**1.1.2 Measuring similarity**

148 The Dynamical Processes of Biodiversity – Case Studies of Evolution and Spatial Distribution

was utilized to evaluate the validity of the dispersal theory (Hubbell 2001) against niche concepts in the Neotropic: Chust et al., (2006), Condit et al., (2002), Duivenvoorden et al., (2002) and Ruokolainen et al., (2002), investigated distance decay in terra firme forests in Panama, Ecuador, and Peru, whereas Duivenvoorden et al., (1995), Ruokolainen et al., (1997), and Tuomisto et al., (2003) concentrated on rain forests in the Amazon basin. However, there are no studies in this regard in the Paleotropic. This hinders the overall generalization of the findings. The data set of the Eastern Ghats provides a unique opportunity as we can calculate distance decay rates in paleotropic ecosystems. Unfortunately there were only few environmental variables recorded. Furthermore these are

Our aim is to develop a spatially explicit, widely applicable method to assess phytodiversity encompassing species richness, spatial and temporal heterogeneity, and functional diversity and to relate it to environmental conditions (including site conditions, disturbance regime and biological richness value). There is an urgent need for standardized and comparable data in order to detect changes of biodiversity. Such methods are required to be representative as well as pragmatic due to the simple fact that there is insufficient time to obtain complete data sets relating to temporal trends. If biodiversity is lost rapidly at the landscape level, frequent re-investigations have to be done in order to detect and analyze such changes. Thus, our objective was to provide a method that allows for the tracking of changes in biodiversity at the landscape scale. Here, we investigate the distance decay of compositional similarity whilst accounting for the change of the relationship between

compositional similarity and its drivers with geographical distance between plots.

distance decay will be relatively low compared to large-scale studies.

favoured compared to competitors (Rawat, 1997).

**1.1.1 Measuring and analyzing similarity** 

Steinitz et al., (2006) highlight that the studies are limited to which extent patterns of distance decay depend on the position along environmental gradients. Accordingly Jones et al., (2006) emphasize that explanatory power of variables might increase with the length of the gradient covered. Based on this, we hypothesize for the present study (Eastern Ghats

1. Similarity in plant species composition decreases continuously with distance. Due to the small scale of the study (largest geographical distance covered is 8 km) the rate of

2. The correlation between compositional similarity of vegetation and the dissimilarity of predictor variables (vegetation type, disturbance - biological richness value, abiotic environmental conditions) is changing with geographical distance between multiple plots. We expect the correlation to increase with spatial scale/distance between sampling units because the sampled environmental gradient is likely to increase as

3. Disturbance is the main driver of vegetation patterns in the regarded transitional ecosystem. We especially consider disturbances because deciduous ecosystems of the Eastern Ghats have a long-lasting disturbance history and stress tolerators are clearly

Compositional similarity or differentiation diversity between sampling plots is an important basis for most numerical analyses in vegetation ecology. It is at the heart of ordination methods and has general importance regarding the testing of ecological theory (Legendre et al., 2005). Moreover, it represents the basis for most of the analyses in the present study and

categorical and assign relatively coarse classes to the plots.

part of Andhra Pradesh, India) as:

well.

**1.1 Previous studies** 

Compositional similarity measures differentiation diversity (Jurasinski and Retzer, 2009) and can be calculated with resemblance measures. These are available in two primary groups (Legendre and Legendre 1998): (1) Quantitative or distance indices are used to calculate the proximity of two samples in data space from quantitative measurements or abundance data. (2) Similarity/Dissimilarity measures can handle binary data as it is typically found in presence/absence data sets. There are a large number of different indices and coefficients available and comparative reviews can be found in Cheetham and Hazel (1969), Hubalek (1982), Janson and Vegelius (1981), Koleff et al., (2003), Shi (1993) and Wolda (1981). All binary similarity/dissimilarity measures are based on the same set of variables. Whether a coefficient measures similarity or dissimilarity depends on the implementation of the formula. However, most of them can easily be transformed from a similarity to a dissimilarity measure by calculating 1-S (with S being similarity, Legendre and Legendre, 1998; Shi, 1993). From some similarity coefficients a dissimilarity measure following the Euclidean metric can be obtained by calculating sqrt 1 − S instead (for details see Legendre and Legendre, 1998).

Indices incorporating the species not present in both of the compared samples (d, see Fig. 1) are controversially discussed (e.g. Simple Matching by Sokal & Michener (1958) or the coefficient of Russel and Rao (1940)). Legendre and Legendre (1998) call these symmetrical (see Janson and Vegelius, 1981 for another definition of symmetry) and discuss the "doublezero problem" at the beginning of their chapter on similarity indices, because it "is so fundamental with ecological data" (Legendre and Legendre, 1998, p. 253): Species are supposed to have unimodal distributions along environmental gradients. If a species is absent from two compared sampling units (which is expressed by zeros in the species matrix), it is not discernable on which end of the gradient the both sites are with respect to a certain environmental parameter (Field, 1969). Both might be above the optimal niche value for that species, or both below, or on opposite tails of the gradient. Thus, the incorporation of unshared species (d) might lead to wrong conclusions when the relation between environment and species composition is under study (Legendre and Legendre, 1998).

Additionally, Shi (1993) states that the status of d in similarity coefficients for paleoecological studies is not clear and cannot be assessed directly because of its great dependence on the less common taxa absent from both sites: "In palaeontology, absence of a taxon, particularly a rare one, may have been derived from differential preservation and/or sampling errors rather than some ecological factors". Finally, Field (1969) shall be cited as he found a very well fitting metaphor: "No marine ecologist would say that the intertidal and abyssal faunas were similar because both lacked the species found on the continental shelf".

Spatial Patterns of Phytodiversity - Assessing Vegetation Using (Dis) Similarity Measures 151

two zones viz., northern Eastern Ghats (Zone-1) and southern Eastern Ghats (Zone-2). Following the gradient in the geology, soil, bioclimatic (temperature and precipitation), altitude and degree of disturbance, the zones were characterized and analysed. For continuous enumeration a total of six transects were laid and data collected was used for analysis. These transects were laid both in zone-1 and zone-2 and distributed evenly in

**4**

**6**

Fig. 2. Location map of the six transects and zones of the northern and southern Eastern

The areas studied are three 0.5-ha plots, which are located in the Sileru-Maredumilli ranges of the Northern Eastern Ghats of Andhra Pradesh (zone-1) and Nallamalais-Seshachalam-Nigidi hill ranges of Southern Eastern Ghats of Andhra Pradesh (zone-2) (Fig. 2). These forests are classified as South Indian Moist Deciduous and Orissa Semi evergreen forests (Champion & Seth, 1968). Three 0.5ha plots area was established at three different sites: Site 1 is located about 2 km from Sukkumamidi, a tribal hamlet in Khammam district, which receives mean annual rainfall about 1200-1400mm and elevation ranging from 400-600m. Site 2 is located about 6 km from Maredumilli tribal village in East Godavari district, receives mean annual rainfall about 1400-1600mm with an elevation of 600-800m. Site 3 is located about 2 km from Lankapakala tribal hamlet in Visakhapatnam district receives mean annual rainfall 1600-1800mm and an elevation of 800-1100m is shown in Table 1. All the three study sites were undisturbed. There are no records on the intensity and the extent of

The next set of sample plots were laid in Southern Eastern of Andhra Pradesh (zone-2). These forests are classified as Tropical Dry Deciduous by Champion & Seth (1968). Three 0.5ha plots area were established at three different sites: Site 4 is located about 3km from

Ghats for random plots in the state of Andhra Pradesh, India

**5**

**1 2**

**Andhra Pradesh**

**3**

Eastern Ghats Boundary Andhra Pradesh Transect site

deciduous ecosystems.

**India**

Site 1: Sukkumamidi Site 2: Maredumulli Site 3: Lankapakala Site 4: Peddacheruvu Site 5: Talakona Site 6: Kuntlapalli

disturbance.

Fig. 1. Illustration of the matching components providing the basis for binary similarity measures a) the number of species shared by two compared units, b) the number of species unique to one of the compared plots, c) the number of species unique to the other one of the compared plots, d) the number of species not found in the two compared plots but in the whole dataset (unshared species).

(Dis) similarity indices, which take unshared species into account, mingle different ideas of differentiation diversity (additive partitioning, multiplicative partitioning, turnover, see Jurasinski, 2007). Furthermore, they tend to be less specific as the values show less variance because d is far bigger than a, b, or c for most of the datasets recorded in the field. "Including double-zeros in the comparison between sites would result in high values of similarity for the many pairs of sites holding only a few species; this would not reflect the situation adequately" (Legendre and Legendre, 1998). Furthermore is the total inventory diversity (gamma) as a background for the calculation of d often difficult to define. When temporal changes are addressed, the question arises, whether the species pool of one time step or the whole species pool as recorded over several time steps should be regarded.
