Effect of Small-Scale Variations in Environmental Factors on the Distribution of Woody Species in Tropical Deciduous Forests of Vindhyan Highlands, India

The aim of this study is to investigate the changes in the composition of mature, naturally established and unmanaged TDF in response to small-scale variations in environmental factors. All woody species with a minimum circumference of 10 cm at 1.37 m height were surveyed in forty-five 20×50 m plots distributed over 5 sites in the TDF of Vindhyan highlands, India. Cluster analysis identified two distinct groups of plots. Group 1 plots had higher soil moisture content (SMC), clay, organic C, total N, total P, and light attenuation than group 2 plots. A total of 48 native species belonging to 25 families were encountered in the sampled area. High eigenvalues for the first two Canonical Correspondence Analysis (CCA) axes indicated the occurrence of species in distinct groups, and significant correlations of the axes with environmental variables indicated the effect of these variables on species grouping. In conclusion, patchiness in the soil resources needs to be considered in restoration efforts. The results of this study are expected to facilitate the decision regarding choice of species in afforestation programmes for restoring the TDF.


Introduction
Tropical forests cover about 30% of the world's land area and 50% of the world's forested area which is around 4 billion ha [1,2]. The dominant vegetation type in tropical forests is the tropical dry forest (TDF) which occupies 42% of the tropical forest area [3]. In India, the share of TDF is 38.2% of the total forest cover [4]. According to Singh et al. [5], the TDF is continuously decreasing in the Vindhyan region and the remnant forest cover exists in the form of noncontiguous patches of varying sizes dominated by single or mixed species. The study of change detection using satellite images (1982)(1983)(1984)(1985)(1986)(1987)(1988)(1989)) of a part of Vindhyan hills identified only 31% of the forested area which remained unchanged since 1982. About 40% of the total forest area was converted from mixed forest with crown cover >50% to mixed forest with crown cover 30-<50%. The rate of conversion from good to poor forest was 6.6% of the forested area each year and savannization in the forest area took place at a rate of 3.3% per year [6]. This calls for massive restoration efforts through reforestation.
Distribution of plant communities, their species, and structural diversity are highly affected by soil water and soil nutrient status [7]. In environments where soil nutrients are abundant, species allocate more to above-ground parts, have more rapid growth rates, and have higher rates of nutrient uptake per gram of root biomass than species from lownutrient environments [8]. Soil organic carbon (C) is a major constituent of soil organic matter which has a major effect on forest productivity and sustainability by influencing soil chemical and biological properties [9]. According to Tateno and Takeda [10], soil nitrogen is strongly linked with forest structure and tree species distribution. In various reports, phosphorus has been estimated to be the principal nutrient limiting tree growth and productivity in tropical forests [7].
De Souza et al. [11] reported significant effect of small variations in soil fertility parameters on the distribution of trees in a Brazilian deciduous forest. Patchiness in the composition of Indian TDF was emphasized by Jha and Singh [12]. However, there was a lack of information on the influence of small-scale variation in the environmental factors on the distribution of the tree species. The aim of this study is to investigate the effect of small-scale variation in selected environmental factors on the composition of mature, naturally established and unmanaged TDF of Vindhyan highlands. To do so, we first applied cluster analysis to separate the experimental plots on the basis of the selected environmental factors. Finally, we investigated the relationship between the mean values of environmental factors and corresponding woody species composition. The results of this study are expected to facilitate the decision regarding choice of species in afforestation programmes for restoring the TDF. They occupy land area of 2555, 394, 1507, 2118, and 199 ha, respectively. The area experiences tropical monsoon climate with three seasons in a year, namely, summer (April to mid-June), rainy (mid-June to September), and winter (November to February). The months of March and October constitute transition periods, respectively, between winter and summer and between rainy and winter seasons. According to the data collected from the meteorological stations of the state forest department for 1980-2010, The mean annual rainfall ranges from 1196 mm (Hathinala) to 865 mm (Kotwa site). About 85% of the annual rainfall occurs during the monsoon season from the southwest monsoon, and the remaining from the few showers in December and in May-June. There is an extended dry period of about 9 months in the annual cycle [12]. The maximum monthly temperature varies from 20 • C in January to 46 • C in June, and the mean minimum monthly temperature reaches 12 • C in January and 31 • C in May.

Data Collection.
At each site, nine plots, each of 1000 m 2 (50 × 20 m) were selected at random for sampling vegetation and soil. Plots were randomly selected to reduce bias caused by within-site differences in soil conditions. Rectangular plots were used because most plant distributions are clumped, and a rectangle can best encompass patches of different species [13]. Further, rectangular quadrats may survey heterogeneity better than square quadrats [14]. The size was decided on the basis of species area curve [15]. In each plot, all woody species with a minimum circumference of 10 cm at 1.37 m height were identified, counted, and measured for basal area. Plant nomenclature was based on "Flora of Madhya Pradesh" 1997 (eds. Mudgal V, Khanna KK, Hajra PK) Botanical Survey of India, Calcutta, India. Soil moisture content (SMC) was measured as percentage by volume by theta probe instrument (type ML 1, Delta-T devices, Cambridge, England) at 10 random locations from all plots. Composite surface soil (0-10 cm) samples were also collected from those locations for chemical analysis. These samples were analysed for texture following Sheldrick and Wang [16]. Analyses of soil samples were also carried out for C [17], total nitrogen (N) [18], and total phosphorus (P) [19] contents. Light attenuation (L) in each plot was measured with a digital lux meter (Lutron LX 101 Lux Meter). It was calculated as follows: Light attenuation Light attenuation is a surrogate for canopy cover; the higher the attenuation, the lower the solar radiation incident on the soil surface and, hence, its effects on SMC.

Data Analyses.
Cluster analysis was performed using the PC-ORD 5 program [20] to identify groups among the sampling plots based on soil physiochemical properties and L. The euclidean distance was used for cluster analysis. The importance value index (IVI) of each species, was calculated as the sum of relative density, relative frequency, and relative basal area [15]. Species richness [21], species evenness [22], Shannon-Wiener diversity index [23], and Simpson's index [24] were also calculated. β-diversity was calculated according to Whittaker [22].
Canonical Correspondence Analysis (CCA) was performed by using the PC-ORD 5 program [20] to correlate environmental variables and vegetation variables [11]. Density data for tree and shrub species were used for the construction of main matrix, and environmental variables were selected as second matrix.

Environmental Variables.
A cluster analysis based on the soil properties and L of the 45 experimental plots showed two distinct groups ( Figure 1). Group 1 consisted of 20 plots (1 to 9 plots of Hathinala; 16 to 18 plots of Gaighat; 19, 20, and 25 to 27 of Harnakachar; 28, 35, and 36 of Ranitali). Group 2 consisted of the remaining 25 plots (10 to 15 of Gaighat; 21 to 24 of Harnakachar; 29 to 34 of Ranitali; 37 to 45 of Kotwa). All the plots of Hathinala were placed in group 1, and all the plots of Kotwa were categorized in group 2 showing a very high difference in environmental variables between the two sites. The largest variation between the two groups was in P, the mean value for group 1 (0.04%) being more than five times the mean value for group 2 (0.006%) ( Figure 2). Significant differences were found in SMC, clay, silt, rockiness, C, N, P, and L between the two groups; Information remaining (%) 1.7E+00 4.6E+03 9.3E+03 1.4E+04 1.9E+04 1 Figure 1: Hierarchical classification of the 45 sampling plots in the forests of Vindhyan highlands based on cluster analysis of some selected environmental variables. Group 1 with 20 plots had high levels of soil moisture, clay, silt, organic C, total N, total P, and canopy light attenuation, and Group 2 with 25 plots had high levels of rockiness and low levels of soil moisture, clay, silt, organic C, total N, total P, and canopy light attenuation.
however, there was no significant difference in sand content ( Figure 2).

Phytosociological Survey.
A total of 4, 680 individuals belonging to 48 species and 25 families were encountered in the 45 experimental plots with a total area of 4.5 ha (Table 1). Leguminosae (10), Rubiaceae (5) (Table 2). Margalef 's species richness and Whittaker's species evenness were also greater in group 1 compared to group 2; whereas β-diversity was maximum in group 2 compared to group 1 ( Table 2). Shannon-Wiener index did not differ significantly between the two groups (Table 2); however, site-level differences were significant and the index ranged from 1.48 at Kotwa to 2.52 at Hathinala. The distribution of tree circumference varied significantly between the two groups for the individuals having stem circumference less than 90 cm ( Figure 3). The number of stems having circumference less than 90 cm was fewer in group 2 compared to group 1 (Figure 3).

Relationships between Tree Communities and Soil Properties.
Results of canonical correspondence analysis using the density data for all species and soil properties and L as environmental variables are presented in Figure 4. Eigenvalues for CCA axes 1, 2, and 3 were 0.443, 0.294, and 0.159, respectively, and the species environment correlations for the axes were 0.889, 0.870, and 0.775, respectively. High eigenvalues for the first two axes indicated the occurrence of species in distinct groups [25], and significant correlations of the axes with environmental variables indicated the effect of these variables on species grouping (Table 3) Adina cordifolia, Albizia odoratissima, Bauhinia racemosa, Bridelia retusa, Cassia fistula, Flacourtia indica, Gardenia latifolia, Grewia serrulata, Miliusa tomentosa, Mitragyna parvifolia, Ougeinia oogenesis, Pterocarpus marsupium, Schleichera oleosa, Schrebera swietenioides, Semecarpus anacardium, Shorea robusta, Terminalia chebula, and Terminalia tomentosa were more frequent in plots with high SMC, clay, N, P, and L (upper left quadrant in Figure 4). Carissa spinarum, Dendrocalamus strictus, Ficus racemosa, Holarrhena antidysenterica, Nyctanthes arbor-tristis, Zizyphus glaberrima, and Zizyphus

Discussion
There was significant difference in SMC in the studied plots of the two groups. Group 1 having high SMC also had greater C, N, P, clay, and L compared to group 2, which had greater silt and rockiness. According to Yang et al. [26], high SMC lowers the rate of organic matter decomposition, which increases C in soils having high moisture content. Yang et al. [27] have also stressed the importance of C and N in structuring the diversity of desert riparian forest of China. Nitrogen mineralization [28] and P availability [29] in soil have also been reported to increase at high SMC. Clay-rich soils have greater availability of water at the surface compared to sand-rich soils where water availability to plants is at greater depths [30]. Clay-rich soils are normally also rich in soil organic matter [31]. According to Lowman [32], L is also predicted to vary according to canopy density and structure. The number of species in semideciduous forests (50-70) and rainforests (100-150) is greater than in tropical deciduous forests [33]. In tropical deciduous forests, species richness is significantly lower due to anthropogenic disturbances, such as burning, grazing, and wood collection [34]. The species diversity is also low in these forests, and only few species show high dominance [35]. According to Gentry [33], the common species in these forests are generally concentrated in few important families, and it is assumed that the abundance of species is mostly governed by ecological factors rather than nonequilibrium chancebased dynamics [11]. Forest structure and species diversity have been widely studied in various regions of the tropics (e.g., [36][37][38][39][40][41]). Gentry [42] recorded 275-283 tree species ha −1 for trees ≥10 cm in diameter at breast height (DBH) at Yanamono and Mishana near Iquitos, Peru. Valencia et al. [43] enumerated 307 species ha −1 for the same DBH in Amazonian Ecuador. These figures formed the world highest record of tree species on a hectare basis for stem ≥10 cm DBH [37]. A direct comparison of tree inventories across tropical forests is difficult due to lack of uniformity in the criteria considered and method employed. In the present study, 48 species were enumerated in the Vindhyan dry tropical forest within the 4.5 ha area for woody species ≥10 cm circumference. The values reported from other large-scale permanent plot inventories (for trees ≥10 cm DBH, i.e., ≥31.4 cm circumference) were as follows: 37 species in Kampong Thom province, Cambodia [44]; 49 species in Vindhyan dry tropical forest, India [45]; 996 species in 52 ha plot of Lambir Hills National Park, Malaysia [46]; 660 species in a 50 ha plot of Pasoh Forest reserve, Malaysia [47]; 229 species in a 50 ha plot in Barro Colorado Island, Panama [48]; 146 species in a 30 ha plot at Varagaliar, Anamalais, Western Ghat, India [37]; 164 species in a 25 ha plot of Sinharaja Biosphere Reserve, Sri Lanka [46].
For tree species having ≥10 cm DBH, Top et al. [44] found a basal area of 23 m 2 ha −1 in Kampong Thom province, Cambodia, and Backéus et al. [49]  The most species-rich family in this study is Leguminosae, and it is also reported by Gentry [33] as a common family in tropical dry forests. Singh et al. [52] and Sagar et al. [45] have reported Acacia catechu, Anogeissus latifolia, Diospyros melanoxylon, Lagerstroemia parviflora, and Hardwickia binata as the most frequent tree species in Vindhyan highlands, and, according to Sagar et al. [45], Shorea robusta is the characteristic species of Hathinala site. The low Shannon-Wiener index for both Groups in this study reflects higher dominance of few species (Terminalia tomentosa, Shorea robusta, Diospyros melanoxylon, Buchanania lanzan, and Lagerstroemia parviflora). Sagar et al. [53] and Raghubanshi and Tripathi [54] have also reported low Shannon-Wiener index in different tree communities of Vindhyan highlands. According to Martijena and Bullock [55], the greater dominance of few species in a forest shows that its formation has taken place under extreme environmental conditions such as shallow soils and low water availability. The lowest Shannon-Wiener index of the driest Kotwa site reflects the dominance of only a few species in the low SMC regime. The significant difference in the distribution of the circumference of woody species in the two groups of plots shows a better regeneration in more moist group 1 plots.
The CCA results in our study showed that the distributional response to small variation in environmental variables differed among the woody species. In other studies, Clark et al. [56] showed the effect of topography and soil properties in the distribution of tree species in a tropical forest of Costa Rica. De Souza et al. [11] analyzed the influence of soil fertility on the distribution of tree species in 1 ha area in a deciduous forest of Brazil and reported that edaphic preferences were not absolute to the extent of determining the presence or absence of a species. Pausas and Austin [57] suggested that, over any large region, the species distribution is likely to be governed by two or more environmental factors and not by a single factor. Russo et al. [58] suggested the demographic responses of the species to resource variations among soil types to play an important role in generating and maintaining the edaphically biased spatial distributions of tree species at Lambir in a Bornean rainforest. In our study involving a total of 4.5 ha, the presence or absence of several species was determined by edaphic preferences.
Our study, thus, indicated that small variations in the environmental variables can determine the distribution of woody species resulting into patchiness in the composition of the TDF. While designing the reforestation programmes for the TDF, the response of different species to the patchiness in edaphic resources has to be kept in mind.