**4.2.3 Slope and aspect**

The variability of slope and aspect has no influence on species composition. This is in contradiction to findings from semi-arid vegetation (Badano *et al.,* 2005; Jurasinski, 2007; Sternberg & Shoshany, 2001). The difference in radiation which in e.g. Mediterranean ecosystems influences a lot of other factors (heat, moisture, evapotranspiration, etc.) is not an issue in tropical systems. Therefore slope and aspect cannot be used as explaining variables for the species composition of the plots. Thus similarity-distance function might predict the slope of a power-law species area curve (Condit *et al.*, 2002). Based on this characteristic the study concludes that it is an appropriate measure of beta-diversity. Already MacArthur (1965) proposed to use species-area curves as an analytical tool to diversity taking the intercept of the curve as a measure of 'alpha-diversity' and the slope parameter as a measure of 'beta-diversity' (see also Caswell & Cohen, 1993; Ricotta *et al.*, 2002). The present study stated that none of the recorded variables provides a good estimator for species richness or species composition. To evaluate the underlying factors many more and preferably continuous environmental variables should be recorded.

#### **4.2.4 Continuous plots for the six sites**

The 6 sites that have been investigated in detail regarding tree species composition can with the help of ordination techniques be grouped. However, this grouping is not very meaningful because most of the sites cannot be grouped regarding their species composition. Only the *sites* 2, 5, and 6 have some more species in common which would allow the specification of a common vegetation type shared by these three sides.

It may be a problem of small sample size that the quadrats of *site* 5 are intermediate in their species composition between quadrats of *site* 3 and *site* 4 (Fig. 16). Otherwise it is astonishing that some of the quadrats in *site* 5 have a lot in common (species wise) with quadrats in *site* 3, which is in the other zone. This leads to the suggestion that the sample is too small for a classification of vegetation types: From the species matrix it is obvious that only few species occur on more than one or on even more quadrats.

The comparison between the wards clustering and the forced grouping into site membership reveals that there are quite some matches. However, this is relatively simplistic because it is not tested whether a quadrat is ordered together with quadrats of its site. This would be a better test, but this is not easily achieved because it is hard to define rules, which can be applied to such an evaluation. If 30 of the quadrats in on site are clustered into one cluster and 30 into other cluster - which is only a very simple case - the problems already start how to evaluate the assignment to several clusters. The simplistic measure already shows that the differentiation in species composition is largely driven by the position in sites. However, on smaller scale vegetation types (or better groupings based on tree species data) might be identifiable.
