**5. Conclusion**

Multivariate statistical analysis techniques were used to establish the relationships between plant diversity, Topography and soil factors. Plant community, structure and biodiversity have been shown to have a high degree of spatial variability that is controlled by both abiotic and biotic factors (Fu et al, 2004).

CCA is the constrained form of CA, and therefore is preferred for most ecological data sets (since unimodality is common). CCA also is appropriate under a linear model, as long as one is interested in species composition rather than absolute abundances (ter Braak and Šmilauer 1998). Correspondence analysis (CA) and canonical correspondence analysis (CCA) are widely used to obtain unconstrained unconstrained or constrained ordinations of species abundance data tables and the corresponding biplots or triplots which are extremely useful for ecological interpretation CA provided a good approximation for species with unimodal distributions along a single environmental gradient. There is a problem with this metric, however: a difference between abundance values for a common species contributes less to the distance than the same difference for a rare species, so that rare species may have an unduly large influence on the analysis (Greig-Smith 1983; ter Braak and Smilauer 1998; Legendre and Legendre 1998).

The most other general ordination technique, nonmetric multidimensional scaling (NMDS), which is based on the rankings of distances between points (Shepard, 1962), circumvents the linearity assumption of metric ordination methods. This method, used in ecological investigations (Kenkel and Orloci, 1986), Comparative studies of ordination techniques have, moreover, demonstrated the superiority of NMDS, and some authors have re commended its use, notwithstanding the computational burden.

The NMDS approach can in fact be tested each time measures of re semblance or dissimilarity are used to classify OTUs, whatever the causes and origins of arrangements found (Guiller et al, 1998).

248 Multivariate Analysis in Management, Engineering and the Sciences

ultimately be governed by this philosophical distinction.

**5. Conclusion** 

abiotic and biotic factors (Fu et al, 2004).

Legendre and Legendre 1998).

– and for such a system, NMDS would be a better *a priori* choice. As De'ath (1999) points out, there are two classes of ordination methods - 'species composition restoration' (e.g. NMDS) and 'gradient analysis' (e.g. DCA). The choice between the methods should

Non-metric multidimensional scaling (NMS) (PC-ORD v. 4.25, 1999) was used to identify environmental variables correlated with plant species composition. A random starting location and Sorensen's distance measurement were used with the NMS autopilot slow and thorough method. Stepwise multiple linear regression (S-PLUS, 2000) was used to select models correlating vegetation cover and structure with environmental factors. Environmental explanatory factors that were not significant contributors (as determined from using stepwise

A Monte Carlo test of 30 runs with randomized data indicated the minimum stress of the 2 axes NMS ordination were lower than would be expected by chance ( p = 0.0968). The final stress and instability of the 2-D solution were 23.71 and 0.00001, respectively. The first ordination axis (NMS1) captured 41.9% of the variability in the dataset and the second (NMS2)

Multivariate statistical analysis techniques were used to establish the relationships between plant diversity, Topography and soil factors. Plant community, structure and biodiversity have been shown to have a high degree of spatial variability that is controlled by both

CCA is the constrained form of CA, and therefore is preferred for most ecological data sets (since unimodality is common). CCA also is appropriate under a linear model, as long as one is interested in species composition rather than absolute abundances (ter Braak and Šmilauer 1998). Correspondence analysis (CA) and canonical correspondence analysis (CCA) are widely used to obtain unconstrained unconstrained or constrained ordinations of species abundance data tables and the corresponding biplots or triplots which are extremely useful for ecological interpretation CA provided a good approximation for species with unimodal distributions along a single environmental gradient. There is a problem with this metric, however: a difference between abundance values for a common species contributes less to the distance than the same difference for a rare species, so that rare species may have an unduly large influence on the analysis (Greig-Smith 1983; ter Braak and Smilauer 1998;

The most other general ordination technique, nonmetric multidimensional scaling (NMDS), which is based on the rankings of distances between points (Shepard, 1962), circumvents the linearity assumption of metric ordination methods. This method, used in ecological investigations (Kenkel and Orloci, 1986), Comparative studies of ordination techniques have, moreover, demonstrated the superiority of NMDS, and some authors have re

commended its use, notwithstanding the computational burden.

captured 31.8%, leading a cumulative 73.7% of variance in dataset explained (Fig.11).

selection at α = 0.05) were excluded from the final model (Davies et al, 2007).

In the biplots, where only the first two axes were used, all methods based upon PCA gave a fair representation of the relative numerical importance of the rare species. The weights in CCA are given by a diagonal matrix containing the square roots of the row sums of the species data table. This means that a site where many individuals have been observed contributes more to the regression than a site with few individuals. CCA should only be used when the sites have approximately the same number of individuals, or when one explicitly wants to give high weight to the richest sites. This problem of CCA was one of our incentives for looking for alternative methods for canonical ordination of community composition data.

For the analysis of sites representing short gradients, PCA may be suitable. For longer gradients, many species are replaced by others along the gradient and this generates many zeros in the species data table. Community ecologists have repeatedly argued that the Euclidean distance (and thus PCA) is inappropriate for raw species abundance data involving null abundances (e.g. Orlóci 1978; Wolda 1981; Legendre and Legendre 1998). For that reason, CCA is often the method favoured by researchers who are analysing compositional data, despite the problem posed by rare species.

De-trended correspondence analysis (DCA) is perhaps the most widely used method of indirect vegetation ordination. But direct ordination of vegetation and environment is achieved with canonical correspondence analysis (CCA). CCA is a relatively new method in which the axes of a vegetative ordination are restricted to linear groups of environmental variables (Zhang et al, 2006)

DCA and CA analyses should be run with the 'downweight rare species' option selected. We generally do not recommend NMS with the Euclidean distance measure; it performed the worst empirically, and has no advantages over the other methods (Culman et al, 2008)

Among the widely used ordination techniques for the plant community analysis Canonical Correspondence (CA) has shown to be superior to others such as PCA (Gauch, 1982). Most community data sets are heterogeneous and contain one or more gradients with lengths of at least two or three half-changes, which makes CA results ordinarily superior to PCA results. However, with relatively homogenous data sets with short gradients, PCA maybe better (Palmer, 1993). Despite the considerable superiority of the CA over PCA, CA is not superior to DCA, which corrects its two major faults such as "arch effect" and "compression of end of first axis" (Gauch, 1982; Kent & Coker, 1992).

For complex and heterogeneous data sets, DCA is distinctive in its effectiveness androbustness (Gauch, 1982). Comparative tests of different indirect ordination techniques have shown that DCA provides a good result (Cazzier & Penny, 2002). This study found that DCA provides better results than CA results (Malik & Husein, 2006).

For example all ordination techniques, used in North East rangeland of Semnan, clearly indicated that gypsum, EC, slope are the most important factors for the distribution of the vegetation pattern.

In the present study, combination of CCA, DCA and RA results showed that *Ar.aucheri-As.spp-Br.to, Artemisia sieberi-Erotia ceratoides, Ar.sieberi-Zy. eurypterum and Zy. eurypterum -Ar. sieberi*  types correlated with A.W2, gr2, O.M2 and clay1 factors and clay in 0-20 depth indicates *Ar.aucheri-As.spp-Br.to* type. *H.strobilaceum* type has strong relationship with soil salinity and heavy texture. This species showed a trend to high soluble rate, salinity and clay percent. S. rosmarinus types indicate soils with light texture and this type directly related to pH and lime percentage while St.barbata-A.aucheri type shows an inverse relation with these factors.

I fact, analysis with DCA gave results similar to CCA, suggesting that there is a relatively strong correspondence between vegetation and environmental factors; with the difference that the DCA is less isolated the site. CCA better shows differences between types. RA shows relationship between sites and factors, like the CCA analysis. RA axis 1 has an eigenvalue of 0.86. RA axis 2 with an eigenvalue of 0.017 is less important. Total variance (inertia) in the species data is 0.8887.In this method eigenvalue of RA axis1 was higher than CCA and DCA axis1. This study reflects that a spatial approach dealing with the most distinctive species of vegetation communities can yield similar results to those obtained with costly physico-chemical analysis and based on complex matrices of plant communities.

Similarity as this study, also Jafari et al (2003) in their study in Hoz-e-Soltan Reigion of Qom Province, showed that PCA analysis indicates that Halocnemum strobilaceum type has direct relationship with Salinity, Lime, pH and Loam.

May this series of papers serve to enhance the understanding and the proper and creative use of ordination methods in community ecology. Finally, understanding relationships between environmental variables and vegetation distribution in each area helps us to apply these findings in management, reclamation, and development of arid and semi-arid grassland ecosystems (Alisauskas, 1998). The ability to factor out covariables and to test for statistical significance further extends the utility of CCA.

Understanding the relationships between ecological variables and distribution of plant communities can provide guidance to sustainable management, reclamation and development of this and similar regions. In this sense, these results increase our understanding of distribution patterns of desert vegetation and related major environmental factors in the North East of Semnan. The results will also provide a theoretical base for the restoration of degenerated vegetation in this area. Understanding the indicator of environmental factors of a given site leads us to recommend adaptable species for reclamation and improvement of that site and similar sites (Zhang et al, 2005)
