Morphological Variation and Ecological Structure of Iroko ( Milicia excelsa Welw . C . C . Berg ) Populations across Different Biogeographical Zones in Benin

Iroko (Milicia excelsa) is a commercially important timber tree species formerly known by local people in Benin. Because of the highly attractive technological properties of its wood and its multipurpose uses, the species was subjected to intensive human pressure. Apart from strong climate oscillation during the Pleistocene, human caused habitat fragmentation through continuous land clearing for agriculture, extensive forests exploitation and urbanization induced the occurrence of many isolated forest plots and trees species among which Milicia excelsa trees. As fragmentation was proved to have deleterious effects on genetic diversity within a species and its morphological structure, it was of interest to investigate the current demographic, morphological and genetic structure of M. excelsa before coming up with conservation strategies. In the current study, morphological variation and ecological structure of M. excelsa populations were assessed in Benin using transect sampling method and multivariate analyses including principal component, cluster and canonical discriminant analyses. On the basis of morphological parameters, M. excelsa individuals and populations were clustered into four and discrimination of groups indicated that most of variations were highly related to edaphic factors and annual rainfall. Erratic diameter distribution was found for many populations although most of them showed bell shaped diameter distribution.


Introduction
Since habitat destruction and fragmentation are the major causes of species extinction [1]; this is the greatest conservation crisis.Indeed, increasing human populations and attendant land-use intensification (e.g., cultivation, grazing, and urban development) resulted in the loss and subdivision of native habitats, increasing species extinction rates, and lowered species diversity within managed ecosystems.Previous to that or concomitantly, several strong climate oscillations which occurred during the Pliocene affected vegetation shape and species distribution all over the world.Indeed, the forest-savannah mosaic vegetation type observed in the Dahomey gap was reported by several authors as the result of mid-Holocene marine transgression, followed by drier and wetter climatic conditions successively [2][3][4].This phenomenon could be referred to as natural habitat loss in terms of changes in landscape composition that might have caused a proportional loss of individuals from the landscape and to natural habitat fragmentation in terms of additional effects resulting from the configuration of habitat (reduction of habitat patch size and isolation of patches [5,6]).Adaptation of a species to these variations may produce different morphological and physiological characteristics, resulting in the development of ecotypes.In addition, such fragmentation of natural plant communities can have deleterious effects on the genetic diversity within a species because there will be a decreasing in levels of gene flow, particularly over longer distances.The subsequent effects of genetic drift in small, isolated populations will lead to loss of diversity, leaving plants less able to adapt to changes in their environment and ultimately increasing the risk of extinction [7].Indeed, many species faced extinction given the current rate of habitat loss and degradation [1,8,9].

International Journal of Forestry Research
Milicia excelsa (formerly Chlorophora excelsa) Welw C.C. Berg.commercially known as iroko, is from Moraceae family, Urticaceae order, Tracheophyta phylum, and planteae kingdom.It is a large deciduous tree up to 30-50 m height, with a diameter of 1.70-2 m, with high crown, umbrella-like and growing from a few thick branches.
Iroko is a hardwood tree of great socioeconomical and cultural importance in Sub-Saharan Africa.It is a dioecious species which occurs in a wide range of climatic and edaphic environment and adapts to various ecological conditions.In Benin, its natural range extends from the south to 10 • 30 N and from east to west of the country.The species has been submitted to its habitat destruction and fragmentation and human pressure through intensive logging and landuse practices.A recent study carried out on timber wood exploitation in Benin [10] revealed the almost nonexistence lumber of iroko in sawmills and wood markets.The remnant iroko trees are sparsely distributed across the landscape, either on farms and public places [11], or in sacred groves [12][13][14][15], owing their existence to traditional ethnobotanic practices of conservation.Such a situation might have affected morphological structure of the populations.
This paper aims to assess ecological structure and morphological variation in Milicia excelsa populations in Benin.

Methods
2.1.Population Samples.Milicia excelsa tree inventory was carried out using a transect methods [16] modified from the Buckland et al.'s [17] plotless distance sampling method for estimating abundance of biological populations.Given that the remnant iroko trees are sparsely distributed across the landscape and relict forest patches, ten transects were used in order to cover as much as possible the species range in Benin.Then, 500 m width transects with variable length, the minimum length being 50 km were laid across different biogeographical zones within the species natural distribution area.Along those transects, all iroko trees were inventoried and registered with the GPS (Global Positioning System).A total of 1,028 iroko trees were measured.
The recorded geographical coordinates were plotted onto Benin map and twelve Milicia excelsa populations were inferred from geographical distribution of iroko trees on the map (Figure 1 and Table 1).

Assessing Structural Characteristics of Iroko Populations.
For each of the censured iroko tree, the diameter at the breast height (DBH), the total height (TH), bole height (BH), and crown diameter (CD) were systematically measured.The crown diameter is an average diameter measured from the crown projection circle on the ground when sun is at zenith.DBH data were used to draw the diameter class distribution of the species for each population.Stand basal area (G) were calculated using the formula: G = (πD 2 i )/4, where D i is the DBH.In addition, trees number per hectare was estimated using the simplified Buckland et al.'s [17] estimator of density (D) expressed as D = n/2Lw, where n is the number of recorded tree, L the total length of transect, and w is the strip width.

Morphometric Analysis.
Eight morphometric parameters were analyzed including bearing and architectural parameters (diameter at the breast height, total height, bole height, and crown diameter) and descriptive parameters of the tree's organ (length and width of the leaf blade, the fruit, and the petiole [18]).The bearing and architectural parameters were measured on all of the censured trees while organ descriptive parameters were measured on 10 leaves and 10 flowers randomly collected from adult trees.A total of 1,880 leaves, 1,340 male flowers and 1,360 female flowers were measured (Table 1).Different ratios such as DBH/TH, DBH/BH, and DBH/CD were used to minimize the effect of the unevenness age of trees.For leaves and flowers measurements, data were averaged over individual tree before undertaking the series of multivariate analyses using appropriate procedures.
Principal component analysis (PCA) was performed on the untransformed morphometric data using the correlation matrix.Two populations were clearly separated, and the third group composed of individuals from the remaining populations was subjected to a partial PCA using the same variables.

Assessing Morphological Variation of Iroko Populations and its Relationship with Environmental Factors.
To evaluate the importance of environmental factors in the morphological variation of iroko population, rainfall data were collected from ASECNA (Agence pour la Securité de la Navigation Aérienne) for the last thirty years and soil physical and chemical characteristics were collected from LSSE (Laboratoire des Sciences de Sol et de l'Eau) for comment.PCA was performed on environmental data including average rainfall (mm) and edaphic variables such as percentage of clay, silt, sand, Carbon, Nitrogen, C/N ratio, organic matter (MO), Ca, Mg, K, Na, total exchangeable bases (TEB), and caption exchange capacity (CEC).Cluster analysis (CA) was performed to examine the morphological similarity, at individual level, between the twelve sampled populations.A total of 134 individuals, which were measured all at once for DBH, total height, leaf dimensions, and flower size, were clustered using PAST software version 1.43 [19].The measure of dissimilarity was Euclidian distance and clustering methods was Unweighted Pair-Group Method using Arithmetic Average (UPGMA).For all of the morphometric variables, Multivariate Analysis of Variance (MANOVA) was used to test for significance of variation among populations and among groups of populations.Multiple regressions were carried out to correlate patterns of populations' differentiation to environmental variables.A Canonical Correspondence Analysis (CCA) was performed on the two sets of variables, the first set containing morphometric variables weighed in the principal components (PC), and the second set composed of the most PC weighed environmental variables.This is a direct gradient analysis which incorporates both ordination and multiple  Diameter at breast height (DBH), total height (TH), bole height (BH), crown diameter (CD), leaf length (LL), leaf width (LW ), male flowers length (MFL), male flower width (MFW), female flowers length (FFL), and female flower width (FFW).
regression techniques to reveal the relationship between tables of multivariate data [20,21].CCA ordination was done on standardized environmental variables.MANOVA and multiple regressions were carried out using SPSS software.PCAs were performed using PAST software version 1.43 [19] while CCA was performed using CANOCO for Windows 4.0 [20].

Principal Component Analysis and Morphological Variation among Individuals.
In principal component analysis, 91.23% of morphological variation was explained by the first two principal axes (Table 2).The first axis, with eigenvalue of 9.55, explained 63.84% of total variation, and the second axis explained 27.8% of variation with eigenvalue of 4.09.Morphological trait coefficient (i.e., eigenvectors) indicated that leaf dimensions (length and width) and the ratio DBH/TH were the loading variables in the first axis while male flowers width and female flowers size (length and width) were the loaded variables onto the second principal axis.Along the principal component axis 1, most of individuals from Niaouli population occupied the right side whereas the mixed group drew aside to the centre and the left side (Figure 2(a)).This first axis differentiated populations on the basis of trees height and leaves width.Along the principal axis 2, most of individuals from Save population occupied the middle lower part while the mixed populations were in the middle around the central point.Partial PCA dispatched the mixed populations into two groups (Figure 2(b)).

Patterns of Morphological Variation among Iroko Populations.
Cluster analysis of iroko individuals revealed four clusters (Figure 3).The first cluster contained most of individuals from Niaouli population.The three others cluster were rather mixed with cluster 2 grouping most of individuals from Save Aplahoue and Lokossa populations, cluster 3 mainly composed of individuals from populations Bohicon Bante and Sakete, and cluster 4 encompassing individuals from Djougou, Bassila, Tamarou, Ketou, and Biro populations.Analysis of variance indicated significant morphological variation among groups of populations (F = 12.28, df = 133, P < .001).Population Niaouli which composed cluster 1 has the highest leaves size (mean LL = 18.93, mean LW = 11.84) and showed highest height growth (average DBH/TH = 0.022).Cluster 2 showed medium leaf size and height growth (mean LL ranged from 15.69 to 16.69, mean LW ranged from 9.64 to 10.41, and Average DBH/TH = 0.03).Cluster 3 has the smallest leaf size and lowest height growth (mean LL varied from 12 to 13.16, mean LW from 8.02 to 8.68, and mean DBH/TH varied from 0.064 to 0.085).Medium leaf size and low height growth were observed for  5), Save (6), Ketou (7), Tamarou (8), Biro (9), Bassila (10), Djougou (11), and Bante (12).
populations in cluster 4 (Mean LL ranged from 13.04 to 14.9, average LW ranged from 7.35 to 7.50, and mean DBH/TH varied from 0.032 to 0.039) except Tamarou population which has small leaf size (LL = 10.37,LW = 6.76).Almost all group of iroko populations produced similar female flowers in terms of flower size except populations Djougou, Tamarou, and Bassila from cluster 4 which showed shortest female flowers (Table 4).

Influence of Environmental Factors on Morphological
Variation in Milicia excelsa.Principal component analysis based on environmental variables revealed that the first two principal components accounted for 80.36% of total variation (Table 3).The order of importance of the various parameters in the first principal component was high (CEC, 0.97; Mg and Ca, 0.96; Sand, −0.95; Na, 0.85; Clay, 0.80, and N, 0.59).This principal component, with eigenvalue of 6.05 explained 65.11% of variation.Rainfall (−0.88) and Silt (0.59) were the variables loaded into the second component which explained 15.25% of variation with eigenvalue of 1.41 (Table 3).Multiple linear regressions of morphological traits on environmental factors indicated that height growth in Milicia excelsa was moderately related to silt content in the soil and rainfall (Table 5) and highly but negatively related to Na content in the soil.Soil texture as clay, silt, and sand amount in soil appeared to be important factors for explaining leaf-size variation among iroko populations.Leaf-length variation was dependent on soil chemical properties such as N, Ca, Mg, Na, Mo, and CEC.Rainfall has moderate but significant influence on flowers-size variation.
CCA combining ordination and multiples regression of morphological traits on environmental variables confirmed that rainfall and edaphic factors significantly affect morphological variation in iroko populations (r = 0.92 and r = 0.87, Table 6).Directions and influences of environmental factors (Figure 4) clearly indicate that variation in the iroko Cluster 1 was explained by soil content of Na and rainfall,  and to a lesser extend by the amount of clay, silt, and N in soil.Morphological variation in cluster 2 (populations of Save, Aplahoue, and Lokossa) and cluster 3 (populations of Bohicon, Bante, and Sakete) was related to amount of Ca, Mg, and clay in the soil and to rainfall and soil's caption exchangeable capacity.Sandy texture and amount of organic matter in soil are the main factors which explained morphological variation in iroko cluster 4 (Figure 4).

Structural Characteristics of Milicia excelsa Populations.
Milicia excelsa populations density varied from 1 stem/ha 2 to 9 stems/ha (see Table 4).The highest average DBH values were recorded for Biro, Bohicon, Niaouli, Tamarou, and Save populations.Stand basal area ranged from 0.19 m 2 /ha to 2.05 m 2 /ha.All of the iroko populations showed bellshaped diameter class distribution except Bassila population (see Figure 5).They typically have fewer number of stem  in the smaller and larger diameter classes and more in the intermediate classes as observed for a well-thinned tree population structure.Among these populations four (Tamarou, Save, Ketou, and Bohicon) showed an erratic structure with large gap in their class distribution.Population of Bassila follows an inverted J shape curve with large number of small trees and small number of large trees.

Morphological Variation in Milicia excelsa.
Our results revealed that height growth and leaf size showed morphological variation in the Milicia excelsa populations.These trends were strongly influenced by environmental factors.Several studies have indicated that morphological variation is apparently the result of an adaptive response to the environment; for example, variation in growth traits and phonological traits is associated with a latitudinal and altitudinal range [22,23] or by contrasting climatic conditions [24].Our results suggested that some characters were variable among populations without showing any geographic trend.The observed trend of morphological variation made mention of adaptation to the contrasting microedaphic conditions prevailing for these groups and this was supported by the significant correlation with soil physicochemical characteristics.The greater discrimination power of adaptation micro edaphic conditions compared to the geographical regions of origin of accession in this study clearly indicated the greater importance of environmental factors (soil texture, soil chemical characteristics, and annual rainfall) than geographical location, in discriminating populations.This corroborated the results of Carter et al. [25] who reported that water stress, together with soil nutritional deficiencies, have led to the development of adaptation aptitudes and hence morphological variation in trees populations.Similar results were found on other plant families (e.g., Casas et al. [26] on Stenocereus stellatus in Central Mexico, Bruschi et al. [24] on Italian populations of Quercus petraea).
The variation observed in Milicia excelsa morphological structure could indirectly reveal the consequences of natural habitat fragmentation and human pressure on ecosystems and tree populations.Indeed, most of the individuals from cluster 1 in this study (Niaouli population) which showed the highest height growth and wider leaf size came from a region which harbors a typical humid semideciduous forest as a proof of its soil nutritional richness.In addition, majority of those sampled trees were from the protected site of a research centre, which has probably experienced various soil fertilizations.Similarly, populations from cluster 2 and 4 have most of their sampled trees being either the oldest one whose growth might have not been submitted to recent human pressure on ecosystems (Save population), or dwelling sites that benefited from positive consequences of natural fragmentations such as (i) humid forest vegetation type and consequently soil type in Bassila and Djougou region due to Atacora mountain and the phenomenon of dense forest species eradiation, and (ii) eastern and western extension of Lama depression which maintained edaphic micro conditions in the surrounded areas in Aplahoue and Lokossa zones in the west and Ketou zone in the east.This explanation is congruent with our CCA results which indicated the amount of clay and silt, MO and CEC as the main explanatory variables of variation in these clusters.In this category, the inclusion of Tamarou and Biro populations harbouring trees with moderate height growth and leaf size, could be explained by the extensive land use and intensive chemical fertilization of soils due to cotton culture in that region.To the contrary, the smallest leaf size and lowest height growth revealed by cluster 3 (Bohicon, Sakete, and Bante populations) could result from intensive land use, lack of soil fertilization and absence of fallow.The flower-size variation observed among populations with     [16] on Bertholletia excelsa from line-transect censuses.The bell-shaped stem class distribution exhibited by iroko populations supported the species temperament as light demanding species are known to show such distribution.They are gap demanding for their regeneration and mortality is higher in earlier stage under closed forest canopy [28][29][30].Combining the species temperament and seeds dispersal pattern of iroko which is a barochore (mature fruits are heavy and drop under the mature trees [31]) and widely disseminated by bat (Eidolon helvum [32]), iroko is expected to show large number of seedling under parent tree but it did not, although inventoried trees are mostly in open areas (in fallow lands, on farm, etc.) and seedling are protected by traditional ethnobotanic practices in Benin [11].This adequate light level and abundant Milicia fruits, but lower than predicted regeneration under and around iroko trees suggest that some factor or combination of factors limits successful regeneration.Indeed, iroko female trees produce abundant fruit but the germination rate is low and decreases very quickly.In addition, the bulk part of fruit found beneath iroko tree is rejecta pellets.According to Taylor et al. [32] the rejecta seeds have a very low percent germination because Eidolon feeds on Milicia fruit by sucking to select the more viable seeds, while immature, deformed or aborted seeds remain in the rejecta pellet.Thomas [33] examining Eidolon excreta, found that its diet during migration was 88.8% Milicia fruits.Nichols et al. [34] reported that fruit fall was heavy beneath female Milicia, but seedlings are found in small clumps, 150 m distanced away from supposed parent trees, which represent sites at which birds or especially bats defecated seeds.The inverted J shaped curve show in Bassila population class distribution can be explained by the heavily selective logging activity in that region which creamed valuable species in term of tree size.
Combining the ecological structure of the remnant iroko populations and the abovementioned morphological variation, it could be inferred that the current trend of those populations mirrored back the adaptation of Milicia excelsa to environmental changes due to its habitat destruction and fragmentation.However fragmentation might not induce only occurrence of microvegetational and edaphic conditions to which species has to adapt but also isolation of individual populations may have been causing decreasing of genetic variability and genetic drift progressively moving towards a discrete extinction of the species.Although morphological traits and ecological structure are known to represent only a small proportion of plants genome because there are influenced by environmental factors [35], morphological variation, and spatial structure may have some genetic basis which could be useful for studies of the developmental mechanisms of plant populations [26,36].Therefore, the results of this study raised urgent needs of genetic variation and population structure assessment in Milicia excelsa species.

Figure 4 :
Figure 4: CCA ordination diagram indicating the influence of rainfall and edaphic factors on the morphological distribution of iroko populations

Table 1 :
Sample-size geographical locations and ecological characteristics of iroko populations sites.

Table 2 :
Results from principal component analysis (PCA) of morphometric traits.

Table 3 :
Results from principal component analysis (PCA) of environmental data.

Table 4 :
Morphological variation of the twelve studied Milicia excelsa populations.

Table 5 :
Coefficient of determination (R 2 ) of the linear regressions of morphometric traits on environmental variables.

Table 6 :
Relationship among morphological variation in iroko populations and environmental factors, derived from CCA.
4.2.Milicia excelsa Populations' Structure and Temperamentof the Species.The stand density calculated for Milicia excelsa populations was low but in the range with that reported by Peres and Baider