 n n n P 0 5 xi xi collection containing N trees, S species, with xi trees belonging to species i (Hurlbert, 1971). H ¼ i¼1 N ln N with xi = number of trees belonging to species i. Fisher’s a represents diversity in a logarithmic series estimating the number of species S within N observed individuals: S = a ln(1 + N/a)(Magurran, 2004). J0 = H0/ln(S)(Pielou, 1966) and z = log(S)/log(cA), i.e. z is the exponent of the Arrhenius model relating the number of species S to the size of the inventory area (A) (Arrhenius, 1921). We used 100 10 m 10 m subdivisions of each 1 ha-subplot to calculate z. We used N = 504, the minimum number of trees inventoried in the subplots in 2009 to calculate r, H0, Fisher’s a, J0 and z (see text).
in poison girdling all trees from non-timber species with a field; for the others, vouchers were taken for further identification DBH P 50 cm (16–26 stems per ha), and in systematically remov- thanks to later herbarium work at the National Herbarium of Yao- ing lianas (Bedel et al., 1998; Ouédraogo et al., 2011). undé (Cameroon) or with images of specimen reference (types or The three treatments (control, logging, logging + thinning) led holotypes) at http://apps.kew.org/herbcat/. to a strong gradient of disturbance intensity among the subplots Of the 28 1-ha subplots included in this study, a total of 17,536 (Table 1; col. 3 to 6). Some of the subplots experienced additional trees were recorded, of which 17,420 (99.3%) were identified to the disturbance events: (i) soon after logging, in one of the 4-ha logged species level, 75 to the genus level and 41 remained unidentified. A plots, all the trees belonging to the early pioneer Musanga cecropio- total of 295 species, 176 genera and 46 families were recorded. ides R. Br. (Urticaceae) were uprooted or cut down to limit Trees identified to the genus level (6 genera) and unidentified trees post-logging competition; (ii) in 1983, a fire ran across eight 1-ha were considered as single species in the analyses, thus yielding subplots within three 4-ha plots (one control and two log- 6 + 1 additional pseudo-species. ged + thinned plots). Because these events were likely to have a Each species was further assigned to one of the three regenera- specific impact on species composition and diversity (e.g. Yedmel tion guilds defined by Hawthorne (1995), with a priority to Haw- et al., 2010, for fire), we removed the 12 affected 1-ha subplots thorne’s assignment when available and completed thanks to local from our analyses. expert knowledge (J.-L. Doucet, comm. pers.) or personal observa- tions (D. Beina). Pioneers (P) are species requiring full light condi- tions all their life cycle long (n = 60 species + 1 taxon identified to 2.3. Species sampling genus); non-pioneer light-demanders (NPLD) are species able to re- cruit in understories but needing light to grow up to the canopy The number of species targeted by botanical surveys at M’Baïki (n = 69 + 2 taxa identified to genus); shade-bearers (SB) are species has progressively increased since 1982, due to evolving research able to recruit and grow in shade conditions (n = 166 + 3 taxa iden- questions: commercial species only have been identified from tified to genus). Unidentified trees remained unassigned. 1982 to 1992, then the set was enlarged. From 2008 to 2010 and The list of species and their assignment to guilds are given in for the first time, all trees P 10 cm DBH were identified by a pro- Appendix A (Table A.1). fessional botanist (D. Beina). Most species were identified in the S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 325
2.4. Quantifying disturbance indices with linear models. We calculated the ilr orthonormal coor- dinates as: We quantified disturbance intensity in the 28 1-ha subplots using four indices (Table 1): (i) the total number of trees P 10 cm x1 ¼ð2=3Þ1=2 log½%P=ð%NPLD %SBÞ1=2 ð1Þ DBH lost during silvicultural operations (Nl, ha 1), calculated as the difference between the number of living trees in 1984 (just be- x2 ¼ð1=2Þ1=2 logð%NPLD=%SBÞð2Þ fore logging operations) and in 1987 (once logging and thinning operations were completed); (ii) the total basal area lost (Gl, (iii) we finally used the inverse of the ilr transform, to illustrate m2 ha 1) between 1984 and 1987; (iii) the total standing volume the relationship between the guilds and the disturbance indices in lost (Vl, m3 ha 1), calculated as the over-bark volume of trees lost the true compositional space. between 1984 and 1987 using volume equations specifically devel- Second, we modeled the relationship between each of the 5 oped in CAR (Tran-Hoang et al., 1991); and (iv) the percentage of standardized diversity indices and disturbance intensity for all spe- trees belonging to pioneer species in 2009 (%P). cies and separately for P, NPLD and SB (5 4 = 20 relationships), The four indices reflected slightly different aspects of the distur- using Linear Mixed Models (LMM, McCulloch et al., 2008). After bance linked to treatments: Nl, Gl and Vl directly quantified the checking that %P was highly correlated to the other disturbance space freed up by logging, thinning and linked damage, with Gl indices (Section 3.1), we kept it as the most appropriate index to and Vl accounting for the mean size of trees lost. The percentage quantify disturbance intensity, first because of its integrative nat- of pioneers %P was a more integrative index, rendering the effects ure, and second because it allowed direct comparisons with other of all types of disturbance throughout recent history (Molino and studies. Relating diversity indices with the other disturbance indi- Sabatier, 2001; Sheil and Burslem, 2003; Slik, 2005; Bongers ces did not change our conclusions (results not shown). et al., 2009). We expected %P to reflect the intensity of disturbance To account for the main trends expected (monotonic increase, linked to treatments, as well as the possible subsequent interac- monotonic decrease, peak at intermediate values), we fitted the tions between anthropogenic and natural disturbances (e.g. trees following general model: may be more exposed to windthrow in logged stands, making 2 the latter more exposed to natural disturbance). Yij ¼ l þ xi b1 þ xi b2 þ dj þ eij ð3Þ
2.5. Quantifying species diversity where Yij denotes a diversity index and xi the disturbance index (%P), l is the intercept, b1 and b2 are two unknown parameters, dj For each 1-ha subplot, we quantified diversity (Table 1) with is the 4-ha plot effect assumed to be a random effect such as 2 three commonly used indices: (i) species richness (S), (ii) Shannon d =(d1,...,d28) is a Gaussian vector ðd Nð0; rdÞÞ, and eij is the 0 2 entropy (H ) and (iii) Fisher’s a, which represents diversity in a log- Gaussian residual error ðe Nð0; re ÞÞ. d and e are assumed to be arithmic series estimating the number of species S within N ob- independent. A random plot effect was used to take into account served individuals (Magurran 2004): S = a ln(1 + N/a). the possible dependencies between 1-ha subplots into a same 4- To further interpret the results, we calculated two additional ha plot (spatial autocorrelation). To select the best model associated indices: (iv) Pielou’s index of evenness (J0) which measures how with each relationship we proceeded as follows: close in terms of number of stems species are; (v) Arrhenius dis- (i) we selected the covariance structure by comparing the two 2 similarity index (z) which is derived from the classic power law full models (all covariates, i.e. xi and xi , included) with and describing the species-area relationship and quantifies the scale- without the random effect. The best model was the one min- independent increase of species richness with the increased sur- imizing the BIC criterion: BIC = 2 log-likeli- veyed area (Harte et al., 1999). To calculate z, we used 10 m hood + log(nobs) npar, where nobs represents the 10 m squares inside each subplot. This index may be interpreted number of observations (nobs = 28 subplots) and npar the 2 2 as a measure of within-subplot heterogeneity in species number of free parameters (l, b1, b2, re , rd if the random 2 distribution. effect is included, or l, b1, b2, re otherwise) (Schwarz, 1978); As all these measures might be biased by differences in stem (ii) given the covariance structure, we selected the best density among plots (Gotelli and Colwell, 2001; Magurran, 2004), expectation model using the Likelihood Ratio test (LR test) we calculated the indices for 1000 samples of a standard number (McCulloch et al., 2008). of trees – the minimum number observed in the 1-ha subplots – randomly selected into each subplot. For S, we used the rarefaction We also calibrated the more flexible non-linear HOF index r (Hurlbert, 1971), or species density, whose expected value (Huisman–Olff–Fresco) models proposed by Oksanen and Minchin could be calculated exactly (see also footnote of Table 1). (2002) and successfully used by Bongers et al. (2009). However, in The 6 indices (S+ the 5 standardized indices) were calculated for our study, they never proved to fit better than the linear models. all species together, and separately for the three guilds (P, NPLD, In the following, models without (respectively with) the ran- 0 SB). dom effect are denoted Mpk (respectively Mpk), where p is the num- ber of fixed covariates included in the expectation model (1 or 2)
2.6. Statistical analysis and k is associated with the maximum power of xi (1 or 2) (see Ta- ble 2 for a summary of all the models tested). First, to test the impact of disturbance on the regeneration guild Finally, to improve our understanding of the relationships evi- composition, we regressed the compositional vector defined for denced, we examined whether disturbance might have a positive each 1-ha plot as (%P, %NPLD, %SB) against each of the three distur- or negative influence on the presence of each species taken individ- bance indices Nl, Gl and Vl. Because the compositional vector was ually in the 28 1-ha subplots. For this purpose, we tested for a sig- constrained to sum 100%, a simple linear model relating each of the nificant difference in %P values between the subplots where the components to each index would have raised biased results. In- target species occurred and the subplots where the target species stead, we used the so-called ‘‘simplicial regression’’ (Egozcue was absent (Wilcoxon signed rank test). et al. (2012). We (i) projected the compositional vector into R2 by All statistical analyses were performed with the R statistical using the isometric log-transformation (ilr); (ii) modeled the rela- software (R Development Core Team, 2009). The diversity indices tionship between the ilr orthonormal coordinates and each of the were calculated with the vegan package, simplicial regressions 326 S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332
Table 2
Models fitted to the Mbaïki data set, using Linear Mixed Models. Yij: diversity index, xi: disturbance index (%P), l: intercept, b1 and b2: unknown parameters. Parameters were 0 estimated with maximum likelihood, and the BIC criterion was used to select the best model (see text for more detailed explanations). M00 and M00 are the null expectation models.
Model (deterministic part) Behavior Denomination Without random effect With random effect
0 Yij = l Flat (null expectation) M00 M00 0 Yij = l + xi b1 Monotonous M11 M11 2 Humpbacked (mode on x = 0), symmetric M 0 Yij ¼ l þ xi b2 12 M12 2 Humpbacked, symmetric M 0 Yij ¼ l þ xi b1 þ xi b2 22 M22
were performed with the compositions package, and the LMM were 3.2. Disturbance effects on species diversity fitted with the lme4 package. We observed the higher species richness (S = 154, r = 141) and diversity (H0 = 4.23, Fisher’s a = 48.3) in one of the control subplots (%P = 8.3%), and the lower species richness (S = 114, r = 109) and 3. Results diversity (H0 = 3.87, Fisher’s a = 32.5) in one of the logged subplots (%P = 14.2%). Evenness (J0) and Arrhenius dissimilarity (z) ranged 3.1. Disturbance effects on regeneration guilds between 0.847 and 0.901, and between 0.910 and 0.946 respec- tively; these extreme values were found in two logged + thinned The gradient of disturbance intensity in the M’Baïki Experiment subplots (%P = 10.7% and %P = 8.7% respectively) (Table 1). was marked (Table 1): the total number of trees lost in the 1-ha 0 We found the null expectation model (M00 or M ) to be the best subplots between 1984 and 1987 (Nl) ranged from 33 to 00 fitted model for 13 of the 20 relationships we examined (Table 3, 131 ha 1 ( 5.9% to 21.3% of the number of trees living in 1984). Fig. 2). For the 7 remaining relationships, the best models fitted The ranges for Gl and Vl were respectively 1.6 to 15.8 m2 ha 1 were monotonic negative (model M11) in 4 cases, and humpbacked ( 5.9% to 44.9% of the basal area in 1984) and 12.5 to (model M12) in 3 cases. The peak of the humpbacked model was 172.5 m3 ha 1 ( 4.7% to 50.8% of the standing volume in 1984). never found within the range of observed %P values, and the diver- The nine subplots which experienced a gain in basal area and sity indices decreased as disturbance intensity increased (Fig. 2). standing volume (i.e. negative values of Gl and Vl ranging from Models including a random effect were found to perform better 1.6 to 0.3 m2 ha 1, and from 8.9 to 1.8 m3 ha 1 respectively) in only 3 cases, all associated with the null expectation model. all belonged to control plots. In most cases, the number of trees We found that disturbance had no significant impact on species also increased in these subplots. density, except for pioneers whose richness significantly decreased In terms of Nl, the highest disturbance was experimented by a when %P increased. In contrast, we found a negative effect of dis- logged subplot, where %P reached 15.9%. In terms of Gl and Vl, dis- turbance on diversity as reflected by H’ and Fisher’s a for all species turbance was the highest in a logged + thinned subplot, where %P and for species of the pioneer guild. We found a negative effect of reached 10.7%. These discrepancies result from the characteristics disturbance on the diversity of shade-bearers, only when measured of the silvicultural treatments: logged subplots experimented a with Fisher’s a. We did not find evidence for any significant effect higher felling intensity, with a higher damage rate in small trees of disturbance on species evenness, except once again for pioneers resulting in higher Nl values. Conversely, in logged + thinned sub- (negative), and we found no significant effect of disturbance on plots, thinning eliminated a number of medium-sized trees with a Arrhenius dissimilarity. The diversity of non-pioneer light low damage rate, resulting in higher Gl and Vl values without nec- demanders did not appear affected by logging and thinning- essarily high Nl values. Nevertheless, the three indices were highly associated disturbance in this forest. positively correlated (Pearson’s coefficients r = 0.83, r p(Nl, Gl) p(Nl, Overall, 158 species (52% of the total number of species re- = 0.82, r = 0.99, all p-values < 0.001). Vl) p(Gl, Vl) corded in the subplots) tended to be absent from the subplots In 2009, the total number of trees per subplot ranged from undergoing the highest levels of disturbance (Appendix A, 504 trees ha 1, in a logged subplot (%P = 18.1%), to 721 trees ha 1, Table A.1, mean value of %P in subplots where the species is pres- in the logged + thinned subplot with the highest Gl and Vl. Tree ent < mean value of %P in subplots were the species is absent), but density was not significantly correlated with either of the distur- this effect was significant only for 13 species. Among the regener- bance indices Nl, Gl or Vl (Pearson’s coefficients, all p-values > 0.1). ation guilds, the number of species tending to be negatively af- In all 1-ha subplots most trees belonged to the SB guild (range: fected by disturbance amounted 53% in SB, 58% in NPLD and 46% 56.3–72.3%), followed by the NPLD and P guilds (17.6–33.0% and in P. The effect was significant for only 6 (0.04% of SB species), 4 5.0–18.1%, respectively; Table 1). (0.06% of NPLD species) and 3 (0.05% of P species) species in the The percentage of pioneer trees (%P) was positively correlated three guilds (Fig. 3 and Appendix A, Table A.1.). with the three disturbance indices Nl, Gl and Vl (Pearson’s coeffi- cients rp(%P, Nl) = 0.82, rp(%P, Gl) = 0.60, rp(%P, Vl) = 0.61, all p-val- ues < 0.001). The simplicial regression showed a highly 4. Discussion significant positive effect of either Nl, Gl and Vl on x1 (i.e. the ratio log[%P/(%NPLD %SB)]), indicating that the greater the number of In this study, we aimed to test whether anthropogenic distur- trees, basal area or standing volume lost, the greater the proportion bances associated with silvicultural operations affect the tree spe- of pioneer trees and the fewer the proportion of the two other cies composition and diversity in a Central African semi-deciduous guilds (Appendix A, Table A.2 and Fig. 1 for the relationships with moist forest, using a long-term controlled experiment. Twenty- Nl). Conversely, there was no significant effect of Nl, Gl or Vl on x2 four years after the application of a range of disturbance (i.e. log(%NPLD/%SB); hence, disturbance was equally detrimental intensities, we found evidence for only few long-lasting effects, to trees of the NPLD and SB guilds. suggesting a high resilience of this forest type. Our results did S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 327
(a) (b) (c)
Fig. 1. Effect of disturbance (number of trees P 10 cm DBH lost between 1984 and 1987) on the proportion of the three regeneration guilds (SB: shade-bearers; NPLD: non- pioneer light-demanders; P: pioneers) in the 1-ha subplots. (a) Relationships are shown for %P, %NPLD and %SB in (a), (b) and (c) respectively. The coordinates of %P, %NPLD and %SB were obtained by using the inverse of the ilr transformation, to take into account the constraint %P + %NPLD + %SB = 1 (see text for details). Note that the three curves are not independent: for each point of the x-axis, the sum of the predictions equals one.
Table 3 Best models found to describe the relationship between the 5 standardized diversity indices and disturbance intensity (measured as the percent of trees belonging to the pioneer guild: %P), for all species and separately for the three regeneration guilds (SB: shade-bearers, NPLD: non-pioneer light demanders, P: Pioneers). Notations and corresponding models are detailed in Table 2. Expectation models differing from the null model are shown in bold, and the corresponding adjusted R-squared and p-values are given. Sign (+)/
( ): for model M11, sign of the slope; for model M12 sign of the trend between the index and %P (the unimodal maximum was never observed within the data range).
Diversity index Best model according to the BIC criterion All species SB NPLD P
Species density (r) M00 M00 M00 M11 ( ) R2 = 0.473 p-Value < 0.001 0 0 Shannon entropy (H ) M12 ( ) M00 M00 M11 ( ) R2 = 0.147 R2 = 0.349 p-Value: 0.025 p-Value < 0.001
Fisher’s a M12 ( ) M11 ( ) M00 M12 ( ) R2 = 0.258 R2 = 0.172 R2 = 0.252 p-Value: 0.003 p-Value : 0.016 p-Value: 0.004 0 Pielou’s index (J ) M00 M00 M00 M11 ( ) R2 = 0.368 p-Value < 0.001 0 0 Arrhenius dissimilarity (z) M00 M00 M00 M00
not conform with the IDH at this local scale: overall, disturbance was similar or larger than the ones reported in other African appeared to have a weak monotonic negative effect on diversity. semi-deciduous forests (329 m2 in southern Ivory Coast, Surprisingly, this effect was mainly due to a clear negative effect Nierstrasz, 1975; 248 m2 in north-east Gabon, Florence, 1981), and on the diversity of the pioneer species, despite an increase in sufficient to favor the recruitment of pioneers (Oldeman, 1974; abundance. Hartshorn, 1978; Bazzaz and Pickett, 1980; Rollet, 1983a,b; Alexandre, 1989; Whitmore, 1998). More than twenty years after the logging operations, these pioneers were still present into the 4.1. Disturbance significantly increased the abundance of pioneers forest overstory, all the more abundant than the initial disturbance was high, thus modifying the overall tree species composition. The anthropogenic disturbance gradient created at M’Baïki ex- The non-pioneer light demanders and shade-bearers decreased tended up to 15.8 m2 ha 1 in terms of basal area removed, largely in proportion in the stands. For SB species, this result is in agree- exceeding the usual range of logging intensity in Central African ment with previous studies conducted in CAR (Hall et al., 2003) forests (3–5 m2 ha 1, corresponding to the removal of 1– and Uganda (Sheil et al., 2000). The result is more surprising for 2 trees ha 1; Fargeot et al., 2004; Durrieu de Madron and Forni, NPLD species, which were expected to benefit from the increase 1997). Twenty-four years after disturbance, however, we could of light availability. However, NPLD species are mostly tall species, not detect any effect on the total number of trees P 10 cm DBH. and at M’Baïki they contributed about 75% and 44% of the logged Similarly, elsewhere in CAR, Hall et al., (2003) found no significant and poisoned trees respectively. This loss of trees was not compen- difference in tree density between unlogged stands and stands that sated over the study period but Ouédraogo et al. (2011) evidenced a were logged 18 years before. positive effect of thinning on the growth and survival of NPLD trees Disturbance strongly favored pioneer trees, as already shown by between 1992 and 2008. Hence, we can expect that stem number of Ouédraogo et al. (2011) and observed in tropical forests worldwide the two non-pioneer guilds will recover within the next years. (Sist and Nguyen-Thé, 2002; Slik et al., 2002; Gourlet-Fleury et al., The strong relationships evidenced between %P and each of the 2004; Berry et al., 2008; Peña-Claros et al., 2008; Swaine and other disturbance indices quantifying the short-term impact of Agyeman, 2008). At M’Baïki, the mean gap size created by P80 cm logging and thinning operations (Nl, Gl and Vl) suggest that this DBH logged trees (350 m2, de Chatelperron and Commerçon, 1986) 328 S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332
(a) (b)
(c) (d)
(e) (f)
(g) (h)
Fig. 2. Anthropogenic disturbance effect on species diversity (see also Table 3 and Appendix A, Tables A.4, A.5, A.6). For all species: (a), (c), (e), (g), (i). Separately for each regeneration guild: (b), (d), (f), (h), (j). SB (black): shade-bearers, NPLD (dark grey): non-pioneer light-demanders, P (light grey): pioneers. Lines illustrate the best model fitted to each data set (horizontal line = null expectation model). Adjusted R-squared and p-values for models differing from the null expectation model are given in Table 3. S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 329
(i) (j)
Fig. 2.(continued)
(a) (b)
(c)
Fig. 3. Potential effect of the intensity of disturbance on the presence/absence of tree species in 28 1-ha subplots of the M’Baïki experimental site. Each point stands for a species. The dotted lines are positioned at the mean-value of %P for the 28 1-ha subplots (%Pmean = 10.7%). Position in the bottom right quadrant (resp. the top left quadrant) indicates that the species tends to be present where the level of disturbance is lower (resp. higher) than elsewhere. Black dots represent species for which a Wilcoxon signed- rank test rejected H0 (there is no significant difference of %P between the two subplots samples: species present/species absent) with a 5% error risk. (a) Shade-bearers. ‘‘antrnan’’: Antrocaryon nannanii, ‘‘dichgla’’: Dichostemma glaucescens, ‘‘drypche’’: Drypetes chevalieri, ‘‘dryppax’’: D. paxii, ‘‘mallopp’’: Mallotus oppositifolius, ‘‘maniaub’’: Manilkara aubrevillei, ‘‘rothlon’’: Rothmannia longiflora. (b) Non-pioneer light demanders. ‘‘chrylac’’: Chrysophyllum lacourtianum, ‘‘chrypru’’: C. pruniforme, ‘‘dacredu’’: Dacryodes edulis, ‘‘erytsua’’: Erythrophleum suaveolens, ‘‘penteet’’: Pentaclethra eetveldeana. (c) Pioneers. ‘‘fernado’’: Fernandoa adolfi-friderici, ‘‘maesemi’’: Maesopsis eminii, ‘‘rauvcaf’’: Rauvolfia caffra,‘‘zantgil’’: Zanthoxylum gilletii.
more integrative variable is relevant for quantifying disturbance in 4.2. Disturbance had a limited impact on tree species diversity logged forests – at least once-logged forests – when the history of logging is unknown. This may prove particularly useful for compar- Mean diversity values at M’Baïki were higher than those found isons, as well as for management purposes. in most of the other African sites studied in similar forests, close to 330 S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 the higher end of the diversity gradient identified by Parmentier Molino and Sabatier (2001), the effect was weak and significant at et al. (2007) (Appendix A, Table A.3). the landscape scale, on a large number (1322) of 1-ha plots. The Anthropogenic disturbance had a weak negative effect on tree range of disturbance (%P 0–80%) and the variability of the re- species diversity: while this effect was significant on Shannon en- sponse of species density on their plots largely encompassed ours: tropy and Fisher’s a, it was not on species density, evenness and the trends that we evidenced at the local scale in our study thus do dissimilarity. Overall, half of the species might have been nega- not conflict with their observations. tively impacted by disturbance to some extent, but only 13 (out of 301) significantly missed from the most disturbed areas. This 4.3. Disturbance differentially affected the diversity of regeneration might result either from a net loss of species due to disturbance, guilds or from trees being non-randomly distributed across subplots when the site was settled in 1982. Unfortunately, we do not have The overall weak monotonic negative effect of disturbance on the detailed botanical knowledge of the subplots before implemen- tree species diversity resulted from a differential response between tation of the treatments, and 12 of the 13 above-mentioned species regeneration guilds. We expected disturbance to favor the most were not known before 1992. However one species, Erythrophleum light-demanding species to the detriment of the most shade- suaveolens, has been recorded since the beginning: among the 11 bearing species, in response to the increased light levels beneath living trees in 2009, 9 were present in 1982 among which 7 were canopies. Actually, we evidenced a slight decrease of the diversity already located in the control plots. of shade-bearers across the disturbance gradient (without signifi- Our findings contrasted with those reported by Molino and cant decrease in species density), while no trend appeared for Sabatier (2001) for species density in French Guiana. Working on non-pioneer light demanders. But surprisingly, we found the diver- the sister experiment of M’Baïki (same treatments: logging and sity of pioneers to significantly decrease (all indices excepted dis- logging + thinning vs controls; similar disturbance intensity: at similarity) as disturbance intensity increased: this finding Paracou, Gl ranged between 0 and 16.5 m2 ha 1 i.e. a loss of be- contrasts with Bongers et al. (2009), who reported a significant po- tween 0% and 51.4% of the initial basal area), those authors evi- sitive monotonic relationship between pioneer species density and denced a peak of species density in places where silvicultural disturbance intensity. disturbances resulted in intermediate values of %P. However, their The negative response for pioneers can be explained by the study differed from ours both by the time after disturbance (10 vs massive recruitment of the early-successional, fast-growing 24 years respectively) and the census threshold DBH (2–10 cm Musanga cecropioides R. Br. (Urticaceae) in the most disturbed vs P10 cm): working sooner after disturbance, and on smaller plots, which rapidly preempted space and resources, leading to a trees gave to them a higher probability to detect clear patterns of strong decrease of both species density –absolute richness did disturbance effects. not decrease – and species evenness (J0) within the guild. While Another explanation may be the difference in disturbance his- this species represented between 7% and 20% of the pioneer tory between African moist forests and their Guianan counterparts. trees P 10 cm DBH in the control plots, it rose to 50% and even African forests have a history of stronger and longer lasting anthro- more in the most disturbed plots. M. cecropioides is widely spread pogenic and/or climatic disturbances than elsewhere within the and encountered throughout the domain of African moist forests, tropical forest biome (Corlett and Primack, 2006). At Paracou, where it colonizes forest gaps, road sides and abandoned fields Freycon et al. (2010) evidenced only little disturbance over the last (Aubréville, 1947). Given its unique growth performance thousand years. Control plots were more strongly dominated by (Aubréville, 1947; Coombe and Hadfield, 1962; Leroy-Deval, shade-bearers, with a far less abundant guild of pioneers than at 1967), it may have outcompeted the other, less fast-growing pio- M’Baïki (75% of SB trees and 2% of P trees – Favrichon 1998, vs neers, which decreased in abundance. We suggest four possible respectively 65.3% and 7.1%). At the M’Baïki site, located approxi- explanations for the discrepancy with the results of Bongers mately 60 km south to the forest-savanna boundary, the occur- et al. (2009): (i) the disturbance events experienced by the plots rence of past large-scale disturbances is suggested by the of M’Baïki could have been stronger and more recent compared dominance of the canopy by emergent P and NPLD trees such as to those experienced by the Ghanaian plots; (ii) M. cecropioides Triplochiton scleroxylon K. Sch. (Malvaceae), Terminalia superba may not be such a good colonizer under the drier conditions char- (Combretaceae) or Entandrophragma spp. (Meliaceae, African acterizing the Ghanaian moist forests (mean annual rainfall round- mahoganies). They can be estimated to be 100–500 years old, from ing 1540 mm in the area studied by Bongers et al. (2009), vs their diameter and mean growth rate, which is consistent with the 1740 mm at M’Baïki), as noticed in Ivory Coast (Kassi and Decocq, DBH-age relationship described by Worbes et al. (2003) in a 2008); (iii) the pioneer species pool might differ between Ghana semi-deciduous forest of Cameroon. These light-demanding spe- and CAR, with a greater number of very fast-growing species in cies require large gaps to regenerate and/or grow up to the canopy Ghana; (iv) finally, as underlined previously, Bongers et al. (2009) and hardly regenerate at M’Baïki (unpublished results). Similar fea- worked with a larger gradient of disturbance and a far greater tures in many moist forests of the region were attributed to the number of plots. The high variability of their species density re- long human occupation and shifting cultivation (Aubréville, sponse might hide any type of relationship in subsamples covering 1947; Letouzey, 1968; White and Oates 1999; van Gemerden a narrower disturbance range. et al., 2003a,b; Brncic et al., 2007), but could also be due to the se- The lack of difference in Arrhenius index across subplots sug- vere drought and fire episodes that accompanied regional rainfall gests that silvicultural operations did not affect the within-plot fluctuations during the last 500 years (Nicholson, 2000; Brncic heterogeneity, irrespective of the set of species considered. Hence, et al., 2009). Recurrent large-scale disturbance and/or relative the spatial mix of tree species characterizing control plots was not short time elapsed since last major disturbance are consistent with altered by silvicultural operations, consistent with the idea that the the particularly high species diversity measured at M’Baïki, M’Baïki forest has been shaped by a long history of disturbance. compared to other sites in the region (Appendix A, Table A.3): they might explain the weak effect on diversity of additional small-scale 5. Conclusion and implications for forest management disturbance, due to the importance of P and NPLD species in the stands. To the best of our knowledge, this study is the first one to exper- While in Ghanaian moist forests Bongers et al. (2009) evidenced imentally assess the impact of anthropogenic disturbance on tree a peak of species density at intermediate disturbance, similarly to diversity in an African semi-deciduous moist forest. Our records S. Gourlet-Fleury et al. / Forest Ecology and Management 304 (2013) 322–332 331 of diversity patterns 24 years after a single major disturbance References event revealed limited impact on tree species diversity, suggesting that this forest type is, at least locally, resilient to silvicultural Alexandre, D.Y., 1989. Dynamique de la régénération naturelle en forêt dense de Côte d’Ivoire. Etudes et thèses, Orstom, Paris. operations (logging and logging + thinning). African moist forests Arrhenius, O., 1921. Species and area. J. Ecol. 9, 95–99. have been repeatedly assumed to be more resilient than their Aubréville, A., 1947a. Les brousses secondaires en Afrique Equatoriale. Côte d’Ivoire American and Asian counterparts, as a result of past episodes that – Cameroun – AEF. Bois et Forêts des Tropiques 2, 24–49. had already removed the most vulnerable species (Balmford, Aubréville, A., 1947b. A propos: ‘‘Les pays tropicaux’’ de Pierre GOUROU – Le Parasolier – La régénération naturelle et l’enrichissement de la forêt 1996). Modern species assemblages have survived not only a mas- équatoriale. Bois et Forêts des Tropiques 4, 20–30. sive contraction in rainforest area during the Pleistocene glacia- Balmford, A., 1996. Extinction filters and current resilience: the significance of past tions, but also a longer period of human impact than elsewhere selection pressures for conservation biology. Trends Ecol. Evol. 11, 193–196. Bawa, K.S., Seidler, R., 1998. Natural forest management and conservation of (Maley and Brenac, 1998; Corlett and Primack, 2006). biodiversity in tropical forests. Conserv. Biol. 12, 46–55. The forest resilience issue is of utmost importance for achieving Bayol, N., Demarquez, B., de Wasseige, C., Eba’a Atyi, R., Fisher, J.-F., Nasi, R., sustainable forest management, especially in the Central African Pasquier, A., Rossi, X., Steil, M., Vivien, C., 2012. Forest management and the timber sector in Central Africa. In: de Wasseige, C., de Marcken, P., Bayol, N., region where about 26% of the forested area, EFBC, 2006) have Hiol Hiol, F., Mayaux, P., Desclée, B., Nasi, R., Billand, A., Defourny, P., Eba’a Atyi, been dedicated to logging (Bayol et al., 2012). Consistently with R. (Eds.), The Forests of the Congo Basin – State of the Forest 2010. Publications our results, low logging intensities (less than one tree per ha, lower Office of the European Union, Luxembourg, pp. 43–62. Bazzaz, F.A., Pickett, S.T.A., 1980. Physiological ecology of tropical succession: a end of our own disturbance gradient) significantly affected tree comparative review. Annu. Rev. Ecol. Syst. 11, 287–310. diversity neither in similar semi-deciduous forests of CAR (Hall Bedel, F., Durrieu de Madron, L., Dupuy, B., Favrichon, V., Maître, H.F., Bar-Hen, A., et al., 2003) and Ivory Coast (Kassi and Decocq, 2008), nor in old- Narboni, P. 1998. Dynamique de croissance dans les peuplements exploités et éclaircis de forêt dense africaine: dispositif de M’Baïki en République growth evergreen forests in Cameroon (van Gemerden et al. Centrafricaine (1982–1995). CIRAD-Forêt, Montpellier. (2003a). In contrast, in the semi-deciduous moist forests of the Berry, N.J., Phillips, O.L., Ong, R.C., Hamer, K.C., 2008. Impacts of selective logging on Omo Forest Reserve in Nigeria, high logging, logging + thinning tree diversity across a rainforest landscape: the importance of spatial scale. and clear-felling + burning disturbance (%P ranging from 16.1% to Landscape Ecol. 23, 915–929. Bobo, K.S., Waltert, M., Sainge, N.M., Njokagbor, J., Fermon, H., Muhlenberg, M., 57.5%, extending the M’Baïki’s disturbance gradient) strongly 2006. From forest to farmland: species richness patterns of trees and decreased tree species diversity (Okali and Ola-Adams, 1987). understorey plants along a gradient of forest conversion in southwestern Together, these studies and ours suggest that in African moist for- Cameroon. Biodivers. Conserv. 15, 4097–4117. Bongers, F., Poorter, L., Hawthorne, W.D., Sheil, D., 2009. The intermediate ests, disturbance has to be strong (i.e. leading to the development disturbance hypothesis applies to tropical forests, but disturbance contributes of a pioneer guild exceeding 20–30%) to cause a significant de- little to tree diversity. Ecol. Lett. 12, 798–805. crease in tree species richness and diversity. Since logging intensity Boulvert, Y., 1986. Carte Phytogéographique de la République centrafricaine (feuille 1 ouest – feuille est) à 1:1,000,000. Editions de I’ORSTOM, Paris. is usually below 2 trees ha in Central Africa, we do not expect Brncic, T.M., Willis, K.J., Harris, D.J., Washington, R., 2007. Culture or climate? any major impact on tree species diversity where large size P The relative influences of past processes on the composition of the lowland and NPLD species are found in the canopy. This intensity might Congo rainforest. Philos. Trans. Roy. Soc. Lond., Ser. B: Biol. Sci. 362, 229– 242. even be too low to ensure the regeneration of these usually eco- Brncic, T.M., Willis, K.J., Harris, D.J., Telfer, M.W., Bailey, R.M., 2009. Fire and climate nomically important species (Makana and Thomas, 2005; Hall, change impacts on lowland forest composition in northern Congo during the 2008; Doucet et al., 2009). However, the long-term impact of last 2580 years from palaeoecological analyses of a seasonally flooded swamp. Holocene 19, 79–89. repeated felling operations (every 30 years in most parts of the Ceccato, P., Bango, E., Ngouanze, F., Damio, T., 1992. Annexe 1: Etude pédologique region), as well as the effects of the depletion of major animal des parcelles d’expérimentation des forêts de Boukoko et La Lolé (M’Baïki, dispersers linked to increased hunting pressure on tree species République Centrafricaine). Projet de recherche forestière. FAC/ARRF/CTFT, diversity still have to be evaluated. Nogent-sur-Marne. Clark, J.A., Covey, K.R., 2012. Tree species richness and the logging of natural forests: A meta-analysis. For. Ecol. Manage. 276, 146–153. Acknowledgements Connell, J.H., 1978. Diversity in tropical rain forests and coral reefs. Science 199, 1302–1310. We thank the ARF Project (Appui la Recherche Forestière) and Coombe, D.E., Hadfield, W., 1962. An analysis of the growth of Musanga cecropioides. J. Ecol. 50, 221–230. its six funding partners: AFD (Agence Française de Développe- Corlett, R.T., Primack, R.B., 2006. Tropical rainforests and the need for cross- ment), CIRAD (Centre de Coopération Internationale en Recherche continental comparisons. Trends Ecol. Evol. 21, 104–110. Agronomique pour le Développement), ICRA (Institut Centrafricain de Chatelperron, G., Commerçon, R., 1986. Mise en exploitation du dispositif de recherche en forêt naturelle dans les forêts de Boukoko et La Lolé en République de Recherche Agronomique), MEFCP (Ministère centrafricain des Centrafricaine. FAC/ARRF/CTFT, Nogent-sur-Marne. Eaux, Forêts, Chasse et Pêche), SCAC/MAE (Service de Coopération Doucet, J.L., Kouadio, Y.L., Monticelli, D., Lejeune, P., 2009. Enrichment of logging et d’Actions Culturelles, Ministère des Affaires Etrangères), and gaps with moabi (Baillonella toxisperma Pierre) in a Central African rain forest. For. Ecol. Manage. 258, 2407–2415. SCAD (Société Centrafricaine de Déroulage) for providing access Durrieu de Madron, L., Forni, E., 1997. Aménagement forestier dans l’Est du to the site and to the data base. We are also grateful to Fidèle Baya Cameroun. Bois Forêts Tropiques 254, 39–50. (MEFCP), Laurent Cerbonney (SCAC/MAE), Emilien Dubiez (SCAC/ EFBC, 2006. Etat des Forêts du Bassin du Congo.
Gotelli, N.J., Colwell, R.K., 2001. Quantifying biodiversity: procedures and pitfalls in Pearce, D., Putz, F.E., Vanclay, J.K., 2003. Sustainable forestry in the tropics: panacea the measurement and comparison of species richness. Ecol. Lett. 4, 379–391. or folly? For. Ecol. Manag. 172, 229–247. Gourlet-Fleury, S., Favrichon, V., Schmitt, L., Petronelli, P., 2004. Consequences of Peña-Claros, M., Peters, E.M., Justinianoa, M.J., Bongers, F., Blate, G.M., Fredericksen, silvicultural treatments on stand dynamics at Paracou. In: Gourlet-Fleury, S., T.S., Putz, F.E., 2008. Regeneration of commercial tree species following Guehl, J.M., Laroussinie, O. (Eds.), Ecology and Management of a Neotropical silvicultural treatments in a moist tropical forest. For. Ecol. Manage. 255, Rainforest. Lessons drawn from Paracou, a long-term experimental research site 1283–1293. in French Guiana. Elsevier, Paris, pp. 254–280. Pielou, E.C., 1966. Species-diversity and pattern-diversity in the study of ecological Hall, J.S., 2008. Seed and seedling survival of African mahogany (Entandrophragma succession. J. Theor. Biol. 10, 370–383. spp.) in the Central African Republic: implication for forest management. For. Plumptre, A.J., 1996. Changes following 60 years of selective timber harvesting in Ecol. Manage. 255, 292–299. the Budongo Forest Reserve, Uganda. For. Ecol. Manage. 89, 101–113. Hall, J.S., Harris, D.J., Medjibe, V., Ashton, P.M.S., 2003. The effects of selective Putz, F.E., Zuidema, P.A., Synnott, T., Pena-Claros, M., Pinard, M.A., Sheil, D., Vanclay, logging on forest structure and tree species composition in a Central African J.K., Sist, P., Gourlet-Fleury, S., Griscom, B., Palmer, J., Zagt, R., 2012. Sustaining forest: implications for management of conservation areas. For. Ecol. Manage. conservation values in selectively logged tropical forests: the attained and the 183, 249–264. attainable. Conserv. Lett. 5, 296–303. Harte, J., McCarthy, S., Taylor, K., Kinzig, A., Fischer, M.L., 1999. Estimating species- R Development Core Team, 2009. R: A language and environment for statistical area relationships from plot to landscape scale using species spatial-turnover computing. R Foundation for Statistical Computing. Vienna, Austria.