1

EDGE EFFECTS IN A FOREST-GRASSLAND MOSAIC IN SOUTHERN INDIA

By

MILIND BUNYAN

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2009

2

© 2009 Milind Bunyan

3

To, Preethi- no one deserves this more (Proverbs 31: 29-30)

4

ACKNOWLEDGMENTS

I would like to thank the chair of my Graduate Committee, Dr. Shibu Jose for giving me

the opportunity to pursue a doctoral degree and for his support, encouragement and patience

especially when things did not go as planned. I would like to thank my committee members, Dr.

Ankila Hiremath, Dr. Kaoru Kitajima, Dr. Alan Long, Dr. Michelle Mack and Dr. George

Tanner for their time, guidance and encouragement through my program. I thank Dr. Ankila

Hiremath in particular, for help with the logistics of research in India. I am especially grateful to

Dr. Robert Fletcher for help with my research and methodology and Aditya Singh without who a

large part of this dissertation might not have existed.

I am grateful to the Karnataka and Kerala Forest Departments for permission to conduct

research in BRT Wildlife Sanctuary and Eravikulam National Park and Pampadum shola

National Park respectively. I would like to thank the staff at Ashoka Trust for Research in

Ecology and the Environment (ATREE), Bangalore and the ATREE Field station at BRT for

their assistance with data collection in BRT. My work in Eravikulam National Park was made

possible in no small part by Senthil Kumar, Ancel and Sarvanan and they are gratefully

acknowledged. Dr. Ganesan’s (ATREE) assistance was invaluable in the identification of plant

specimens.

I would like to thank my family in India for their love and support throughout the program.

None of this would have been possible without the endless supply of coffee, food and love from

my wife Preethi and right near the end, smiles and kisses from my one-year old, Rahael.

5

TABLE OF CONTENTS page

ACKNOWLEDGMENTS ...............................................................................................................4

LIST OF TABLES ...........................................................................................................................7

LIST OF FIGURES .........................................................................................................................8

ABSTRACT .....................................................................................................................................9

CHAPTER

1 THE SHOLA-GRASSLAND ECOSYSTEM MOSAIC: A SYNTHESIS OF EXISTING LITERATURE ....................................................................................................11

Introduction .............................................................................................................................11 The Shola-grassland Ecosystem Mosaic ................................................................................12

What is not a Shola? ........................................................................................................13 Flora ........................................................................................................................................13 Fauna .......................................................................................................................................15 Hydrology ...............................................................................................................................16 Soils and Nutrient Cycling ......................................................................................................17 The Shola-grassland Edge ......................................................................................................18 Conclusion ..............................................................................................................................19

2 MICRONENVIRONMENT EDGE EFFECTS IN THE SHOLA GRASSLAND ECOSYSTEM MOSAIC ........................................................................................................25

Introduction .............................................................................................................................25 Methods ..................................................................................................................................29

Study Area .......................................................................................................................29 Sampling Protocol ...........................................................................................................31 Shola Plots .......................................................................................................................31 Grassland Plots ................................................................................................................32

Data Analyses .........................................................................................................................32 Results .....................................................................................................................................34

Comparing Shola and Grassland Plots ............................................................................34 Microenvironment Gradients in Shola Fragments ..........................................................35

Discussion ...............................................................................................................................36 Seasonal and diurnal variation .........................................................................................38 Comparison between One-edge and Multiple Edge Models ...........................................39

Conclusion ..............................................................................................................................40

3 SPECIES DISTRIBUTION PATTERNS IN SHOLA FRAGMENTS ..................................48

Introduction .............................................................................................................................48

6

Methods ..................................................................................................................................51 Study Area .......................................................................................................................51 Sampling Protocol ...........................................................................................................53

Data Analyses .........................................................................................................................54 Vegetation Structure and Composition ...........................................................................54 Edge-interior Gradients in Species Distributions ............................................................55

Results .....................................................................................................................................56 Vegetation Structure and Composition ...........................................................................56 Edge-interior Gradients in Species Distributions ............................................................57

Discussion ...............................................................................................................................58 Conclusion ..............................................................................................................................61

4 EFFECT OF TOPOGRAPHY ON THE DISTRIBUTION OF SHOLA FRAGMENTS: A PREDICTIVE MODELING APPROACH ........................................................................72

Introduction .............................................................................................................................72 Methods ..................................................................................................................................74

Study Area .......................................................................................................................74 Imagery Pre-processing and Data Preparation ................................................................75 Predictive Modeling ........................................................................................................76

Results .....................................................................................................................................77 Combined Model .............................................................................................................77 Data Subsets ....................................................................................................................78

Discussion ...............................................................................................................................78 Comparison of Data Subsets ...........................................................................................80 Implications for the Shola-grassland Edge ......................................................................80

Conclusion ..............................................................................................................................81

5 SUMMARY AND CONCLUSION .......................................................................................91

APPENDIX ....................................................................................................................................95

LIST OF REFERENCES .............................................................................................................111

BIOGRAPHICAL SKETCH .......................................................................................................123

7

LIST OF TABLES

Table page 1-1 Faunal species richness in tropical montane forests ..........................................................22 

1-2 Surface soil chemical characteristics in tropical montane forests .....................................23 

2-1 Location of tropical montane forest fragments across study sites in the Western Ghats, southern India. ........................................................................................................42 

2-2 Plant-essential nutrient concentrations in tropical montane forest and grassland surface soils in the Western Ghats, southern India. ...........................................................42 

2-3 Principal components scores for soil nutrient variables in tropical montane forest fragments and grassland soils ............................................................................................43 

2-4 Edge-interior gradients in tropical montane forest fragments in the Western Ghats, southern India .....................................................................................................................43 

3-1 Location and vegetation characteristics of tropical montane forest fragments studied across the Western Ghats, southern India. .........................................................................63 

3-2 Species richness, dominance and diversity estimates in the shola-grassland ecosystem mosaic in the Western Ghats, southern India. ..................................................64 

3-3 Similarity and complementarity between tropical montane forest (shola) fragments in the Western Ghats, southern India. ....................................................................................65 

3-4 Pearson correlation coefficients for environmental variables with axis scores of NMS ordination. ..........................................................................................................................66 

4-1 Topographical variables used to predict presence of shola fragments in the Western Ghats, southern India. ........................................................................................................83 

4-2 Topographical parameter coefficients and AuC estimates for predictive models. ............83 

A-1 Overstory species presence matrix along an edge-interior gradient across nine fragments in the shola-grassland ecosystem mosaic in the Western Ghats, southern India. ..................................................................................................................................96 

A-2 Understory species presence matrix along an edge-interior gradient across nine fragments in the shola-grassland ecosystem mosaic in the Western Ghats, southern India .................................................................................................................................103 

8

LIST OF FIGURES

Figure page 1-1 Alternate stable state diagram for the shola-grassland ecosystem mosaic. .......................24 

2-1 Location of study sites in the Western Ghats. ....................................................................44 

2-2 Schematic representation of estimation of distance to additional edges. ...........................45 

2-3 Effect of the shola-grassland edge on microenvironment variables in the shola-grassland ecosystem mosaic ..............................................................................................46 

2-4 Edge-interior gradients in microenvironment and soil nutrient parameters in Pampadum shola National Park. Error bars represent one standard error. ........................47 

3-1 Species accumulation curves, species richness estimates and rare species (singletons and doubletons) for overstory and understory species.. .....................................................67 

3-2 Mean understory density along the edge-interior gradient in tropical montane forest (shola) fragments.. .............................................................................................................68 

3-3 Edge-interior gradients in diameter distribution of overstory tropical montane forest (shola) species across nine fragments.. ..............................................................................69 

3-4 Non-metric multi-dimensional scaling (NMS) ordination of overstory data.. ...................70 

3-5 Non-metric multi-dimensional scaling (NMS) ordination of understory data. ..................71 

4-1 Location of study sites in the Western Ghats. ....................................................................84 

4-2 Histogram of Intercept coefficients for Eravikulam National Park and Nilgiris data subsets.. ..............................................................................................................................85 

4-3 Histogram of Elevation coefficients for Eravikulam National Park and Nilgiris data subsets.. ..............................................................................................................................86 

4-4 Histogram of Slope coefficients for Eravikulam National Park and Nilgiris data subsets. ...............................................................................................................................87 

4-5 Histogram of Northness coefficients for Eravikulam National Park and Nilgiris data subsets.. ..............................................................................................................................88 

4-6 Histogram of Eastness coefficients for Eravikulam National Park and Nilgiris data subsets. ...............................................................................................................................89 

4-7 Histogram of Curvature of the Slope coefficients for Eravikulam National Park and Nilgiris data subsets. ..........................................................................................................90

9

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

EDGE EFFECTS IN A FOREST-GRASSLAND MOSAIC IN SOUTHERN INDIA

By

Milind Bunyan

August 2009 Chair: Shibu Jose Co-chair: Alan Long Major: Forest Resources and Conservation

Tropical montane forests in the Western Ghats in southern India consist of dense, insular

fragments in a matrix of grasslands separated by an abrupt, natural, edge. I studied edge effects

in the shola-grassland ecosystem mosaic across nine fragments in three study sites in the Western

Ghats. I measured microenvironment and soil variables and overstory species in 10×5 m plots

along an edge-interior gradient at 5 m intervals. Understory density and richness was recorded

from 5×5 m sub plots. Conventional distance to one-edge models indicated edge-interior

gradients in relative humidity (p = 0.018), magnesium (p = 0.027) and potassium (p = 0.008) in

large fragments. In small fragments, gradients in air temperature (p = 0.03), light transmittance

(p = 0.007) and soil moisture (p = 0.0002) were observed as a function of distance to multiple

edges.

We recorded 111 species (77 overstory; 83 understory) across nine fragments but did not

observe any edge-interior trends in overstory density or dominance. Similarly, no edge-interior

gradients were observed in understory density and structure. Non-metric multidimensional

scaling techniques for overstory and understory revealed greater variation among fragments than

could be attributed to edge-related within fragment variation. Overstory composition in mid-

elevation fragments differed significantly from high-elevation fragments while understory

10

vegetation varied as a function of fragment size. Our data indicate that mid-elevation fragments

should not be considered with high elevation fragments in future studies.

We also used a multiple logistic regression model to predict the presence of shola

fragments across two study sites in the Western Ghats. Elevation, slope, aspect (expressed as

eastness and northness), slope curvature and wetness index were used to predict the presence of

shola fragments. We observed that shola fragments were more likely to occur on northern and

western aspects than southern and eastern aspects. Shola fragments were also most likely to

occur on wet, steep slopes. The stability of the shola-grassland edge appears to be driven by fire

rather than frost while exposure to wind might be a driving factor also.

The shola-grassland ecosystem mosaic offers insights into fragmentation related patterns in

small patches and recommendations are made for future investigations.

11

CHAPTER 1 THE SHOLA-GRASSLAND ECOSYSTEM MOSAIC: A SYNTHESIS OF EXISTING

LITERATURE

Introduction

Tropical montane forests (alternatively called tropical montane cloud forests or simply

cloud forests) represent some of the most threatened ecosystems globally. Tropical montane

forests (TMF) are characterized and defined by the presence of persistent cloud cover. A

significant amount of moisture may be captured through the condensation of cloud-borne

moisture on vegetation distinguishing TMF from other forest types. Bruijzneel and Hamilton

(2000) described five kinds of TMF. Four of these, i.e. lower montane forest, lower montane

cloud forest, upper montane cloud forest and subalpine cloud forest, are based on elevation and

tree height whereas the last one an azonal low elevation dwarf cloud forest.

Elevations at which TMF are found, vary with mountain range size and insularity or

proximity to coast. Due to the mass-elevation effect (also known as the Massenerhebung effect),

larger mountain ranges permit the extension of the altitudinal range of plant species. Similarly,

higher humidity levels near coastal mountains enable the formation of clouds at lower altitudes.

On insular or coastal mountain ranges, TMF has been reported from elevations as low as 500m

(Bruijzneel and Hamilton, 2000). As elevation increases, tree height in TMF reduces and leaf

thickness and complexity in tree architecture increases. Other distinctive features of TMF are the

prolific growth of epiphytes and mosses and the lack of vertical stratification. TMF soils are

typically clay-rich, have low pH, abundant organic matter and are often nutritionally poor. TMF

are characterized by high levels of endemism driven by the limited availability of habitat (Bubb

et al. 2004). Located in the headwater catchments of seasonal or perennial streams, TMF

provides often undervalued ecosystem services to downstream communities.

12

Within the Western Ghats-Sri Lanka (WGSL) biodiversity hotspot (Myers et al. 2000),

TMF occurs as a mosaic of forests (locally and hereafter sholas) and grasslands and is commonly

referred to as the shola-grassland ecosystem. With limited exceptions (Werner 1995,

Somanathan and Borges 2000), data from the shola-grassland ecosystem mosaic are rarely

included in biome-wide popular (Bruijzneel and Hamilton 2000) as well as academic (Hamilton

et al. 1995, Bubb et al. 2004) synopses of scientific literature. As such, this document aims to

provide a synthesis of current research and the state of knowledge of the shola-grassland

ecosystem from peer-reviewed literature published on tropical montane forests in the WGSL

biodiversity hotspot. Additionally, a synopsis of research on the sholas of Kerala was also

reviewed (Nair et al. 2001).

The Shola-grassland Ecosystem Mosaic

The Western Ghats located in the WGSL hotspot are a 1600 km long mountain (160,000

km2) chain in southern India. Located above 1700 m, the shola-grassland ecosystem mosaic

consists of rolling grasslands with shola fragments restricted to sheltered folds and valleys in the

mountains separated from the grasslands with a sharp edge. Since, sholas frequently have

persistent cloud cover they can be classified as lower montane cloud forest or upper montane

cloud forest depending on elevation (Bruijzneel and Hamilton, 2000). Ecologists and foresters

have been puzzled over the pattern of the shola-grassland ecosystem mosaic for decades. While

some of the earliest scientific descriptions of the shola-grassland ecosystem described the mosaic

as dual climax (Ranganathan 1938), proponents of the single climax concept (Clements 1936)

argued that the forests represented a biotic (Bor 1938, Noble 1967) or edaphic climax (Jose et al.

1994). A δC13 analysis of peat samples from shola fragments in the Nilgiris indicated that shola

and grasslands have undergone cyclical shifts in dominant vegetative cover. Arid conditions

from 20,000-16000 yr BP led to predominance of C4 vegetation. This was followed by a wetter

13

phase which peaked around 11,000 yr BP leading to a dominance in C3 vegetation. The

weakening of the monsoon around 6000 yr BP led to the expansion of the C4 vegetation again

and the establishment of the current pattern, although a brief warm, wet phase around 600-700yr

BP also occurred (Sukumar et al. 1993).

What is not a Shola?

Arguably, the shola-grassland ecosystem mosaic is among the most distinct ecosystem

types in the WGSL biodiversity hotspot. Although, sholas are typically seen at elevations ≥ 1700

m, sholas at elevations as low as 1050 m have been studied by ecologists (Sudhakara 2001). In

the Anamalais and Nilgiris, the shola-grassland mosaic is characteristically patchy. Often though,

shola fragments are linear strips that may or may not be contiguous with lowland evergreen

forest which contain a different suite of species. While species dominance patterns are distinct

from lowland forest, sholas of different regions exhibit little similarity in species composition.

Yet, physiognomic characteristics of sholas are consistent. Sholas consistof profusely branched,

stunted trees (rarely exceeding 15 m) with prolific epiphytic growth. In order to distinguish shola

from non-shola forest types, despite the varied conditions under which they are found, I propose

that ecologically, a shola be defined as a high elevation (≥1700 m) stunted forest with distinct

physiognomy. Studies on sholas at elevations below 1700 m should be restricted to shola

fragments surrounded by grasslands. Indeed, in plots located at lower elevations, Sudhakara et al.

(2001) recorded families uncommon to sholas but common to lowland forests (Bombaceae,

Clusiaceae, Dichapetalaceae).

Flora

Since Thomas and Palmer (2007) provide a comprehensive review of current research on

grasslands in the shola-grassland ecosystem mosaic, I will restrict this section to reviewing work

on the shola vegetation only. Shola fragments contain species of both tropical and temperate

14

affinities. Also, the grasslands of the Western Ghats show more biogeographic similarity with

Western Himalayan species than TMF in Sri Lanka (Karunakaran et al. 1998).

Phytogeographical analysis of shola genera reveals that genera found on the fringes of shola

fragments and as isolated trees on grasslands are typically temperate (Rubus, Daphiphyllum and

Eurya) or sub-tropical (Rhododendron, Berberis, Mahonia are Himalayan) in origin. Species

within shola fragments on the other hand are IndoMalayan or Indian (rarely Paleotropical) in

origin (Suresh and Sukumar 1999, Nair and Menon 2001). Overstory species in the shola are

dominated by members of Lauraceae, Rubiaceae, Symplocaceae, Myrtaceae, Myrsinaceae and

Oleaceae while dicotyledonous understory species are dominated by Asteraceae, Fabaceae,

Acanthaceae, (Davidar et al. 2007, Swarupanandan et al. 2001). Dominant monocot species in

the understory include members of Poaceae, Orchidaceae & Cyperaceae (Swarupanandan et al.

2001). Along edge-interior gradients in shola fragments, species were found to be significantly

influenced by soil moisture (overstory and understory) and soil nitrogen (understory only) (Jose

et al. 1996). However this study was based on observations from a single shola patch.

Based on our knowledge of species-area curves, we might expect that the limited

availability of suitable habitat for shola species within the shola-grassland ecosystem mosaic

would limit α-diversity. However estimates for α-diversity are highly variable. Estimates for

Shannon-Weiner’s diversity index (H') range from 4.71 (Sudhakara 2001) to 0.87(Murali et al.

1998). Estimates for endemism are also highly variable- from 19.5 % (Suresh and Sukumar

1999) to 83.3 % (Nair and Menon 2001). Historically, the regeneration of arborescent flora in

shola fragments had been expressed as a concern (Vishnu-Mittre and Gupta 1968). A series of

studies now indicate that shola species show adequate regeneration under natural conditions

15

(Jose et al. 1994, Swarupanandan et al. 2001). Additionally, germination rates as high as 95%

have been recorded for shola species in germination trials (Srivastava, 2000).

As with other TMF, shola fragments exhibit prolific epiphytic growth. Studies in TMF

have shown that epiphytic species may constitute up to 25% of all biomass in tropical montane

forests (Foster 2001). They also provide microhabitats for invertebrates and amphibians

(Benzing 1998), store significant amounts of water (Sugden 1981) and influence nutrient cycling

(Benzing 1998). However, given the extent of scientific literature on arborescent flora in the

shola (Jose et al. 1996, Suresh and Sukumar 1999, Davidar et al. 2007, also see Nair et al. 2001),

very limited work exists on epiphytes in the shola-grassland ecosystem mosaic (Abraham 2001).

Similarly, very few studies have quantified productivity in the shola-grassland ecosystem

mosaic. In a comparison of net primary productivity (NPP) patterns of exotic plantations and

native shola forest, NPP and biomass of older exotic plantations (Eucalytpus globulus and Pinus

patula) were significantly higher than that of shola species. However this was at the cost of

lowering of NPP and biomass in the understory in exotic plantations, possibly due to allelopathic

inhibition (Jeeva and Ramakrishnan 1997).

Fauna

The shola-grassland ecosystem mosaic provides habitat for many faunal species of

conservation concern including the tiger (Panthera tigris tigris), dhole (Cuon alpinus), gaur (Bos

gaurus gaurus), Nilgiri langur (Trachypithecus johnii ) and Nilgiri marten (Martes gwatkinsii).

Endemic to the ecosystem-mosaic is the Nilgiri tahr (Niligiritragus hylocrius) which has been

studied meticulously over the years (Davidar 1971, Rice 1984, Rice 1988, Mishra and Johnsingh

1998, Daniels 2006). Although considered a flagship species for the ecosystem, uncertainty over

16

population estimates persists (Daniels et al. 2008) even as the population shows a declining trend

(Alembath and Rice 2008).

Faunal species too have been observed to mirror the shola-grassland ecosystem mosaic

pattern through habitat preferences. Small mammal communities in the Nilgiris (two species of

the nine recorded), showed a high degree of preference for either shola or grassland despite a

lack of resource-driven interspecific competition. However, these patterns were obscured in

exotic plantations (Shanker 2001). Strong habitat selection patterns have also been observed in

avian species in the shola-grassland ecosystem mosaic. Habitat suitability models for the Nilgiri

laughing thrush (Garrulax cachinnans) indicate that habitat use typically restricted to shola cover

might extend to exotic plantations (unsuitable habitat) when located near shola fragments (Zarri

et al. 2008). Other avian species such as the black and orange flycatcher (Ficedula nigrorufa)

have also been known to show a strong preference for shola cover.

Other than those mentioned above, inventories have also been conducted on amphibian,

avian, invertebrate and fish species (Dinesh et al. 2008, also see Nair et al. 2001). However with

the exception of the Nilgiri tahr, the body of scientific literature on faunal species in the shola-

grassland ecosystem mosaic is limited. (See Table 1-1)

Hydrology

Globally, tropical montane forests (alternatively tropical montane cloud forests) have been

shown to significantly influence ecosystem hydrology and biogeochemistry (Bruijnzeel and

Hamilton, 2000). In addition to providing cover and reducing erosion potential, net precipitation

(precipitation reaching the ground) under tropical montane forests is often greater than 100%

(and as high as 180%). This has been attributed to condensation of wind-driven fog on tree

crowns (termed fog drip). In areas of low precipitation such as the Canary Islands, interception

of cloud water can double annual precipitation (Gioda et al., 1995). Protection of tropical

17

montane habitat serves the purpose of hydrological regulation for downstream consumers also.

This is especially significant in the Western Ghats where major rivers originating in the shola-

grassland ecosystem mosaic provide hydrological services to consumers. A study by

Krishnaswamy et al. (2006) demonstrated that individual rainfall events could contribute as

much as 20-30% of the annual sediment load of 239-947 Mg km-2 (Krishnaswamy et al. 2006).

Other studies report significantly lower sediment load estimates (30-97 Mg km-2 year-1) from

other areas (Thomas and Sankar 2001, Sahoo et al. 2006) with as much as 90% of the annual

runoff occurring during the SW Monsoon (Thomas and Sankar 2001).

Soils and Nutrient Cycling

Soils in the shola-grassland ecosystem mosaic are granitic or metamorphic gneisses in

origin. They are of varying depth, ranging from deep (Ranganathan 1938) to shallow, stony soils

(Gupta and Shankarnarayan 1962). Typically, soils are shallower in the grasslands as compared

to shola soils and are more prone to soil moisture loss. During the dry season, shola soils have

been shown to retain as much as twice the soil moisture in the surrounding grasslands (Thomas

and Sankar 2001). Shola and grassland soils also differ nutritionally. Total N, available P and K

are higher in the sholas as compared to adjoining grasslands. Though this could be attributed to

higher litter decomposition and nutrient recycling rates in the sholas, these differences are rarely

significant. Jose et al. (1994) report organic carbon content in shola surface soils that are

comparable to those recorded in TMF in Ecuador. These values though are much higher than

those recorded by other authors for surface soils under varied types of cover in the shola-

grassland ecosystem mosaic (See Table 1-2). No soil-depth related trends have been reported for

plant essential micronutrients (Cu, Mn, Zn and Fe) in sholas or adjacent grasslands although

differences between sholas and adjacent grasslands have been observed (Nandakumar et al.

2001). Shola soils have higher soil nutrient pools than those under exotic plantations of blue gum

18

(E. globulus) or tea (Camellia sinensis) (Jeeva and Ramakrishnan 1997, Venkatachalam et al.

2007). Nutrient cycling under natural shola vegetation has also been described as steady state

and less likely to suffer losses to leaching since the return through leaf litter is low (Jeeva and

Ramakrishna 1997).

Trends in biogeographical affinities have also been recorded for soil microflora. Though

most soil fungal species recorded in soils in the shola-grassland ecosystem mosaic are

cosmopolitan in distribution, the prevalence of the genus Penicillium is characteristic of

temperate forests (Sankaran and Balasundaram 2001). Soil fungal species diversity is

comparable between shola fragments and grasslands (H'SHOLA=4.18, H'GRASSLAND=4.18) albeit

highly habitat specific (Sankaran and Balasundaram 2001). Shola soils also had significantly

higher soil bacterial and actinomycetes populations than grassland or plantation soils while

fungal populations were highest in grassland soils. Plantation soils under tea, blue gum and black

wattle (Acacia mearnsii) were also consistently observed to have lower soil microbial biomass

than soils under native vegetation (Venkatachalam et al. 2007).

The Shola-grassland Edge

The current dynamic equilibrium between insular shola fragments and grasslands is

indicative of the existence of alternate stable states enforced by environmental parameters (May

1977, Beisner et al. 2003). A change in parameters causes a shift in dominant cover (see Figure

1-1). Applied to the shola-grassland ecosystem mosaic, these parameters might include frost

(Ranganathan 1938, Meher-Homji 1965, 1967), fire (Bor 1938), grazing (Bor 1938, Noble

1967), soil nutrient status (Jose et al. 1994), soil depth (Ganeshaiah per comm.), wind

(Balasubramanian and Kumar 1999) and illegal harvesting (Gupta 1962). The persistence of the

mosaic in areas relatively free of anthropogenic grazing and illegal harvesting in some protected

areas make these two parameters tenuous for explaining the pattern. Although some authors have

19

observed grassland soils to be shallower than shola soils (Jose et al. 1994); others (Nandakumar

et al 2001) did not find a consistent trend. Moreover, shola species have been observed growing

on shallow soils too (Ranganathan 1938). Grassland fires are used as a management tool in the

shola-grassland ecosystem mosaic to reduce fuel loads (Karunakaran et al. 1998) and in some

instances a protective belt is cleared of vegetation around the shola before the grasslands are

fired to preclude fire from the shola (pers observ). These fires could act as an effective deterrent

in the colonization of the grasslands by shola species. Additionally, a study on vegetation fires

during the dry season (February-June) of 2006 revealed that tropical montane forests in the

Indian subcontinent accounted for 8.07% (92 fires) of all fires (Vadrevu et al. 2008). Although

current understanding points to fire as the dominant factor responsible for the maintenance of the

edge, ambiguity remains.

Conclusion

Historically, the shola-grassland ecosystem mosaic has undergone extensive habitat loss.

Plantations of exotic tree species were established in the grasslands aimed at augmenting timber

production as early as 1843 (Palanna 1996) with further introductions in 1870 in the Palni hills

(Srivastava 2001). Plantation programs were expanded under colonial rule to establish extensive

tea plantations in the mosaic. Post independence, tree plantation programs also received national

(federal) budgetary support (Raghupathy and Madhu 2007). Significant populations of invasive

shrubs and herbs (Eupatorium glandulosum, Ulex europaeus and Cytisus scoparius) in the shola-

grassland ecosystem mosaic were reported by early researchers (Bor 1938, Agarwal 1961). This

list continues to expand as new exotic species (e.g. Calceolaria mexicana, Erigeron

mucronatum) have recently been reported from the ecosystem (Seshan 2006). Threats to the

mosaic today include the harvesting of shola species to meet biomass and fuelwood requirements

and cattle grazing (Sekar 2008). In areas adjoining settlements, these threats can be significantly

20

amplified altering patterns in species richness and dominance (Chandrasekhara et al. 2001). The

WGSL biodiversity hotspot is likely to undergo extinctions in plant and vertebrate species due to

the limited availability of habitat (Brooks et al. 2002). Further, globally, TMF experience higher

annual loss in habitat than any other tropical forest biome (FAO 1993). As Sukumar et al. (1995)

suggest, climate change is expected to alter the dynamic equilibrium between the forest and

grassland through a reduction in the incidence of frost coupled with the strengthening of the

monsoon which would select for C3 species. Responding to these threats appropriately requires

the application of current state of knowledge coupled with an identification of gaps in our

knowledge.

Studies detailing the response of vegetation to the edge remain limited. To my knowledge

only one study to date quantifies edge effects in the shola-grassland ecosystem mosaic (Jose et

al. 1996). Unlike the sholas, fragmentation in other tropical montane forests (such as the

neotropics) is often a result of recent anthropogenically induced pressures (e.g. fire, conversion

to pasture). As such, edge effect studies in the shola-grassland ecosystem might be especially

insightful since species in older fragments have had time to equilibrate with fragmentation-

induced pressures (Turner et al. 1996, Harper et al. 2005). Fragmentation studies often observe a

proportional increase in area under edge influence with diminishing fragment size. Small

fragments may then be dominated by edge effect and lack an ‘interior’ or ‘core’, making them

susceptible to complete collapse (Gascon et al. 2000). An edge effect study in the shola-

grassland ecosystem would help us understand patterns in small fragments since shola fragments

in the shola-grassland ecosystem mosaic are often small (~1 ha).

In this study I investigate the effect of the shola-grassland edge on microenvironment

gradients in the ecosystem mosaic. I then assess patterns in species distributions and structure

21

between and within fragments studied. I expect these to provide an understanding of patterns in

smaller, older fragments. In mountainous terrain (where the shola-grassland ecosystem mosaic is

found), topography can significantly determine factors influencing vegetation (such as the

occurrence fire and frost). Topography can thus serve as a surrogate for these variables. I

therefore investigate the topographical variables determining the presence of shola fragments in

the mosaic. Through this, I seek to determine processes driving the shola-grassland edge.

22

Table 1-1. Faunal species richness in tropical montane forests Cover class Site Elevation Taxa Species

richness Species diversity (H')

Percent Endemism Source

Tropical montane forest/ Shola

Mannavan shola

1600-1700 Birds 30 - 20 Nameer (2001) 2000-2100 Birds 40 - 23 Nameer (2001)

Kerala - Fish 24 - - Ghosh (2001) Chembra 1700 Insects 81 4.22 - Mathew et al. (2001) Nilgiris 1800-2500 Small mammals 8 - - Shanker (2001) Nilgiris 2000-2050 Bacteria 93.22 - - Venkatachalam et al. (2007) Nilgiris 2000-2050 Fungi 7.78 - - Venkatachalam et al. (2007)

Grassland Nilgiris 1800-2500 Small mammals 3 - - Shanker (2001) Nilgiris 2000-2050 Bacteria 30.31† - - Venkatachalam et al. (2007) Nilgiris 2000-2050 Fungi 8.89* - - Venkatachalam et al. (2007)

Plantation (Mixed)

Nilgiris 1800-2500 Small mammals 3 - - Shanker (2001) Nilgiris 2000-2050 Bacteria 37.53† - - Venkatachalam et al. (2007) Nilgiris 2000-2050 Fungi 7.66* - - Venkatachalam et al. (2007)

Plantation (Tea)

Nilgiris 1800-2050 Small mammals 4 - - Shanker (2001) Nilgiris 2000-2050 Bacteria 18.54† - - Venkatachalam et al. (2007) Nilgiris 2000-2050 Fungi 5.78* - - Venkatachalam et al. (2007)

†Bacterial biomass (× 107colony forming units -1soil), *Fungal biomass (×105colony forming units-1soil)

23

Table 1-2. Surface soil chemical characteristics in tropical montane forests

Cover class Site Soil nutrients

Authors pH C N P K Ca Mg % kg ha-1

Tropical montane forest/ Shola

Brahmagiri 5.60* 2.80* 0.18*‡ 0.02* 0.14* 0.02* Thomas and Sankar (2001)

Nilgiris 5.44# 1.65# 306.91# 7.27# 138.43# 183.40§a 6.8§a Venkatachalam et al. (2007)

Eravikulam - 22.48# 1.21#‡ 0.02#‡ 0.01#‡ - - Jose et al. (1994) Ecuador (1960 m) 4.60+ 39.00+ 2.10+‡ 0.87+‡ 0.35+‡ 0.36+‡ 0.14+‡ Wilcke et al. (2008)

Ecuador (2090 m) 3.90+ 48.50+ 1.80+‡ 0.57+‡ 0.11+‡ 0.51+‡ 0.06+‡ Wilcke et al. (2008)

Ecuador (2450 m) 4.40+ 35.60+ 1.20+ ‡ 0.34+‡ 0.11+‡ 0.18+‡ 0.03+‡ Wilcke et al. (2008)

Grassland

Brahmagiri 5.00* 2.40* 0.04*‡ 0.01*‡ 0.03*‡ 0.02*‡ Thomas and Sankar (2001)

Nilgiris 4.04# 0.87# 132.92# 1.84# 70.68# - - Venkatachalam et al. (2007)

Eravikulam 18.88# Jose et al. (1994) Plantation (Eucalyptus globulus)

Nilgiris - - 97.50§ 10.60§ 74.50§ 123.60§ 45.70§ Jeeva and Ramakrishnan (1997)

Plantation (Pinus patula)

Nilgiris - - 188.50§ 22.70§ 109.80§ 158.80§ 127.90§ Jeeva and Ramakrishnan (1997)

Plantation (Mixed) Nilgiris 4.45# 1.10# 199.99# 3.67# 88.65# - - Venkatachalam et al. (2007)

Plantation (Tea) Nilgiris 4.06# 0.98# 205.32# 4.14# 100.19# - - Venkatachalam et al. (2007)

§0-10cm, *0-15cm, #0-20cm, +undefined (depth of O horizon), aJeeva and Ramakrishnan 1997, ‡ percent concentration

24

Figure 1-1. Alternate stable state diagram for the shola-grassland ecosystem mosaic. A change

in parameter (e.g. frost, fire) can cause a shift in communities. An increase in fire occurrence can move the dominant community (ball) from shola (A) to grassland (B). A reversal of the parameter can cause the community to return to its original state (along the dashed line).

(B) 

(A) 

25

CHAPTER 2 MICRONENVIRONMENT EDGE EFFECTS IN THE SHOLA GRASSLAND ECOSYSTEM

MOSAIC

Introduction

The creation of edges in forested habitat is a significant threat to the health and

maintenance of forested ecosystems. The term edge-effect was first used by Aldo Leopold

(1933) who observed higher species diversity near habitat edges. This was followed up by other

reports of increased edge-related species diversity (Johnston 1947). Land managers were advised

to create edges even while research sought to determine the optimum pattern of edge creation

(Yahner 1988). However, studies soon revealed altered ecological processes (Gates and Gysel

1978), species distribution patterns (Anderson et al 1977) and ecosystem structure (Laurance et

al. 1998, Gascon et al. 2000) near edges. At a newly created edge there are immediate changes in

the flow of energy, materials and organisms (termed ecological flows) across the edge as

communities gain access to resources hitherto separated. Species then respond to the altered

levels of resources (resource mapping) by adapting their distributions. Finally, this causes a

change in the interactions between species (Ries et al. 2004). Feedbacks between these

mechanisms (ecological flows, increased access, resource mapping and species interactions)

occur iteratively as primary and secondary responses with increasing edge age.

Numerous studies have quantified the mechanisms and patterns observed at edges as they

relate to fragmentation of forested habitats (for a review see Harper et al. 2005). The

characteristics of an edge are determined to a large extent by the influence of the edge on

overstory structure and composition and its interaction with the abiotic environment. Increased

light transmittance at newly created edges alters microenvironment conditions resulting in

elevated levels of air temperature and reduced relative humidity. Soil characteristics are altered

26

too as elevated soil temperatures, nutrient cycling and litter decomposition rates and diminished

soil moisture content are seen near edges.

As the edge evolves, changes in the abiotic and edaphic variables are reflected by changes

in the biological community. In fragments, this often results in a deepening of area under edge

influence (Gascon et al. 2000, Harper and Macdonald 2002). Interactions with other

anthropogenically induced threats such as fire, logging, introduction of exotic invasives and the

elimination of predators can result in the complete collapse of fragments in threatened

ecosystems (Curran et al. 1999, Gascon et al. 2000). Further, meso-scale climate phenomena

such as the El-Niño Southern Oscillation can amplify susceptibility to threats such as fires

(Cochrane, 2001). Following the creation of an edge, edges may either deepen (as discussed

above) or regenerate as organisms adapt to the physiological constraints of the edge environment

(Laurance et al. 2002). However at habitat edges that are maintained, a dense, vertical layer of

vegetation can develop, sealing off the forest interior from adjacent communities which leads to

the development of a short, sharp edge (Camargo and Kapos 1995, Didham and Lawton 1999).

These edges are more likely to develop when the contrast between adjacent habitats (patch

contrast) is high such as a forest-grassland edge and can affect a fragment’s persistence (Gascon

et al. 2000, Harper et al. 2005). In tropical montane forests too (TMF), the response to habitat

edges may differ based on edge type. For example, López-Barrera et al. (2006) observed light

transmittance and soil moisture changed abruptly across a hard edge in a Mexican TMF.

Studies on microenvironment and soil edge gradients are limited in TMF (Jose et al. 1996,

López-Barrera et al. 2006) as compared to those of lowland forests (Williams-Linera 1990,

Camargo and Kapos 1995, Williams-Linera et al. 1998, Sizer and Tanner 1999, Didham and

Lawton 1999). This is surprising considering the disproportionate endemism associated with

27

TMF (Hamilton et al. 1995). Moreover, microenvironment edge-interior gradients in TMF can

operate at different scales than lowland forests. Microenvironment edge-interior gradients in

lowland forests typically vary between 40-60m from the forest edge (Kapos 1989, Williams-

Linera 1990, Didham and Lawton 1999) with edge-effects being reported as much as 100m away

from the edge (Laurance et al. 2002). In comparison, edges in tropical montane forests and

patterns are complex and even contradictory. Edge-interior gradients in TMF have been observed

to be as short as 15-30m (Jose et al. 1996), while other studies have reported an insignificant

effect of patch contrast on seedling growth and defoliation (López-Barrera et al. 2006). Some

edges may even display ‘reverse’ edge-effects with lower nest predation rates being recorded

near the edge than the forest interior (Carlson and Hartman 2001) unlike patterns observed for

other forest edges.

Conventional edge-effect studies though (irrespective of biome) typically consider the

effect of distance from the nearest edge. This offers the least complex framework to demonstrate

the response of and interaction between abiotic and biotic response variables to (initially) edge

creation and (later) edge habitat. However, it limits the extrapolation of edge effects across

spatial scales (Ries et al. 2004). Although extensive literature exists on edge-effects as a function

of distance from the nearest edge, limited studies exist on multiple edge effects (Malcolm 1994,

Fernández et al. 2002, Zheng and Chen 2003, Fletcher 2005). Interestingly, even large scale

manipulative studies on fragments of various sizes do not consider the effect of additional edges.

This is despite the fact that edge effects can penetrate as much as 100m away from the edge.

(BDFFP, Laurance et al. 2002). The incorporation of multiple edges has been shown to generate

stronger edge effects than models using distance to nearest edge alone. Further these models

become increasingly significant in small fragments that may be dominated by edge habitat

28

(Gascon et al. 2000). Studies have used different techniques to investigate the effect of multiple

edges on response variables. Malcolm (1994) modeled the edge as a collection of points

integrating the effect of each of these points along the edge (point edge effects). However, as

Fletcher (2005) observed, point edge effects offer a robust, yet abstract estimate of edge effects

since point edge effects are hard to measure. Fletcher (2005) also observed that bobolink

(Dolichonyx oryzivorus) distributions were influenced by multiple edges, but the response to

these edges was determined by scale considerations (i.e. multiple edge effects were less

important at a landscape scale). Increasing the number of edges in edge-effect models has

statistical implications too. As the number of parameters increases, the power of the model

reduces (the probability of Type II error increases). Mancke and Gavin (2000) suggested the use

of an integrative index that combines the use of distance to edges in four directions through a

complex ruleset to ensure orthogonal separation in irregular fragments. They suggest though that

simple orthogonal separation can be ensured by estimating distance to 2nd, 3rd and 4th edges in

three cardinal directions with respect to the edge-interior transect.

Within the Western Ghats-Sri Lanka (WGSL) biodiversity hotspot in southern India, TMF

(locally and hereafter termed as sholas) are seen at elevations ≥1700 m and consist of insular

forest fragments in a matrix of grasslands. Shola fragments are typically small (~1 ha) and are

separated from the surrounding grasslands by a sharp edge which is natural in origin. This study

was conducted on the shola-grassland edge in the Western Ghats within the WGSL hotspot. I

quantified effects of the shola-grassland edge on microenvironment and soil variables at a

regional scale in the shola-grassland ecosystem mosaic. I tested for differences in

microenvironment and edaphic variables between shola fragments and grasslands and to

determine if the edge is controlled by these variables. I also tested for gradients in these variables

29

as a function of distance from the edge and by doing so, tested for differences between small and

large fragments. I hypothesize that the small (1-15 ha) fragments will be dominated by edge

habitat (i.e. weak or absent edge-interior gradients) while large fragments (> 50 ha) will exhibit

strong edge-interior gradients. I also tested for edge-effects in grassland plots along a grassland-

edge-shola interior transect and hypothesize that grassland plots too will be affected by distance

from the shola-grassland edge.

Methods

Study Area

The study area is located in the Western Ghats, a region of high conservation importance

(Myers et al. 2000) in southern India. Three study sites were selected in the Western Ghats in

southern India; Biligiri Rangaswamy Temple (BRT) Wildlife Sanctuary in the state of Karnataka

and Eravikulam National Park and Pampadum shola National Park in the state of Kerala ( Fig 2-

1). BRT Wildlife Sanctuary (540km2), is located at the eastern-most edge of the Western Ghats

(11°43’N to 12°08’ and 77° to 77°16’E). Soils in BRT have been classified as Vertisols and

Alfisols and described as shallow to moderately shallow with well drained gravelly clay soils on

the hills and ridges (Krishnaswamy 2004, NATMO 2009). BRT is exposed to both southwestern

and northeastern monsoons with a dry season from November to May. Rainfall ranges between

898 mm to 1750 mm and strong spatial trends have been observed based on topography and

altitude (Krishnasmamy et al. 2004). Vegetation in the sanctuarty is represented by a diversity of

forest types ranging from dry scrub forests to dry and moist deciduous forests, evergreen forests

and the high-altitude TMF-grassland (or shola-grassland) ecosystem mosaic. Within BRT, three

shola fragments were identified as study sites (Figure 2-1, Table 2-1). Accessibility and the

limited availability of insular shola fragments restricted the number of fragments available in

BRT. Although shola vegetation is more extensive in the southern portion of the sanctuary,

30

(particularly near Jodigere shola) most of these areas were considered inappropriate for the

purpose of this study since they are contiguous with lowland forest and do not represent true

forest fragments.

Eravikulam National Park (ENP) is located in the Anaimalai Hills (10°05’ to 10°20’N and

77°0’ to 77°10’E) within the state of Kerala (Fig 2-1). The park (119 km2) consists of a base

plateau at an elevation of 2000m surrounded by peaks with a maximum elevation of 2695 m.

While the mean maximum temperature recorded in lower elevation tea estates outside ENP

varies between 21.9 °C and 22.7 °C, mean maximum temperature within ENP is as low as 16.6

°C. Similarly, mean minimum temperatures within ENP are lower than those recorded in the tea

estates (13.3 °C). The warmest month in ENP is May with a mean maximum temperature of 24.1

°C and January is the coldest month with a mean mimimum temperature 3 °C (Rice 1984). The

soil had been classified as Alfisols and described as Arachean igneous in origin consisting of

granites and gneisses (NATMO 2009). Soils are sandy clay, have moderate depth (30-100 cm)

and are acidic (ph 4.1-5.3). ENP receives rainfall approximately 5200 mm of rainfall annually

from both the southwest and the northeast monsoon with the former contributing as much as 85

% of the annual rainfall. Although contiguous rainforest formations can also be found at lower

elevations, vegetation in the park is predominantly shola-grassland ecosystem mosaic consisting

of rolling grasslands interspersed with dense, insular shola fragments. Five shola fragments were

selected within ENP (Table 2-1).

Pampadum shola National Park (PSNP) located within Idukki district in the state of Kerala

consists of a large shola patch (1.32 km2) contiguous with lowland forest. Although little is

known about this newly notified park (located approximately at 77°15’E and 10°08’N) its

proximity to the larger Mannavan shola (77°09’-77°12’N and 10°10’-10°12’S) provides insights

31

into broad climatic trends. Mean annual temperature in Mannavan shola has been reported to be

approximately 20°C with mean minimum temperature in the coldest months (December and

January) reaching 5°C. Mean annual precipitation in Mannavan shola ranges between 2000-3000

mm. Data from PSNP were collected to compare patterns observed against smaller fragments

chosen in BRT and ENP.

Sampling Protocol

To investigate edge-interior gradients in shola fragments, transects were laid out

perpendicular to the shola-grassland edge. Plots (10×5 m) were established along these transects

with the first plot being established at the edge itself. The plot was laid out with the longer side

(10 m) running parallel to the edge to capture maximum variability along the edge-interior

gradient. Successive plots along transects were separated by 5 m and extended till the middle of

the patch or to a distance of 35 m from the edge, whichever was shorter. Replicate transects

within a patch were separated by at least 35 m though most often greater than 50 m. The number

of replicate transects (minimum 1, maximum 4) were determined by the size of the fragment.

Shola Plots

Within each 10×5m plot, microenvironment and soil variables were measured. Soil

temperature was measured using a Barnant® Type K thermocouple thermometer (BC Group

International Inc, St Louis, MO) and air temperature and relative humidity were measured using

a digital thermohygrometer (Oakton Instruments, Vernon Hills, IL). Photosynthetically active

radiation (PAR) was measured above and below shola canopy using an AccuPAR LP-80

Ceptometer (Decagon Devices, Pullman, WA). Light transmittance (τ) was then calculated as the

ratio between above and below canopy PAR. Since the shola-grassland ecosystem is

characterized by highly variable light environment, above canopy PAR was measured in the

adjoining grassland prior to measurement of below canopy PAR in the shola plots. From each

32

10×5 m plot, one surface soil (0-30 cm) sample was collected using a soil corer and immediately

double sealed in Ziplock® bags. Soil cores were collected from all but one fragment

(Karadishonebetta, BRT) Gravimetric water content was measured by oven-drying soil samples

for 24 hours at 105°C. Soil samples were analyzed for soil organic carbon (Walkley-Black

method), total nitrogen (micro-Kjeldahl distillation method), and extractable phosphorus (Bray-

Kurtz method). Available potassium and sodium were estimated using the Flame photometric

method (Jackson 1962). Calcium and Magnesium were estimated by EDTA titration (Allen

1974). DTPA extractable iron, manganese, zinc, copper, boron and molybdenum were estimated

using an Atomic Absorption Spectrophotometer (Soultanpour and Schwab 1977).

Grassland Plots

In four of the shola fragments studied, transects were also established in the grassland to

study effects of the shola-grassland edge on the grassland. Since these transects were established

as matching pairs to the shola-transects, they originate at the same point as the shola transect

with replicate transects at least 35 m apart. Microenvironment and soil samples were collected

and soil samples analyzed using protocols described above from two plots at distances of 5 m

and 15 m from the shola-grassland edge to test for gradients along grassland-edge-shola interior

transects.

Data Analyses

In small forest fragments, it was hypothesized that plots would be influenced by more than

one edge. These plots were spatially analyzed using ArcGIS 9.2 (ESRI Labs, Redlands CA) to

determine distance to 2nd, 3rd and 4th nearest edges from each plot (Fig 2-2A). These edges were

determined by measuring the distance to the edge in orthogonal directions, thus achieving a

measure of distance to edge in each cardinal direction. This approach offers an optimum

combination of incorporating information from additional edges while ensuring orthogonal

33

separation (Mancke and Gavin 2004). The spatial analysis of additional edges also revealed

anomalies in the estimation of distance to the nearest edge in small fragments. In some instances,

distances to the edge assumed to be the nearest during data collection would be larger than

distances to the 2nd (on one occasion the 3rd nearest) edge. Although uncommon, these distance

to nearest edge estimates were then revised prior to data analysis (Fig 2-2B). Data obtained from

the large shola fragment (PSNP) were also analyzed separately from other fragments in order to

test for differences between large and small fragments.

An ANOVA was used to test for differences between shola and grassland

microenvironment and soils along grassland-edge-shola interior transects in the fragments (n = 4)

with both shola and grassland plots. Tukey’s HSD was used for multiple comparisons (family-

wise error rate p < 0.05) to test for differences across the six distance classes (two in the

grassland and four in the shola fragments). Additionally, a t-test was used to test for differences

in soil microenvironment conditions between grassland and shola soils. For these analyses, soil

nutrient concentrations were arcsine-square root transformed in order to achieve normality. A

Principal component analysis (PCA) was used to quantify the response of soil nutrient variables

to the shola-grassland edge. PCA is often used to define the underlying structure between

correlated variables and reduce the number of variables to a smaller number of more meaningful

factors and are particularly appropriate when the number of variables is large. Component scores

(or loadings) were used to test for a) differences between shola and grassland soils (n = 4) using

a one-way ANOVA and b) for edge-effects on soil nutrient concentrations (using a simple linear

regression) in shola fragments (n = 8).

A regression analysis was used to study the effect of the forest (shola)-grassland edge on

microenvironment (n = 9) and soil variables (n = 8) in the sholas. A simple linear regression was

34

also used to quantify the response of microenvironment and edaphic variables as a function of

distance from the nearest edge (one-edge model) while a multiple linear regression was used to

quantify the effect of multiple (nearest, 2nd nearest, 3rd nearest and 4th nearest) edges on response

variables (multiple edge model) in small fragments (>1 ha- 13 ha). Response variables were

regressed against distance estimates and log-transformed distance estimates. Log-transformed

distance estimates weight plots near edges higher since I have an a priori expectation that edge-

interior gradients in shola fragments will be short though sharp. All data analysis was performed

using SAS (SAS Institute, Cary NC). Given the small sample size I used a conservative estimate

of p = 0.05 in order to reduce Type I errors.

Results

Comparing Shola and Grassland Plots

Air temperature (p = 0.031) and soil temperature (p < 0.020) declined significantly along

the grassland-edge-shola interior transects (Fig 2-3). Although grassland plots have significantly

lower humidity (p = 0.034) than shola plots no gradients were observed along grassland-edge-

shola interior transects (p = 0.47). Nutritionally, shola and grassland soils differed little (Table 2-

2) and no trends along the grassland-edge-shola interior transect were observed. The use of

principal component analysis (PCA) identified four factors as significant using the latent root

criterion (eigenvalue ≥ 1.0) and explained 73% of the variation (Table 2-3). A varimax rotation

of the selected components was used to maximize orthogonal separation between components

and reduce loading of soil nutrients on more than one principal component (or cross-loading). An

ANOVA of the selected soil nutrient principal components revealed that on the first principal

component (PC1), grassland and shola soils differed marginally (p = 0.054). Organic carbon and

sodium load heavily on PC1 but vary inversely with calcium and magnesium which also load

heavily on PC1 (Table 2-3). None of the other principal components (PC2, PC3 and PC4) varied

35

between habitat type (p ≥ 0.35). Further none of the factors varied along the two-sided grassland-

edge-shola interior transect (p ≥ 0.22).

Microenvironment Gradients in Shola Fragments

In the large shola fragment (PSNP), relative humidity and soil temperature decreased

linearly with increasing distance from edge (Table 2-4, Figure 2-4). While relative humidity

decreased strongly (p = 0.01) with increasing distance from edge and soil temperature decreased

strongly weakly (p = 0.06), no trends were observed in air temperature (p = 0.74) and light

transmittance (p = 0.79). However, light transmittance decreased to 3(±0.3) % of overstory light

conditions within 5 m of the edge. Although most soil nutrients did not vary as a function of

distance from the nearest edge, available potassium increased (p = 0.0089) with increasing

distance from edge (Table 2-3, Fig 2-2). Contrary to other studies that have reported strong non-

linear relationships in soil variables (Jose et al. 1996), I found no evidence of non-linearity.

Adding a quadratic (distance to nearest edge) term did not significantly improve the fit of

relative humidity (p = 0.018) and reduced the significance of estimates of available potassium (p

= 0.030) and other significant variables (p ≥ 0.101). A factor analysis of soil nutrient variables

with quartimax rotation for PSNP soils revealed two factors as significant that explained 72% of

the variation (data not presented). While the first factor loads heavily on organic carbon, total

nitrogen, available phosphorous and calcium, the second factor loads heavily on manganese.

However a linear regression of factors against both factors did not vary with distance from the

nearest edge (p ≥ 0.13).

Small forest fragments though showed no trends in microenvironment (air temperature,

relative humidity, light transmittance) or soil variables (soil temperature, macronutrient and

micronutrient concentrations) as a function of distance of nearest edge (single edge model).

Responses were weaker still when log-distances were used (p ≥ 0.242). The incorporation of

36

information through distance estimates for additional (2nd, 3rd and 4th nearest) edges through the

multiple edge model resulted in a significant improvement of the regression models. With

increasing distance from edge, significant patterns were observed as air temperature (p = 0.0320)

and light transmittance (p = 0.0072) decreased. Weaker trends were observed with relative

humidity increasing and soil temperature decreasing with increasing distance from edge (Table

2-4). Soil macronutrient variables were not correlated with edge distance using either the one-

edge or multiple edge models. Factor analysis of soil nutrient variables produced three factors

(and explained 55%) of the variation). Again, none of these factors were correlated with either

the one-edge or multiple edge models. A canonical correlation was also used to test all soil

nutrient variables against all distances in order to test for correlations between the variables.

However, the analysis failed to find any significant correlations.

Discussion

The microenvironment in grasslands was significantly different from that of shola

fragments. However, no significant differences were observed between grassland and shola

fragment soils. Similar trends in soil response variables have been reported by other authors too

(Jose et al. 1994, Nandakumar et al. 2001). Our estimates for soil nutrient variables for sholas

and grasslands (Table 2-1) are comparable to others from the ecosystem mosaic (Thomas and

Sankar 2001). However, significantly higher estimates for soil organic carbon and soil moisture

have also been reported from the ecosystem mosaic (Jose et al. 1994). Factor analysis of soil

variables indicates that soil organic carbon and calcium load had positive and negative loading

on the first factor (which explained 20% of the variation in the dataset). This is typical of forest

soils that have higher soil organic carbon while basic cations (Ca2+) are lower (as they are lost to

leaching) than grassland soils.

37

Although two-sided edges are less investigated, their proportion vis-à-vis one sided edge

studies has increased considerably (Fonseca and Joner 2007). Their relevance to matrix-based

approaches to understanding edge-effects and restoration ecology is also being increasingly

recognized (Ries and Sisk 2004, Ries et al. 2004, Fonseca and Joner 2007). This could be

significant in the shola-grassland ecosystem mosaic where the matrix surrounding forest

fragments (typically grassland), has been shown to be hostile to animal taxa (Shanker 2001, Zarri

et al. 2008). In our study I found evidence of two-sided edges in microenvironment response

variables. However this pattern was absent in soil response variables with the exception of weak

trends in soil organic carbon. Information on two-sided edges in the shola-grassland ecosystem

mosaic can also assist in ecosystem restoration efforts by providing requisite conditions for re-

establishment of shola fragments in areas converted to exotic tree plantations.

Soils in the shola-grassland ecosystem mosaic had high levels of DTPA extractable iron.

Although much higher than another estimate from the mosaic (Nandakumar et al. 2001) it is

comparable to those observed from TMF-grassland edge in Sri Lanka experiencing dieback

(Ranasinghe et al. 2007). Although Ranasinghe et al. (2007) suggested that the dieback of TMF

could be related to iron toxicity induced diminished nitrogen absorption for soils with iron levels

that are higher (200-400 ppm). I did not observe such high levels of DTPA extractable iron in

our study. However, unlike TMF-grassland soils in Sri Lanka, shola fragment and grassland soils

alike did not record high DTPA extractable manganese or copper levels.

Large shola fragments exhibit different patterns from smaller shola fragments. In

Pampadum shola distance from the nearest edge was sufficient to explain trends in

microenvironment and edaphic variables. However, distance from the nearest edge could not

sufficiently explain the soil variables.

38

Seasonal and diurnal variation

Since individual transects were sampled over the course of a single day, a diurnal variation

in relative humidity and air temperature can be expected. As the day progresses, ambient air

temperatures would increase leading to a lowering of relative humidity. An inverse relationship

between relative humidity and air temperature, it can be argued, could be an artifact of either

distance from the forest fragment (shola)-grassland edge or diurnal variation. However, in our

study, air temperature and relative humidity were not inversely related. In smaller fragments, we

did indeed observe such trends (air temperature decreased with increasing distance while relative

humidity increased). However in the large fragment (PSNP), relative humidity decreased with

increasing distance from edge (and as the day progressed) while air temperature showed no edge-

interior trends. Such changes in relative humidity trends as a function of distance from the edge

could also be an artifact of seasonality. Sampling of the large fragment (PSNP) was done in the

wet season and often under constant cloud cover leading to higher relative humidity near the

shola-grassland edge and lower humidity inside (Figure 2-4). Since the sampling of small

fragments was carried out in both wet and dry seasons, trends are more intuitive (increasing

humidity with increasing distance from edge) albeit weakened by wet season data.

Our estimate of light transmittance (3% of overstory light conditions) is significantly lower

than light transmittance at the edge (12%) reported by Jose et al. (1996) although similar to

estimates from old growth rainforest-pasture edges in Mexico (4.7%, Williams-Linera et al.

1998). Other studies too have reported an abrupt change in light conditions across a high contrast

edge (López-Barrera et al. 2006). Shola fragments thus represent deeply shaded habitats and low

depth of influence (DEI) with regard to light transmittance. Analysis of the large shola fragment

(PSNP) revealed trends (although weak) widely reported from other edge-effect studies (e.g. soil

39

temperature) both in tropical montane forests and lowland tropical forests (Jose et al. 1996,

Laurance et al. 2002).

The present study investigates edge-interior gradients in microenvironment and soil

nutrient availability across nine tropical montane (shola) fragments. To my knowledge only other

study in the shola-grassland ecosystem investigates edge-interior gradients (Jose et al. 1996).

Although the study reported distinct edge-interior gradients for both microenvironment and

edaphic variables, this dataset is based on data from a single fragment.

Comparison between One-edge and Multiple Edge Models

Studies on edge-effects across biomes have demonstrated the effect of distance from edge

on microenvironment and edaphic variables. In tropical forests, edge effects have been

particularly worrisome as microenvironment gradients extend as much as 100m away from the

edge of forested habitat (Laurance et al. 2002). As a corollary to this, smaller fragments have

been hypothesized to be dominated by edge habitat (Gascon et al. 2000). Most studies in tropical

fragments though measure the response variables as a function of distance from nearest edge and

do not incorporate the effect of distance from additional edges (however, see Malcolm 1994). In

the shola-grassland ecosystem mosaic, large shola fragments exhibited different patterns from

smaller shola fragments. In Pampadum shola distance from the nearest edge was sufficient to

explain trends in microenvironment and edaphic variables. However in smaller forest fragments,

microenvironment and edaphic variables did not show a response with distance from the nearest

edge suggesting that fragments are dominated by edge habitat. However, the incorporation of

additional edges as predictor variables in the model significantly improved model fit. Moreover

multiple edge-models were more parsimonious (lower AIC values) than one-edge models (Table

2-4). To my knowledge, no other study from the shola-grassland ecosystem and indeed other

other tropical montane forest fragments test for multiple edge effects. The incorporation of

40

multiple edge-effect models might better explain the absence of edge interior gradients in other

edge effect studies also (Williams-Linera et al. 1998).

Delineating edges and determining edges using spatial analysis techniques may not always

be feasible though. In our study, insular forest fragments are isolated from other forest types by a

high contrast grassland edge. This made determination of distance to additional edges relatively

easy. When edges separate similar habitat types (such as a native forest-tree plantation edge)

distinguishing habitat types might require the use of other techniques (landcover/landuse

classification).

Conclusion

The present study describes the response of microenvironment and edaphic variables to

distance from a tropical montane forest (shola)-grassland edge. Nutritionally, grassland and shola

soils differed little. Since such results have been consistently observed across multiple study sites

in the ecosystem mosaic, an edaphically controlled shola-grassland edge appears unlikely. The

influence of other driving variables such as fire or frost needs further investigation. I observed

that conventional one-edge models sufficiently explained variation trends in microenvironment

variables along the edge-interior gradient in large fragments. As with other studies on small

fragments though, I observed no edge effects with the use of a conventional one-edge model.

However, the inclusion of multiple edges in small fragments significantly improved model fit. I

can conclude that small fragments observed to be dominated by edge habitat may in fact

resemble larger fragments with the inclusion of multiple edges. Our models did not evaluate non-

linear effects which often better explain patterns in edge-interior gradients (Malcolm 1994). The

incorporation of such non-linear models in the system might further improve model fit. Finally,

further research is required in investigating the effect of multiple edge models in predicting edge

41

effects across fragment sizes, edge types and biomes in order to improve our understanding of

edge effects.

42

Table 2-1. Location of tropical montane forest (shola) fragments across study sites in the Western Ghats, southern India.

Fragment (shola name)

Fragment code

Study site Sampling season

Longitude (°E)

Longitude (°N)

Patch area (ha)

Karadishonebetta KR BRT Dry - - - Gummanebetta GU BRT Dry 77° 10' 38" 12° 1' 35" 1.73 Jodigere JG BRT Dry 77° 11' 12" 11° 47' 6" 0.20 Suicide point SP ENP Wet 77° 1' 33" 10° 8' 23" 1.24 Chinnanaimudi CA ENP Wet 77° 3' 24" 10° 9 '4" 1.66 Kolathan KS ENP Wet 77° 4' 31" 10° 12' 57" 1.58 Pusinambara PS ENP Wet 77° 4' 3" 10° 12' 56" 1.75 Mukkal mile MM ENP Wet 77° 5' 7" 10° 13' 51" 13.42 Pampadum PP PSNP Wet 77° 15' 5" 10° 7' 28" 132.00 Table 2-2. Plant-essential nutrient concentrations in tropical montane forest (shola) and

grassland surface soils in the Western Ghats, southern India. Values are means (±1 SE) expressed as g kg-1 unless noted otherwise.

Soil nutrient parameter Shola Grassland Macronutrients Soil organic carbon 11.28* (0.293) 10.52* (0.625) Nitrogen 0.53 (0.0090) 0.55 (0.0152) Phosphorous 0.144 (0.0036) 0.143 (0.0060) Potassium 0.135 (0.00033) 0.133 (0.00699) Calcium 0.139 (0.0033) 0.145 (0.0052) Magnesium 0.142 (0.0033) 0.134 (0.0069) Micronutrients Iron 0.121(0.0016) 0.124 (0.0040) Boron 0.0057 (0.00013) 0.0058 (0.00029) Manganese 0.00013 (0.0000046) 0.00014 (0.0000117) Zinc 0.048 (0.0011) 0.050 (0.0017) Copper 0.000056 (0.0000014) 0.000061 (0.0000029) Molybdenum 0.000138 (0.0000045) 0.000134 (0.0000111) * P < 0.05

43

Table 2-3. Principal components scores for soil nutrient variables in tropical montane forest

(shola) fragments and grassland soils (n = 4) in the Western Ghats, southern India (eigenvalue ≥ 1). Bold values indicate the component on which the variable loads after a varimax rotation was applied.

Variable PC1 PC2 PC3 PC4 Eigenvalue 2.6339 1.7456 1.5472 1.1782 Proportion variance explained 0.2634 0.1746 0.1547 0.5927 Cumulative variance explained - 0.4380 0.5927 0.7105 OGC 0.8230 -0.2062 0.2420 -0.1481 NIT -0.2197 0.8292 -0.0645 -0.0431 PHO -0.0569 0.7863 0.1720 0.3743 POT -0.1527 0.0845 -0.0660 0.8147 CAL -0.6289 0.0689 0.5139 0.2867 MAG -0.7399 -0.0993 0.3650 -0.311 SOD 0.6805 -0.1299 0.1163 -0.1697 MAN 0.0357 0.6671 0.1054 -0.5675 MOL 0.0254 0.1001 0.8919 0.0920 COP -0.0454 0.0006 -0.6009 0.2412 OGC: Organic carbon, NIT: Total Nitrogen, PHO: Available phosphorous, POT: Potassium, CAL: Calcium, MAG: Magnesium, SOD: Sodium, MAN: Manganese, MOL: Molybdenum, COP: Copper Table 2-4. Edge-interior gradients in tropical montane forest (shola) fragments in the Western

Ghats, southern India. Small fragments (n = 7) varied between 0.2-13 ha whereas the large fragment (n = 1) was 132 ha.

Variable

Fragment size Large Small Predictor variables

r2 p Predictor variables

r2 p

Air temperature (°C) 0.017 0.7492 logdist, logdist2

0.174 0.0320

Relative humidity (%) logdist 0.443 0.0181 0.142 0.0625 Light transmittance (%) 0.006 0.7978 logdist,

logdist2 0.268 0.0072

Soil temperature (°C) 0.302 0.0637 0.180 0.0695 Total nitrogen 0.010 0.7568

0.0637 0.1068

Magnesium logdist 0.3990 0.0276 0.002 0.7418 Potassium (%) logdist 0.511 0.0089 0.007 0.5864 Avail. Phosphorous (%) 0.316 0.0570 0.114 0.1979 Calcium 0.067 0.4149 0.114 0.3305 Significant relationships are in bold (p < 0.05), number of predictor variables used in the model was chosen based on Akaike information criterion (AIC). Predictor variables are only displayed for significant models.

44

Figure 2-1. Location of study sites in the Western Ghats. Insular forest fragments were selected from three sites- Biligiri Rangaswamy Temple Wildlife Sanctuary (BRT), Eravikulam National Park (ENP) and Pampadum shola National Park (PSNP). Three fragments were studied in BRT (A), Gummanebetta (GU), Jodigere (JG) and Karadishonebetta (KR). Five fragments were studied in Eravikulam National Park (B), Chinnanaimudi (CA), Kolathan (KS), Mukkal Mile (MM), Pusinambara (PS) and Suicide point (SP). See Table 2-1 for details on.fragments. For inset maps, 1cm equals 4 kilometers

45

Figure 2-2. Schematic representation of estimation of distance to additional edges. (A) d1, d2, d3 and d4 represent distance estimates to 1st, 2nd 3rd and 4th nearest edge. (B) In small fragments, estimated distances to additional edges (gray letters) differed from actual distances (black letters). These distance estimates were revised prior to data analysis. Black dots represent the plot center and d1-d4 lines represent distance to nearest, 2nd, 3rd, and 4th nearest edges respectively.

A B 

46

Figure 2-3. Effect of the shola-grassland edge (dashed line) on microenvironment variables in the shola-grassland ecosystem mosaic. (A) Soil temperature. (B) Air temperature. (C) relative humidity. Different letters indicate significant differences as determined by Tukey’s HSD comparison (p < 0.05). Error bars represent one standard error.

47

2.5 12.5 22.5 32.5

Mag

nesi

um (%

)

0.010

0.012

0.014

0.016

0.018

Rel

ativ

e hu

mid

ity (%

)

50

60

70

80

90

100

Pota

ssiu

m (%

)

0.008

0.010

0.012

0.014

0.016

0.018

0.020

A

B

C

Distance to edge (m)

Figure 2-4. Edge-interior gradients in microenvironment and soil nutrient parameters in Pampadum shola National Park (~132 ha). Error bars represent one standard error.

48

CHAPTER 3 SPECIES DISTRIBUTION PATTERNS IN SHOLA FRAGMENTS

Introduction

The creation of an edge in forested habitat alters both abiotic and biotic components of

ecosystems. Immediate changes to microenvironment conditions at the edge— higher air and soil

temperatures, lowered relative humidity and soil moisture— are a result of elevated insolation

and turbulence caused by an alteration of the boundary interface. Changes are manifested in

altered ecosystem processes: evapotranspiration, litter decomposition, nutrient cycling and

dispersal. These ecosystem processes interact iteratively through feedback loops with the biotic

component of ecosystems causing a shift in species composition through changes in recruitment,

growth, reproduction and mortality. Although edges affect the biotic and abiotic components, the

influence of abiotic processes diminishes as species adapt to altered conditions (Laurance et al.

2002). Biological responses to edge creation include elevated tree mortality (Williams-Linera

1990) and reduced canopy and sub-canopy cover (Laurance 1991). This opening of the canopy

manifests itself in increased seed rain from generalist species at the edge and an increase in

density of liana and understory species (Janzen 1983, Laurance 1991). These shifts in biota last

longer than microenvironment response to edge creation and may even be permanent. Spatially

too, the biotic response to edge creation is of a much larger magnitude than those of

microenvironment edge-interior gradients (i.e., high DEI, Harper et al. 2005). Long-term, large

scale manipulative experiments on Amazonian fragments reveal that while microenvironment

gradients extend up to 100m from the forest edge, edge effects on tree mortality can penetrate as

much as 300m from the edge (Laurance et al. 2002). Larger estimates of species response to

forest disturbance (~10 km) have also been attributed to edge creation (Curran et al. 1999).

49

However these might be an artifact of the collapse of predator populations in fragments rather

than edge effects per se (Ickes and Williamson 2000, Laurance 2000).

In tropical montane forests (TMF), species richness and vegetation structure can change

rapidly over short distances (Nadkarni and Wheelwright 2000). β-diversity was observed to

increase with an increase in elevation in TMF fragments in Mexico indicating that fragment

complementarities increased with increasing elevation (Williams-Linera et al. 2002). Species

within TMF fragments are often dispersal limited, such that late successional species probably

regenerate from seeds originating within the patch itself (del Castillo and Rios 2008). As a result

TMF fragments maybe characterized by high endemism and the number of unique species

assemblages is negatively correlated with fragment size (Young and Leon 1995, Ganeshaiah et

al. 1997). Smaller fragments are thus more likely to form unique habitats and, from a

conservation standpoint, are equally as important as larger fragments. Studies on species

response to a forest-grassland (or forest-pasture) edge in fragmented TMF ecosystems reveal

contrasting trends. For example, Jose et al. (1996) found trends in species distributions as a

function of distance from the edge in TMF fragments in southern India with sub-tropical and

temperate species near the edge while species with tropical affinities were located in interior

plots. On the other hand, studies in other TMF fragments report weak or complex distance to

edge trends in species diversity (Oosterhoorn and Kappelle 2000) and stem density (Young and

Keating 2001).Often though, TMF fragments may be a greater source of variation than distance

to edge per se (Williams-Linera 2003). Reduced canopy heights have also been recorded from

edge habitats in Costa-Rican TMF fragments. Combined with increased sub-canopy heights this

resulted in a simplification of vertical structure of edge habitat as compared to fragment interiors

(Oosterhoorn and Kappelle 2000).

50

The theory of island biogeography (IBT, Macarthur and Wilson 1963, 1967) offers an

elegant and robust explanation of the process of colonization and speciation on islands.

Numerous studies since have sought to apply the theory to fragmented and insular terrestrial

habitats (for a review see Laurance 2008). However, as Laurance (2008) indicates, system

properties associated with terrestrial fragments (e.g. non-random habitat conversion, edge effects

and matrix hostility) render IBT inadequate in explaining the effects of habitat fragmentation. In

terrestrial ecosystems, unlike true oceanic islands, the surrounding matrix has a significant

influence on the magnitude and direction of species responses to edge effects. In a study of

Amazonian forest fragments, fragments bordering cattle pastures (a high-contrast edge) had

higher tree mortality rates than low-contrast forest-secondary regrowth edges (Mesquita et al.

1999). In the Western Ghats, TMF (hereafter shola) occurs as insular forest fragments in a matrix

of grasslands separated by a stable, natural, high contrast edge. Since the edge is natural in

origin, we can expect equilibrium in species distribution patterns as a function of edge-related

processes. Although the hostility of the matrix, as observed in the response of species to the edge

(Shanker 2001) is often suggestive of true oceanic islands, it is unlikely that IBT alone will

sufficiently explain patterns observed in the ecosystem.

The study of created (or anthropogenic) edges in forested habitats often has direct

management implications. However, natural edges (such as those observed in the shola-grassland

ecosystem mosaic) allow us to observe patterns and processes operating at edges while isolating

human influence. Edge effects in naturally patchy ecosystems though may be suppressed or even

absent (Pavlacky and Anderson 2007). In a study of nest predation on TMF in Tanzania, contrary

to expectation, higher nest predation rates were reported for nests away from forest edges

(Carlson and Hartman 2001). Natural fragments may also maintain a high diversity of tree

51

species while vertebrate (especially mammal) diversity may not be comparable (Bowman and

Woinarski, 1994). Finally, edge-effect related processes are also dependent on patch size. As

fragment size reduces the proportional area under edge influence increases. This effectively

reduces the core-area of a fragment that land managers and conservationists are advised to

maximize by selecting for larger fragments (Gascon et al. 2000). However, smaller fragments

may serve as long-term refugia for threatened species (Turner and Corlett 1996), differ

compositionally from original forests (Tabarelli et al. 1999) and even contain unique species

assemblages not associated with larger fragments (Ganeshaiah et al. 1997).

This study was conducted in the shola-grassland ecosystem mosaic in southern India. I

quantified species distribution patterns as a function of distance from the edge. I hypothesize that

there will be edge-interior gradients in species distributions and expect abundance and

dominance patterns to be correlated with microenvironment and edaphic variables. Since edges

offer increased light transmittance to understory vegetation I expect edges to have greater

understory structural complexity and higher overstory basal area than interior plots. Further,

since the edge is a high contrast edge, I expect higher densities near the edge. Finally, since data

is available at a regional scale, I anticipate compositional differences between overstory species

in sub-montane and montane forests.

Methods

Study Area

Three study sites were selected in the Western Ghats in southern India within the Western

Ghats-Sri Lanka (WGSL) biodiversity hotspot; Biligiri Rangaswamy Temple (BRT) Wildlife

Sanctuary in the state of Karnataka and Eravikulam National Park and Pampadum shola National

Park in the state of Kerala ( Fig 2-1). BRT Wildlife Sanctuary (540km2), located at the eastern-

most edge of the Western Ghats (11°43’N to 12°08’ and 77° to 77°16’E), occurs at the

52

confluence of the Eastern Ghats and the Western Ghats. Soils in BRT have been classified as

Vertisols and Alfisols and described as shallow to moderately shallow with well drained gravelly

clay soils on the hills and ridges (Krishnaswamy 2004, NATMO 2009). Rainfall in BRT—

which is exposed to both southwestern and northeastern monsoons —ranges between 898 mm to

1750 mm and strong spatial trends have been observed based on topography and altitude

(Krishnasmamy et al. 2004). The dry season extends from November to May with February

recorded as the driest month. Vegetation in the sanctuary is represented by a diversity of forest

types ranging from dry scrub forests, dry and moist deciduous forests, evergreen forests and the

high-altitude TMF-grassland (or shola-grassland) ecosystem mosaic. Within BRT, three shola

fragments were identified as study sites (Figure 2-1, Table 3-1). Accessibility and the limited

availability of insular shola fragments restricted the number of fragments available in BRT.

Although shola vegetation is more extensive in the southern portion of the sanctuary,

(particularly near Jodigere shola) most of these areas were considered inappropriate for the

purpose of this study since they are contiguous with lowland forest and do not represent true

forest fragments.

Eravikulam National Park (ENP) is located in the Anaimalai Hills (10°05’ to 10°20’N and

77°0’ to 77°10’E) within the state of Kerala (Fig 2-1). The park (119 km2) consists of a base

plateau at an elevation of 2000m surrounded by peaks with a maximum elevation of 2695 m. The

mean maximum temperature recorded within ENP is 16.6 °C while the mean minimum is 6°C.

The warmest month in ENP is May with a mean maximum temperature of 24.1 °C and January

is the coldest month with a mean mimimum temperature 3 °C (Rice 1984). The soil had been

classified as Alfisols and described as Arachean igneous in origin consisting of granites and

gneisses (NATMO 2009). Soils are sandy clay, have moderate depth (30-100 cm) and are acidic

53

(ph 4.1-5.3). ENP receives rainfall approximately 5200 mm of rainfall annually from both the

southwest and the northeast monsoon with the former contributing as much as 85 % of the

annual rainfall. Although contiguous rainforest formations can also be found at lower elevations,

vegetation in the park is predominantly shola-grassland ecosystem mosaic consisting of rolling

grasslands interspersed with dense, insular shola fragments. Five shola fragments were selected

within ENP (Table 3-1).

Pampadum shola National Park located within Idukki district (approximately at 77°15’E

and 10°08’N) in the state of Kerala consists of a large shola patch (1.32 km2) contiguous with

lowland forest. Mean annual temperature in the adjacent Mannavan shola has been reported to be

approximately 20°C with mean minimum temperature in the coldest months (December and

January) reaching 5°C. Mean annual precipitation in Mannavan shola ranges between 2000-3000

mm. Data from PSNP were collected to compare patterns observed against smaller fragments

chosen in BRT and ENP.

Sampling Protocol

To investigate edge-interior gradients in species distributions in shola fragments, transects

were laid out perpendicular to the shola-grassland edge. Overstory structure was characterized in

terms of basal area and density of tree species in 10×5 m plots along each transect with the first

plot being established at the edge itself. Plots were laid out with the longer side (10 m) running

parallel to the edge to capture maximum variability along the edge-interior gradient. In each plot,

trees ≥ 5 cm dbh were identified to species and measured. Overstory species were assigned to

one of 5 diameter classes based on dbh (5-10 cm, 11-20 cm, 21-40 cm and > 40 cm). Nested 5×5

m sub-plots were established to identify and measure understory vegetation. For the purpose of

this study all vegetation ≥ 50cm in height was defined as understory vegetation. Understory

individuals were assigned to one of five height classes (50-99 cm, 100-149 cm, 150-199 cm,

54

200-249 cm and > 250 cm) based on maximum height (Hmax). Specimens were collected for

overstory and understory individuals that could not be identified confidently and were later

identified with the help of taxonomists at Ashoka Trust for Ecology and the Environment,

Bangalore. I also collected data on microenvironment (soil and air temperature, relative

humidity, light transmittance, soil moisture) and soil macronutrient (organic carbon, total

nitrogen, available phosphorous, potassium) variables from 10×5 m plots. Details on techniques

used to record microenvironment data and collect soil nutrient samples are presented earlier in

Chapter 2.

Successive plots along transects were separated by 5 m and extended till the middle of the

patch or to a distance of 35 m from the edge, whichever was shorter. Edge effects in shola

fragments are negligible at this distance (Jose et al. 1996). Thus, data are available from four

classes- 2.5 m, 12.5 m, 22.5 m and 32.5 m (as measured from the shola-grassland edge to the

mid-point of the plot)- along the edge-interior gradient. Replicate transects within a patch were

separated by at least 35 m though most often greater than 50 m.

Data Analyses

Vegetation Structure and Composition

I assessed plant species richness between fragments and completeness of species

inventories using a sample-based rarefaction (species accumulation curves) approach as

calculated by EstimateS (Colwell 2006). Species accumulation curves were calculated using the

Mao-Tau function (Colwell et al. 2004, Mao et al. 2005). Although the Mao-Tau function does

not require resampling runs to calculate species accumulation curves (Sobs), the sampling order

was randomized over 100 runs to compute singletons (species occurring in one sample) and

doubletons (species occurring in two samples). Species diversity was assessed using Fisher’s

alpha index and richness was estimated using the nonparametric Bootstrap estimator (Smith and

55

van Belle 1984) since it is well suited to quadrat data (Chao 2005). Moreover, it provided the

most stable and least biased estimate of species richness for overstory and understory species.

Species diversity was assessed across all fragments using Shannon-Weiner index (H') (Magurran

1988). Beta-diversity was computed using the Chao Abundance-based Sørensen similarity index.

These indices correct for the undersampling bias associated with traditional similarity indices

(e.g., Sørensen similarity index) and are especially suitable for use when rare species are

numerous (Chao et al. 2005). Fragment complementarity (C) between pairs of fragments was

assessed by calculating the Marczewski-Steinhaus distance (Colwell and Coddington 1994).

Complementarity varies from zero (when all species are common to both fragments) to unity

(when no species are common to both fragments). Regression analysis was used to test for trends

in overstory dominance and vertical structure of overstory (dbh distribution) and understory

(Hmax) as a function of distance to the shola-grassland edge. Soil samples were analyzed for

macronutrient concentration using standard techniques (see Chapter 2). A stepwise regression

was used to determine the microenvironment and edaphic variables determining overstory

dominance.

Edge-interior Gradients in Species Distributions

I quantified the variation in species compositional patterns as a function of distance to the

shola-grassland edge across fragments using non-metric multidimensional scaling (NMS) as

implemented in PC-ORD (McCune and Mefford 2006). NMS was run in the ‘medium’ mode

(200 iterations, 0.0001 instability criterion, 50 runs and four starting axes) of the autopilot

procedure using the Sørensen distance measure to calculate dissimilarity coefficients. A Monte-

Carlo randomization test (50 runs) was used to test the best solution against a randomized

dataset. A three dimensional solution yielding a stress significantly lower than obtained by

chance was selected for each of the overstory (p = 0.019) and understory datasets (p = 0.019). A

56

joint plot was constructed to explore the relationship between species and environmental

variables (soil temperature, air temperature, relative humidity, soil moisture, soil organic carbon,

total nitrogen, available phosphorous and potassium). Axis scores were correlated with

environmental measures to assess the ecological significance of axis in ordination space. For the

understory dataset light transmittance was included as an environmental variable.

Results

Vegetation Structure and Composition

I recorded 2379 individuals representing 111 species from 49 families in the nine

fragments studied. Fragment size varied from 0.2 ha to 132 ha (with most fragments ~1 ha). The

most abundant families are Lauraceae (19 species), Rubiaceae (18 species), Myrsinaceae (7

species), Myrtaceae and Oleaceae (5 species each). Species accumulation curves failed to reach

an asymptote for overstory (Figure 3-1a) and understory (Figure 3-1b) species. Curves for

singletons (species occurring once) did not decline for overstory species but declined marginally

for understory species indicating adequate sampling for rare species in understory vegetation.

Shannon-Weiner diversity indices (H') were 3.96 and 3.57 for overstory and understory species

respectively (Table 3-2). The Bootstrap estimator indicated that 12 additional overstory species

and 13 additional understory species are estimated as not being represented in the dataset

assuming that the selected fragments were representative of the shola ecosystem (Figure 3-1a, 3-

1b). On a landscape scale, shola fragments had high variation between fragments. This was

especially true for overstory species with complete separation often observed between mid

elevation and high elevation fragments (Chao-Sørsensen index: 0.0). For understory species

mean complementarity was lower (C = 0.85) than overstory (C = 0.90) and only one pair of

fragments (JG-MM) showed complete separation.

57

Edge-interior gradients in vegetation were largely absent from shola fragments. Although

edge plots (2.5m) on average had higher understory density than interior plots (Fig 3-2), no

statistically significant trends were observed along the edge-interior gradient (p = 0.190).

Structurally, there were no edge-interior trends across overstory dbh classes (p ≥ 0.202) and

trends were inconsistent. With increasing distance from edge, the total number of individuals

across diameter classes decreased in some fragments while other fragments saw an increase in

total number of individuals (Figure 3-3). Others still appeared to have a bimodal distribution

(MM, GU). Only in the largest fragment (PP) were the largest trees (> 40cm dbh) present across

all distances. In the understory, while the density of individuals in the smallest height class (50-

99cm) declined weakly (p = 0.079) along the edge-interior gradient, such trends were absent

across other height classes (p ≥ 0.581). Mean plot basal area for overstory species across

fragments varied between 15.5m2ha-1 and 78.5m2ha-1 (Table 3-1). However, no trends were

observed in overstory basal area along the edge-interior gradient (p = 0.897). Stepwise regression

revealed that soil temperature was significantly related to the variation in overstory dominance (p

= 0.039). However, this variation may be an effect of rather than a cause of the variation in

overstory dominance. No other microenvironment or soil variables sufficiently explained the

variation in overstory dominance.

Edge-interior Gradients in Species Distributions

For the overstory dataset NMS ordination produced three significant axes which accounted

for 57% of the variation between species in the ordination space and the original space. A stable

solution (instability = 0.00005) was reached in 200 iterations. NMS separated the plots

reasonably well in ordination space (Figure 3-4). The alignment of overstory plots was better in

terms of fragments than distance to edge classes with lower elevation shola fragments occupying

the ‘south eastern’ portion of the ordination cloud. Axis 2 showed a strong correlation with

58

relative humidity (r = 0.536) and soil moisture (r= 0.560) and slightly weaker correlations with

soil organic carbon (r = 0.458), total nitrogen (r = 0.467), available phosphorous (r = 0.350) and

potassium (r = 0.414). Axis 2 thus appears to represent soil nutrient availability well. Axis 3 was

weakly correlated with soil temperature (r = 0.284) and air temperature (r = 0.245) alone while

Axis 1 was not correlated with any of the environmental variables (Table 3-4).

NMS ordination indicated three axes as significant which accounted for 59.3% of the

variation between understory species. A stable solution was reached in 169 iterations (instability

< 0.00001). For understory and overstory datasets, final stress (which is a measure of the

dissimilarity between plots in ordination space and original space) was relatively high (18.13 for

overstory; 16.79 for understory). Similar to overstory, understory plots aligned better as

fragments than distance classes. However the separation between shola fragments based on

elevation was not seen. A separation between plots based on fragment size was observed with

larger shola fragments (PP and MM) occupying south-eastern portion of the ordination cloud

(Figure 3-5). Axis 1 was strongly correlated with the environmental variables relative humidity

(0.447) and soil moisture (r = 0.452) and weakly with soil organic carbon (r = 0.328). Axis 2 was

weakly correlated with air temperature (r = 348), soil temperature (r = 0.320) total nitrogen (r =

0.374) and available phosphorous (r = 0.383) while correlations with Axis 1 were weak.

Discussion

Patterns in small fragments may contrast from those observed in larger fragments and

small fragments may be dominated by edge habitat (Tabarelli et al. 1999, William-Linera et al.

1998). In this study, I observed an absence of edge-interior gradients in overstory dominance and

understory density in tropical montane forest (shola) fragments. Although this is typical of a high

contrast edge (low DEI, Harper et al. 2005), shola fragments are not ‘sealed’ by a wall of

vegetation across strata as seen at other high contrast edges (Strayer et al. 2003, Laurance et al

59

2002). I also observed that vertical complexity of the understory did not vary along the edge-

interior gradient. In contrast, a study on Costa Rican tropical montane forest fragments observed

a decrease in density and simplification in understory structure with increasing distance from

edge (Oosterhoorn and Kappelle 2000). Departures in my dataset from patterns observed in

neotropical montane fragments could be due to differences in the origin of tropical montane

fragments studied. While neotropical fragments (such as those in Costa Rica) are often

anthropogenic in origin, tropical montane forest fragments in the Western Ghats are natural in

origin (Sukumar et al. 1993). Other naturally fragmented tropical montane forests too have been

observed to be characterized by atypical edge effects (Carlson and Hartman 2001). With

increasing fragment age, species equilibrate with the microenvironment leading to a softening of

edge effects (Harper et al. 2005). Structurally, mean plot basal area ranged between 15.5 m2ha-1

to 78.5 m2ha-1, with high elevation shola fragments generally supporting a much higher basal

area (>50 m2ha-1) than mid-elevation fragments. The basal area values reported are within the

range of those reported for other TMF fragments (Williams-Linera 2002) but higher than those

reported from other shola fragments (Jose et al. 1994). The variation in overstory basal area

could be significantly, albeit weakly, explained only by soil temperature. Soil temperature has

been shown to be a good indicator of the penetration of edge effect on microclimatic variables

since it is coupled with changes in insulating overstory vegetation (Laurance and Yensen 1991).

I found 111 species from 49 families across the nine fragments studied which is higher

than other studies from shola fragments (Jose et al. 1994, Davidar et al. 2007) possibly because

data were collected from mid-elevation as well as high elevation shola fragments. Bootstrap

species richness estimators also indicated that 12 overstory species and 13 understory species

were not represented in our dataset. My estimates for species diversity (H') are within the range

60

of reported values for shola fragments (Jose et al. 1994, Menon 2001) and other tropical montane

forest fragments (Oosterhoorn and Kappelle 2000).

Ordination of species data indicates that no edge-interior gradients were apparent in shola-

(TMF) fragments for understory and overstory data. However, a separation of plots based on

elevation was noticed among overstory plots. Plots at elevations lower than 1500 m (BRT) were

distinct from those at elevations greater than 1700 m in species ordination space. This is also

supported by patterns observed in the diameter distribution of overstory species along the edge-

interior gradient (Figure 3-3). Data from mid-elevation shola fragments (GU, JG, KR) reveal

that, the smallest trees (5-10cm dbh) were often absent across all distances leading to a distinct

separation between overstory and understory strata. Such differentiation is typically not

associated with tropical montane forest (Shola) fragments in the Western Ghats and is indeed

distinct from patterns observed in high elevation fragments in this study. Although this is not

surprising, numerous studies have characterized lower elevation forest fragments as shola or

tropical montane forests both within BRT (Ganeshaiah et al. 1997, Murali et al. 1998) and in

other medium elevation ecosystems within the Western Ghats (Sudhakara, 2001). While medium

elevation forests in BRT have previously been recognized as distinct from lowland evergreen

forests (Murali et al. 1998), this study clearly demonstrated their separation from ‘true’ high

elevation montane forest in the Western Ghats.

Soil nutrient availability was well represented in the overstory ordination cloud and

assumed greater significance for high elevation fragments (Figure 3-4). Globally, tropical

montane forests are found on nutritionally poor soils (Hamilton et al. 1995). In my study too, soil

nutrient availability became significant only in high elevation shola fragments. The separation of

fragments in general and mid-elevation fragments from high elevation fragments in particular is

61

also expressed in fragment complementarity scores and the Chao Abundance-based Sørensen

similarity indices for our overstory dataset. While across the dataset fragment complementarity is

high, separation between mid-elevation fragments and high elevation fragments is often

complete (C = 1.00).

For understory vegetation, fragment size was a greater source of variation than elevational

differences. While Axis 1 was correlated with relative humidity, soil moisture and organic

carbon, Axis 2 was weakly correlated with air temperature and negatively with soil temperature

and available phosphorous. Thus, in larger fragments microenvironment and soil nutrient

availability determined the distribution of species whereas in smaller fragments only air

temperature appeared to determine variation in species distributions. Light transmittance was

very weakly correlated with Axis 1. Since shola fragments are deeply shaded, this clearly

demonstrates that understory species are well adapted to low light conditions as seen in other

tropical montane forest fragments (Williams-Linera 2003)

Conclusion

This study, to my knowledge, represents the first attempt to investigate edge-interior

gradients in species richness and distributions across multiple tropical montane forest (shola)

fragments in the Western Ghats. Previous studies have investigated similar gradients in a single

fragment (Jose et al. 1996). Patterns observed in species density and structure are indicative of a

stable, mature shola-grassland edge. Across fragments, differences in overstory species

distributions were seen based on elevation while differences in understory species distributions

were based on fragment size. In both instances species distributions were not determined by edge

effects. Such distinct elevation based separation of fragments highlights the need to sub-classify

tropical montane forests as seen in other tropical montane forest ecosystems (Bruijzneel and

Hamilton 2000). From a management standpoint, the high degree of complementarity between

62

fragments exhibited by the ecosystem emphasizes the need to conserve the present extent of the

landscape and restore disturbed fragments where possible. My designed preponderance to

smaller fragments limits extensive characterization of fragments using fragment metrics (e.g.

core area, edge/area ratio, edge length). Although such an analysis would add invaluable

knowledge to the existing study, investigative efforts are probably best directed to filling the

current void of process based studies in the ecosystem. As stable, natural fragments in a hostile

grassland matrix, the study of shola fragments offers tremendous opportunities to study

fragmentation patterns and processes.

63

Table 3-1. Location and vegetation characteristics of tropical montane forest (shola) fragments studied across the Western Ghats, southern India.

Fragment (shola name)

Fragment code

Study site

Fragment area (ha)

Elevation (m)

Species richness Species diversity (H')

Overstory basal area (m2 ha-1)

Understory density (indiv ha-1) OS* US† OS* US†

Karadishonebetta KR BRT 1311 9 15 2.09 2.40 15.47 2120 Gummanebetta GU BRT 1.73 1433 11 21 2.00 2.52 44.47 2288 Jodigere JG BRT 0.20 1550 3 9 0.62 1.32 27.06 4450 Suicide point SP ENP 1.24 1810 17 20 2.62 2.83 38.59 1900 Chinnanaimudi CA ENP 1.66 2000 26 24 2.96 1.60 78.51 6956 Kolathan KS ENP 1.58 2205 16 21 2.41 2.37 69.06 5070 Pusinambara PS ENP 1.75 2222 17 14 2.64 1.99 59.43 7900 Mukkal mile MM ENP 13.42 2066 15 22 2.40 2.64 67.96 7574 Pampadum PP PNSP 132.00 2005 23 22 2.70 2.12 70.10 17300 *OS: overstory species (≥ 5 cm dbh) †US: understory species (≥ 50 cm in height but ≤ 5 cm dbh)

64

Table 3-2. Species richness, dominance and diversity estimates in the shola-grassland ecosystem mosaic in the Western Ghats, southern India.

Area

Elevation range (m)

Basal Area (m2ha-1)

Density (indiv ha-1)

Species richness

Species diversity (H')

Reference

Overstory ENP+BRT+PSNP 52.29 1711 77 3.96 This study Eravikulam 48.00‡ 1884 53 4.86* Jose et al. 1994 Mannavan shola 1550-1750 372.44 16710 55 4.71 Sudhakara 2001 Pampadum 1750-1950 114.94 21894 44 4.53 Sudhakara 2001 Eravikulam 2100 59.37 2.50 Swarupanandan et al. 2001 Silent Valley 1950-2300 - - - 2.73 Nair and Baburaj 2001 Eravikulam 2050-2200 - - - 2.71 Nair and Baburaj 2001 Kukkal RF 58.47 451 45 12.1† Davidar et al. 2007 Understory BRT+ENP+PSNP 1500-2200 - 6173 83 3.57 This study Eravikulam - - 65 4.86* Jose et al. 1994 †Fisher’s Alpha, *H' calculated for cumulatively for overstory and understory strata, ‡Overstory ≥ 10cm gbh, Overstory ≥ 5cm dbh for our study, ≥10 cm for other studies, understory ≥ 50cm for our study. BRT, ENP, PSNP represent BRT Wildlife Sanctuary, Eravikulam National Park and Pampadum shola National Park respectively.

65

Table 3-3. Similarity and complementarity between tropical montane forest (shola) fragments in the Western Ghats, southern India. The Chao- Sørsensen abundance based index varies from 0 (completely dissimilar) to 1 (completely similar) whereas Complementarity varies from 0 (all species shared) to 1 (no shared species).

Fragments Chao-Sørsensen index Complementarity GU JG SP CA KS PS MM PP† GU JG SP CA KS PS MM PP† Overstory KR 0.78 0.11 0 0.06 0 0.05 0 0.10 0.66 0.90 1.00 0.97 1.00 0.95 1.00 0.96 GU 0.59 0 0.05 0.06 0 0.03 0.19 0.72 1.00 0.97 0.96 1.00 0.96 0.93 JG 0 0 0.08 0 0 0.04 1.00 1.00 0.94 1.00 1.00 0.96 SP 0.82 0.23 0.28 0.22 0.29 0.61 0.90 0.86 0.85 0.88 CA 0.08 0.44 0.20 0.41 0.97 0.86 0.92 0.80 KS 0.60 0.56 0.20 0.73 0.81 0.91 PS 0.42 0.19 0.76 0.91 MM 0.13 0.94 Understory KR 0.76 0.63 0.31 0.05 0.54 0.03 0.03 0.03 0.61 0.66 0.90 0.94 0.79 0.97 0.97 0.94 GU 0.95 0.10 0.05 0.25 0.15 0.07 0.02 0.69 0.94 0.92 0.87 0.95 0.97 0.95 JG 0.05 0.06 0.20 0.12 0 0.18 0.96 0.90 0.85 0.92 1.00 0.89 SP 0.76 0.12 0.11 0.41 0.50 0.70 0.86 0.92 0.80 0.80 CA 0.10 0.11 0.06 0.18 0.88 0.84 0.85 0.78 KS 0.55 0.19 0.17 0.75 0.87 0.90 PS 0.56 0.23 0.65 0.83 MM 0.49 0.84 Karadishonebetta (KR), Gummanebetta (GU) and Jodigere (JG) sholas were located with BRT Wildlife Sanctuary (BRT). Suicide point (SP), Chinnanaimudi (CA), Kolathan (KS), Pusinambara (PS) and Mukkal mile sholas were located within Eravikulam National Park (ENP). Pampadum shola (PP) is located within Pampadum shola national park (PSNP). †PP represents the single large fragment (~132 ha) as most other fragments were smaller (0.2-13 ha).

66

Table 3-4. Pearson correlation coefficients (r2) for environmental variables with axis scores of

NMS ordination. Correlation coefficients accounting for more than 30 % of the variation in individual factors are highlighted in bold.

Environmental variable Axis1 Axis 2 Axis 3 Overstory Soil temperature -0.077 0.150 0.284 Air temperature -0.124 0.231 0.245 Relative humidity 0.158 0.536 0.295 Soil moisture P 0.020 0.560 0.232 Organic carbon -0.019 0.458 0.341 Total nitrogen -0.013 0.467 0.317 Available phosphorous -0.001 0.350 0.194 Available potassium 0.207 0.414 0.160 Understory Light transmittance -0.268 0.138 -0.124 Soil temperature -0.082 -0.320 -0.211 Air temperature -0.005 0.348 -0.246 Relative humidity 0.447 -0.219 0.137 Soil moisture 0.452 -0.252 0.001 Organic carbon 0.328 -0.285 -0.164 Total nitrogen 0.270 -0.374 -0.010 Available phosphorous 0.155 -0.383 -0.086 Available potassium 0.244 -0.291 0.009

67

Ric

hnes

s

0

20

40

60

80

100

120Sobs (Mao Tau) BootstrapSingletonsDoubletons

Cumulative number of plots

0 10 20 30 40

Ric

hnes

s

0

20

40

60

80

100

Figure 3-1. Species accumulation curves (Sobs), species richness estimates (Bootstrap) and rare species (singletons and doubletons) for (a) overstory and (b) understory species from tropical montane forest (shola) fragments in the Western Ghats, southern India. Species data were pooled across 4 distances within the 9 fragments (n=35).

a

b

68

Distance from edge (m)

2.5 12.5 22.5 32.5

No.

of i

ndiv

idua

ls

0

10

20

30

40

5050-99100-149 150-199 200-249 > 250

Figure 3-2. Mean understory density (number of individuals/plot) along the edge-interior gradient in tropical montane forest (shola) fragments (n = 9). Classes represent maximum height (Hmax) of understory individuals recorded in centimeters.

69

No.

of i

ndiv

idua

ls (n

)0

2

4

6

8

Distance to edge (m)2.5 12.5 22.5

No.

of

indi

vidu

als (

n)

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

No.

of i

ndiv

idua

ls (n

)

0

4

8

12

Distance to edge (m)2.5 12.5 22.5 32.5

No.

of i

ndiv

idua

ls (n

)

0

4

8

12

16

No.

of i

ndiv

idua

ls (n

)

0

2

4

6

No.

of i

ndiv

idua

ls (n

)

0

4

8

12

16

5-10 cm10-20 cm20-40 cm>40 cm

No.

of i

ndiv

idua

ls (n

)

0

6

12

18

No.

of i

ndiv

idua

ls (n

)

0

6

12

18

Distance to edge (m)2.5 12.5 22.5 32.5

No.

of i

ndiv

idua

ls (n

)

0

4

8

12

GU

JG

SP

CA PP

MM

KS

PS

KR

Figure 3-3. Edge-interior gradients in diameter distribution of overstory tropical montane forest (shola) species across fragments. Individuals are for each plot averaged across transects within a patch. Karadishonebetta (KR), Gummanebetta (GU) and Jodigere (JG) sholas were located within BRT Wildlife Sanctuary. Suicide point shola (SP), Chinnanaimudi (CA), Pusinambara (PS), Kolathan (KS) and Mukkal mile sholas are located within Eravikulam National Park while Pampadum shola (PP) was located within Pampadum shola National Park

70

Figure 3-4. Non-metric multi-dimensional scaling (NMS) ordination of overstory data. KR, GU

and JG represent mid elevation shola (TMF) fragments while the other fragments represent high elevation shola fragments. Points represent plots showing their distribution across fragments. Arrows indicate the strength and direction of correlations with environmental variables (moist: soil moisture, nit: total nitrogen, orgc: organic carbon, relhum: relative humidity) on the two axes. Letters denote fragment names as in Table 3-1 and numbers indicate distance from edge classes.

71

Figure 3-5. Non-metric multi-dimensional scaling (NMS) ordination of understory data. Points

represent plots showing their distribution across fragments. Arrows indicate the strength and direction of correlations with environmental variables (moist: soil moisture, relhum: relative humidity)on the two axes. Letters denote fragment names as in Table 3-1 and numbers indicate distance from edge classes.

72

CHAPTER 4 EFFECT OF TOPOGRAPHY ON THE DISTRIBUTION OF SHOLA FRAGMENTS: A

PREDICTIVE MODELING APPROACH

Introduction

The influence of topography on vegetation across biomes is well known. For example,

increasing elevation brings with it a change in plant communities. For a given latitude, a change

in elevation implies a shift ‘upward’ in species composition and assemblages progressively to

alpine or boreal communities. Although the influence of other topographical variables on

processes has been known for a while (e.g. aspect, Merz 1953), research continues to be directed

at it. Recently, topographical variables have been used to determine species richness (Hofer et al.

2008), regional biodiversity patterns (Coblentz and Ritters 2004), forest canopy health (Stone et

al. 2008), species distributions (Warren 2008) and gradients in exotic species (Qian 2008). As

Coblentz and Riitters (2004) identify, this is driven in part by the global availability of high

resolution elevation data (e.g., GLCF, www.landcover.org) coupled with the availability of high

speed computing platforms. Within the Western Ghats, the potential of geographic information

systems (GIS) and remotely sensed data in tropical ecosystems (such as the Western Ghats

hotspot) from the perspective of a landscape was recognized early (Menon and Bawa 1997).

Studies have since utilized these tools to highlight gaps in the network of conservation areas

(Ramesh et al. 1997), quantify threats (Barve et al. 2005) and prioritize areas for conservation

(Das et al. 2006)

Topographical based analyses are especially insightful in mountainous regions,

presumably due to the extent of topography related heterogeneity. In a study on tropical montane

forests (TMF) in Ethiopia, topography related variation in soil physicochemical characteristics

and proximity to water accounted for differences in forest communities (Aerts et al. 2006).

Similarly, a study in an Indonesian TMF revealed that topography induced variation in relative

73

humidity and wind velocity influenced species presence (Sri-Ngernyuang et al. 2003). While it is

well known that the abiotic environment (e.g., temperature, soil moisture) of a plant is an

important determinant in it’s ability to establish, grow and reproduce (Gurevitch et al. 2002), an

extensive determination of the abiotic environmental variables at the scale of a landscape (or

larger) may not be feasible. However, the relative ease of determination of topographical

variables using extensive, large scale elevation data makes these variables a suitable surrogate

for abiotic environmental conditions.

In tropical montane forests (TMF) topography can result in the creation and maintenance

of abrupt boundaries or edges. For instance, the Cordilleras in the Dominican Republic exhibit a

discrete TMF-pine forest ecotone maintained by a combination of fire and frost (Martin et al.

2007). Similarly, abrupt boundaries have been observed for tree islands in the Ecuadorian Andes

(Coblentz and Keating 2008). In the Western Ghats, tropical montane forests (or sholas) consist

of insular forest fragments in a matrix of grasslands. Carbon isotope analysis of peat samples

from the ecosystem mosaic reveal a cyclical shift in dominant vegetation type (forest or

grassland) corresponding to glacial cycles (Sukumar et al. 1993). The edge is thus assumed to be

natural in origin although its persistence may be due to a combination of natural (fire, frost,

edaphic) and anthropogenic (fire) factors. Shola fragments are characterized by high species

diversity (α-diversity) and endemism (Jose et al. 1994, Nair and Menon 2001). In addition,

fragments have been shown to have a high degree of complementarity (high diversity among

fragments, Chapter 3). Since patterns in plant distributions are non-random (Grieg-Smith 1979),

predicting their occurrence is dependent on quantifying ecological variables controlling

incidence (Franklin 1995). In this study, the influence of topographical variables on determining

the presence of shola fragments was tested. A multiple logistic regression approach was used to

74

predict presence or absence of insular fragments in the matrix of grasslands in two disjunct study

sites in southern India using elevation data.

Since the southwest monsoon contributes the bulk of the rainfall in both our study sites, I

hypothesize that the presence of fragments will be negatively correlated with eastern aspects

(hereafter Eastness). Further, I expect weak or no correlations with northern aspects (hereafter

Northness), since N-S aspects are not expected to be significant at the low latitudes of our study

sites. Preliminary field observations indicate that shola fragments are more likely to be found on

steeper, wetter slopes. Accordingly, I hypothesize that presence of shola fragments will be

positively correlated with wetness index (Moore et al. 1993, Gessler et al. 1995) and curvature of

the slope (McNab 1989)

At a regional scale, I expect differences between study sites at a macro scale (as defined by

Delcourt and Delcourt 1998). Although topographical variables are assumed to be acting

similarly on the same ecosystem mosaic, differences in site history may result in differences in

the response to topographical variables (Bader and Ruijten 2008).

Methods

Study Area

Two distinct areas were selected for the purpose of our study. Eravikulam National Park

(ENP) is located in the Anaimalai Hills within the state of Kerala (Fig 4-1). The park consists of

a base plateau at an elevation of 2000m surrounded by peaks with a maximum elevation of

2695m at Anaimudi. The soil had been classified as Vertisols and described as Arachaen igneous

in origin and consisting of granites and gneisses. Soils are sandy clay, have moderate depth (30-

100cm) and are acidic (ph 4.1-5.3). The mean maximum temperature recorded within ENP is

16.6 °C while the mean minimum temperature is 6°C. ENP soils had been classified as Alfisols

consist of granites and gneisses (NATMO 2009) are sandy clay, have moderate depth (30-100

75

cm) and are acidic (ph 4.1-5.3). ENP receives rainfall approximately 5200 mm of rainfall

annually from both the southwest and the northeast monsoon with the former contributing as

much as 85 % of the annual rainfall with a brief dry period (January to April). Although

contiguous rainforest formations can also be found at lower elevations, vegetation in the park is

predominantly shola-grassland ecosystem mosaic consisting of rolling grasslands interspersed

with dense, insular shola fragments. Additionally, exposed rock formations and cliffs and lower

elevation forests occupy 5.9% and 8.5% of the total area of ENP (Menon 2001).

The second study area is the Nilgiris (NIL) which is located in the state of Tamilnadu and

includes the Mukurthi National Park in the Nilgiri Biosphere Reserve. Details on the second

study site can be found in Zarri et al. (2008). Together our two study sites represent the two

largest contiguous extents of the high elevation shola-grassland ecosystem mosaic.

Imagery Pre-processing and Data Preparation

Elevation was derived by digitizing topographical maps at a scale of 1:50,000 scale (58

F/4, 58 F/3, Survey of India) at 20 m intervals for Eravikulam National Park (ArcGIS 9.2) to

generate a digital elevation model (DEM). Remotely sensed ETM+ imagery (acquisition date:

January 14, 2001) was used to identify and digitize shola fragments in Eravikulam National Park.

Images were pan-sharpened and a PCA low-pass filter was used to enable clearer delineation of

shola fragments from grasslands (ERDAS Imagine 9.2). Similarly a DEM (at 30 m contour

intervals) and a landscape map were generated for the Nilgiri Hills (see Zarri et al. 2008 for

details). In both these areas, these contour intervals represent the finest scale of elevation

information available. In addition to elevation, all other topographic covariate layers were

developed using DEMs (Table 4-1). A grid of random points was generated for Eravikulam

(1341 points) and the Nilgiris (2188 points). Shola points were identified by overlaying the

points with digitized shola fragments and declaring all other points as non-shola. In order to

76

further distinguish shola fragments from non-shola forest types, all points below 1700m were

masked and excluded from further data analysis. Estimates of topographic covariates were

extracted for these points (ENP: 401 shola, 940 non-shola; NIL: 666 shola, 1522 non-shola).

The DEM(s) were used to derive the following topographical variables: slope, aspect

(eastness and northness), curvature of the slope and wetness index (Table 4-1). The wetness

index (also termed Compound Topographical Index) combines topographical variables with

resource variables (moisture) while the other variables represent direct topographical variables.

Curvature of the slope was calculated using ArcGIS 9.2. A positive curvature indicates that the

surface is downwardly convex at that cell while a negative curvature indicates that the surface is

upwardly concave at that cell. A value of zero indicates that the surface is flat.

Predictive Modeling

In order to quantify topography dependent variables that influence the distribution of

sholas in the ecosystem mosaic 3 models were generated. For the first model, the data for the two

areas were combined and a dummy variable was introduced to test for differences between sites.

Since these two areas represent disjunct ecosystems (distance between ENP-NIL ~ 120km), the

dummy variable would test for regional differences. A multiple logistic regression model was

used for deriving estimates of the coefficients for topographical parameters and area under the

receiver operating characteristic curve (AuC). The AuC value is the proportion of randomly

drawn pairs of shola and non-shola points that are correctly classified using the multiple logistic

regression model. A value of 0.50 therefore indicates a random process (i.e. there is an equal

probability of the point being shola or non-shola) whereas a value ≥ 0.50 indicates a non-random

process. Seventy-five percent of the data were randomly selected for training the model and 25%

were retained as a cross-validation set. This process was repeated to 1000 iterations and a dataset

containing parameter coefficients and AuC estimates for each iteration was created. Two-tailed t-

77

tests were used to test for a) AuC estimates to be significantly different from random (AuC >

0.50) and b) significance of parameter coefficients (βi > 0). We used R programming language

throughout (R, 2008).

Two additional models were developed using our data subsets (ENP and NIL) to cross

validate each other. This was done to obtain conservative estimates of our parameter coefficients

and increase the robustness of our model. For the second model, the ENP data subset was used to

train the model and this was used to predict the presence of shola fragments in NIL. Similarly,

for the third model, the NIL data subset was used to train the model and this was used to predict

the presence of shola fragments in ENP. As with Model 1 (the combined model) a multiple

logistic regression model was used and repeated to 1000 iterations on the second and third model

to obtain topographical coefficients and AuC estimates for NIL and ENP respectively. AuC

values from Models 2 and 3 were then compared using a t-test (p < 0.05). When the two models

are compared, the data subset for the model with a significantly larger AuC was declared as a

better predictor of the presence of shola fragments based on topographical variables. Two-tailed

t-tests were used to compare the strength and direction of topographical parameter coefficients

across Models 2 and 3.

Results

Combined Model

For the combined model, topographical variables significantly predict whether the

vegetation type is shola or non-shola. The distribution of shola fragments was significantly

different from random as indicated by the area under the receiver operating curve (AuC = 0.714,

SE 0.0005, p < 0.0001). All topographical variables were highly significant (p < 0.0001, Table 4-

2). Coefficients for northness were positive while those of eastness were negative as related to

the presence of shola fragments (p < 0.0001). Coefficients for wetness index, slope and elevation

78

were positive and negative for curvature of the slope (for all p < 0.0001). Significant differences

were also seen between the study sites (p < 0.0001) with the Nilgiris (NIL) data subset predicting

the presence of shola fragments in Eravikulam (ENP) better than the use of the ENP dataset to

predict shola fragments in the Nilgiris (p < 0.0001).

Data Subsets

The Nilgiris (NIL) data subset was able to distinguish shola fragments from non-shola in

Eravikulam better (AuCNIL = 0.624) than vice versa (AuCERV= 0.709). While all parameter

coefficients from both models differed in magnitude, they did not differ in direction (Table 4-2,

Figure 4-2 to Figure 4-7). For both models, aspect (as quantified in eastness and northness)

strongly determined the presence of shola fragments and to a lesser extent wetness index (βCTI)

and curvature of the slope (βCUR).

Discussion

Results from our multiple logistic regression models clearly indicate the influence of

topography induced variation on the location of tropical montane forest (shola) fragments

relative to the surrounding non-shola matrix. AuC estimates provide statistical validation for the

strong non-random pattern of shola fragments in the mosaic. The effect of elevation on

determining the presence of shola fragments was unexpected in all our models a priori. This

could be due to our exclusion of points below 1700 m. However, at elevations below 1700 m,

separation of shola vegetation from evergreen forest may require a more intensive landcover

classification model which was beyond the scope of this investigation. Shola fragments below

1700 m (mid-elevation sholas) have also been shown to differ compositionally and structurally

from high elevation shola fragments and might warrant a formal partitioning into forest sub-

types (Chapter 3). Digitized shola fragments in ENP and field observations indicate the presence

of fragments (as large as 2 ha) at 2500 m and absence only from the highest peaks, possibly due

79

to the extremely steep terrain on those peaks (e.g. Anaimudi). The topographical limitation on

elevation in the Western Ghats prohibits the formation of true tree line as seen in other tropical

montane forest ecosystems (Troll 1973, Bader and Ruijten 2008). This might explain the lack of

strength associated with elevation in our models.

Shola fragments were also likely to be found on northern and western aspects. The

proportion of rainfall attributed to the southwestern monsoon in Eravikulam (ENP) and the

Nilgiris (NIL) might explain the preponderance of shola fragments on western slopes which

support my hypothesis. The dependence of shola fragments on the availability of soil moisture is

also reflected in the direction of the response to the wetness index (Compound Topographical

Index, CTI). The response to wetness index is significant as topographical variables that combine

a measure of topography with resource variable (such as CTI) have been shown to predict

patterns in vegetation better than those that measure topography alone (Franklin et al. 2000). The

significant bias of shola fragments to north-facing slopes was unexpected and contrary to our

hypothesis. Aspect induced variations in solar exposure along a north-south gradient are unlikely

to be significant at the latitude associated with the study sites (10°05’-11°32’). During the

months corresponding to the southwestern monsoon (May-October) significant differences in

wind exposure are likely as southern aspects (exposed to monsoonal winds) are likely to have

higher exposure to wind than northern aspects. This might limit the establishment of tree cover

(shola fragments). However evidence for wind-exposure related restriction of shola fragments in

this study is tenuous at best. Bader & Ruijten (2008) suggest that in addition to wind, diurnal

variation in radiation received might cause aspect-related differences in the tree line in Andean

tropical montane forest They surmise that more radiation is received earlier in the day (i.e.

eastern aspects) which ‘causes photoinhibition in maladapted plants’. Lack of data on diurnal

80

variation in intensity of radiation from our study however precludes me from making such

conclusions. However, this might be worth investigating in the system.

Curvature of the slope (or Terrain shape index) is a measure of the shape of the landform.

For shola fragments, presence of shola fragments was significantly linked to negative

coefficients which are indicative of concave slopes (see Equation 4-1). This is in keeping with

field observations of sholas and provides proof for my hypothesis. Similarly, the coefficient for

slopes was positive, i.e. sholas were present on steeper, more inaccessible terrain.

Comparison of Data Subsets

The use of the Nilgiris (NIL) dataset to predict the presence of shola fragments in

Eravikulam (ENP) produced a better model than vice versa. However, the coefficient of

topographical parameters was stronger for the NIL model. This can be explained by differences

in the history of the study sites as Bader and Ruitjen (2007) suggested from their study in

Andean tropical montane forests. Among sites in this study, ENP represents the better protected

of the two sites having received formal protection from colonial times and was used as a game

reserve by colonial settlers. Further, unlike tropical montane forests in other parts of the Western

Ghats (including the Nilgiris) colonial management of the grasslands (and indeed even the

sholas) in ENP did not involve extensive afforestation with exotic plantations. While the

influence of topographical variables on shola presence/absence in ENP (and consequently using

that data subset to predict presence/absence of fragments) is more indicative of processes with

limited human influence, both sites are not completely free of human influence.

Implications for the Shola-grassland Edge

Historically, theories on the persistence of the shola-grassland edge in the ecosystem

mosaic include frost, fire and edaphic limitation. While shola and grassland soils have been

shown to exhibit little variation in soil physicochemical properties and soil depth (Table 1-1,

81

Chapter 2), topographical variables such as wetness index can also provide insights into patterns

of fire and frost. Areas with a high wetness index also indicate wetter areas that are likely to be

precluded by fire, protecting tropical montane fragments. Studies have shown that patterns in

water accumulation and frost occurrence are often similar (Blennow 1998). Areas with a high

wetness index (depressions etc.) are likely to have higher occurrence of frosts reducing the

potential for shola fragment persistence. This is in contrast to the shola-grassland ecosystem

mosaic where sholas are located within depressions. Thus, the positive correlation of wetness

index with the presence of shola fragments indicates that, in addition to the strong linkage to

hydrological regulation, among fire and frost, fire rather than frost might be responsible for the

maintenance of the edge. Patterns in exposure to wind too might be responsible for the edge.

Conclusion

A common problem associated with predictive modeling is the lack of discrete boundaries

between habitat patches and high variation within patches (Hofer et al. 2008). The sharp edge

between shola fragments and grasslands in the shola-grassland ecosystem mosaic can therefore

provide significant insights without the limitations associated with other systems. Tropical

montane forest (shola) fragments in the Western Ghats have a high variation among fragments.

Further edge-interior studies reveal that fragments show little variation along edge-interior

gradients (or within fragments, see Chapter 3). This study provides unambiguous support for the

maintenance of the edge by fire rather than frost in addition to strong linkages to hydrological

regulation. Predictive models are useful tools for restoration efforts and at larger scales can

provide inputs into existing restoration efforts. The use of topographical variables to predict the

presence of shola fragments represents a subset of static, environmental sorting variables

influencing the distribution of shola fragments. The inclusion of dynamic variables such as fire

82

history and regime and biotic interactions (e.g. topography induced dispersal limitation) might

provide further insights and enhance our understanding of the shola-grassland ecosystem mosaic.

83

Table 4-1. Topographical variables used to predict presence of shola fragments in the Western Ghats, southern India.

Model parameter Code Description Elevation ELE Slope SLO Northness(NOR) NOR sin (Aspect) Eastness (EAS) EAS cos (Aspect) Curvature of the slope CUR Wetness index CTI ln (As / (tan(β)) †‡ Location Code LOC Boolean (ENP = 1, NIL = 0) As: Flow accumulation grid, β: Slope (radians), † Moore et al. 1993., ‡Gessler et al. 1995 Table 4-2. Topographical parameter coefficients and AuC estimates for predictive models.

Coefficients represent mean values from 1000 runs. Training datasets for the models were: Pooled data from the Eravikulam (ENP) and Nilgiris (NIL) data subsets for the Combined model; ENP for the NIL model and NIL for the ENP model. All parameter coefficients were significant at p < 0.0001.

Model output Model type (number)

Combined NIL ENP (1) (2) (3)

AuC 0.71 0.62 0.70 Parameter coefficients Intercept (INT) -8.2730 -9.6763 -6.7185 Elevation (ELE) 0.0032 0.0038 0.0024 Slope (SLO) 0.0234 0.0459 -0.0023 Northness (NOR) 0.3895 0.5703 0.3327 Eastness (EAS) -0.1324 -0.2673 -0.1080 Curvature of the slope (CUR) -0.0416 -0.0634 -0.0365 Wetness index (CTI) 0.0582 0.0215 0.0914 Location (LOC) -0.2786 - -

84

Figure 4-1. Location of study sites in the Western Ghats. Two study sites were selected for the

study are (A), Nilgiris (NIL) and (B), Eravikulam National Park (ENP), For inset maps, 1cm equals 5.5 kilometers

85

Figure 4-2. Histogram of Intercept coefficients (INTC) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets.

Superimposed curves represent normal curves.

86

Figure 4-3. Histogram of Elevation coefficients (DEM) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets.

Superimposed curves represent normal curves.

87

Figure 4-4. Histogram of Slope coefficients for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

88

Figure 4-5. Histogram of Northness coefficients for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

89

Figure 4-6. Histogram of Eastness coefficients for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

90

Figure 4-7. Histogram of Curvature of the Slope coefficients (CURV) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

91

CHAPTER 5 SUMMARY AND CONCLUSION

Tropical montane forest (sholas) in the Western Ghats in southern India consist of insular

forest fragments in a matrix of grasslands. The boundary between the sholas and the grasslands is

thought to be natural in origin and is relatively stable. As with other tropical montane forest

(TMF) systems, the sholas consist of dense stands of stunted trees (~15m). Physiognomically,

branch architecture is twisted or gnarled and branches are often covered in epiphytic growth.

Given the absence of anthropogenically induced fragmentation in the mosaic, the primary aim of

this study was to investigate edge effects in shola fragments. Nine fragments across three study

sites were selected in the Western Ghats to represent a range of elevations (1400m-2200m) and

fragment sizes (0.2ha-132ha). Data on microenvironment (soil and air temperature, relative

humidity, light transmittance. soil moisture), soil macronutrient (organic carbon, total nitrogen,

available phosphorous, potassium calcium and magnesium) and micronutrient (boron, copper,

molybdenum, manganese, iron and zinc) variables were collected. Data were also collected on

overstory (richness and dominance) and understory (richness and vertical structure) vegetation.

Data were collected along grassland exterior-edge-fragment interior gradients for

microenvironment variables and edge-interior gradients for species variables.

Edge effect studies typically consider the effect of the response of a variable (for e.g., soil

moisture) as a function of distance to the nearest edge. This is driven in part by the nature of

conventional edge effect studies which quantify the effect of the creation of an edge between

original (often forested) habitat and converted habitat (e.g., agriculture, road). However, studies

have shown that the response to an edge can be better explained when distance to multiple edges

is considered. The inclusion of distance to multiple edges in edge models becomes increasingly

important in highly fragmented ecosystems and/or small fragments. Since the study was

92

designed to study edge effects in small fragments distance-to-edge in three additional directions

(2nd nearest, 3rd nearest and 4th nearest edge) mutually orthogonal to our original distance were

obtained. The use of the distance to nearest edge alone found no evidence of edge-interior

gradients similar to patterns observed in other small fragments. However the inclusion of

distance to multiple edges significantly improved the predictability of our models in small

fragments. Highly significant patterns for light transmittance (p = 0.0072) and soil moisture (p =

0.0002) and weaker patterns for air temperature (p = 0.032) were observed as a function of

distance to multiple edges. For larger fragments though distance to the nearest edge was

sufficient to observe gradients in relative humidity and potassium. With the exception of soil

temperature no exterior-edge-shola interior gradients in microenvironment and soil nutrient

variables were observed. Further, nutritional differences between grassland and shola fragment

soils did not differ statistically as has been observed in other studies.

No gradients in overstory dominance and vertical complexity in understory as a function of

distance to edge were observed. Nonmetric multidimensional scaling (NMS) ordination

techniques revealed a higher variation in species distributions between fragments than within

fragments along the edge-interior gradient. Compositionally, significant differences were

observed between mid-elevation fragments and high elevation fragments for overstory species.

Since species found in high elevation (≥ 1700 m) fragments are more representative of tropical

montane forest habitat the exclusion of fragments ≤ 1700 m from future tropical montane forest

studies is recommended possibly designating these areas as premontane habitat. However, this

requires a formal exercise to clearly delineate the two habitat subtypes.

The apparent non-random pattern of shola fragments in the ecosystem mosaic also

prompted a study on the effect of topographical variables in determining the presence (or

93

absence) of shola species. For the purpose of this study the two largest high-altitude extents of

the shola-grassland ecosystem mosaic- Eravikulam National Park (ENP) and the Nilgiris (NIL)

were used. Elevation, slope, curvature of the slope (or Terrain Shape index), aspect (expressed as

Eastness and Northness) and wetness index were selected as the topographical variables used to

predict shola presence. Three models were used- a combined model, and two models using ENP

and NIL individually as the training datasets-to test hypotheses. Although all topographical

variables quantified were highly significant (p < 0.0001), the response to aspect was unexpected.

While the preference of shola fragments for western aspects captures monsoonal related trends

(and hence soil moisture) the preference for northern aspects was contrary to expectation. Due to

the limitations of the dataset I am unable to make conclusive arguments for the existence of this

pattern. Soil moisture related variability is also captured in the response to wetness index with

fragments preferring wetter sites.

The persistence and relative stability of the shola-grassland edge continues to puzzle

ecologists and foresters. Our predictive model provides support for fire and hydrological

regulation as driving forces maintaining the edge as shola fragments preferred wetter habitats

(which are less likely to burn). Since research has shown that the occurrence of frosts mirrors

patterns in the flow of water, wetter areas are more likely to have ground frosts. This would

inhibit the establishment of shola, which is in contrast to our observations. We can conclude that

it is unlikely that frost maintains the edge. Maintenance of the edge due to edaphic differences is

unlikely since nutritionally, shola and grassland habitat do not differ significantly.

The shola-grassland ecosystem mosaic offers insights into fragmentation-related processes

in the absence of human influence. The persistence of small fragments (~ 1ha) in the mosaic also

needs to be highlighted. However, more needs to be learnt about processes driving the observed

94

patterns. In addition to the need for basic hydrological studies, data are also lacking about the

role of the shola epiphytic community in rainfall interception and nutrient cycling observed from

other tropical montane forest ecosystems (e.g., Nadkarni et al. 2000). Physiological data for

shola species are conspicuous by their absence as are experimental studies on seedling

establishment essential for ecosystem restoration efforts (however, see Sekar 2008). Information

on genetic diversity of plant species within the mosaic is scarce (Jain et al. 2000, Deshpande et

al. 2001) and studies need to address this given the patchy nature of sholas both at a mosaic level

and at a regional level.

I conclude by highlighting the dearth of process oriented studies in the ecosystem and

suggest that further research needs to be directed to it.

95

APPENDIX

96

Table A-1. Overstory species (≥ 5 cm dbh) presence matrix along an edge-interior gradient across nine fragments in the shola-grassland ecosystem mosaic in the Western Ghats, southern India. Karadishonebetta (KR), Gummanebetta (GU) and Jodigere (JG) sholas were located within BRT Wildlife Sanctuary (BRT). Suicide point shola (SP), Chinnanaimudi (CA), Pusinambara (PS), Kolathan (KS) and Mukkal mile (MM) sholas were located within Eravikulam National Park (ENP). Pampadum shola (PSNP) is represented by a single fragment. Presence across 4 distances (2.5 m, 12.5 m, 22.5 m and 32.5 m marked 1, 2, 3 and 4 respectively) is indicated by gray squares.

BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4ACSI ACSP ALCO ALSP APLA ARBL BESP CATR CETI CHMA CHRO CIMA CISP CISU DESP DEVE

ACSI: Acacia sinuata; ACSP: Actinodaphne spp.; ALCO: Allophylus cobbe; ALSP: Albizzia spp.; APLA: Apodytes laerithanniana; ARBL: Ardisia blatteri; BESP: Beilschmiedea spp.; CATR: Canthium travancorium; CETI: Celtis timorensis; CHMA: Chionanthus malabarica; CHRO: Chrysophyllum roxburghii; CIMA: Cinnamomum malabarica; CISP: Cinnamomum spp.; CISU: Cinnamomum sulphuratum; DESP: Derris spp.; DEVE: Debregeasia velutina.

97

Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4DIOP ELSE EMBA EMRI EMSP EUSP EXRO FASP GASP GOCO HYAL ILSP ILTH ISSP IXNO IXSP LABL LASP LGSP LIDE LISP

DIOP: Dioscorea oppositifolia; ELSE: Elaeocarpus serratus; EMBA: Embelia basaal; EMRI:Embelia ribes; EMSP: Embelia spp.; EUSP: Eugenia spp.; EXRO: Excoecaria robusta; FASP: Fagraea spp.; GASP: Garcinia spp.; GOCO: Gomphandra coriaceae; HYAL: Hydnocarpus alpina; ILSP: Ilex spp.; ILTH: Ilex thwaitesii; ISSP: Isonandra spp.; IXNO: Ixora notoniana; IXSP: Ixora spp.; LABL: Lasianthus blumeanus; LASP: Lasianthus spp.; LGSP: Ligustrum spp.; LIDE: Litsea deccanensis; LISP: Litsea spp.

98

Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4LIWI MAPE MALE MAPH MAAR MASP MESI NECA NEFI NEFO NESC NESP NEZE OLGL PEMA PHWI PISP PINE PSCO

LIWI: Litsea wightiana; MAPE: Maesa perrottetiana; MALE: Mahonia leschenaultii; MAPH: Mallotus philippensis; MAAR: Mastixia arborea; MASP: Mastixia spp.; MESI: Meliosma simplicifolia; NECA: Neolitsea cassia; NEFI: Neolitsea fischeri; NEFO: Neolitsea foliosa; NESC: Neolitsea scrobiculata; NESP: Neolitsea spp.; NEZE: Neolitsea zeylanica; OLGL: Olea glandulifera; PEMA: Persea macrantha; PHWI: Phoebe wightii; PISP: Piper spp.; PINE: Pittosporum neelegherrense; PSCO: Psychotria congesta.

99

Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4PSNI PSSP RHNI RUCO SAFO SCRO SCWA SCCR STAL STIS SYMO SYCA SYDE SYSP TAAS TESP TRCO TUNE VALE VANE VASP

PSNI: Psychotria niligiriensis; PSSP: Psychotria spp.; RHNI: Rhododendron nilagaricum; RUCO: Rubia cordifolia; SAFO: Saprosoma foetens ssp. ceylanicum; SCRO: Schefflera rostrata; SCWA: Schefflera wallichiana; SCCR: Scolopia crenata; STAL: Strobilanthes integrifolius; STIS: Strobilanthes isophyllus; SYMO: Symplocos monatha; SYCA: Syzygium carophyllatum; SYDE: Syzygium densiflorum; SYSP: Syzygium spp.; TAAS: Tarenna asiatica; TESP: Ternstroemia spp.; TRCO: Tricalysia apiocarpa; TUNE: Turpinia nepalensis; VALE: Vaccinium leschenaultii; VANE: Vaccinium neilgherrense; VASP: Vaccinium spp.

100

Table A-1. Continued Sps. PSNP 1 2 3 4 ACSI ACSP ALCO ALSP APLA ARBL BESP CATR CETI CHMA CHRO CIMA CISP CISU DESP DEVE DIOP ELSE EMBA EMRI EMSP EUSP EXRO FASP GASP GOC HYAL ILSP ILTH

ACSI: Acacia sinuata; ACSP: Actinodaphne spp.; ALCO: Allophylus cobbe; ALSP: Albizzia spp.; APLA: Apodytes laerithanniana; ARBL: Ardisia blatteri; BESP: Beilschmiedea spp.; CATR: Canthium travancorium; CETI: Celtis timorensis; CHMA: Chionanthus malabarica; CHRO: Chrysophyllum roxburghii; CIMA:Cinnamomum malabarica; CISP: Cinnamomum spp.; CISU: Cinnamomum sulphuratum; DESP: Derris spp.; DEVE: Debregeasia velutina; DIOP: Dioscorea oppositifolia; ELSE: Elaeocarpus serratus; EMBA: Embelia basaal; EMRI: Embelia ribes; EMSP: Embelia spp.; EUSP: Eugenia spp.; EXRO: Excoecaria robusta; FASP: Fagraea spp.; GASP: Garcinia spp.; GOCO: Gomphandra coriaceae; HYAL: Hydnocarpus alpina; ILSP: Ilex spp.; ILTH: Ilex thwaitesii

101

Table A-1. Continued Sps. PSNP 1 2 3 4 ISSP IXNO IXSP LABL LASP LGSP LIDE LISP LIWI MAPE MALE MAPH MAAR MASP MESI NECA NEFI NEFO NESC NESP NEZE OLGL PEMA PHWI PISP PINE PSCO PSNI

ISSP: Isonandra spp.; IXNO: Ixora notoniana; IXSP: Ixora spp.; LABL: Lasianthus blumeanus; LASP: Lasianthus spp.; LGSP: Ligustrum spp.; LIDE: Litsea deccanensis; LISP: Litsea spp.; LIWI: Litsea wightiana; MAPE: Maesa perrottetiana; MALE: Mahonia leschenaultii; MAPH: Mallotus philippensis; MAAR: Mastixia arborea; MASP: Mastixia spp.; MESI: Meliosma simplicifolia; NECA: Neolitsea cassia; NEFI: Neolitsea fischeri; NEFO: Neolitsea foliosa; NESC: Neolitsea scrobiculata;NESP: Neolitsea spp.; NEZE: Neolitsea zeylanica; OLGL: Olea glandulifera; PEMA: Persea macrantha; PHWI: Phoebe wightii; PISP: Piper spp.; PINE: Pittosporum neelegherrense; PSCO: Psychotria congesta; PSNI: Psychotria niligiriensis.

102

Table A-1. Continued Sps. PSNP 1 2 3 4 PSSP RHNI RUCO SAFO SCRO SCWA SCCR STAL STIS SYMO SYCA SYDE SYSP TAAS TESP TRCO TUNE VALE VANE VASP

PSSP: Psychotria spp.; RHNI: Rhododendron nilagaricum; RUCO: Rubia cordifolia; SAFO: Saprosoma foetens ssp. ceylanicum; SCRO: Schefflera rostrata; SCCR: Scolopia crenata; SCWA: Schefflera wallichiana; STAL: Strobilanthes integrifolius; STIS: Strobilanthes isophyllus; SYMO: Symplocos monatha; SYCA: Syzygium carophyllatum; SYDE: Syzygium densiflorum; SYSP: Syzygium spp.; TAAS: Tarenna asiatica; TESP: Ternstroemia spp.; TRCO: Tricalysia apiocarpa; TUNE: Turpinia nepalensis; VALE: Vaccinium leschenaultii; VANE: Vaccinium neilgherrense; VASP: Vaccinium spp.

103

Table A-2. Understory species presence matrix along an edge-interior gradient across nine fragments in the shola-grassland

ecosystem mosaic in the Western Ghats, southern India. All vegetation greater than 50 cm in height but less than 5 cm dbh was defined as understory. Karadishonebetta (KR), Gummanebetta (GU) and Jodigere (JG) sholas were located within BRT Wildlife Sanctuary (BRT). Suicide point (SP), Chinnanaimudi (CA), Pusinambara (PS), Kolathan (KS) and Mukkal mile (MM) sholas were located within Eravikulam National Park (ENP). Pampadum shola (PSNP) is represented by a single fragment. Presence across 4 distances (2.5 m, 12.5 m, 22.5 m and 32.5 m marked 1, 2, 3 and 4 respectively) is indicated by gray squares.

Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4AGAD ALVE ARBL ARSP ASSP ATMO BESP CASP CATR CAPA CETI CHMA CHOD CIMA CISP CISU

AGAD: Ageratina adenophora; ALVE: Alstonia venenata; ARBL: Ardisia blatteri; ARSP: Ardisia spp.; ASSP: Asparagus spp.; ATMO: Atalantia monophylla; BESP: Beilschmiedea spp.; CASP: Calamus spp.; CATR: Canthium travancorium; CAPA: Cassine paniculata; CETI: Celtis timorensis; CHMA: Chionanthus malabarica; CHOD: Chromolaena odorata; CIMA: Cinnamomum malabarica; CISP: Cinnamomum spp.; CISU: Cinnamomum sulphuratum

104

Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4CIIN COAR COSP DESP DIOP ELKO ELSE EMBA EMRI EUSP EXRO FASP GAGU GASP GLSP HEDY ISSP IXSP LACA LABL LASP LIOL

CIIN: Cipadessa indica; COAR: Coffea arabica; COSP: Coleus spp.; DESP: Derris spp.; DIOP: Dioscorea oppositifolia; ELKO: Elaegnus kologa; ELSE: Elaeocarpus serratus; EMBA: Embelia basaal; EMRI: Embelia ribes; EUSP: Eugenia spp.; EXRO: Excoecaria robusta; FASP: Fagraea spp.; GAGU: Garcinia gummi-gutta; GASP: Garcinia spp.; GLSP: Glochidion spp.; HEDY: Hedyotis spp.; ISSP: Isonandra spp.; IXSP: Ixora spp.; LACA: Lantana camara; LABL: Lasianthus blumeanus; LASP: Lasianthus spp.; LIOL: Ligustrum oleaceae

105

Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4LGSP LIDE LISP LIWI MAIN MAPE MALE MAPH MAAR MESI MEUM NEFI NESC NEZE NONI PEMA PHWI PISP POSP POAC PSCO PSNI

LGSP: Ligustrum spp.; LIDE: Litsea deccanensis; LISP: Litsea spp.; LIWI: Litsea wightiana; MAIN: Maesa indica; MAPE: Maesa perrottetiana; MALE: Mahonia leschenaultii; MAPH: Mallotus philippensis; MAAR: Mastixia arborea; MESI: Meliosma simplicifolia; MEUM: Memecylon umbellatum; NEFI: Neolitsea fischeri; NESC: Neolitsea scrobiculata; NEZE: Neolitsea zeylanica; NONI: Nothapodytes nimmoniana; PEMA: Persea macrantha; PHWI: Phoebe wightii; PISP: Piper spp.; POSP: Polygonum spp.; POAC: Polyscias acuminata; PSCO: Psychotria congesta; PSNI: Psychotria niligiriensis

106

Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4PSSP RUSP SCCA SCCR SORO STGU STAL STIS STSP SYCO SYMO SYCU SYSP TASP TESP TOAS TRAP unBA unFE unSH VASP VIPU WETH

PSSP: Psychotria spp.; RUSP: Rubus spp.; SCCA: Schefflera capitata; SCCR: Scolopia crenata; SORO: Solanum robustum; STGU: Sterculia guttata; STAL: Strobilanthes integrifolius; STIS: Strobilanthes isophyllus; STSP: Strobilanthes spp.; SYCO: Symplocos cochinchinensis; SYMO: Symplocos monatha; SYCU: Syzygium cuminii; SYSP: Syzygium spp.; TASP: Tarenna spp.

107

Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4TESP TOAS TRAP unBA unFE unSH VASP VIPU WETH

TESP: Ternstroemia spp.; TOAS: Toddalia asiatica; TRAP: Tricalysia apiocarpa; unBA: Unidentified bamboo; unFE: Unidentified Fern; unSH: Unidentified shrub; VASP: Vaccinium spp.; VIPU: Viburnum punctatum; WETH: Wendlandia thyrsoidea

108

Table A-2. Continued Sps PSNP 1 2 3 4 AGAD ALVE ARBL ARSP ASSP ATMO BESP CASP CATR CAPA CETI CHMA CHOD CIMA CISP CISU CIIN COAR COSP DESP DIOP ELKO ELSE EMBA EMRI EUSP EXRO FASP GAGU GASP GLSP

AGAD: Ageratina adenophora; ALVE: Alstonia venenata; ARBL: Ardisia blatteri; ARSP: Ardisia spp.; ASSP: Asparagus spp.; ATMO: Atalantia monophylla; BESP: Beilschmiedea spp.; CASP: Calamus spp.; CATR: Canthium travancorium; CAPA: Cassine paniculata; CETI: Celtis timorensis; CHMA: Chionanthus malabarica; CHOD: Chromolaena odorata; CIMA: Cinnamomum malabarica; CISP: Cinnamomum spp.; CISU: Cinnamomum sulphuratum; CIIN: Cipadessa indica; COAR: Coffea arabica; COSP: Coleus spp.; DESP: Derris spp.; DIOP: Dioscorea oppositifolia; ELKO: Elaegnus kologa; ELSE: Elaeocarpus serratus; EMBA: Embelia basaal; EMRI: Embelia ribes; EUSP: Eugenia spp.; EXRO: Excoecaria robusta; FASP: Fagraea spp.; GAGU: Garcinia gummi-gutta; GASP: Garcinia spp.; GLSP: Glochidion spp.

109

Table A-2. Continued Sps PSNP 1 2 3 4 HEDY ISSP IXSP LACA LABL LASP LIOL LGSP LIDE LISP LIWI MAIN MAPE MALE MAPH MAAR MESI MEUM NEFI NESC NEZE NONI PEMA PHWI PISP POSP POAC PSCO PSNI

HEDY: Hedyotis spp.; ISSP: Isonandra spp.; IXSP: Ixora spp.; LACA: Lantana camara; LABL: Lasianthus blumeanus; LASP: Lasianthus spp.; LIOL: Ligustrum oleaceae; LGSP: Ligustrum spp.; LIDE: Litsea deccanensis; LISP: Litsea spp.; LIWI: Litsea wightiana; MAIN: Maesa indica; MAPE: Maesa perrottetiana; MALE: Mahonia leschenaultii; MAPH: Mallotus philippensis; MAAR: Mastixia arborea; MESI: Meliosma simplicifolia; MEUM: Memecylon umbellatum; NEFI: Neolitsea fischeri; NESC: Neolitsea scrobiculata; NEZE: Neolitsea zeylanica; NONI: Nothapodytes nimmoniana; PEMA: Persea macrantha; PHWI: Phoebe wightii; PISP: Piper spp.; POSP: Polygonum spp.; POAC: Polyscias acuminata; PSCO: Psychotria congesta; PSNI: Psychotria niligiriensis

110

Table A-2. Continued Sps PSNP 1 2 3 4 PSSP RUSP SCCA SCCR SORO STGU STAL STIS STSP SYCO SYMO SYCU SYSP TASP TESP TOAS TRAP unBA unFE unSH VASP VIPU WETH

PSSP: Psychotria spp.; RUSP: Rubus spp.; SCCA: Schefflera capitata; SCCR: Scolopia crenata; SORO: Solanum robustum; STGU: Sterculia guttata; STAL: Strobilanthes integrifolius.; STIS: Strobilanthes isophyllus; STSP: Strobilanthes spp.; SYCO: Symplocos cochinchinensis; SYMO: Symplocos monatha; SYCU: Syzygium cuminii; SYSP: Syzygium spp.; TASP: Tarenna spp.; TESP: Ternstroemia spp.; TOAS: Toddalia asiatica; TRAP: Tricalysia apiocarpa; unBA: Unidentified bamboo; unFE: Unidentified Fern; unSH: Unidentified shrub; VASP: Vaccinium spp.; VIPU: Viburnum punctatum; WETH: Wendlandia thyrsoidea

111

LIST OF REFERENCES

Abraham, T.K. 2001. Meliolaceous and soil fungi and lichen flora of the sholas of Munnar and Waynad. Pages 118-135 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Aerts, R., Overtveld, K.V., Haile, M., Hermy, M., Deckers, J., Muys, B. 2006. Species composition and diversity of small Afromontane forest fragments in northern Ethiopia. Plant Ecology 187: 127-142.

Agrawal, S.C., Madan, U.S., Chinnamani, S., Rege, N.D. 1961. Ecological studies in the Nilgiris. Indian Forester 87: 376-389.

Alembath, M., Rice, C. 2008. Nilgiritragus hylocrius. In: IUCN 2008. 2008 IUCN Red List of Threatened Species. www.iucnredlist.org. Downloaded on 30 January 2009.

Allen S. E. 1974. Chemical analysis of ecological materials. Blackwell Scientific, Oxford.

Anderson, S.H.,Mann, K. Shugart,H.H. 1977. The effect of transmission-lines corridors on bird populations. American Midland Naturalist 97: 216-221.

Bader, M.Y., Ruijten, J.J.A. 2008. A topography-based model of forest cover at the alpine tree line in the tropical Andes. Journal of Biogeography 35: 711-723.

Balasubramanian, K., Kumar, K.K. 1999. The riddle of the shola. Evergreen 42: 1-5.

Barve, N., Kiran, M.C., Vanaraj, G., Aravind, N.A., Rao, D., Shaanker, R.U., Ganeshaiah, K.N., Poulsen, J.G. 2005. Measuring and mapping threats to a wildlife sanctuary in southern India. Conservation Biology 19: 122-130.

Beisner, B.E., Haydon, D.T., Cuddington, K. 2003. Alternative stable states in ecology. Frontiers in Ecology and the Environment 1: 376-382.

Benzing, D.H. 1998. Vulnerabilities of tropical forests to climate change: the significance of resident epiphytes. Climatic Change 39: 519-540.

Blennow, K. 1998. Modelling minimum air temperature in partially and clear felled forests. Agricultural and Forest Meteorology 91: 223-235.

Bor, N.L. 1938. The vegetation of the Nilgiris Indian Forester 64: 600-609.

Brooks, T.M., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A.B., Rylands, A.B., Konstant, W.R., Flick, P., Pilgrim, J., Oldfield, S., Magin, G., Hilton-Taylor, C. 2002. Habitat loss and extinction in the hotspots of biodiversity. Conservation Biology 16: 909-923.

112

Bruijnzeel, L.A., Hamilton, L.S. 2000. Decision time for cloud forests: IHP Humid tropics programme series No. 13. UNESCO, Paris, France.

Bowman, D.M.J.S., Woinarski, J.C.Z. 1994. Biogeography of Australian monsoon rainforest mammals: Implications for the conservation of rainforest mammals. Pacific Conservation Biology 1: 98-106.

Bubb, P. May, I. Miles, L., Sayer, J. 2004. Cloud Forest Agenda. UNEP-WCMC, Cambridge UL. Online at http://www.unep-wcmc.org/resources/publications/UNEP_WCMC_bio_series/20.htm

Camargo, J.L.C., Kapos, V. 1995. Complex edge effects on soil moisture and microclimate in central Amazonian forest. Journal of Tropical Ecology 11: 205-221.

Carlson, A., Hartman, G. 2001. Tropical forest fragmentation and nest predation- an experimental study in an Eastern Arc montane forest, Tanzania. Biodiversity and Conservation 10: 1077-1085.

Clements, F.E. 1936. Nature and structure of climax. Journal of Ecology 24: 252-284.

Chandrasekhara, U.M. Muraleedharan, P.K., Sibichan, V.2001. Disturbed sholas of Kerala and strategies for its conservation and management. Pages 395-437 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Chandrasekharan, C. 1962. Forest types of Kerala. Indian Forester 88: 837-847.

Chao, A. 2005. Species richness estimation. Pages 7909-7916 in Balakrishnan, N., Read, C. B., Vidakovic, B., editors. Encyclopedia of Statistical Sciences. Wiley, New York, USA.

Chao, A. Chazdon, R.L., Colwell, R.K., Shen, T.J. 2005. A new statistical approach for assessing similarity of species composition with incidence and abundance data. Ecology Letters 8: 148-159.

Coblentz, D., Keating, P.L 2008. Topographic controls on the distribution of tree islands in the high Andes of south-western Ecuador. Journal of Biogeography 35: 2026-2038.

Coblentz, D.D., Riitters, K.H. 2004. Topographic controls on the regional scale biodiversity of the south-western USA. Journal of Biogeography. 31: 1125-1138.

Cochrane, M.A. 2001. Synergistic interactions between habitat fragmentation and fire in evergreen tropical forests. Conservation Biology 15: 1515-1521.

Colwell, R.K. 2006. EstimateS: Statistical estimation of species richness and shared species from samples. Version 8. Persistent URL <purl.oclc.org/estimates>.

Colwell, R. K., Coddington, J. A. 1994. Estimating terrestrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society (Series B) 345: 101-118.

113

Colwell, R.K., Mao, C.X., Chang, J. 2004. Interpolating, extrapolating and comparing incidence-based species accumulation curves. Ecology 85: 2717-2727.

Curran, L.M. Caniago, I., Paoli, G.D., Astianti, D., Kusneti, M., Leighton, M., Nirarita, C.E., Haeruman, H. 1999. Impact of El-Niño and logging on canopy recruitment in Borneo. Science 286: 2148-2188.

Daniels 2006. The Nilgiri Tahr: An endemic south Indian mountain goat. Macmillan India, New Delhi, India.

Daniels, R.J.R.; Easa, P.S.; Alembath, M. 2008. Distribution and status of the endangered Nilgiri tahr. Current Science 94: 10-11.

Das, A., Krishnaswamy, J., Bawa, K.S., Kiran, M.C., Srinivas, V., Kumar, N.S., Karanth, K.U. 2006. Prioritization of conservation areas in the Western Ghats, India. Biological Conservation 133: 16-31.

Davidar, E.R.C. 1971. A note on the status of the Nilgiri Tahr in the grass hills in the Anamalais. Journal of the Bombay Natural History Society 68: 347-354.

Davidar, P., Mohandass, D., Vijayan, S.L. 2007. Floristic inventory of woody plants in a tropical montane forest in the Palni hills of the Western Ghats, India. Tropical Ecology 48: 15-25.

del Castillo, R.F., Rios, M.A.P. 2008. Changes in seed rain during secondary succession in a tropical montane forest region in Oaxaca, Mexico. Journal of Tropical Ecology 28: 433-444.

Delcourt, H.R. Delcourt, P.A. 1988. Quaternary landscape ecology: Relevance scales in space and time. Landscape Ecology 2: 32-44.

Deshpande, A.U., Apte, G.S., Bahulikar, R.A., Lagu, M.D., Kulkarni, B.G., Suresh, H.S., Singh, N.P., Rao, M.K.V., Gupta, V.S., Pant, A., Ranjekar, P.K. 2001. Genetic diversity across plant populations of three montane plant species from the Western Ghats, India revealed by intersimple sequence repeats. Molecular Ecology 10: 2397-2408.

Didham, R.K., Lawton, J.H. 1999. Edge structure determines the magnitude of changes in microclimate and vegetation structure in tropical forest fragments. Biotropica 31: 17-30.

Dinesh, K.P., Radhakrishnan, C., Bhatta, G. 2008. A new species of Nyctibatrachus Boulenger (Amphibia: Anura: Nyctibatrachidae) from the surroundings of Bhadra Wildlife Sacntuary, Western Ghats, India. Zootaxa 1914: 45-56.

FAO (Food and Agricultural Organization). 1993. Forest resources assessment 1990: tropical countries. FAO Forestry Paper No. 112. Rome

Fernández, C., Acosta, F.J., Abellá, G., López, F., Díaz, M. 2002. Complex edge effect fields as additive processes in patches of ecological systems. Ecological Modelling 149: 273-273.

114

Fletcher, R.J. 2005. Multiple edge effects and their implications in fragmented landscapes. Journal of Animal Ecology 74: 342-352.

Fonseca, C.R., Joner,F.Two sided edge effect studies and the restoration of endangered ecosystems. Restoration Ecology 15: 613-619.

Foster, P. 2001. The potential negative impacts of global climate change on tropical montane cloud forests. Earth Science Reviews 55: 73-106.

Franklin, J. 1995. Predictive vegetation mapping, geographic modelling of biospatial patterns in relation to environmental gradients. Progress in Physical Geography 19:474–499.

Franklin, J., McCullough, P., Gray, C. 2000. Terrain variables used for predictive mapping of vegetation communities in southern California. Pages 331-353 in: Wilson, J., Gallant, J., editors. Terrain analyses: principles and applications, John Wiley and sons, New York, USA.

Ganeshaiah, K.N., Umashaanker, R.U., Bawa, K.S. 1997. Diversity of species assemblages of islands: Predictions and their test using tree species composition of shola fragments. Current Science 73: 188-194.

Gardner, F.P., Pearce, R.B., Mitchell, R.L. 1984. Physiology of crop plants. Iowa State University Press, Ames, IA, USA.

Gascon, C.G., Williamson, B., da Fonseca, G.A.B. 2000. Receding forest edges and vanishing reserves. Science 288: 1356-1358.

Gates, J.E., Gysel, L.W. 1978. Avian nest dispersion and fledgling success in field-forest ecotones. Ecology 59: 871-883.

George, M., Mohandas, K., Brijesh, C.M. 2001. Insect fauna of the sholas of Idukki and Wayanad districts. Pages 317-340 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Gessler, P.E., Moore, I.D. McKenzie, N.J., Ryan, P.J. 1995. Soil-landscape modeling and spatial prediction of soil attributes. International Journal of GIS. 9: 421-432.

Ghosh, D.S. 2001. Fish diversity and associated species in the sholas of Munnar, Idukki District. Pages 353-364 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Gioda, A. Maley, J., Guasp, R.E., Baladón, A.A. 1995. Some low elevation fog forests of dry environments: application to African palaeoenvironments. Pages 156-164 in Hamilton, L.S., Juvik, J.O., Scatena, F.N., editors. Ecological studies 110: Tropical Montane Cloud Forests. Springer-Verlag, New York, USA.

115

Greig-Smith, P. 1979. Pattern in vegetation. Journal of Ecology 67:755–779.

Gupta, R.K. 1962. Studies in some shola forests on the Palni hills near Kodaikanal. Indian Forester 88: 848-853.

Gupta, R.K., Shankarnarayan, K.A. 1962. Ecological status of the grasslands in south India. Tropical Ecology 3: 75-78.

Gurevitch, J., Scheiner, S.M., Fox G.A. 2002. The Ecology of plants Sinauer Associates, Sunderland, Massachusetts, USA.

Hamilton, L.S., Juvik, J.O., Scatena, F.N. 1995. Ecological studies 110: Tropical Montane Cloud Forests. Springer-Verlag, New York, USA.

Harper, K.A., MacDonald, S.E. 2002. Structure and composition of edges next to regenerating clear-cuts in mixed-wood boreal forest. Journal of Vegetation Science 13: 535-546.

Harper, K.A., Macdonald, S.E., Burton, P.J., Chen, J.Q., Brosofske, K.D., Saunders, S.C., Euskirchen, E.S., Roberts, D., Jaiteh, M.S., Esseen, P.A.. 2005. Edge influence on forest structure and composition in forest fragments. Conservation Biology 19: 768-782.

Hofer, G., Wagner, H.H., Herzog, F., Edwards, P.J. 2008. Effects of topographic variability on the scaling of plant species richness in gradient dominant landscapes. Ecography 31: 131-139.

Ickes, K., Williamson, B.G. 2000. Edge effects and ecological processes: are they on the same spatial scale? Trends in Ecology and Evolution 15: 373.

Jain, A., Pandit, M.K., Elahi, S., Jain, A., Bhaskar, A., Kumar, V. 2000. Reproductive behaviour and genetic variability in geographically isolated populations of Rhododendron arboreum (Ericaceae). Current Science 79: 1377-1381.

Jackson, M.L. 1962. Soil chemical analysis. Prentice-Hall, Englewood Cliffs, NJ, USA.

Janzen, D.H. 1983. No park is an island- Increase in interference from outside as park size decreases. Oikos 41: 402-410.

Jeeva, V., Ramakrishnan, P.S. 1997. Studies on impact of plantation forestry in Nilgiri Hills of the Ghats on soil quality and nutrient cycling. Tropical Ecology 38: 215-235.

Johnston, V.R. 1947. Breeding birds of the forest edge in Illinois. Condor 2: 45-53.

Jose, S. Sreepathy, A. Mohan Kumar, B., Venugopal, V.K. 1994. Structural, floristic and edaphic attributes of the shola-grassland forests of Eravikulam in peninsular India. Forest Ecology and Management 65: 279-291.

116

Jose, S. Gillespie, A.R., Goerge, S.J., Kumar, M.K. 1996. Vegetation response along edge-interior gradients in a high altitude tropical forest in peninsular India. Forest Ecology and Management 87: 51-62.

Kapos, V. 1989. Effects of isolation on the water status of forest patches in the Brazilian Amazon. Journal of Tropical Ecology 5: 173-185.

Karunakaran, P.V., Rawat, G.S., Uniyal, V.K. 1998. Ecology and conservation of the grasslands of Eravikulam National Park and the Western Ghats. Wildlife Institute of India, Dehradun, India.

Kruskal, J.B. 1964. Nonmetric multi-dimensional scaling: a numerical method. Psychometrika 29: 115-129.

Krishnaswamy, J., Kiran, M.C., Ganeshaiah, K.N. 2005. Tree model based eco-climatic vegetation classification and fuzzy mapping in diverse tropical deciduous ecosystems using multi-date NDVI. International Journal of Remote Sensing 25: 1185-1205.

López-Barrera, F. Newton, A., Manson, R. 2005.Edge effects in a tropical montane forest mosaic: experimental tests of post-dispersal acorn removal. Ecological Research. 20: 31-40.

Laurance, W.F. 1991. Edge effects in tropical forest fragments: application of a model for the design of nature reserves. Biological Conservation 57: 205-219.

Laurance, W.F. 2000. Reply from W.F. Laurance. Trends in Ecology and Evolution 15: 373.

Laurance, W.F. 2008. Theory meets reality: How habitat fragmentation research has transcended island biogeographic theory. Biological Conservation 141: 1731-1744.

Laurance,W.F., Ferreira,L.V. Rankin-De Merona, J.M., Laurance, S.G., Hutchings, R.W. Lovejoy, T.E. 1998. Effects of forest fragmentation on recruitment patterns in Amazonian tree communities. Conservation Biology 12: 460-464.

Laurance, W.F., Lovejoy, T.E., Vasconcelos, H.L., Bruna, E.M., Didham, R.D., Stouffer, P.C., Gascon, C., Bierregaard, R.O., Laurance, S.G., Sampaio, E.G. 2002. Ecosystem decay of Amazonian forest fragments: a 22-Year investigation. Conservation Biology 16: 605-618.

Laurance, W.F., Yensen, 1991. Predicting the impacts of edge effects in fragmented habitats. Biological Conservation 69: 23-32.

Leopold,A. 1993.Game management. Scribner Sons, New York, USA.

López-Barrera, F., Manson, R.H. González-Espinosa, M., Newton, A.C. 2006. Effects of the type of montane forest edge on oak seedling establishment along forest-edge-exterior gradients. Forest Ecology and Management 225: 234-244.

117

López-Barerra, F. Newton, A., Manson, R. 2005.Edge effects in a tropical montane forest mosaic: experimental tests of post-dispersal acorn removal. Ecological Research. 20: 31-40.

Macarthur, R.H., Wilson, E.O. 1963. An equilibrium theory of insular zoogeography. Evolution 17: 373-387.

Macarthur, R.H., Wilson, E.O. 1967. The Theory of Island Biogeography. Princeton University Press, Princeton, New Jersey, USA.

Magurran, A.E. 1988. Ecological diversity and its measurement. Croom Helm, London, UK.

Malcolm, J.R. 1994. Edge effects in central Amazonian forest fragments. Ecology 75: 2438-2445.

Mancke, R.G.,Gavin, T.A.2000. Breeding bird density in woodlots: Effects of depth and buildings at the edges. Ecological Applications 10: 598-611.

Mao, C.X., Colwell, R.K., Chang, J, 2005. Estimating species accumulation curves using mixtures. Biometrics 61: 433-441.

Martin, P.H., Sherman, R.E., Fahey, T.J. 2007. Tropical montane forest ecotones: climate gradients, natural disturbance, and vegetation zonation in the Cordillera Central, Dominican Republic. Journal of Biogeography 34: 1792-1806.

Mather, P.M. 1976. Computational methods of multivariate analysis in physical geography. J Wiley and sons, London, UK.

May, R.M. 1977. Thresholds and breakpoints in ecosystems with multiplicity of states. Nature 269: 471-477.

McCune, B., Mefford, M.J. 2006. PC-ORD. Multivariate analysis of ecological data. Version 5. MjM Software, Glenden Beach, Oregon, USA.

McNab, W.H. 1989. Terrain Shape Index: Quantifying effect of minor landforms on tree height. Forest Sciance 35: 91-104.

Meher-Homji, V.M. 1965. Ecological status of the montane grasslands of the south Indian hills: A phytogeographic reassessment. Indian Forester 91: 210-215.

Meher-Homji, V.M. 1967. Phytogeography of the south Indian hill stations. Bulletin of the Torrey Botanical Club 94: 230-242.

Menon, A.R.R. 2001.Mapping and analysis of the shola-grassland vegetation of Eravikulam, Idukki district. Pages 95-115. in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

118

Menon, S., Bawa, K.S. 1997. Applications of geographic information systems, remote-sensing and a ladscape ecology approach to biodiversity conservation in the Western Ghats. Current Science 73: 134-145.

Merz, R.W. 1953. Site-index estimates made easy. Journal of Forestry 51: 749-750.

Mesquita, R.C.G., Delamônica, P., Laurance, W.F. 1999. Effect of surrounding vegetation on edge-related mortality in Amazonian forest fragments. Biological Conservation 91: 129-134.

Mishra, C., Johnsingh, A.J.T. 1998. Population and conservation status of the Nilgiri tahr Hemitragus hylocrius in Anamalai hill, south India. Biological Conservation 86: 199-206.

Moore, I.D., Gessler, P.E., Nielsen, G.A., and Petersen, G.A. 1993. Terrain attributes: estimation methods and scale effects. Pages 189-214 in: Jakeman, A.J., Beck, M.B. McAleer, M., editors. Modeling Change in Environmental Systems. Wiley, London, UK.

Murali, K.S., Shetty, R.S., Ganeshaiah, K.N., Shaanker, R.U. 1998. Does forest type classification reflect spatial dynamics of vegetation? An analysis using GIS techniques. Current Science 75: 220-227.

Myers, N., Mittermeier, R.A.,Mittermeier, C.G., da Fonseca, G.A.B., Kent, J. Biodiversity hotpots for conservation priorities. Nature 403: 853-858.

Nadkarni, N.M. Wheelwright, N.T. 2000. Monteverde: Ecology and Conservation of a Tropical Montane Cloud Forest. Oxford University Press, New York, USA.

Nadkarni, N.M., Lawton, R.O., Clark, K.L., Matelson, T.J., Schaefer, D. 2000. Ecosystem Ecology and Forest Dynamics. Pages 303-350 in Nadkarni, N.M. Wheelwright, N.T., editors. Monteverde: Ecology and Conservation of a Tropical Montane Cloud Forest. Oxford University Press, New York, USA.

Nair, K.K.N., Menon, A.R.R. 2001. Endemic arborescent flora of the sholas of Kerala and its population and regeneration status. Pages 209-236 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K. 2001. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Nameer, P.O. 2001. Richness of avifauna in the sholas of Munnar, Idukki District. Pages 365-389 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Nandakumar, V., Rajendran, P., Narendra Babu, K. 2001. Characterization of soils in the sholas of Idukki and Wayanad districts Pages 25-70 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

119

NATMO.2009. Soil map of India (National Atlas and Thematic Mapping Organization). Available online at www.indiabiodiversity.org. Last accessed on May 04th 2009.

Noble, W.A. 1967. The shifting balance of grasslands, shola forests and planted trees on the upper Nilgiris, southern India. Indian Forester 93: 691-693.

Oosterhoorn, M. Kappelle, M. 2000. Vegetation structure and composition along an edge-interior gradient in a Costa Rican montane cloud forest. Forest Ecology and Management 126: 291-307.

Palanna, R.M. 1996. Eucalyptus in India. In Kashio, M. White, K. (editors) Reports submitted to the regional expert consultation on Eucalyptus. Vol II. Available online at http://www.fao.org/docrep/005/ac772e/ac772e06.htm#bm06

Pavlacky, D.C., Anderson, S.H. 2007. Does avian species richness in natural patch mosaics follow the forest fragmentation paradigm? Animal Conservation 10: 57-68.

Qian, H. 2008. A latitudinal gradient of beta diversity for exotic vascular plant species in North America. Diversity and Distributions 14: 556-560.

R Development Core Team (2008). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org.

Ramesh, B.R., Menon, S., Bawa, K.S. 1997. A vegetation based approach to biodiversity gap analysis in the Agastyamalai region, Western Ghats, India. Ambio 26: 529-536.

Ranasinghe, P.N., Dissanayake, C.B., Samarasinghe, D.V.N., Galappatti, R. The relationship between soil geochemistry and die back of montane forests in Sri Lanka: a case study. Environmental Geology 51: 1077-1088.

Ranganathan, C.R. 1938.Studies in the ecology of the shola grassland vegetation of the Nilgiri Plateau. Indian Forester 64: 523-541.

Raghupathy, R., Madhu, M. 2007. The natural grasslands of the Nilgiris: Past and present scenario. Range Management and Agroforestry 28: 271-273.

Rice, C. 1984. The behavior and ecology of the Nilgiri Tahr (Hemitragus hylocrius Ogilby, 1838). Dissertation. Texas A & M University, College Station, Texas, USA.

Rice, C. 1988. Habitat, population dynamics and conservation of the Nilgiri Tahr, Hemitragus hylocrius. Biological Conservation 44: 137-156.

Ries, L. Fletcher, R.J., Battin, J. 2004. Ecological responses to habitat edges: Mechanisms, models and variability explained. Annual Review of Ecology Evolution and Systematics 35: 491-522.

Ries, L., Sisk, T.D. 2004. A predictive model of edge effects. Ecology 85: 2917-2926.

120

Sahoo, D.C., Sharda, V.N., Jayakumar, M., Tripathi, K.P., Padmanabhan, M.V., Raghunath, B., Chandran, B. 2006. Hydrology of small watersheds in high hills of Nilgiris. Indian Journal of Soil Conservation 34: 97-101.

Sankaran, K.V., Balasundaram, M. 2001. Soil microflora of the sholas of Eravikulam National Park, Idukki district. Pages 151-178 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Schaller, G.B. 1970. Mountain monarchs: Wild sheep and goats of the Himalayas. The University of Chicago Press, Chicago, USA.

Sekar, T. 2008. Observations on survival and growth of different Shola species under a Shola afforestation programme in Nilgiris District, Tamil Nadu. Indian Forester 134: 451-457.

Seshan, S. 2006. Notes from the edge. Available online at : http://www.gbsanctuary.org/articles/Notes_from_the_Edge.pdf

Shanker, K. 2001. The role of competition and habitat in structuring small mammal communities in a tropical montane forest ecosystem in southern India. Journal of Zoology 253: 15-24.

Sizer, N., Tanner, E.V.J. 1999. Responses of woody plant seedlings to edge formation in forest fragments in central Amazonia. Biological Conservation 91: 135-142.

Smith, E.P., van Belle, G. 1984. Nonparametric estimation of species richness. Biometrics 40: 119-129.

Somanathan, H., Borges, R.M. 2000. Influence of exploitation on population structure, spatial distribution and reproductive success if dioecious species in fragmented cloud forest in India. Biological Conservation 80: 9-15.

Soltanpour, P.M., Schwab, A.P. 1977. A new soil test for simultaneous extraction of macro and micro nutrient in alkaline soils. Communications in Soil Science Plant Analysis 8: 195-207.

Sugden, A.M. 1981. Aspects of ecology of vascular epiphytes in two Colombian cloud forests: II. Habitat preferences of Bromeliaceae in the Serrania de Macuira. Selbyana 5: 264-273.

Suresh, H.S., Sukumar, R. 1999. Phytogeographical affinities of Flora of Nilgiri Biospehere Reserve. Rheedea 9: 1-21.

Sukumar, R. Suresh, H.S., Ramesh, R. 1995. Climate change and its impact on tropical montane ecosystems in southern India. Journal of Biogeography 22: 533-536.

Sri-Ngernyuang, K., Kanzaki, M., Mizuno, T., Noguchi, H., Teejuntuk, S., Sungpalee, C., Hara, M., Yamakura, T., Sahunalu, .P, Dhanmanonda, P., Bunyavejchewin, S. 2003. Habitat fragmentation of Lauraceae species in a tropical lower montane forest in northern Thailand. Ecological Reseach 18:1—14.

121

Srivastava, R.K. 2000. Seed germination tests of pioneer shola species in Nilgiris. Annals of Forestry 8: 293-294.

Srivastava, R.K. 2001. Biotic pressure and entry of exotics in shola grassland ecosystem of upper Palnis. Indian Journal of Forestry 24: 324-327.

Still, C.J., Foster, P.N. Schneider, S.H. 1998. Simulating the effects of climate change on tropical montane forests. Nature 398: 608-610.

Stone, C., Kathuria, A., Carney, C., Hunter, J. 2008. Forest canopy health and stand structure associated with bell miners (Manorina melanophrys) on the central cpast of New South Wales. Australian Forestry 71: 294-302.

Strayer, D.L., Power, M.E., Fagan, W.F., Pickett, S.T.A., Belnap, J. 2003. A classification of ecologi al boundaries Bioscience 53: 723-729.

Sudhakara, K. 2001. Inventory and computerized herbarium of higher plants in the sholas of Munnar, Idukki district. Pages 179-208 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Svenning, J-C. 2001. Environmental heterogeneity, recruitment limitation and the mesoscale distribution of palms in a tropical montane forest (Maquipucuna, Ecuador). Journal of Tropical Ecology 17: 97-113.

Swarupanandan, K., Sashidharan, N., Chacko, K.C., Basha, S.C. 2001. Pages 237-258 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Thomas, S.M. Palmer, M.W. 2007. The montane grasslands of the Western Ghats, India: Community ecology and conservation. Community Ecology 8: 67-73.

Thomas, T.P., Sankar, S. 2001. The role of sholas in maintaining watercourses in the High Ranges of Kerala. Pages 71-115 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Swarupanandan, K., Sasidharan, N., Chacko, K.C., Basha, S.C. 2001. Floristic and ecological studies on the sholas of Idukki district. Pages 259-286 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India.

Tabarelli, M., Mantovani, W., Peres, C.A. 1999. Effects of habitat fragmentation on plant guild structure in the montane Atlantic forest of southeastern Brazil. Biological Conservation 91: 119-127.

Troll, C. 1973. The upper timberlines in different climatic zones. Arctic and Alpine Research 5: A3-A18.

122

Turner, I.M., Corlett, R.T. 1996. The conservation value of small, isolated fragments of lowland tropical rain forest. Trends in Ecology and Evolution 11: 330-333.

Warren, R.J. 2008. Mechanisms driving understory herb distributions across slope aspects: as derived from landscape position. Plant Ecology 198: 297-308.

Williams-Linera, G. 1990. Vegetation structure and environmental conditions of forest edges in Panama. Journal of Ecology 78: 356-373.

Williams-Linera, G., Domínguez, Garcia-Zurita, M.E. 1998. Microenvironment and floristics of different edges in a fragmented tropical rainforest. Conservation Biology 12: 1092-1102.

Williams-Linera, G. 2002. Tree species richness, complementarity, disturbance and fragmentation in a Mexican tropical montane cloud forest. Biodiversity and Conservation 11: 1825-1843.

Williams-Linera, G. 2003. Temporal and spatial phonological variation of understory shrubs in a tropical montane forest. Biotropica 35: 28-36.

Venkatachalam, S., Kalaiselvi, T., Neelakantan and Gunasekaran, S. 2007. A comparative study on soil microflora and nutrient status of sholas and adjoining vegetation. Indian Journal of Forestry 30: 135-140.

Vishnu-Mittre, Gupta, H.P. 1968. A living fossil community in south Indian hills. Current Science 37: 671-672.

Werner, W.L. 1995. Biogeography and ecology of the upper montane rain forest of Sri Lanka (Ceylon). Pages 343-352 in Hamilton, L.S., Juvik, J.O., Scatena, F.N., editors. Ecological studies 110: Tropical Montane Cloud Forests. Springer-Verlag, New York, USA.

Yahner, R.H.1988. Changes in communities near edges. Conservation Biology 2: 333-339.

Young, K.R., Keating, P.L. 2001. Remnant forests of Volcán Cotacachi, Northern Ecuador. Arctic, Antarctic and Alpine Research 33: 165-172.

Young, K.R., Leon, B. 1995. Distribution and conservation of Peru’s montane forests: Interactions between the biota and human society. Pages 362-276 in Hamilton, L.S., Juvik, J.O., Scatena, F.N., editors. Tropical Montane Cloud Forests. Springer-Verlag, New York, USA.

Zarri, A. A., Rahmani, A.R., Singh, A., Khushwaha, S.P.S. 2008. Habitat suitability assessment for the endangered Nilgiri laughing thrush: A multiple logistic regression approach. Current Science 94: 1487-1494.

Zheng,D.,Chen, J. 2000. Edge effects in fragmented landscapes: a generic model for delineating area of edge influences (D-AEI). Ecological Modelling 132: 175-190.

123

BIOGRAPHICAL SKETCH

Milind Christopher Bunyan was born in Bangalore, India in 1979. The pursuit of an

undergraduate degree in Environmental Science at St. Joseph’s College of Arts and Science,

Bangalore, which he received in 1999, was more by chance than by design. Numerous hikes later

though, he discovered an interest in the natural world. He went on to receive a Master’s in Forest

Ecology and Management from the Forest Research Institute, Dehradun, India in 2001. After a

brief stint as a Research Assistant with Wildlife Conservation Society, India, he worked as a

Research Associate in the Ashoka Trust for Research in Ecology and the Environment (ATREE)

between 2002 and 2004. It was at ATREE that his interest in Ecology encouraged him to pursue

a doctoral degree. Under the supervision of Shibu Jose, he earned his PhD from the School of

Forest Resources and Conservation, at the University of Florida in Gainesville, Florida studying

edge effects in tropical montane forest fragments in southern India.