This study is part of the ErA Net BiodivERsA CoForChange project which involves the following companies: Alpicam, Bois et Placages et Lopola, Danzer, Congolaise Industriale des Bois, Industries Forestieres de Batalino, Likouala Timber, Mokabi SA. and Vicwood. The authors thank the timber companies Alpicam, BPL, Danzer, DLH, IFB, Likouala Timber, Rougier, SEFCA, SCAD, SCAF and Vicwood for authorizing access to the inventory data, and the field teams who conducted the inventories. The consulting firm Forest Resource Management (FRM), facilitated contacts and exchange with several logging companies, participated in data collection and data compiling, and provided their inventory data files. There are no patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials, as detailed online in the guide for authors.
Analyzed the data: AF FM. Contributed reagents/materials/analysis tools: VF JLD NF GC MRM. Wrote the paper: AF BE MS SGF.
Understanding the factors that shape the distribution of tropical tree species at large scales is a central issue in ecology, conservation and forest management. The aims of this study were to (i) assess the importance of environmental factors relative to historical factors for tree species distributions in the semi-evergreen forests of the northern Congo basin; and to (ii) identify potential mechanisms explaining distribution patterns through a trait-based approach.
We analyzed the distribution patterns of 31 common tree species in an area of more than 700,000 km2 spanning the borders of Cameroon, the Central African Republic, and the Republic of Congo using forest inventory data from 56,445 0.5-ha plots. Spatial variation of environmental (climate, topography and geology) and historical factors (human disturbance) were quantified from maps and satellite records. Four key functional traits (leaf phenology, shade tolerance, wood density, and maximum growth rate) were extracted from the literature. The geological substrate was of major importance for the distribution of the focal species, while climate and past human disturbances had a significant but lesser impact. Species distribution patterns were significantly related to functional traits. Species associated with sandy soils typical of sandstone and alluvium were characterized by slow growth rates, shade tolerance, evergreen leaves, and high wood density, traits allowing persistence on resource-poor soils. In contrast, fast-growing pioneer species rarely occurred on sandy soils, except for
The results indicate strong environmental filtering due to differential soil resource availability across geological substrates. Additionally, long-term human disturbances in resource-rich areas may have accentuated the observed patterns of species and trait distributions. Trait differences across geological substrates imply pronounced differences in population and ecosystem processes, and call for different conservation and management strategies.
Identifying the factors that shape species distributions is a central issue in ecology. Since species distribution patterns underlie community composition, diversity and ecosystem function, an understanding of the factors forming these patterns is necessary for projecting consequences of climate and land use changes, and for designing effective conservation and forest management strategies.
Many tropical tree species are differentially distributed with respect to environmental factors at continental, regional and local scales
Few studies have yet related tropical tree species distribution patterns at large scales with processes that potentially determine them
The aims of this study were (i) to assess the role of environmental and historical factors in shaping the distribution patterns of common tree species in the semi-evergreen forests of the northern Congo basin taking spatial auto-correlation into account; and (ii) to identify potential mechanisms (e.g. environmental filtering or historical factors) explaining distribution patterns through a trait-based approach. At regional scale, we expected tropical tree species with similar distribution patterns to converge in strategy because of establishment, survival and/or growth barriers imposed by environmental and/or historical factors.
In this study, we used abundance data for 31 common tree species in 56,445 0,5-ha plots spread over more than 700,000 km2 at the northern edge of the Congo basin and combined them with the spatial variation of factors potentially driving the tree distribution patterns as well as with values for functional traits indicative of resource use strategy. This is one of the first quantitative data sets on tropical forest composition assembled at this scale.
The study area was distributed over south-eastern Cameroon, southern Central African Republic and northern Republic of Congo (
Plot scores on the first three compositional axes of a correspondence analysis of the plots (n = 56,445)×species (n = 31) abundance matrix were mapped (A, C, E). Variance explained by each axis is given in brackets. Solid lines represent country borders, names of main cities are indicated on the third map. Barplots give species scores across each compositional axis (B, D, F), with bar shading indicating the four species groups with contrasting distribution patterns that were identified by a cluster analysis (
In the study area, forest inventories were conducted by logging companies to quantify timber resources and their spatial variation. Data from forest inventories have so far been little exploited in the Congo basin despite the consistency of protocols and the vast spatial area covered but see
Species (Family) | n | Freq (%) | Leaf pheno |
Shade tol |
WD (g.cm−3) | Growth (cm.yr−1) |
7 783 | 11.49 | Ever | P | 0.725 | 0.796 | |
3 869 | 5.89 | Deci | P | 0.378 | 0.955 | |
4 383 | 6.65 | Deci | NPLD | 0.782 | 1.273 | |
1 319 | 2.13 | Ever | NPLD | 0.876 | 0.700 | |
1 29 | 2.05 | Deci | P | 0.382 | 1.114 | |
2 71 | 4.39 | Deci | P | 0.418 | 1.114 | |
5 262 | 7.46 | Deci | P | 0.275 | 1.910 | |
14 133 | 16.28 | Ever | ST | 0.446 | 0.796 | |
2 385 | 3.77 | Deci | P | 0.565 | 1.114 | |
11 56 | 15.28 | Ever | ST | 0.658 | 0.637 | |
8 383 | 12.68 | Deci | NPLD | 0.461 | 0.955 | |
7 133 | 11.10 | Deci | NPLD | 0.572 | 0.796 | |
28 271 | 33.80 | Deci | NPLD | 0.573 | 1.273 | |
2 203 | 3.65 | Deci | NPLD | 0.521 | 0.955 | |
14 923 | 18.95 | Deci | ST | 0.484 | 1.114 | |
6 755 | 9.25 | Ever | ST | 0.527 | 0.796 | |
9 859 | 12.95 | Ever | ST | 0.568 | 0.637 | |
9 807 | 9.48 | Ever | P | 0.864 | 0.637 | |
3 903 | 5.48 | Deci | NPLD | 0.440 | 1.353 | |
5 624 | 8.16 | Ever | ST | 0.586 | 0.509 | |
4 613 | 7.01 | Deci | P | 0.547 | 1.273 | |
993 | 1.59 | Deci | P | 0.725 | 1.114 | |
36 382 | 21.9 | Ever | P | 0.250 | 3.820 | |
2 888 | 4.56 | Ever | P | 0.627 | 1.253 | |
13 798 | 19.63 | Ever | NPLD | 0.715 | 0.955 | |
1 68 | 2.68 | Deci | ST | 0.614 | 0.780 | |
12 65 | 18.01 | Deci | NPLD | 0.587 | 2.324 | |
29 384 | 32.96 | Ever | NPLD | 0.414 | 1.114 | |
46 424 | 41.73 | Ever | ST | 0.744 | 0.637 | |
36 743 | 30.34 | Deci | P | 0.450 | 2.228 | |
18 483 | 13.12 | Deci | P | 0.327 | 1.910 |
Total stem number (n) and frequency of occurrence (Freq, % of plot presences) were calculated in the 56,445 0.5-ha plots. Leaf phenology (Leaf pheno) and shade tolerance (Shade tol) were extracted from
Abbreviations for leaf phenology and shade tolerance correspond to:
Deci: deciduous species; Ever: evergreen species.
P: pioneer species; NPLD: non-pioneer light demanding species; ST: shade tolerant species.
We assigned the values of five environmental variables pertaining to climate (annual rainfall and dry season length), topography (slope and altitude) and geology (8 distinct substrates) to each plot. Climatic data were obtained by spatial extrapolation of METEOSAT records (P. Mayaux, pers. com.). According to these data, mean annual rainfall varied between 1200 mm and 1700 mm over the study area. While the METEOSAT records tended to underestimate rainfall in comparison with available ground measurements, they were suitable for comparisons among sites. The climate in the study area is characterized by a pronounced dry season of varying duration. The length of the dry season (in months) with rainfall less than 50 mm was used to quantify the seasonality of rainfall, which ranged from 0 to 3 months. Topography was characterized by slope (0–112%) and altitude (270–1070 m), obtained from the Shuttle Radar Topography Mission (SRTM). A homogenized geological map of the study area was obtained through the comparative analysis of three national maps
To assess recent human disturbance (i.e. in the 20th century), we extracted information on land cover through digitalizing 15 topographical and land-use maps (1∶200 000 scale) by the French national mapping agency (IGN), published between 1959 and 1964 (maps are available in the IGN cartographic library, Paris, France). Most of the plots experienced no or minor human disturbances, and were classified as dense forest (97.2%), seasonally flooded forest (85 plots, 0.2%) or swamps (477 plots, 0.9%). The remaining plots were classified on the maps as degraded forest (180 plots, 0.3%) or savannas and plantations (843 plots, 1.5%), and we considered them as disturbed for the purpose of this study. Additionally, we included buffer zones in which forest plots were considered to have been prone to recent human disturbance: 15, 10, 5 and 3 km around main towns, main villages, secondary villages and hamlets respectively, and of 1 km and 500 m around main roads and tracks, respectively (6,780 plots, 12%). Combining information on land cover from old maps and on proximity to road and villages from recent maps (buffer zones), a total of 7,382 plots (13%) were classified as disturbed. For maps of environmental and historical factors see
We compiled information on four key functional traits of tropical trees: leaf phenology, shade tolerance, wood density and maximum growth rate. Information on leaf phenology was compiled from
We investigated the main axes of variation in species composition and underlying patterns of species distribution with correspondence analysis (CA) of the plots (n = 56,445)×species (n = 31) abundance matrix. CA, also known as reciprocal averaging, produces a simultaneous ordination of plots and species
To identify the main determinants of species composition and patterns of species distribution, we first used linear models relating plot scores on the first three compositional axes to the set of explanatory variables pertaining to environment and history of human disturbances. As to be expected, the residuals of the linear models remained highly spatially autocorrelated for the three axes (with values of Moran's I of 0.23, 0.13 and 0.18, respectively). We then used simultaneous autoregressive (SAR) models to deal with spatial autocorrelation. We specifically used spatial error models, which are recommended for spatially autocorrelated distribution data
Finally, to quantify the relative impact of each environmental and historical variable on the three compositional axes, we used a simple forward approach based on the Likelihood Ratio test (LR test) and the BIC to select explanatory variables sequentially
To assess differences in species composition among geological substrates, we compared plot scores on the first compositional axis for pairs of substrates using pairwise Wilcoxon tests at P<0.001. Additionally, we analyzed differences in stand structure (i.e. total plot basal area) among substrates, to control for a possible confounding effect of stand age. We tested the relationship between species distribution patterns (species scores on the first compositional axis) and key functional traits. We used Kruskall-Wallis one-way analysis of variance for categorical traits (leaf phenology and shade tolerance) and Spearman correlation coefficients for quantitative traits (wood density and growth rate). We tested for significant differences at P<0.05 in species scores among shade tolerance categories using pairwise Wilcoxon tests. All analyses were performed within the R environment
The 31 focal species represented on average 17.5% of the stems and 25.3% of basal area in the plots. These common tree species occurred in immense areas of African rainforest, similar to the situation in upper Amazonian forests
We identified three dominant and independent geographical gradients of tree species composition (
In agreement with the patterns described above, a cluster analysis of the species scores identified four contrasting distribution patterns (
A hierarchical cluster analysis on the Euclidean distances between species scores and an average agglomeration method was used to identify groups of species according to their distribution patterns, i.e. position across the compositional axes. The grey boxes indicate the cut-off level used to identify the four groups.
Geological substrate, climate and the recent history of human disturbance were the main drivers of species distribution patterns in the study area, which were strongly spatially autocorrelated (
Model |
BIC | LR test | df | P-value | λ | P-value (λ) |
Axis 1∼1 | 94 545 | |||||
Axis 1∼geology | 94 529 | −92.3 | 7 | <2.2×10−16 | 0.90 | <2.2×10−16 |
Axis 2∼1 | 123 664 | |||||
Axis 2∼dry season | 123 575 | −99.9 | 1 | <2.2×10−16 | ||
Axis 2∼dry season+geology | 123 555 | −97.0 | 7 | <2.2×10−16 | 0.77 | <2.2×10−16 |
Axis 3∼1 | 92 536 | |||||
Axis 3∼disturbance | 92 463 | −83.9 | 1 | <2.2×10−16 | ||
Axis 3∼disturbance+dry season | 92 450 | −23.8 | 1 | 0.4×10−5 | 0.83 | <2.2×10−16 |
We used spatial error models to identify the most important factors (environment or history) for species composition taking spatial autocorrelation into account. To select explanatory variables sequentially we used a simple forward approach. At each step from the null model, we used the Likelihood Ratio test (LR test) to assess the significance of adding a new explanatory variable in the model and the Bayesian Information Criteria (BIC) to select the most important variable. Best spatial models (lowest BIC) are given for the three compositional axes. The value and significance of the spatial autoregression coefficient (λ) is also given for the best spatial models.
For convenience, the spatial term (λW
Species distribution and community composition were strongly affected by the underlying geological substrate (
Plot scores on the first compositional axis were used to describe the main variation in species composition (A). Structural differences were assessed as the % deviation from mean plot basal area (B). Significant differences between substrates at P<0.001 in paired Wilcoxon tests are indicated by different letters.
A large sandstone plateau covering ca. 25,300 km2 in the centre of the study area played a dominant role in shaping distribution patterns (
The strong association between tree species distributions and geological substrates, on which soils develop that have distinct chemical and structural properties and consequently different nutrient and water availability, suggests that soil resource availability is important for these patterns. Dominant soils on both the sandstone and alluvium substrates were characterized as sandy loam
Previous experimental and observational data on the two species with the most contrasting distribution patterns (
In contrast to geology, climate and disturbance had significant but lesser influence on species distribution patterns, determining the remaining variations in the distribution of two pioneer species
These results were in marked contrast with other studies showing strong relationships between tree species distribution and rainfall
Ancient human disturbances (>100 y, not accounted for in the disturbance variable used here) may have accentuated the observed patterns of species distributions across geological substrates. Specifically, forests growing on sandy soils may have experienced little human disturbances in the past because local people may have preferred fertile sites for slash and burn agriculture. Such differential land use would have favoured the abundance of fast growing pioneer species on fertile soils, and the persistence of slow growing species on less fertile sandy soils. Footprints of ancient human disturbances have indeed been demonstrated in tropical forests worldwide
We examined whether species with contrasting distribution patterns exhibited characteristic suites of functional traits. We found that species scores on axis one, indicating species association with geological substrates, differed between species with different traits (
Species scores on the first compositional axis were used as an indicator of species association with the geological substrate. Relationships between species scores and functional traits were assessed for leaf phenology (A), shade tolerance (B), wood density (C), and maximum annual growth rate (D). Different lower case letters above the boxplots indicate significant differences (P<0.05) in paired comparisons using Wilcoxon tests. Regression lines were plotted for quantitative traits. Symbol shading indicates the four species groups with contrasting distribution patterns: light and dark grey symbols indicate species positively or negatively associated with the sandstone substrate, respectively, while black and white symbols indicate the two pioneer species
The four traits examined were highly correlated with each other (
Evergreen leaves tend to be favoured under low soil fertility and/or high soil moisture conditions, because they limit the loss of nutrients, but may increase water demand through year-round activity
Wood density of tropical trees has been shown to decrease with both soil fertility and rainfall
Pioneer species, except
Species associated with sandy soils exhibited low potential growth rates. Traits that enable plants to exploit low-resource environments, such as dense wood and evergreen leaves (see above) are physiologically linked to low growth rates
The differential distribution of the four correlated traits, which are indicative of the trait syndromes of rapid acquisition
The data are also consistent with a resource-related demographic trade-off, where low mortality rates (associated with low growth rates) allow persistence on low-resource soils, while high growth rates (associated with high mortality rates) give species an advantage on high-resource soils
Overall, our results show that species with similar distribution patterns converge in strategy, suggesting that species that do not possess adequate functional traits are not able to survive, grow or reproduce in the community
The observed trait distributions have consequences for ecosystem function, suggesting high biomass and carbon storage, but low productivity and turn-over rates of the forests on resource-poor sandy soils
Maps of environmental and historical factors. Spatial variation of five environmental factors: annual rainfall (A), dry season length (B), slope (C) and altitude (D) and geology (E), and one historical factor (recent human disturbance, F) were quantified from maps and satellite records. Climate and topography correspond to satellite records (METEOSAT and SRTM, respectively) while geology is a synthesis of three national maps. The recent history of disturbance combines information on forest cover from old maps with data on proximity to road and villages from recent maps (see
(TIF)
Results of the pairwise relationships between the four functional traits. To test the correlation between pairs of functional traits, we used Spearman correlation coefficient (rS) for quantitative traits, Kruskal-Wallis chi-squared test (K-W χ2) for a mix of a quantitative and a categorical trait; and chi-squared (χ2) test for categorical traits.
(DOCX)
We thank the timber companies Alpicam, BPL, Danzer, DLH, IFB, Likouala Timber, Rougier, SEFCA, SCAD, SCAF and Vicwood for authorizing access to the inventory data, and the field teams who conducted the inventories. Acknowledgements are also due to the forest managers who compiled and referenced the data, and to the consulting firm Forest Resource Management (FRM), which facilitated contacts and exchange with several logging companies, participated in data collection and data compiling, and provided their inventory data files. We thank Philippe Mayaux (DG-Joint Research Center in ISPRA) who kindly provided the climate data and Liza Comita for constructive comments on the manuscript.