Version classiqueVersion mobile

Forest landscapes of the southern western Ghats, India

 | 
B.R. Ramesh
, 
Rajan Gurukkal

Part II. Analysing The Landscape: The Anamalai Region (Nelliyampathi Hills)

4. Stand and Floristic Structural Variability of Natural Vegetation

P.V. Karunakaran, Pierre Couteron, B.R. Ramesh et S. Aravajy

Texte intégral

1The conservation and management of forest vegetation and related biodiversity, insist upon proper knowledge of spatial variation of vegetation features within the area under consideration. In fact, it is the awareness that spatial heterogeneity of vegetation is constantly interacting with landscape dynamics that has given rise to landscape ecology. Describing and characterizing this variation of vegetation is thus a prerequisite, which has to be carried out at several complementary scales, with use of scale-specific vegetation variables or attributes.

2In the chapter 3, a broad scale delineation and mapping of land cover / land use types has been achieved with the help of remote sensing images and to an adequate typology of vegetation, each type being indirectly defined from simple qualitative attributes, mostly pertaining to the level of standing biomass and to the phenology of the dominant trees. In fact, it is the tight relationship between such attributes and the radiometric properties of the vegetation cover, which permitted an efficient mapping of the land cover /land use types.

3However, additional and precise information regarding stand structure and species composition of the various types of forests is needed, in order to meet the requirements of conservation and sustainable management of plant diversity. This kind of information can only be acquired by means of a carefully planned field sampling, based on field plots of suitable size, and aimed at systematically describing vegetation via a set of attributes (i.e., stand and floristic variables) which are to be consistently assessed at plot level. Field-assessed vegetation attributes are subsequently used to classify plots into vegetation groups that can be related to broad scale vegetation types, and also that can be analyzed in the light of extrinsic ecological variables, such as altitude or bioclimate or with respect to regimes of anthropogenic disturbances. In fact, it is important to assess the relative influence of natural vs human-induced factors on the structure and composition of forest vegetation, in order to get an insight on the dynamical trajectories of the observable vegetation types in the landscape units under focus.

Methods

Sampling design

4The vegetation map (1: 250,000 scale) prepared by the French Institute (Ramesh et al, 2002) was considered as the baseline data for stratifying the study area for sampling. It was decided to sample on a systematic basis of 0.1 % of the total extent of natural vegetation (135202 ha). Thus, an estimated minimum number of 135 sampling plots of 0.1 ha were required to study spatial and floristic structures of different vegetation types in the study area.

5The entire study area was divided into 2 km x 2 km grids and sampling grids were randomly selected proportionately to each vegetation type. For each sampled field location, a square plot called ‘X’ (Figure 4.1) measuring 31.5 m x 31.5 m (0.1 ha) was laid for collecting data on trees and lianas with ≥ 30 cm at 130 cm height (girth at breast height- GBH). The X plot was further divided into 9 subplots of 10.5 m x 10.5 m and three diagonal subplots (named as Y1, Y2 and Y3) were selected to measure the individuals between 10 and 30 cm GBH. In each of these diagonal plots, at the centre, micro plots of 1 m x 1m (Z1 to Z3) were laid to record the regeneration and abundance of herbaceous elements of the vegetation. Apart from GBH, height of individual trees was also recorded in all the plots (small ones were measured by graduated poles and big ones with ocular estimation). The geographical coordinates, altitude, aspect and slope were also recorded for each plot.

Figure 4.1 Sampling design

6In addition to sampling in forest areas, grassland (Nilgiri tahr habitat) and reed patches were also sampled using smaller plots (1 m2 and 100 m2 respectively) laid along different transects. In the grassland, the grass species and its percentage cover was recorded. In the reed sampling plots, the number of clumps and culms in each clump were recorded. Figure 4.2 shows the distribution of plots in the study area.

Figure 4.2 Distribution of sample plots in th landscape units

7Out of the 147 plots laid, 136 are exclusively for assessing stand and floristic structure of the forest vegetation and the remaining are for describing characteristics of reed area and grasslands. Among the 136 plots, 94 were in evergreen forests and 42 in deciduous (including semi-evergreen).

Statistical analysis of stand and floristic structure:

8The description and analysis of the vegetation was based on stand and floristic variables, which are classically used in forest ecology and vegetation science. The stand variables aim at depicting the structure of the forest, in terms of the density of trees and basal area. These parameters were assessed by taking in to account tree height and girth frequencies at different threshold levels. At plot level, the basal area is computed by adding individual basal area values (i.e. times the squared girth) found for each tree present in the plot, and is further rescaled in m2/ha.

9Floristic structure was investigated in terms of species diversity, which can be expressed by means of several popular diversity indices, that is species richness and Shannon-Weiner and Simpson-Gini’s indices (Magurran, 1988; Legendre and Legendre, 1998). The enumeration of species present in each plot (species richness) is the simplest of the diversity indices. The two other indices integrate the relative abundance of species in the plots to provide complementary information on vegetation composition.

10A Windows-based computer application, Greenbase (developed by the French Institute of Pondicherry), was used to calculate various variables and indices from the database (under MS-Access®) incorporating field data.

11Multivariate analysis was used to draw synthetic insights from the variations of many inter-dependent variables. We analyzed the data matrix having plots as rows and stand/floristic variables as columns using Principal Component Analysis (PCA; Legendre and Legendre, 1998), which is a wellknown standard for defining a limited number of synthetic, composite variables (called principal components) to summarize the variation of vegetation characteristics observed in the field plots. Principal components can then be used to ordinate plots along a limited number of meaningful gradients and to group plots based on stand and floristic attributes. These groups were further analyzed with respect to variations of vegetation attributes, using analysis of variance (ANOVA; Legendre and Legendre, 1998), while correlations between principal components and ecological variables, such as topography and bioclimate, have been computed to enhance the ecological interpretation. PCA and subsequent statistical analyses have been carried out using XLSTAT®.

Results

Stand structure

12The average basal area of the pooled data was 53.8 m2/ha and the maximum-recorded value in an individual plot of 0.1 ha was 109.5 m2/ha (Chimmony Wildlife Sanctuary) and the minimum was 18.6 m2/ha (Vazhachal division). The mean basal area of girth class <30-60cm> was 3.2 m2/ha, <60-120 cm> was 9.3 m2/ha and >120 cm is 40.6 m2/ha which indicated the importance of the contribution of tall trees to the standing volume of the forests.

13The mean density of the pooled data was 490 individuals per ha whereas the maximum density was 990 individuals and the minimum was 120 individuals per ha. The maximum density was found in the medium elevation evergreen forests and minimum in the degraded forests of the lower altitude. This corresponded to evergreen forests of Vazhachal and Parambikulam and degraded forests of Chalakudy and Malayattoor divisions.

Figure 4.3 Height class density of trees in the 0.1 ha plots

14The height of the trees was grouped into five classes (<8, 8-16, 16-24, 24-34 and >34 m). Higher density was noticed in the height class 8-16 m and lowest in the group >34 (Figure 4.3). The highest densities for tall trees were found in the Malayattoor division and Chimmony Wildlife Sanctuary (100 and 60 trees per ha respectively) and densities for medium sized (8-16m) peaked in Sholayar region of Vazhachal division, which may indicate the intensity of logging in the past and the recouping of the forests.

Plant diversity and endemism

15Totally 436 taxa were encountered in the sample plots (X, Y, & Z) in which X-plots were having 294 species, of this 22 species were shrubs. The Z plots, which represent herbaceous elements and regeneration, were having more number of species (374).

16In X-plot, 60 % of the total individuals were contributed by 35 species. Among these, Xylia xylocarpa (n=337) and Palaquium ellipticum (n=302) counted the most, followed by Reinwardtiodendron anamallayanum (n=260) and Terminalia paniculata (n=190). Most of the individuals of Xylia were recorded in woodlands of primary and secondary moist deciduous formations. The preponderance of Palaquium may be due to the large extent of medium elevation evergreen forests in the landscape units. When all the plots and subplots were pooled (X, Y, and Z), Reinwardtiodendron anamallayanum (n=1059) and Palaquium ellipticum (n=762) contributed a maximum to the stand structure of evergreen vegetation.

17The number of species and their individual contribution to the total frequency, for the entire X plots was highly skewed (Figure 4.4). In the pooled data 173 species contributed <10 individuals and very few species (3 nos.) were represented by >200 individuals. There were only 15 species that contributed more than 100 individuals in the stand.

Figure 4.4 Number of species and their contribution in the plots

18Further, the phenology of the species revealed that evergreen species were far higher in number (83 %) compared to deciduous (17 %). Among the evergreen species, primary and secondary ones contributed 54 % and 29 % respectively (Figure 4.5).

Figure 4.5 Repartition of species represented in the plots based on
phenological and ecological characteristics

19The species richness in the pooled data varied from 4 to 35 species per X-plot and the mean number was 19 species. The lowest number (4) of species was encountered in the deciduous forests (Thattekkad, Chalakudy, Vazhachal and Malayattoor) and highest (35) in the medium elevation forests of the Vazhachal and Malayattoor divisions.

20The Simpson’s index (1-D) varied from 0.48 to 0.98 (the value ranges between 0 and 1). Plots with high dominance index (i.e. low Simpson’s diversity) were found in Malayattoor, Nemmara and Vazhachal divisions and those with low dominance were found in parts of Parambikulam and Chalakudy divisions. The Shannon index (H’) that varied from 0.86 to 3.29 showed a similar trend, with values higher in Vazhachal and lower in Chalakudy divisions.

21The data collected on vegetation from the ‘X’ plots revealed that the percentage of endemism varied from 0 to 79 and the mean percentage was 33. Plots with more endemic species fall in the divisions such as Malayattoor, Vazhachal and Chimmony. There were 27 plots without any endemic species.

Definition of vegetation types in relation to stand and floristic structure

22Principal Component analysis (PCA) was used to summarize the information provided by both stand and floristic variables. After preliminary analyses, we retained only eight variables to keep the analysis truly informative by avoiding excessive redundancy between closely related variables, which would be likely to dominate the multivariate synthesis.

23Prior to the multivariate approach, we first analyzed pair wise correlations between the individual variables, and it appeared that almost all the pairs of variables were significantly correlated except endemism with secondary evergreen species density (Table 4.1). Several correlations were nevertheless low (less than 0.4). Among the positive correlations, the most significant were between the diversity indices (species richness, Simpson (1- D) and the Shannon (H’)), and between the percentage of endemic species and the proportion of evergreen species (% of density). The only pair of variables that showed negative correlation is between secondary species density and basal area.

Table 4.1 Correlation between different variables used in PCA (S = species richness, BA = Basal area, 1-D = Simpson’s index, H’ Shannon index,
End % = percent of endemic species, Sec and Eg are proportions of secondary evergreen and evergreen species in terms of tree density, respectively)
Bold indicates significant values

24These relationships between vegetation variables were then summarized by PCA. The principal components, or axes F1 and F2 accounted for 60 and 16 percent of variation respectively (Figure 4.6). The first axis ordinated the plots on the basis of a phenological gradient, leading from evergreen to semi evergreen and deciduous forests. Endemism, species richness and diversity indices also contributed considerably on this first axis.

Figure 4.6 Correlations between the variables and
the two main PCA axes

25The density of secondary species and basal area, both contributed significantly to the second axis but with opposite loadings. Other variables such as diversity indices (Shannon and Simpson) and species density had little significance on the second axis. Basal area is known to diminish in disturbed forest concomitantly with an increase in the abundance of secondary species. This allowed us to interpret the second axis as expressing a degradation gradient, according to which plots belonging to the same phenological class aligned themselves.

26The correlation between the PCA factor scores (F1 and F2) and the physical and bioclimatic variables, such as altitude and rainfall, showed low significance or practically none (-0.136; p=0.050). In other words it meant that in these areas the influence of physical variables is strongly blurred by anthropogenic factors.

27We applied the K-means clustering algorithm (Legendre and Legendre, 1998) to scores on axes F1 and F2, that the phenological and degradation gradients, to objectively delineate five groups of plots (Figure 4.7).

Figure 4.7 Scatter plot showing the five vegetation groups obtained by
applying K-means clustering on PCA scores

28The plots relating to evergreen forests were divided into three categories, one (EG3) corresponding to heavily disturbed locations (high presence of secondary species and low basal area), and vice versa for another one (EG1), while an intermediate third group (EG2) related to situations marked by medium levels of disturbance and degradation. The remaining two groups corresponded to semi evergreen (SE) and moist deciduous forests (DE), respectively.

29The vegetation groups derived from PCA more or less matched the classification of major natural vegetation types in the land cover and land use map. The three evergreen forest groups - EG1, EG2 and EG3, were mostly found in dense, disturbed and highly disturbed to secondary evergreen forests respectively. As there was no obvious difference in structure between primary and secondary moist deciduous formation, all the deciduous formations were grouped under DE.

Characterization and analysis of vegetation groups obtained from PCA

30We first considered the three evergreen groups (excluding semi-evergreen) to carry out correlation analyses on the selected variables. There was a rather strong negative correlation between secondary species density and both the percentage of endemic species and the basal area (Figure 4.8). This indicates that when disturbance increases, not only does endemic species get reduced but standing biomass (represented by basal area) also becomes considerably low. On the other hand, there was no clear correlation between total species richness and secondary species density (r = -0.19; not shown) and species richness and basal area were not correlated (Figure 4.9). Hence, in our study area, total species richness remained only weakly related to the perturbation gradient, even when the three evergreen groups were independently considered from the deciduous and semi-evergreen groups. On the other hand, there was a significant, yet fuzzy relationship between species richness and the percentage endemics in the evergreen plots (Figure 4.9).

Figure 4.8 Correlation analysis between secondary species density, the proportion of endemic species and basal areas for three evergreen types (EG1, EG2, EG3).

31We subsequently investigated inter-groups, variations of the principal vegetation variables, which have been plotted on Figure 4.10. These variations were highly significant for all variables according to one-way ANOVA tests (p < 0.001).

Figure 4.9 Correlation analysis between the species richness, the proportion of endemic species and basal areas for three evergreen types (EG1, EG2, EG3).

32We subsequently investigated inter-groups, variations of the principal vegetation variables, which have been plotted on Figure 4.10. These variations were highly significant for all variables according to one-way ANOVA tests (p < 0.001).

33Basal area was significantly higher in EG1 (evergreen undisturbed) with no significant variation among other vegetation types. Species richness reached similar values for the three evergreen groups and significantly lower values in semi-evergreen and deciduous groups. The other diversity indices (H’ and 1-D, not shown in the Figure 4.10) were strongly correlated with species richness and displayed the same pattern of variation. The contribution of endemics to species richness was the highest (ca. 50 %) in both EG1 and EG2 but significantly lower in EG3 and very low in SE and DE. As expected, the proportion of trees belonging to secondary species peaked for EG3, while the proportion of evergreen trees did not discriminate between EG1, EG2 and EG3. On the other hand, the absolute number of deciduous species clearly separated the disturbed evergreen type (EG3) from EG1 and EG2, confirming that an influx of both secondary and deciduous species was observed in disturbed evergreen forests. It is furthermore clear from these last results that the weak difference in diversity indices between EG3 vs EG1 and EG2 was the consequence of compensation by secondary and deciduous species for the loss of primary evergreen species due to disturbance.

Figure 4.10 Variations of stand and floristic variables according to vegetation groups. Error bars indicate 95 % confidence interval around the mean.

34In terms of dominant species (with respect to the density of trees with GBH > 30), EG1 was characterized by Palaquium ellipticum, Reinwardtiodendron anamallayanum, Myristica dactyloides and Cullenia exarillata and, EG2 by Reinwardtiodendron anamallayanum, Dimocarpus longan, Macaranga peltata and Polyalthia fragrans. Hence, in EG2 plots, P. ellipticum and of C. exarillata, which were known to characterize the Medium Elevation Evergreen Forests (MEEF, Pascal 1988) were not among the most abundant species. This notable difference with EG1 can be directly explained by the lower elevation of EG2 plots, which was mainly observed under 700 m elevation (Figure 4.11), which defines the lower boundary of the MEEF (Pascal 1988). An indirect explanation is that, within their ecological range of occurrence, P. ellipticum and C. exarillata, were generally targeted by former logging activities, which had historically a more pronounced impact at low elevation.

Figure 4.11 Distribution of vegetation groups with respect to elevation and mean annual rainfall. Each group is represented by the average position of its constituting plots (cross) and by an ellipsoid whose amplitude is twice the standard error of the mean.

35Dominant species in EG3 were Ixora brachiata, Aporusa lindleyana, Baccaurea courtallensis, Vitex altissima, Pterygota alata, Vernonia arborea, Turpinia malabarica and Acrocarpus fraxinifolius, most of them being secondary species.

36In the semi-evergreen type (SE) the list of dominant species was Terminalia paniculata, Lagerstroemia microcarpa, Aporusa lindleyana, Litsea laevigata, Croton malabaricus, Xylia xylocarpa and Olea dioica. And deciduous species ranked among the most abundant ones and accounted indeed for up to 68 % of the total number of species. (But less than 40 % of the total number of trees; Figure 4.10)

37In the deciduous types, the dominant species were Xylia xylocarpa, Lagerstroemia microcarpa, Dillenia pentagyna and Terminalia paniculata, which are all deciduous species. Associated evergreen species, which accounted for less than 20 % of the trees, were all secondary in nature.

Ecological characterization and spatial distribution of the vegetation groups obtained from the PCA.

38Groups were plotted according to the average altitude and mean annual rainfall values of their constituting plots (Figure 4.11). These two ecological variables displayed rather weak correlations with the PCA axes underlying the definition of the groups. The three evergreen groups clearly differentiated from the semi-evergreen and deciduous groups on the basis of mean annual rainfall. In fact, the three groups corresponded to average rainfall values of ca. 3800 mm/year and the corresponding “confidence ellipsoids” were limited above 3600 mm/year and did not encompass average values found for the two remaining groups. On the other hand, the three evergreen group appeared to follow an altitudinal gradient, the most disturbed among them (EG3) being found in the lowest elevation while the less disturbed was mostly observed within the highest altitudinal range sampled. Comparatively, the two remaining, non-evergreen groups were mostly sampled at elevations below 600 m. Those two types did not significantly differentiate from each other on the basis of mean annual rainfall and elevation, although most of the plots sampled in the driest situations (Parambikulam area) belonged to the deciduous type (DE).

39The frequency of the five vegetation groups was compared to each of the two landscape units (LUs) under study (Figure 4.12). From such an analysis it is apparent that the groups relating to the most severe levels of degradation (i.e., EG3, SE and partially DE) were over-represented in LU13. Conversely more than 40 % of the plots belonging to those groups have been sampled in LU13 whereas the area of this unit is no more than 30 % of the entire study area (both LUs). This is also explicit from the map presenting the spatial distribution of the types (Figure 4.13).

Figure 4.12 Distribution of plots in different landscape units

40Good patches of evergreen forests (EG1) characterized by high species richness, endemism, presence of large primary trees (indicated by high basal area), were found on the plateaus and/or moderately dissected hills, principally in the upper region of Vazhachal adjoining with Parambikulam. They mostly corresponded to medium elevation evergreen forests. Even though such canopy forests were also present in the Malayattoor division (Anakkulam and Variayam area), the undergrowths were generally scanty due to cardamom cultivation in the recent past. All over the study area, large-scale infrastructure development associated with the construction of dams (e.g., Sholayar) has been a cause of localized disturbance with an apparent effects on vegetation (Figure 4.13). Elements of evergreen species persisted in spite of highly degraded patches of forests close to the reservoirs, dams and power generation stations.

41Most of the plots relating to substantial levels of forest degradation (EG3, SE and partially DE) were found in the highly dissected lower reaches (mainly in Chimmony, Chalakudy, Thattekkad and parts of Malayattoor divisions) of the study area. These areas have been more exposed to maninduced degradations of various kinds due to their geomorphologic features more amenable to human penetration and infrastructure creation (large valleys and associated pedological features, moderate slopes) and also due to their position at the contact of the densely populated areas of the plain. On the other hand, the corresponding belt of degraded forests and plantations has, to some extent, played the role of a buffer zone, limiting disturbance on the less degraded forests observed in the center of LU16. This belt of degraded forest areas also relates to lower average rainfall, which may be, through fire hazards and occasional drought, an additional factor of sustenance of degraded formation, once the natural evergreen forest cover has been artificially opened or destroyed. It may nevertheless be underlined that annual rainfall received in LU13 and related locations are neatly above the minimal level (ca. 2500 mm; Pascal 1988) considered sufficient to allow the existence and perpetuation of the evergreen forest in the absence of serious anthropogenic disturbance. Parambikulam, part of which exposes towards the leeward side of the Western Ghats and receives less rainfall (1500-2000 mm), is probably the only area where dominance of deciduous species can be safely considered as a direct influence of the bioclimate.

Figure 4.13 The distribution of different vegetation groups in the LUs

42Similarly the area under constant interaction of people for the collection of reeds and other resources are also plotted with degraded formations of evergreen forests even in the high rainfall area (e.g., Malayattoor and parts of Vazhachal divisions). In Malayattoor division fairly undisturbed patches of forests are restricted towards Anakkulam and adjoining areas.

Conclusion

43The two landscape units under investigation mostly belong to the windward side of the Western Ghats and are endowed with annual rainfall permitting the existence of evergreen forest vegetation. However, a gradient expressing the relative importance of evergreen vs deciduous species in the field plots was the most prominent result yielded by the multivariate analysis of the vegetation variables (Figure 4.6). This is due to the fact that, in the landscape units as in most of the Western Ghats, anthropogenic disturbances are generally favorable to the establishment of deciduous species and detrimental to the persistence of many evergreen species. The dominance of deciduous species can be considered as natural only for those plots located in areas receiving less than 2000-2500 mm (mainly Parambikulam, on the leeward side) rainfall. Another adverse consequence of degradation is the relative decline of species endemic to the Western Ghats (Ahmedullah and Nayar, 1986), which jeopardizes one of the main goals of biodiversity management, that is the in situ conservation of endemic species.

44However, floristic changes actually depend on complex synergies between disturbance regimes and site characteristics, and not only deciduous species but also secondary evergreen species were found abundant in some disturbed areas (Figure 4.10), especially in places where annual rainfall is high, say more than 3500 mm/year (Figure 4.11). Interestingly, the multivariate analysis on vegetation variables yielded two distinct gradients (Figure 4.6), expressing the relative abundance of evergreen vs deciduous and of primary vs secondary evergreen species. While the first gradient reflected both disturbance and bioclimatic variations, the second gradient strictly expressed the diminution of human pressure with altitude (Figure 4.6; Figure 4.11).

45In fact, most of the less disturbed patches of evergreen forest were found at elevation above 600 m, in LU16, where relief features and remoteness from the largest settlements have limited the natural resource harvesting and ensuing degradation. On the contrary, the different categories of degraded plots were systematically found at low elevation, especially in LU13, which has been the most exposed to anthropogenic degradations of various kinds due to topographic features favorable to human penetration and proximity to densely populated hinterland.

46This conclusion corroborates the fundamental assumption underlying the present book about the pertinence of geomorphologic features to define landscape units as a basis for the study of man-nature interactions. Such interactions are of course, complex, diverse and varying in space and time. But they are strongly constrained by the most prominent natural factors, and especially by climate and geomorphology. And this makes such factors a relevant entry for a holistic appraisal of man-vegetation interactions. Furthermore, as forest resource harvesting is made of diversified and intricate activities, which are therefore difficult to follow through time, their cumulated impact over years is due to leave a decipherable footprint on vegetation structure and species composition. Consequently, pertinent analyses of vegetation variables can valuably contribute to characterize the dynamics of man-nature relationships in the Western Ghats.

Auteurs

French Institute of Pondicherry
11, St. Louis Street
Pondicherry 605 001
INDIA

© Institut Français de Pondichéry, 2007

Conditions d’utilisation : http://www.openedition.org/6540

Acheter

Rechercher dans OpenEdition Search

Vous allez être redirigé vers OpenEdition Search