<%BANNER%>

Edge Effects in a Forest-Grassland Mosaic in Southern India

Permanent Link: http://ufdc.ufl.edu/UFE0024634/00001

Material Information

Title: Edge Effects in a Forest-Grassland Mosaic in Southern India
Physical Description: 1 online resource (123 p.)
Language: english
Creator: Bunyan, Milind
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: edges, forest, fragments, ghats, logistic, metric, montane, multidimensional, multiple, non, regression, scaling, shola, tropical, western
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre: Forest Resources and Conservation thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: 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 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.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Milind Bunyan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Jose, Shibu.
Local: Co-adviser: Long, Alan J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024634:00001

Permanent Link: http://ufdc.ufl.edu/UFE0024634/00001

Material Information

Title: Edge Effects in a Forest-Grassland Mosaic in Southern India
Physical Description: 1 online resource (123 p.)
Language: english
Creator: Bunyan, Milind
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2009

Subjects

Subjects / Keywords: edges, forest, fragments, ghats, logistic, metric, montane, multidimensional, multiple, non, regression, scaling, shola, tropical, western
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre: Forest Resources and Conservation thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: 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 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.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Milind Bunyan.
Thesis: Thesis (Ph.D.)--University of Florida, 2009.
Local: Adviser: Jose, Shibu.
Local: Co-adviser: Long, Alan J.
Electronic Access: RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2010-02-28

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2009
System ID: UFE0024634:00001


This item has the following downloads:


Full Text

PAGE 1

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 FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2009

PAGE 2

2 2009 Milind Bunyan

PAGE 3

3 To, Preethino one deserves th is more (Proverbs 31: 29-30)

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank the chai r of m y Graduate Committee, Dr. Shibu Jose for giving me the opportunity to pursue a doctoral degree a nd 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. Al an Long, Dr. Michelle Mack and Dr. George Tanner for their time, guidance and encouragem ent through my program. I thank Dr. Ankila Hiremath in particular, for help wi th the logistics of research in India. I am especially grateful to Dr. Robert Fletcher for help with my resear ch and methodology and Ad itya 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 Erav ikulam National Park and Pampadum shola National Park respectively. I would like to tha nk 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 coll ection 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. Ganesans (ATREE) assistance wa s 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 coff ee, food and love from my wife Preethi and right near the end, smile s and kisses from my one-year old, Rahael.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4 LIST OF TABLES ...........................................................................................................................7 LIST OF FIGURES .........................................................................................................................8 ABSTRACT ...................................................................................................................... ...............9 CHAP TER 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 EFFECT S IN THE S HOLA 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 Fragm ents ..........................................................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

PAGE 6

6 Methods ..................................................................................................................................51 Study Area .......................................................................................................................51 Sampling Protocol ...........................................................................................................53 Data Analyses .........................................................................................................................54 Vegetation Structur e and Com position ...........................................................................54 Edge-interior Gradients in Species Distributions ............................................................55 Results .....................................................................................................................................56 Vegetation Structur e and Com position ...........................................................................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

PAGE 7

7 LIST OF TABLES Table page 1-1 Faunal species richness in tropical montane forests .......................................................... 221-2 Surface soil chemical characterist ics in tropical montane forests ..................................... 232-1 Location of tropical montane forest fr agments across study sites in the Western Ghats, southern India. ........................................................................................................422-2 Plant-essential nutrient concentrations in tropical montane forest and grassland surface soils in the Western Ghats, southern India. ........................................................... 422-3 Principal components scores for soil nutri ent variables in trop ical montane forest fragments and grassland soils ............................................................................................432-4 Edge-interior gradients in tropical montan e forest fragments in the Western Ghats, southern India ................................................................................................................ .....433-1 Location and vegetation characteristics of tropical montane forest fragments studied across the Western Ghats, southern India. .........................................................................633-2 Species richness, dominance and divers ity estimates in the shola-grassland ecosystem mosaic in the Wester n Ghats, southern India. ..................................................643-3 Similarity and complementarity between tropical montane forest (shola) fragments in the Western Ghats, southern India. .................................................................................... 653-4 Pearson correlation coefficients for envir onmental variables with axis scores of NMS ordination. ................................................................................................................... .......664-1 Topographical variables used to predict pr esence of shola fragments in the Western Ghats, southern India. ........................................................................................................834-2 Topographical parameter coefficients a nd AuC estimates for predictive models. ............ 83A-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. ..................................................................................................................................96A-2 Understory species presence matrix al ong an edge-interior gradient across nine fragments in the shola-grassland ecosystem mosaic in the Western Ghats, southern India .................................................................................................................................103

PAGE 8

8 LIST OF FIGURES Figure page 1-1 Alternate stable state diagram for the shola-grassland ecosystem m osaic. ....................... 242-1 Location of study sites in the Western Ghats. .................................................................... 442-2 Schematic representation of estimation of distance to additional edges. ........................... 452-3 Effect of the shola-grassland edge on microenvironment variables in the sholagrassland ecosystem mosaic .............................................................................................. 462-4 Edge-interior gradients in microenvi ronment and soil nutrient parameters in Pampadum shola National Park. Error ba rs represent one st andard error. ........................473-1 Species accumulation curves, species richness estimates and rare species (singletons and doubletons) for overstory a nd understory species.. ..................................................... 673-2 Mean understory density along the edge-int erior gradient in tr opical montane forest (shola) fragments.. ........................................................................................................... ..683-3 Edge-interior gradients in diameter dist ribution of overstory tropical montane forest (shola) species across nine fragments.. ..............................................................................693-4 Non-metric multi-dimensional scaling (NMS) ordination of overstory data.. ................... 703-5 Non-metric multi-dimensional scaling (NMS) ordination of understory data. .................. 714-1 Location of study sites in the Western Ghats. .................................................................... 844-2 Histogram of Intercept coefficients for Eravikulam National Park and Nilgiris data subsets.. ..................................................................................................................... .........854-3 Histogram of Elevation coefficients for Eravikulam National Park and Nilgiris data subsets.. ..................................................................................................................... .........864-4 Histogram of Slope coefficients for Er avikulam National Park and Nilgiris data subsets. ...................................................................................................................... .........874-5 Histogram of Northness coefficients for Er avikulam National Park and Nilgiris data subsets.. ..................................................................................................................... .........884-6 Histogram of Eastness coefficients for Er avikulam National Park and Nilgiris data subsets. ...................................................................................................................... .........894-7 Histogram of Curvature of the Slope coefficients for Eravikulam National Park and Nilgiris data subsets. ..........................................................................................................90

PAGE 9

9 Abstract of Dissertation Pres ented 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 consis t 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 va riables and overstory sp ecies in 10 m plots along an edge-interior gradient at 5 m intervals. Understory dens ity and richness was recorded from 5 m sub plots. Conven tional distance to one-edge m odels indicated edge-interior gradients in relative humidity (p = 0.018), magne sium (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 unders tory) 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 re vealed greater variation among fragments than could be attributed to edge-related within fr agment variation. Overst ory composition in midelevation fragments differed significantly from high-elevatio n fragments while understory

PAGE 10

10 vegetation varied as a function of fragment size. Our data indicat e that mid-elevation fragments should not be considered with high elev ation fragments in future studies. We also used a multiple logistic regressi on 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 curv ature and wetness index were used to predict the presence of shola fragments. We observed that shola fragme nts were more likely to occur on northern and western aspects than southern a nd 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 in sights into fragmentation related patterns in small patches and recommendations are made for future investigations.

PAGE 11

11 CHAPTER 1 THE SHOLA-GRASSLAND ECOSYSTEM MOSAIC: A SYNTHESIS OF EXISTING LITERATURE Introduction Tropical m ontane forests (alternatively called tropical montane cloud forests or simply cloud forests) represent some of the most th reatened ecosystems globally. Tropical montane forests (TMF) are characterized and defined by the presence of persistent cloud cover. A significant amount of moisture may be captu red through the condens ation of cloud-borne moisture on vegetation distinguis hing TMF from other forest t ypes. Bruijzneel and Hamilton (2000) described five kinds of TMF. Four of these, i.e. lowe r 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 w ith mountain range size and insularity or proximity to coast. Due to the mass-elevation e ffect (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 enab le 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 increas es. Other distinctive f eatures 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 dr iven by the limited availability of habitat (Bubb et al. 2004). Located in the headwater catchme nts of seasonal or perennial streams, TMF provides often undervalued ecosystem se rvices to downstream communities.

PAGE 12

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 gra sslands and is commonly referred to as the shola-grassland ecosy stem. With limited exceptions (Werner 1995, Somanathan and Borges 2000), data from the s hola-grassland ecosystem mosaic are rarely included in biome-wide popular (Bruijzneel a nd Hamilton 2000) as well as academic (Hamilton et al. 1995, Bubb et al. 2004) synopses of scientific lite rature. As such, this document aims to provide a synthesis of current research and the state of know ledge of the shola-grassland ecosystem from peer-reviewed literature publis hed 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 W estern 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 sholagrassland ecosystem mosaic for decades. While some of the earliest scientific de scriptions of the shola-grassla nd ecosystem described the mosaic as dual climax (Ranganathan 1938), proponents of the single climax concept (Clements 1936) argued that the forests represente d a biotic (Bor 1938, Noble 1967) or edaphic climax (Jose et al. 1994). A C13 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

PAGE 13

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 brie f warm, wet phase around 600-700yr BP also occurred (Sukumar et al. 1993). What is not a Shola? Arguably, the shola-grassland ecosystem mosa ic is among the most distinct ecosystem types in the WGSL biodiversity hotspot. Although, sholas are t ypically 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 characteristic ally 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 growt h. 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 fore st with distinct physiognomy. Studies on sholas at elevations below 1700 m should be re stricted to shola fragments surrounded by grasslands. Indeed, in plots located at lowe r elevations, Sudhakara et al. (2001) recorded families uncommon to sholas but common to lowland forests (Bombaceae, Clusiaceae, Dichapetalaceae) Flora Since Thomas and Palm er (2007) provide a co mprehensive 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 c ontain species of both tropical and temperate

PAGE 14

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 s hola 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 ot her hand are IndoMalayan or Indi an (rarely Paleotropical) in origin (Suresh and Sukumar 1999, Nair and Men on 2001). Overstory species in the shola are dominated by members of Lauraceae, R ubiaceae, 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 unde rstory) 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 curv es, we might expect that the limited availability of suitable habitat for shola specie s within the shola-grassland ecosystem mosaic would limit -diversity. However estimates for -diversity are highly variable. Estimates for Shannon-Weiners diversity index (H') range fr om 4.71 (Sudhakara 2001) to 0.87(Murali et al. 1998). Estimates for endemism are also highly variablefrom 19.5 % (Suresh and Sukumar 1999) to 83.3 % (Nair and Menon 2001). Historically, the regenerati on of arborescent flora in shola fragments had been expressed as a conc ern (Vishnu-Mittre and G upta 1968). A series of studies now indicate that shol a species show adequate rege neration under natural conditions

PAGE 15

15 (Jose et al. 1994, Swarupanandan et al. 2001). A dditionally, germination rates as high as 95% have been recorded for shola species in germination trials (Srivastava, 2000). As with other TMF, shola fragments exhibi t prolific epiphytic gr owth. Studies in TMF have shown that epiphytic species may constitute up to 25% of all bioma ss in tropical montane forests (Foster 2001). They also provide micr ohabitats 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, Da vidar et al. 2007, also see Nair et al. 2001), very limited work exists on epiphytes in the shola-grassland ecosyste m mosaic (Abraham 2001). Similarly, very few studies have quantified pr oductivity in the shol a-grassland ecosystem mosaic. In a comparison of net primary productiv ity (NPP) patterns of e xotic plantations and native shola forest, NPP and bioma ss 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 pr ovides 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-mos aic is the Nilgiri tahr ( Niligiritragus hylocrius ) which has been studied meticulously over th e years (Davidar 1971, Rice 1984 Rice 1988, Mishra and Johnsingh 1998, Daniels 2006). Although considered a flagship species for the ecosystem, uncertainty over

PAGE 16

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 mi rror the shola-grassland ecosystem mosaic pattern through habita t preferences. Small mammal communities in the Nilgiris (two species of the nine recorded), showed a high degree of pref erence 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 selec tion 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 habita t use typically restri cted to shola cover might extend to exotic plantations (unsuitable habitat) when locat ed 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, inventor ies 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 sc ientific literature on fa unal species in the sholagrassland ecosystem mosaic is limited. (See Table 1-1) Hydrology Globally, tropical m ontane forests (alternativel y tropical montane cloud forests) have been shown to significantly influence ecosystem hydrology and biogeochemistry (Bruijnzeel and Hamilton, 2000). In addition to providing cover a nd 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 attribut ed to condensation of wind-driven fog on tree crowns (termed fog drip). In areas of low preci pitation such as the Canary Islands, interception of cloud water can double annual precipitation (G ioda et al., 1995). Pr otection of tropical

PAGE 17

17 montane habitat serves the purpose of hydrological regulation for downs tream consumers also. This is especially significant in the Western Ghats where major rivers originating in the sholagrassland ecosystem mosaic provide hydrological services to consumers. A study by Krishnaswamy et al. (2006) dem onstrated that indivi dual rainfall events could contribute as much as 20-30% of the annual se diment load of 239-947 Mg km-2 (Krishnaswamy et al. 2006). Other studies report signif icantly 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-g rassland ecosystem mosaic are granitic or me tamorphic 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 moistu re loss. During the dry season, shola soils have been shown to retain as much as twice the so il moisture in the surround ing grasslands (Thomas and Sankar 2001). Shola and grassla nd soils also differ nutritionally. Total N, available P and K are higher in the sholas as compared to adjoin ing grasslands. Though this could be attributed to higher litter decomposition and nutrien t recycling rates in the sholas these differences are rarely significant. Jose et al. (1994) report organic ca rbon content in shola surface soils that are comparable to those recorded in TMF in Ec uador. These values though are much higher than those recorded by other authors for surface soil s under varied types of cover in the sholagrassland 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 adja cent grasslands although differences between sholas and adjacent grassl ands have been observe d (Nandakumar et al. 2001). Shola soils have higher soil nutrient pools than those under exotic plantations of blue gum

PAGE 18

18 ( E. globulus ) or tea ( Camellia sinensis ) (Jeeva and Ramakrishnan 1997, Venkatachalam et al. 2007). Nutrient cycling under natural shola vegeta tion has also been described as steady state and less likely to suffer losses to leaching since th e return through leaf li tter is low (Jeeva and Ramakrishna 1997). Trends in biogeographical affi nities have also been reco rded 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 Balasundara m 2001). Soil fungal sp ecies diversity is comparable between shola fragments and grasslands (H'SHOLA=4.18, H'GRASSLAND=4.18) albeit highly habitat specific (Sankaran and Balasunda ram 2001). Shola soils also had significantly higher soil bacterial and actinom ycetes populations than grassl and or plantation soils while fungal populations were highest in grassland soils. Plan tation 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 curren t dynamic equilibrium between insular shola fragments and grasslands is indicative of the existence of alternate stable st ates 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-gras sland 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 ha rvesting (Gupta 1962). The persistence of the mosaic in areas relatively free of anthropogenic grazing and illegal harves ting in some protected areas make these two parameters tenuous for e xplaining the pattern. Alth ough some authors have

PAGE 19

19 observed grassland soils to be sh allower than shola soils (Jose et al. 1994); others (Nandakumar et al 2001) did not find a consis tent trend. Moreover, shola spec ies 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 vege tation 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 rev ealed that tropical montane forests in the Indian subcontinent accounted for 8.07% (92 fires) of all fires (Vadre vu et al. 2008). Although current understanding points to fire as the dominan t factor responsible for the maintenance of the edge, ambiguity remains. Conclusion Historically, the shola-grassland ecosystem mosaic has undergone ex tensive habitat loss. Plantations of exotic tree species were established in the grasslands aimed at augmenting timber production as early as 1843 (Pala nna 1996) with further introducti ons in 1870 in the Palni hills (Srivastava 2001). Plantation program s were expanded under colonial rule to establish extensive tea plantations in the mosaic. Post independence, tree plantati on programs also received national (federal) budgetary support (R aghupathy and Madhu 2007). Signifi cant populations of invasive shrubs and herbs (Eupatorium glandulosum Ulex europaeus and Cytisus scoparius) in the sholagrassland ecosystem mosaic were reported by ear ly 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 th e ecosystem (Seshan 2006). Threats to the mosaic today include the harves ting of shola species to meet biomass and fuelwood requirements and cattle grazing (Sekar 2008). In ar eas adjoining settlements, thes e threats can be significantly

PAGE 20

20 amplified altering patterns in sp ecies richness and dominance (C handrasekhara et al. 2001). The WGSL biodiversity hotspot is like ly to undergo extinctions in plan t and vertebrate species due to the limited availability of habitat (Brooks et al. 2002). Further, globall y, TMF experience higher annual loss in habitat than any other tropical fo rest 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 co upled 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-grassl and 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 anthropoge nically induced pressures (e.g. fire, conversion to pasture). As such, edge eff ect studies in the shola-grassla nd ecosystem might be especially insightful since species in older fragments have had time to equilibrate with fragmentationinduced pressures (Turner et al. 1996, Harper et al. 2005). Fragmentation studies often observe a proportional increase in area unde r 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 sholagrassland ecosystem would help us understand patterns in small fr agments since shola fragments in the shola-grassland ecosystem mosaic are often small (~1 ha). In this study I investigate the effect of the shola-gra ssland edge on microenvironment gradients in the ecosystem mosaic. I then assess patterns in species distributions and structure

PAGE 21

21 between and within fragments studied. I expect these to provide an unde rstanding of patterns in smaller, older fragments. In mountainous terrain (where the shola-grassla nd ecosystem mosaic is found), topography can significantly determine f actors influencing vege tation (such as the occurrence fire and frost). Topography can thus serve as a surrogate for these variables. I therefore investigate the topographi cal variables determining the pr esence of shola fragments in the mosaic. Through this, I seek to determine processes driving the shola-grassland edge.

PAGE 22

22 Table 1-1. Faunal species richne ss 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 (5colony forming units-1soil)

PAGE 23

23 Table 1-2. Surface soil chemical characte ristics 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# 183.40a 6.8a 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

PAGE 24

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)

PAGE 25

25 CHAPTER 2 MICRONENVIRONMENT EDGE EFFECTS IN THE S HOLA GRASSLAND ECOSYSTEM MOSAIC Introduction The crea tion 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 dive rsity (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 pattern s (Anderson et al 1977) and ecosy stem structure (Laurance et al. 1998, Gascon et al. 2000) near ed ges. At a newly created edge there are immediate changes in the flow of energy, materials and organisms (t ermed 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 adapti ng their distributions. Finally, this causes a change in the interactions between species (Ries et al. 2004). Feedbacks between these mechanisms (ecological flows, increased acce ss, resource mapping and species interactions) occur iteratively as primary and secondary responses with increasing edge age. Numerous studies have quantified the mechanis ms and patterns observe d at edges as they relate to fragmentation of fo rested 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 compositi on and its interaction with the abiotic environment. Increased light transmittance at newly cr eated edges alters microenviro nment conditions resulting in elevated levels of air temperat ure and reduced relative humidity. Soil characteristics are altered

PAGE 26

26 too as elevated soil temperatures, nutrient cycl ing and litter decomposition rates and diminished soil moisture content are seen near edges. As the edge evolves, changes in the abioti c 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 e xotic invasives and the elimination of predators can result in the comp lete collapse of fragments in threatened ecosystems (Curran et al. 1999, Gascon et al. 200 0). Further, meso-scale climate phenomena such as the El-Nio Southern Oscillation can am plify susceptibility to threats such as fires (Cochrane, 2001). Following the creation of an e dge, edges may either deepen (as discussed above) or regenerate as organism s adapt to the physiological cons traints of the edge environment (Laurance et al. 2002). However at habitat edges that are maintaine d, a dense, vertical layer of vegetation can develop, sealing off the forest in terior 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 th e contrast between adjacent habitats (patch contrast) is high such as a forest-grassland edge and can affect a fragments persistence (Gascon et al. 2000, Harper et al. 2005). In tropical montane forests too (T MF), the response to habitat edges may differ based on edge type. For exampl e, Lpez-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 grad ients are limited in TMF (Jose et al. 1996, Lpez-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

PAGE 27

27 TMF (Hamilton et al. 1995). Moreover, microenvi ronment edge-interior gradients in TMF can operate at different scales than lowland forest s. Microenvironment edge -interior gradients in lowland forests typically vary between 40-60m from the fore st edge (Kapos 1989, WilliamsLinera 1990, Didham and Lawton 1999) with edge-e ffects 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), wh ile other studies have reported an insignificant effect of patch contrast on seedling growth and defoliation (Lpez-Barrera et al. 2006). Some edges may even display reverse edge-effects w ith lower nest predati on 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 biom e) typically consider the effect of distance from the nearest edge. This o ffers 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 ed ge-effects as a function of distance from the nearest edge, limited studi es exist on multiple edge effects (Malcolm 1994, Fernndez et al. 2002, Zheng and Chen 2003, Fletch er 2005). Interestingly, even large scale manipulative studies on fragments of various sizes do not consider th e 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 distan ce to nearest edge alone Further these models become increasingly significant in small fragments that may be dominated by edge habitat

PAGE 28

28 (Gascon et al. 2000). Studies have used different techniques to investigate the effect of multiple edges on response variables. Malcolm (1994) m odeled the edge as a collection of points integrating the effect of each of these points alon g the edge (point edge effects). However, as Fletcher (2005) observed, point edge effects offer a robust, yet abstract es timate of edge effects since point edge effects are ha rd to measure. Fletcher ( 2005) also observed that bobolink ( Dolichonyx oryzivorus ) distributions were influenced by mu ltiple 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 edge s in edge-effect models has statistical implications too. As the number of parameters incr eases, the power of the model reduces (the probability of Type II error increases). Mancke a nd Gavin (2000) suggested the use of an integrative index that co mbines the use of distance to ed ges 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 resp ect to the edge-interior transect. Within the Western Ghats-Sri Lanka (WGSL) bi odiversity 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. S hola 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 th e Western Ghats within the WGSL hotspot. I quantified effects of the shola-grassland edge on microenviro nment and soil variables at a regional scale in the shola-gr assland ecosystem mosaic. I tested for differences in microenvironment and edaphic variables betwee n shola fragments and grasslands and to determine if the edge is controlled by these variables. I also tested for gradients in these variables

PAGE 29

29 as a function of distance from the edge and by doi ng 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 gradie nts) while large fragments (> 50 ha) will exhibit strong edge-interior gradients. I also tested for edge-effects in grassland plots along a grasslandedge-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 W estern Ghat s, a region of high conservation importance (Myers et al. 2000) in southern India. Three study sites were se lected 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 21). BRT Wildlife Sanctuary (540km2), is located at the eastern-m ost edge of the Western Ghats (11N to 12 and 77 to 77E). Soils in BRT have been classi fied 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 fr om November to May. Rainfall ranges between 898 mm to 1750 mm and strong sp atial trends have been obser ved 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 deci duous forests, evergreen forests and the high-altitude TMF-grassland (or shola-gr assland) 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 re stricted the number of fragments available in BRT. Although shola vegetation is more extensive in the southern por tion of the sanctuary,

PAGE 30

30 (particularly near Jodigere shol a) most of these areas were co nsidered 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 to 10N and 77 to 77E) 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 temper atures 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 m ean mimimum temperature 3 C (Rice 1984). The soil had been classified as Alfisols and descri bed 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 rain forest formations can also be found at lower elevations, vegetation in the park is predominan tly shola-grassland ecosystem mosaic consisting of rolling grasslands interspersed with dense, in sular shola fragments. Five shola fragments were selected within ENP (Table 2-1). Pampadum shola National Park (PSNP) located w ithin 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 77E and 10N) its proximity to the larger Mannavan shola (77-77N and 10-10S) provides insights

PAGE 31

31 into broad climatic trends. Mean annual temperat ure in Mannavan shola has been reported to be approximately 20C with mean minimum temperature in the coldest months (December and January) reaching 5C. Mean annual precipitat ion in Mannavan shola ranges between 2000-3000 mm. Data from PSNP were collected to compar e patterns observed against smaller fragments chosen in BRT and ENP. Sampling Protocol To investigate edge-interior gradients in shola fragm ents, transects were laid out perpendicular to the shola-grassland edge. Plots (10 m) were established along these transects with the first plot being established at the edge its elf. The plot was laid out with the longer side (10 m) running parallel to the edge to capture maximum variab ility along the edge-interior gradient. Successive plots along tr ansects were separated by 5 m and extended till the middle of the patch or to a distance of 35 m from the e dge, whichever was shorte r. 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 W ithin each 10m plot, microenvironmen t 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 (Oakt on Instruments, Vernon Hills, IL). Photosynthetically active radiation (PAR) was measured above and be low shola canopy using an AccuPAR LP-80 Ceptometer (Decagon Devices, Pull man, WA). Light transmittance ( ) 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 ca nopy PAR was measured in the adjoining grassland prior to measurement of below canopy PAR in the shola plots. From each

PAGE 32

32 10 m plot, one surface soil (030 cm) sample was collected usi ng a soil corer and immediately double sealed in Ziplock bags. Soil cores were collected from all but one fragment (Karadishonebetta, BRT) Gravimetric water cont ent was measured by oven-drying soil samples for 24 hours at 105C. Soil samples were analyz ed for soil organic carbon (Walkley-Black method), total nitrogen (micro-Kjeldahl disti llation method), and extr actable phosphorus (BrayKurtz 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, c opper, boron and molybdenum were estimated using an Atomic Absorption Spectrophotometer (Soultanpour and Schwab 1977). Grassland Plots In four of the shola fragm ents studied, transects were also es tablished in the grassland to study effects of the shola-grassland edge on the gr assland. Since these transects were established as matching pairs to the shola-transects, they or iginate 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 grassla nd-edge-shola interior transects. Data Analyses In sm all 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 e dge in orthogonal direct ions, thus achieving a measure of distance to edge in each cardinal direction. This approach offers an optimum combination of incorporating information fr om additional edges while ensuring orthogonal

PAGE 33

33 separation (Mancke and Gavin 2004). The spatial an alysis 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 neares t 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. Tukeys HSD was used for multiple comparisons (familywise error rate p < 0.05) to test for differen ces across the six distan ce classes (two in the grassland and four in the shola fragments). Additi onally, a t-test was used to test for differences in soil microenvironment conditions between grassl and 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 sma ller 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 nut rient concentrations (using a simple linear regression) in shola fragments (n = 8). A regression analysis was used to study the e ffect of the forest (shola)-grassland edge on microenvironment (n = 9) and soil variables (n = 8) in the sholas. A simple linear regression was

PAGE 34

34 also used to quantify the response of microenvi ronment 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 fragme nts (>1 ha13 ha). Response variables were regressed against distance estimates and log-transformed distance estimates. Log-transformed distance estimates weight plots n ear edges higher since I have an a priori expectation that edgeinterior gradients in shola fragments will be sh ort 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 tem perature (p = 0.031) and soil temper ature (p < 0.020) declined significantly along the grassland-edge-shola interior transects (Fig 2-3). Although gr assland plots have significantly lower humidity (p = 0.034) than shola plots no gradients were observed along grassland-edgeshola interior transects (p = 0.47). Nutritionally, shola and grassland soils differed little (Table 22) 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 variat ion (Table 2-3). A varimax rotation of the selected components was used to ma ximize orthogonal separation between components and reduce loading of soil nutrients on more than one principal co mponent (or cross-loading). An ANOVA of the selected soil nutrient principal com ponents revealed that on the first principal component (PC1), grassland and shola soils diffe red 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 pr incipal components (PC 2, PC3 and PC4) varied

PAGE 35

35 between habitat type (p 0.35). Further none of the factors va ried along the two-sided grasslandedge-shola interior transect (p 0.22). Microenvironment Gradients in Shola Fragments In the large shola fragment (PSNP), relativ e hum idity and soil temperature decreased linearly with increasing distan ce 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 obs erved in air temperatur e (p = 0.74) and light transmittance (p = 0.79). However, light transmittance decreased to 3(.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 pota ssium increased (p = 0.0089) with increasing distance from edge (Table 2-3, Fi g 2-2). Contrary to other studi es that have reported strong nonlinear 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 si gnificantly improve the fit of relative humidity (p = 0.018) and reduced the signif icance of estimates of available potassium (p = 0.030) and other significant variables (p 0.101). A factor analysis of soil nutrien t variables with quartimax rotation for PSNP soils revealed two factors as significant that explained 72% of the variation (data not presented). While the fi rst factor loads heavily on organic carbon, total nitrogen, available phosphorous and calcium, th e 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 transmitta nce) or soil variables (soil temperature, macronutrient and micronutrient concentrations) as a function of distance of neares t edge (single edge model). Responses were weaker still when log-distances were used (p 0.242). The incorporation of

PAGE 36

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 patter ns were observed as air temperature (p = 0.0320) and light transmittance (p = 0.0072) decreased. W eaker trends were observed with relative humidity increasing and soil temperature decreasi ng with increasing distance from edge (Table 2-4). Soil macronutrient variables were not correlated w ith edge distance using either the oneedge 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 dist ances in order to test for corre lations between the variables. However, the analysis failed to fi nd any significant correlations. Discussion The m icroenvironment in grasslands was signi ficantly different from that of shola fragments. However, no signifi cant differences were observed between grassland and shola fragment soils. Similar trends in soil response variables have b een reported by other authors too (Jose et al. 1994, Nandakumar et al. 2001). Our es timates for soil nutrient variables for sholas and grasslands (Table 2-1) are comparable to others from the ecosystem mosaic (Thomas and Sankar 2001). However, significan tly higher estimates for soil organic carbon and soil moisture have also been reported from the ecosystem mosa ic (Jose et al. 1994). Factor analysis of soil variables indicates that soil or ganic carbon and calcium load had positive and negative loading on the first factor (which explaine d 20% of the variation in the data set). This is typical of forest soils that have higher soil orga nic carbon while basic cations (Ca2+) are lower (as they are lost to leaching) than grassland soils.

PAGE 37

37 Although two-sided edges are less investigated, th eir proportion vis--vis one sided edge studies has increased considerably (Fonseca and Joner 2007). Their relevance to matrix-based approaches to understanding edge-effects and re storation 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), ha s been shown to be hostile to animal taxa (Shanker 2001, Zarri et al. 2008). In our study I f ound evidence of two-sided edges in microenvironment response variables. However this pattern wa s absent in soil response variab les 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 restorati on efforts by providing requi site conditions for reestablishment 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-grassl and edge in Sri Lanka experiencing dieback (Ranasinghe et al. 2007). Although Ranasinghe et al. (2007) suggest ed 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-gra ssland soils in Sri Lanka, shola fragment and grassland soils alike did not record high DTPA extract able manganese or copper levels. Large shola fragments exhibit different patterns from sm aller shola fragments. In Pampadum shola distance from the nearest e dge was sufficient to explain trends in microenvironment and edaphic variables. Howeve r, distance from the nearest edge could not sufficiently explain the soil variables.

PAGE 38

38 Seasonal and diurnal variation Since individual transects were sam pled over th e course of a single da y, 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 te mperature, it can be argued, c ould be an artifact of either distance from the forest fragment (shola)-grassland edge or diur nal variation. However, in our study, air temperature and relative hu midity 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 pr ogressed) while air temperature showed no edgeinterior trends. Such changes in relative humidity trends as a function of distance from the edge could also be an artifact of seasonality. Sampli ng of the large fragment (PSNP) was done in the wet season and often under constant cloud cover l eading to higher relative humidity near the shola-grassland edge and lowe r 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 e dge) 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%) repo rted by Jose et al. (1996) although similar to estimates from old growth rainforest-pasture ed ges in Mexico (4.7%, Williams-Linera et al. 1998). Other studies too have reported an abrupt change in li ght conditions across a high contrast edge (Lpez-Barrera et al. 2006). Shola fragments t hus 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 othe r edge-effect studies (e.g. soil

PAGE 39

39 temperature) both in tropical montane forests and lowland tropical fo rests (Jose et al. 1996, Laurance et al. 2002). The present study investigates edge-interio r gradients in microenvironment and soil nutrient availability across nine tropical montane (shola) fragments. To my knowledge only other study in the shola-grassland ecosystem investigat es edge-interior gradients (Jose et al. 1996). Although the study reported distin ct 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 dem onstrated the effect of distance from edge on microenvironment and edaphic variables. In tropical forests, edge effects have been particularly worrisome as microenvironment grad ients extend as much as 100m away from the edge of forested habitat (Laura nce 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 respon se 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 mo saic, large shola fragments exhibited different patterns from smaller shola fragments. In Pampadum shola di stance from the nearest edge was sufficient to explain trends in microenvironment and edaphic va riables. However in smaller forest fragments, microenvironment and edaphic vari ables did not show a response w ith distance from the nearest edge suggesting that fragments ar e dominated by edge habitat. However, the incorporation of additional edges as predictor va riables 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 fo r multiple edge effects. The incorporation of

PAGE 40

40 multiple edge-effect models might better explain the absence of edge interior gradients in other edge effect studies also (W illiams-Linera et al. 1998). Delineating edges and determining edges using spatial analysis techniques may not always be feasible though. In our study, insular forest frag ments are isolated from other forest types by a high contrast grassland edge. This made determin ation of distance to a dditional edges relatively easy. When edges separate simila r habitat types (such as a native forest-tree plantation edge) distinguishing habitat types mi ght require the use of other techniques (landcover/landuse classification). Conclusion The present study describes th e response of m icroenvironment and edaphic variables to distance from a tropical montane forest (shola)-gr assland edge. Nutritiona lly, grassland and shola soils differed little. Since such results have been consistently observed across multiple study sites in the ecosystem mosaic, an edaphically controll ed 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 e xplained 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 sm all 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 nonlinear effects which often better explain patterns in edge-interior gradients (Malcolm 1994). The incorporation of such non-linear m odels in the system might further improve model fit. Finally, further research is required in investigating the e ffect of multiple edge models in predicting edge

PAGE 41

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

PAGE 42

42 Table 2-1. Location of tropical montane forest (shola) frag ments 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 nutri ent concentrations in tropical montane forest (shola) and grassland surface soils in the Western Ghats, southern India. Values are means ( 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

PAGE 43

43 Table 2-3. Principal components sc ores for soil nutrient variable s in tropical montane forest (shola) fragments and grassland soils (n = 4) in the Western Gh ats, southern India (eigenvalue 1). Bold values indicate the component on which the variable loads after a varimax rota tion 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, P HO: Available phosphorous, POT: Potassium, CAL: Calcium, MAG: Magnesium, SOD: S odium, MAN: Manganese, MOL: Molybdenum, COP: Copper Table 2-4. Edge-interior gradient s in tropical montane forest (s hola) 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 0.39900.0276 0.002 0.7418 Potassium (%) logdist 0.5110.0089 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 criterio n (AIC). Predictor vari ables are only displayed for significant models.

PAGE 44

44 Figure 2-1. Location of study sites in the Western Ghats. Insular forest fragments were selected from three sitesBiligiri 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 studie d 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

PAGE 45

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 esti mates 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

PAGE 46

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

PAGE 47

47 2.512.522.532.5 Magnesium (%) 0.010 0.012 0.014 0.016 0.018 Relative humidity (%) 50 60 70 80 90 100 Potassium (%) 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 grad ients in microenvironment and soil nutrient parameters in Pampadum shola National Park (~132 ha). Error bars represent one standard error.

PAGE 48

48 CHAPTER 3 SPECIES DISTRIBUTION PATTERNS IN SHOLA FRAGMENTS Introduction The creation of an edge in forested habitat alters both abiotic and biotic com ponents of ecosystems. Immediate changes to microenvironm ent conditions at the ed ge higher air and soil temperatures, lowered relative humidity and soil moisture are a result of elevated insolation and turbulence caused by an al teration of the boundary interface. Changes are manifested in altered ecosystem processes: evapotranspiration, litter decomposition, nut rient cycling and dispersal. These ecosystem processes interact it eratively through feedback loops with the biotic component of ecosystems causing a shift in specie s 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 in clude elevated tree mortality (Williams-Linera 1990) and reduced canopy and sub-canopy cover (L aurance 1991). This opening of the canopy manifests itself in increased seed rain from gene ralist species at the edge and an increase in density of liana and understory species (Janzen 1983, Laurance 1991). These sh ifts 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 gr adients (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 attr ibuted to edge creation (Curran et al. 1999).

PAGE 49

49 However these might be an artifact of the colla pse 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 (N adkarni and Wheelwright 2000). -diversity was observed to increase with an increase in elevation in TM F fragments in Mexico indicating that fragment complementarities increased with increasing el evation (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 en demism and the number of unique species assemblages is negatively correlated with fr agment size (Young and Leon 1995, Ganeshaiah et al. 1997). Smaller fragments are thus more lik ely 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 spec ies distributions as a function of distance from the edge in TMF fragme nts in southern India with sub-tropical and temperate species near the edge while species w ith tropical affinities were located in interior plots. On the other hand, studi es in other TMF fragments repor t weak or complex distance to edge trends in species diversity (Oosterhoor n 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 ver tical structure of edge habitat as compared to fragment interiors (Oosterhoorn and Kappelle 2000).

PAGE 50

50 The theory of island biogeography (IBT, Macarthur and Wilson 1963, 1967) offers an elegant and robust explanation of the proce ss of colonization and speciation on islands. Numerous studies since have sought to apply the theory to frag mented and insu lar terrestrial habitats (for a review see La urance 2008). However, as Laur ance (2008) indicates, system properties associated with terr estrial fragments (e.g. non-random habitat conversion, edge effects and matrix hostility) render IBT in adequate in explaining the effect s of habitat fragmentation. In terrestrial ecosystems, unlike true oceanic isla nds, the surrounding matrix has a significant influence on the magnitude and direction of speci es 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 re growth edges (Mesquita et al. 1999). In the Western Ghats, TMF (h ereafter shola) occurs as insula r forest fragments in a matrix of grasslands separated by a stab le, 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 hostil ity of the matrix, as observed in th e response of species to the edge (Shanker 2001) is often suggestive of true ocean ic 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 pa tchy 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 ma y also maintain a hi gh diversity of tree

PAGE 51

51 species while vertebrate (especially mammal) diversity may not be comparable (Bowman and Woinarski, 1994). Finally, edge-eff ect related processes are also dependent on patch size. As fragment size reduces the propor tional 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 (Gas con et al. 2000). However, smaller fragments may serve as long-term refugia for threaten ed 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-grassla nd ecosystem mosaic in southern India. I quantified species distribution patterns as a functi on of distance from the edge. I hypothesize that there will be edge-interior gradients in speci es distributions and expect abundance and dominance patterns to be correl ated with microenvironment and edaphic variables. Since edges offer increased light transmittance to understory vegetation I expect e dges to have greater understory structural complexity and higher overstory basal area th an 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 co mpositional differences between overstory species in sub-montane and montane forests. Methods Study Area Three study sites were selected in the W estern Ghats in southern India within the Western Ghats-Sri Lanka (WGSL) biodiversity hotspot; Biligiri Rangaswamy Temple (BRT) Wildlife Sanctuary in the state of Karn ataka and Eravikulam National Park and Pampadum shola National Park in the state of Kerala ( Fi g 2-1). BRT Wildlife Sanctuary (540km2), located at the easternmost edge of the Western Ghats (11N to 12 and 77 to 77E), occurs at the

PAGE 52

52 confluence of the Eastern Ghats and the Western Gh ats. 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 nor theastern 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 exte nds from November to May with February recorded as the driest month. Vegetation in the sa nctuary is represented by a diversity of forest types ranging from dry scrub forests, dry and mo ist 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 shol a) most of these areas were co nsidered 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 to 10N and 77 to 77E) within the state of Kerala (Fig 2-1). The park (119 km2) consists of a base plateau at an elevation of 2000m surrounded by p eaks with a maximum elevation of 2695 m. The mean maximum temperature record ed within ENP is 16.6 C while the mean minimum is 6C. 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 temp erature 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 ar e sandy clay, have moderate depth (30-100 cm) and are acidic

PAGE 53

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 fo rmer 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 sholagrassland ecosystem mosa ic consisting of rolling grasslands interspersed with dense, insular shol a fragments. Five shola fragments were selected within ENP (Table 3-1). Pampadum shola National Park located within Idukki district (appr oximately at 77E and 10N) 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 20C with mean minimum temperature in the coldest months (December and January) reaching 5C. Mean annual precipitat ion in Mannavan shola ranges between 2000-3000 mm. Data from PSNP were collected to compar e patterns observed against smaller fragments chosen in BRT and ENP. Sampling Protocol To investigate edge-interior grad ients in species distributions in shola fragm ents, transects were laid out perpendicular to the shola-grasslan d edge. Overstory structur e was characterized in terms of basal area and density of tree species in 10 m plots along each tr ansect with the first plot being established at the edge itself. Plots were laid out wi th the longer side (10 m) running parallel to the edge to capture maximum variabil ity along the edge-interior gradient. In each plot, trees 5 cm dbh were identified to species and measured. Overstor y 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 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,

PAGE 54

54 200-249 cm and > 250 cm) based on maximum hei ght (Hmax). Specimens were collected for overstory and understory individuals that could not be iden tified confidently and were later identified with the help of taxonomists at Ashoka Trust for Ecology and the Environment, Bangalore. I also collected data on microenviro nment (soil and air temperature, relative humidity, light transmittance, soil moisture) and soil macronutrient (organic carbon, total nitrogen, available phosphorous, pot assium) variables from 10 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 ex tended 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 classes2.5 m, 12.5 m, 22.5 m a nd 32.5 m (as measured from the shola-grassland edge to the mid-point of the plot)along the e dge-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 asses sed 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). Alt hough the Mao-Tau function does not require resampling runs to calc ulate 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 tw o samples). Species diversity was assessed using Fishers alpha index and richness was estimated using th e nonparametric Bootstrap estimator (Smith and

PAGE 55

55 van Belle 1984) since it is well suited to quadr at data (Chao 2005). Moreover, it provided the most stable and least biased estimate of speci es richness for overstory and understory species. Species diversity was assessed across all fragme nts using Shannon-Weiner index (H') (Magurran 1988). Beta-diversity was computed using the Ch ao Abundance-based Srensen similarity index. These indices correct for the unders ampling bias associated with traditional similarity indices (e.g., Srensen similarity index) and are especially suitable fo r use when rare species are numerous (Chao et al. 2005). Fragment comple mentarity (C) between pairs of fragments was assessed by calculating the Marczewski-Stei nhaus distance (Colwell and Coddington 1994). Complementarity varies from zero (when all sp ecies 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 vert ical structure of overstory (d bh distribution) and understory (Hmax) as a function of distance to the sholagrassland edge. Soil samples were analyzed for macronutrient concentration using standard tech niques (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 com positiona l patterns as a functi on 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 st arting axes) of the autopilot procedure using the Srensen distance measure to calculate dissimilarity coefficients. A MonteCarlo randomization test (50 runs) was used to test the best solution against a randomized dataset. A three dimensional so lution yielding a stress signif icantly lower than obtained by chance was selected for each of the overstory (p = 0.019) and understory datasets (p = 0.019). A

PAGE 56

56 joint plot was constructed to explore the re lationship between species and environmental variables (soil temperature, ai r temperature, relative humidity, soil moisture, soil organic carbon, total nitrogen, available phosphorous and potassi um). Axis scores were correlated with environmental measures to assess the ecological si gnificance of axis in ordination space. For the understory dataset light transmittance was included as an environmental variable. Results Vegetation Structure and Composition I reco rded 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 spec ies), Rubiaceae (18 species), Myrsinaceae (7 species), Myrtaceae and Oleaceae (5 species each). Sp ecies 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 declin e for overstory species but declined marginally for understory species indicating adequate sampling for rare speci es in understory vegetation. Shannon-Weiner diversity indices (H') were 3.9 6 and 3.57 for overstory and understory species respectively (Table 3-2). The Bootstrap estimato r indicated that 12 addi tional overstory species and 13 additional understory spec ies are estimated as not bein g represented in the dataset assuming that the selected fragments were repres entative of the shola ecosystem (Figure 3-1a, 31b). On a landscape scale, shola fragments had high variation between fragments. This was especially true for overstory species with complete separati on often observe d between mid elevation and high elevation fr agments (Chao-Srsensen 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.

PAGE 57

57 Edge-interior gradients in vegetation were la rgely 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 acro ss overstory dbh classes (p 0.202) and trends were inconsistent. With increasing distan ce from edge, the total number of individuals across diameter classes decreased in some frag ments 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) we re the largest trees (> 40cm dbh) present across all distances. In the understory, wh ile the density of individuals in the smallest height class (5099cm) 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 al ong the edge-interior gradient (p = 0.897). Stepwise regression revealed that soil temperature was significantly rela ted 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 microenvironmen t or soil variables sufficiently explained the variation in overstory dominance. Edge-interior Gradients in Species Distributions For the overstory dataset NMS ordination produc ed three significant axes which accounted for 57% of the variation between species in the o rdination 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). Th e alignment of overstory plots was better in terms of fragments than distance to edge classes with lower el evation shola fragments occupying the south eastern portion of the ordination cloud. Axis 2 showed a strong correlation with

PAGE 58

58 relative humidity (r = 0.536) and so il moisture (r= 0.560) and slight ly 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 repres ent soil nutrient availab ility well. Axis 3 was weakly correlated with soil temperature (r = 0.28 4) and air temperature (r = 0.245) alone while Axis 1 was not correlated with any of th e environmental variables (Table 3-4). NMS ordination indicated three axes as significant which accounted for 59.3% of the variation between understory specie s. A stable solution was reached in 169 iterations (instability < 0.00001). For understory and oversto ry datasets, final stress (w hich 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, underst ory plots aligned better as fragments than distance classes. However th e separation between shola fragments based on elevation was not seen. A separation between pl ots 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 environmen tal 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 sm all fragments may contrast fr om 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 gradie nts 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 co ntrast edges (Strayer et al. 2003, Laurance et al

PAGE 59

59 2002). I also observed that vertic al complexity of the understory did not vary along the edgeinterior gradient. In contrast, a study on Costa Rican tropical montane forest fragments observed a decrease in density and simplification in unde rstory structure with increasing distance from edge (Oosterhoorn and Kappelle 2000). Departures in my dataset from patterns observed in neotropical montane fragments c ould be due to differences in the origin of tropical montane fragments studied. While neotropical fragment s (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 frag mented tropical montane forests too have been observed to be characterized by atypical edge effects (Car lson 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 ba sal 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 w eakly, explained only by soil te mperature. Soil temperature has been shown to be a good indicator of the penetr ation of edge effect on microclimatic variables since it is coupled with change s in insulating overstory vegeta tion (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 we ll as high elevation shol a fragments. Bootstrap species richness estimators also indicated that 12 overstory species a nd 13 understory species were not represented in our dataset. My estimates for species diversity (H') are within the range

PAGE 60

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 indi cates 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. Plot s 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 di stribution of overstory species along the edgeinterior 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 st rata. 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 elevatio n 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 (Ganes haiah et al. 1997, Murali et al. 1998) and in other medium elevation ecosystems within th e 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 th eir separation from true high elevation montane forest in the Western Ghats. Soil nutrient availability was well represen ted in the overstory ordination cloud and assumed greater significance for high elevati on 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-elev ation fragments from high elevati on fragments in particular is

PAGE 61

61 also expressed in fragment complementarity scores and the Chao Abundance-based Srensen similarity indices for our overstory dataset. While across the dataset fragment complementarity is high, separation between mid-el evation fragments and high elevation fragments is often complete (C = 1.00). For understory vegetation, fragment size was a greater source of variat ion than elevational differences. While Axis 1 was correlated with relative humidity, soil moisture and organic carbon, Axis 2 was weakly correlat ed with air temperature and ne gatively with soil temperature and available phosphorous. Thus, in larger fr agments microenvironment and soil nutrient availability determined the di stribution of species whereas in smaller fragments only air temperature appeared to determine variation in species distributions. Light transmittance was very weakly correlated with Ax is 1. Since shola fragments ar e 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 m y knowledge, represents the fi rst 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 simila r gradients in a single fragment (Jose et al. 1996). Patt erns observed in species density and structure are indicative of a stable, mature shola-grassland edge. Across fr agments, 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 distributi ons 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 trop ical montane forest ecosystems (Bruijzneel and Hamilton 2000). From a management standpoint, the high degree of complementarity between

PAGE 62

62 fragments exhibited by the ecosystem emphasizes the need to conserve the present extent of the landscape and restore disturbed fragments wh ere possible. My designed preponderance to smaller fragments limits extensive characteriza tion of fragments using fragment metrics (e.g. core area, edge/area ratio, edge length). Although such an an alysis would add invaluable knowledge to the existing study, investigative efforts are probably best directed to filling the current void of process based stud ies in the ecosystem. As stable, natural fragments in a hostile grassland matrix, the study of shola fragme nts offers tremendous opportunities to study fragmentation pattern s and processes.

PAGE 63

63 Table 3-1. Location and vegetation characteri stics of tropical montane fo rest (shola) fragments studi ed 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 PNSP132.00 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)

PAGE 64

64 Table 3-2. Species richness, dominance a nd 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 Fishers 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 Pampa dum shola National Park respectively.

PAGE 65

65 Table 3-3. Similarity and complementarity between tropical montane forest (shola) frag ments in the Western Ghats, southern Ind ia. The ChaoSrsensen abundance based index varies from 0 (completely dissimilar) to 1 (completely similar) whereas Complementarity varies from 0 (all speci es shared) to 1 (no shared species). Fragments Chao-Srsensen 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 Understor y 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), Gummanebe tta (GU) and Jodigere (JG) s holas were located with BRT W ildlife Sanctuary (BRT). Suicide point (SP), Chinnanaimudi (CA), Ko lathan (KS), Pusinambara (PS) and Mukkal mile s holas 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).

PAGE 66

66 Table 3-4. Pearson correlation coefficients (r2) for environmental variable s with axis scores of NMS ordination. Correlation coefficients accounting for more than 30 % of the variation in individual factor s 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 Understor y 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

PAGE 67

67 Richness 0 20 40 60 80 100 120 Sobs (Mao Tau) Bootstrap Singletons Doubletons Cumulative number of plots 01 02 03 04 0 Richness 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) ove rstory and (b) understory species from tropical montane forest (shola) fragments in the Western Ghats, southern India. Species data were pooled across 4 distan ces within the 9 fragments (n=35). a b

PAGE 68

68 Distance from edge (m) 2.5 12.5 22.5 32.5 No. of individuals 0 10 20 30 40 50 50-99 100-149 150-199 200-249 > 250 Figure 3-2. Mean understory density (number of individuals/plot) along the edge-interior gradient in tropical montan e forest (shola) fragments (n = 9). Classes represent maximum height (Hmax) of understory individuals recorded in centimeters.

PAGE 69

69 No. of individuals (n) 0 2 4 6 8 Distance to edge (m) 2.512.522.5 No. of individuals (n) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 No. of individuals (n) 0 4 8 12 Distance to edge (m) 2.512.522.532.5 No. of individuals (n) 0 4 8 12 16 No. of individuals (n) 0 2 4 6 No. of individuals (n) 0 4 8 12 16 5-10 cm 10-20 cm 20-40 cm >40 cm No. of individuals (n) 0 6 12 18 No. of individuals (n) 0 6 12 18 Distance to edge (m) 2.512.522.532.5 No. of individuals (n) 0 4 8 12 GU JG SP CA PP MM KS PS KR Figure 3-3. Edge-interior gradie nts in diameter distri bution 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), Kola than (KS) and Mukkal mile sholas are located within Eravikulam National Park while Pampadum shola (PP) was located within Pampadum shola National Park

PAGE 70

70 Figure 3-4. Non-metric multi-dimensional scal ing (NMS) ordination of overstory data. KR, GU and JG represent mid elevation shola (T MF) 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 (m oist: soil moisture, nit: total nitrogen, orgc: organic carbon, relhum: relative hum idity) on the two axes. Letters denote fragment names as in Table 3-1 and number s indicate distance from edge classes.

PAGE 71

71 Figure 3-5. Non-metric multi-dimensional scalin g (NMS) ordination of understory data. Points represent plots showing their distributi on 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 indica te distance from edge classes.

PAGE 72

72 CHAPTER 4 EFFECT OF TOPOGRAPHY ON THE DI ST RIBUTION OF SHOLA FRAGMENTS: A PREDICTIVE MODELING APPROACH Introduction The influence of topography on vegetation acr oss 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 speci es composition and assemb lages progressively to alpine or boreal communities. Although the in fluence of other topogr aphical 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 us ed to determine species richness (Hofer et al. 2008), regional biodiversity patte rns (Coblentz and Ritte rs 2004), forest canopy health (Stone et al. 2008), species distributions (Warren 2008) and gradients in exotic species (Qian 2008). As Coblentz and Riitters (2004) identif y, this is driven in part by the global availability of high resolution elevation data (e.g., GLCF, www.landcover.org) coupled with the av ailability of high speed computing platforms. Within the Western Ghats, the potential of geographic information systems (GIS) and remotely sensed data in tr opical ecosystems (such as the Western Ghats hotspot) from the perspective of a landscape was recognized early (M enon and Bawa 1997). Studies have since utilized thes e tools to highlight gaps in th e 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 especia lly insightful in mountainous regions, presumably due to the extent of topography re lated heterogeneity. In a study on tropical montane forests (TMF) in Ethiopia, topography related va riation in soil physicoc hemical 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

PAGE 73

73 humidity and wind velocity influe nced species presence (Sri-Nge rnyuang 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 its abi lity 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 re lative ease of determination of topographical variables using extensive, large scale elevation data makes these variable s 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 th e Western Ghats, tropical montan e forests (or sholas) consist of insular forest fragments in a matrix of gra sslands. Carbon isotope analysis of peat samples from the ecosystem mosaic reveal a cyclical shift in dominant vegeta tion type (forest or grassland) corresponding to glacia l cycles (Sukumar et al. 1993). Th e edge is thus assumed to be natural in origin although its pe rsistence may be due to a combin ation of natural (fire, frost, edaphic) and anthropogenic (fire) factors. S hola fragments are charac terized by high species diversity ( -diversity) and endemism (Jose et al. 1994, Nair and Menon 2001). In addition, fragments have been shown to have a high de gree of complementarity (high diversity among fragments, Chapter 3). Since patterns in plan t distributions are non-ra ndom (Grieg-Smith 1979), predicting their occurrence is dependent on quantifying ecolo gical variables controlling incidence (Franklin 1995). In this study, the influence of t opographical variables on determining the presence of shola fragments was tested. A multiple logistic regression approach was used to

PAGE 74

74 predict presence or absen ce of insular fragments in the matrix of grasslands in two disjunct study sites in southern India using elevation data. Since the southwest monsoon cont ributes the bulk of the rainfa ll 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 observat ions indicate that s hola 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 be tween study sites at a m acro scale (as defined by Delcourt and Delcourt 1998). A lthough 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 pur pose 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 su rrounded by peaks with a maximum elevation of 2695m at Anaimudi. The soil had be en classified as Vertisols a nd described as Arachaen igneous in origin and consisting of granites and gneisses. Soils are sandy clay, have moderate depth (30100cm) 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

PAGE 75

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 northea st 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 eco system mosaic consisting of ro lling grasslands interspersed with dense, insular shola fragme nts. 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 Nilg iris (NIL) which is located in the state of Tamilnadu and includes the Mukurthi National Pa rk in the Nilgiri Biosphere Reserve. Details on the second study site can be found in Zarri et al. (2008). Together our tw o study sites represent the two largest contiguous extents of the high elevation shola-grasslan d 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 elev ation model (DEM). Remotely sensed ETM+ imagery (acquisition date: January 14, 2001) was used to identify and digiti ze 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 Im agine 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 in tervals represent the finest scale of elevation information available. In addi tion to elevation, all other t opographic 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). Sh ola points were identified by overlaying the points with digitized shol a fragments and declaring all other points as non-shola. In order to

PAGE 76

76 further distinguish shola fragme nts from non-shola forest type s, all points below 1700m were masked and excluded from further data analys is. Estimates of topograp hic 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 follo wing topographical variab les: 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) wh ile 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 va riables 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 in troduced to test for differences between sites. Since these two areas represent disjunct ecosy stems (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 indicat es a random process (i.e. there is an equal probability of the point being shol a or non-shola) whereas a value 0.50 indicates a non-random process. Seventy-five percent of the data were randomly selected for tr aining the model and 25% were retained as a cross-validati on 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-

PAGE 77

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 ( i > 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 conser vative 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 fragme nts 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 di rection of topographical parameter coefficients across Models 2 and 3. Results Combined Model For the combined model, topographical va riables significantly predict whether the vegetation type is shola or non-shola. The dist ribution of shola fragments was significantly different from random as indicated by the area u nder the receiver operating curve (AuC = 0.714, SE 0.0005, p < 0.0001). All topographical variables were highly significant (p < 0.0001, Table 42). Coefficients for northness were positive while those of eastness were negative as related to the presence of shola fragments (p < 0.0001). Coef ficients for wetness index, slope and elevation

PAGE 78

78 were positive and negative for curvature of the slope (for all p < 0.0001) Significant differences were also seen between the study si tes (p < 0.0001) with the Nilgiris (N IL) 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 subs et 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 di ffered in magnitude, they did not differ in direction (Table 4-2, Figure 4-2 to Figure 4-7). For both models, aspect (as quantif ied in eastness and northness) strongly determined the presence of shola frag ments and to a lesser extent wetness index ( CTI) and curvature of the slope ( CUR). Discussion Results from our multiple logistic regression models clearly indicate the influence of topography induced variation on th e location of tropical montan e forest (shola) fragments relative to the surr ounding non-shola matrix. AuC estimates pr ovide 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 wa s unexpected in all our models a priori. This could be due to our exclusion of points belo w 1700 m. However, at el evations below 1700 m, separation of shola vegetation from evergreen fo rest 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 show n to differ compositionally and structurally from high elevation shola fragments and might warrant a formal partitioning into forest subtypes (Chapter 3). Digitized shola fragments in ENP and field obs ervations indicate the presence of fragments (as large as 2 ha) at 2500 m and ab sence only from the highest peaks, possibly due

PAGE 79

79 to the extremely steep terrain on those peaks (e.g. Anaimudi). The topographical limitation on elevation in the Western Ghats pr ohibits the formation of true tr ee 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 southw estern monsoon in Eravikulam (ENP) and the Nilgiris (NIL) might explain the preponderance of shola frag ments on western slopes which support my hypothesis. The dependence of shola fragme nts on the availability of soil moisture is also reflected in the direction of the res ponse to the wetness index (Compound Topographical Index, CTI). The response to wetness index is si gnificant as topographical variables that combine a measure of topography with resource variable (such as CTI) have b een shown to predict patterns in vegetation better than those that measure topography al one (Franklin et al. 2000). The significant bias of shola fragme nts to north-facing slopes was une xpected and contrary to our hypothesis. Aspect induced variat ions in solar exposure along a north-south gradient are unlikely to be significant at the latitu de associated with the study sites (10-1132). During the months corresponding to the southwestern monsoon (May-October) significant differences in wind exposure are likely as southe rn aspects (exposed to monsoona l winds) are likely to have higher exposure to wind than nor thern aspects. This might limit the establishment of tree cover (shola fragments). However evidence for wind-expos ure related restriction of shola fragments in this study is tenuous at best. Bader & Ruijten (20 08) 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 ea rlier in the day (i.e. eastern aspects) which causes phot oinhibition in maladapted plants Lack of data on diurnal

PAGE 80

80 variation in intensity of radi ation from our study however pr ecludes 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 frag ments was significantly linked to negative coefficients which are indicative of concave slopes (see Equation 41). 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 s ites 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 forest s in other parts of the Western Ghats (including the Nilgiris) co lonial management of the gra sslands (and indeed even the sholas) in ENP did not involve extensive affo restation with exotic plantations. While the influence of topographical variab les on shola presence/absence in ENP (and consequently using that data subset to predict presence/absence of fr agments) 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 limitati on. While shola and gras sland soils have been shown to exhibit little va riation in soil physicochemical prope rties and soil depth (Table 1-1,

PAGE 81

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 tr opical montane fragments. Studies have shown that patterns in water accumulation and frost occurrence are of ten similar (Blennow 1998). Areas with a high wetness index (depressions etc.) are likely to ha ve 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 a ddition to the strong linkage to hydrological regulation, among fire a nd frost, fire rather than fros t might be responsible for the maintenance of the edge. Patterns in exposure to wind too might be resp onsible for the edge. Conclusion A common problem associated w ith predictive modeli ng is the lack of discrete boundaries between habitat patches and high variation within patc hes (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 limitati ons associated with other systems. Tropical montane forest (shola) fragments in the West ern Ghats have a high variation among fragments. Further edge-interior studies reveal that frag ments show little varia tion 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 fr ost in addition to str ong linkages to hydrological regulation. Predictive models are useful tools fo r restoration efforts and at larger scales can provide inputs into existing rest oration efforts. The use of topogr aphical variables to predict the presence of shola fragments represents a subs et of static, environmental sorting variables influencing the distribution of s hola fragments. The inclusion of dynamic variables such as fire

PAGE 82

82 history and regime and biotic interactions (e.g. to pography induced dispersal limitation) might provide further insights and enha nce our understanding of the shol a-grassland ecosystem mosaic.

PAGE 83

83 Table 4-1. Topographical variable s 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( )) Location Code LOC Boolean (ENP = 1, NIL = 0) As: Flow accumulation grid, : Slope (radians), Moore et al. 1993., Gessler et al. 1995 Table 4-2. Topographical parameter coefficien ts 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

PAGE 84

84 Figure 4-1. Location of study site s 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

PAGE 85

85 Figure 4-2. Histogram of Inter cept coefficients (INTC) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

PAGE 86

86 Figure 4-3. Histogram of Eleva tion coefficients (DEM) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

PAGE 87

87 Figure 4-4. Histogram of Slope coefficients for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

PAGE 88

88 Figure 4-5. Histogram of Northne ss coefficients for Eravikulam National Park (ERV ) and Nilgiris (NIL) data subsets. Superimpos ed curves represent normal curves.

PAGE 89

89 Figure 4-6. Histogram of Eastne ss coefficients for Eravikulam National Park (ERV ) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

PAGE 90

90 Figure 4-7. Histogram of Curvat ure of the Slope coefficients (C URV) for Eravikulam National Park (ERV) and Nilgiris (NIL) data subsets. Superimposed curves represent normal curves.

PAGE 91

91 CHAPTER 5 SUMMARY AND CONCLUSION Tropical m ontane 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 relativel y 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 br anches are often covere d 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 sh ola fragments. Nine fragments across three study sites were selected in the West ern 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), so il macronutrient (organic carbon, total nitrogen, available phosphorous, potassium calcium and magnesium) and micronutrient (boron, copper, molybdenum, manganese, iron and zinc) variables we re collected. Data we re also collected on overstory (richness and dominance) and understory (richness and vertical st ructure) vegetation. Data were collected along grassland exterior-edge-fragment interior gradients for microenvironment variables and edge-inter ior gradients for species variables. Edge effect studies typically co nsider the effect of the respons e 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 th e effect of the creati on of an edge between original (often forested ) habitat and converted ha bitat (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 multi ple edges in edge models becomes increasingly important in highly fragmented ecosystems a nd/or small fragments. Since the study was

PAGE 92

92 designed to study edge effects in small fragments distance-to-edge in thre e additional directions (2nd nearest, 3rd nearest and 4th nearest edge) mutually orthogona l to our original distance were obtained. The use of the dist ance to nearest edge alone found no evidence of edge-interior gradients similar to patterns observed in othe r small fragments. However the inclusion of distance to multiple edges significantly improved the predictability of our models in small fragments. Highly significant patt erns for light transmittance (p = 0.0072) and soil moisture (p = 0.0002) and weaker patterns for air temperature (p = 0.032) we re observed as a function of distance to multiple edges. For larger frag ments though distance to the nearest edge was sufficient to observe gradients in relative humid ity and potassium. With the exception of soil temperature no exterior-edge-shola interior grad ients 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 vertic al complexity in understory as a function of distance to edge were observed. Nonmetric multidimensional scaling (NMS) ordination techniques revealed a higher vari ation in species dist ributions between fragments than within fragments along the edge-interior gradient. Compositionally, significant differences were observed between mid-elevation fragments and hi gh elevation fragments for overstory species. Since species found in high elevation ( 1700 m) fragments are more representative of tropical montane forest habitat th e exclusion of fragments 1700 m from future tropical montane forest studies is recommended possibly de signating 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 fr agments in the ecosystem mosaic also prompted a study on the effect of topographical variables in determin ing the presence (or

PAGE 93

93 absence) of shola species. For the purpose of this st udy the two largest highaltitude extents of the shola-grassland ecosystem mo saicEravikulam 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 se lected 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 tr aining datasets-to te st hypotheses. Although all topographical variables quantified were highly significant (p < 0.0001), the res ponse to aspect was unexpected. While the preference of shola fragments for west ern 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 vari ability is also captured in th e response to wetness index with fragments preferring wetter sites. The persistence and relative stability of th e shola-grassland edge continues to puzzle ecologists and foresters. Our predictive model provides supp ort for fire and hydrological regulation as driving forces maintaining the edge as shola fragments pr eferred wetter habitats (which are less likely to burn). Si nce 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 co ntrast 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 grassl and habitat do not differ significantly. The shola-grassland ecosystem mosaic offers insights into fragmentation-related processes in the absence of human influence. The persisten ce of small fragments (~ 1ha) in the mosaic also needs to be highlighted. However, more needs to be learnt about proces ses driving the observed

PAGE 94

94 patterns. In addition to the need for basic hydrol ogical studies, data are also lacking about the role of the shola epiphytic community in rainfa ll 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 absen ce as are experimental studies on seedling establishment essential for ecosy stem restoration efforts (howev er, see Sekar 2008). Information on genetic diversity of plant spec ies 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 pr ocess oriented studies in the ecosystem and suggest that further research needs to be directed to it.

PAGE 95

95 APPENDIX

PAGE 96

96 Table A-1. Overstory species ( 5 cm dbh) presence matrix along an edge-interior gradient acro ss nine fragments in the sholagrassland ecosystem mosaic in the West ern Ghats, southern India. Karadishonebetta (KR), Gummanebetta (GU) and Jodigere (JG) sholas were located with in BRT Wildlife Sanctuary (BRT). Suicide point shola (SP), Chinnanaimudi (CA), Pusinambara (PS), Kolathan (KS) and Mukkal mile (MM) sholas were located within Eravikul am National Park (ENP). Pampadum shola (PSNP) is represented by a single fragment. Presence across 4 dist ances (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 12341234123 4 123412341234 ACSI 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.

PAGE 97

97 Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 DIOP 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.

PAGE 98

98 Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 LIWI 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.

PAGE 99

99 Table A-1. Continued BRT ENP Sps. KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 PSNI 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.

PAGE 100

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

PAGE 101

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.

PAGE 102

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.

PAGE 103

103 Table A-2. Understory species presence ma trix along an edge-interior gradient acros s 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. Karadishonebet ta (KR), Gummanebetta (GU) and Jodigere (JG) shol as were located within BRT Wildlife Sanctuary (BRT). Suicide point (SP), Chinnanaimudi (CA), Pusina mbara (PS), Kolathan (KS) and Mukkal mile (MM) sholas were located within Eravikulam National Park (ENP). Pamp adum 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 ma rked 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 12341234123 4 123412341234 AGAD 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

PAGE 104

104 Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 CIIN 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

PAGE 105

105 Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 LGSP 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

PAGE 106

106 Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 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.

PAGE 107

107 Table A-2. Continued Sps. BRT ENP KR GU JG SP CA PS KS MM 1 2 3 1 2 3 4 12341234123 4 123412341234 TESP 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

PAGE 108

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.

PAGE 109

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

PAGE 110

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

PAGE 111

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 Biodi versity. Kerala Forest Research Institute, Peechi, India. Aerts, R., Overtveld, K.V., Haile, M., Herm y, M., Deckers, J., Muys, B. 2006. Species composition and diversity of small Afromontan e 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 IU CN Red List of Threatened Species. www.iucnredlist.org Downloaded on 30 January 2009. Allen S. E. 1974. Che mical analysis of ecol ogical materials. Blackwell Scientific, Oxford. Anderson, S.H.,Mann, K. Shugart,H.H. 1977. The ef fect of transmission-lines corridors on bird populations. American Midla nd Naturalist 97: 216-221. Bader, M.Y., Ruijten, J.J.A. 2008. A topography-base d 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. A lternative stable states in ecology. Frontiers in Ecology and the Environment 1: 376-382. Benzing, D.H. 1998. Vulnerabiliti es of tropical forests to clim ate change: the significance of resident epiphytes. Clim atic 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 Fo rester 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., Oldf ield, S., Magin, G., Hilton-Taylor, C. 2002. Habitat loss and extinction in the hotspots of biodivers ity. Conservation Biology 16: 909923.

PAGE 112

112 Bruijnzeel, L.A., Hamilton, L.S. 2000. Decisi on time for cloud forest s: IHP Humid tropics programme series No. 13. UNESCO, Paris, France. Bowman, D.M.J.S., Woinarski, J.C.Z. 1994. Bi ogeography 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.unepwc mc.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. Journa l of Tropical Ecology 11: 205-221. Carlson, A., Hartman, G. 2001. Tropical forest fragmentation and nest predationan experimental study in an Ea stern 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., Sibi chan, 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 richne ss estimation. Pages 7909-7916 in Balakrishnan, N., Read, C. B., Vidakovic, B., editors. Encyclopedia of Sta tistical 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 inci dence and abundance data. Ecology Letters 8: 148-159. Coblentz, D., Keating, P.L 2008. Topographic controls on the distri bution 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 c ontrols on the regional scale biodiversity of the south-western USA. Journa l of Biogeography. 31: 1125-1138. Cochrane, M.A. 2001. Synergistic interactions between habitat fragmentation and fire in evergreen tropical forests. C onservation Biology 15: 1515-1521. Colwell, R.K. 2006. EstimateS: Statistical estimati on of species richness and shared species from samples. Version 8. Persistent URL . Colwell, R. K., Coddington, J. A. 1994. Estimati ng terrestrial biodiversit y through extrapolation. Philosophical Transactions of the R oyal Society (Series B) 345: 101-118.

PAGE 113

113 Colwell, R.K., Mao, C.X., Chang, J. 2004. Inter polating, extrapolating and comparing incidencebased 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-Nio 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 st atus of the Nilgiri Tahr in th e grass hills in the Anamalais. Journal of the Bombay Natura l History Society 68: 347-354. Davidar, P., Mohandass, D., Vijayan, S.L. 2007. Fl oristic inventory of woody plants in a tropical montane forest in the Palni hi lls of the Western Ghats, Indi a. 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: 433444. 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 montan e plant species from the West ern Ghats, India revealed by intersimple sequence repeats. Molecular Ecology 10: 2397-2408. Didham, R.K., Lawton, J.H. 1999. Edge structur e determines the magnitude of changes in microclimate and vegetation structure in tropica l forest fragments. Biotropica 31: 17-30. Dinesh, K.P., Radhakrishnan, C., Bhatta, G. 2008. A new species of Nyctibatrachus Boulenger (Amphibia: Anura: Nyctibatr achidae) from the surroundings of Bhadra Wildlife Sacntuary, Western Ghats, India. Zootaxa 1914: 45-56. FAO (Food and Agricultural Orga nization). 1993. Forest resour ces assessment 1990: tropical countries. FAO Forestry Paper No. 112. Rome Fernndez, C., Acosta, F.J., Abell, G., Lpez, F ., Daz, M. 2002. Complex edge effect fields as additive processes in patches of ecological systems. Ecological Modelling 149: 273-273.

PAGE 114

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. Restora tion 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, geogra phic modelling of bios patial patterns in relation to environmental gradients. Progress in Physical Geography 19:474. 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: prin ciples and applications, J ohn 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 tr ee 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 disp ersion 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 Ke rala: Environment and Biodiversit y. 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. Intern ational Journal of GIS. 9: 421-432. Ghosh, D.S. 2001. Fish diversity and associated sp ecies in the sholas of Munnar, Idukki District. Pages 353-364 in Nair, K.K.N., Khanduri, S.K., Ba lasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research Institute, Peechi, India. Gioda, A. Maley, J., Guasp, R.E., Baladn, A.A. 1995. Some low elevatio n 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.

PAGE 115

115 Greig-Smith, P. 1979. Pattern in vegetation. Journal of Ecology 67:755. Gupta, R.K. 1962. Studies in some shola forest s on the Palni hills near Kodaikanal. Indian Forester 88: 848-853. Gupta, R.K., Shankarnarayan, K.A. 1962. Ecological status of the grassl ands 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 a nd composition of edges next to regenerating clear-cuts in mixed-wood bor eal 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 frag ments. 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 gr adient dominant land scapes. Ecography 31: 131139. 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., Bhas kar, A., Kumar, V. 2000. Reproductive behaviour and genetic variability in geogra phically 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 islandIncr ease in interference from outside as park size decreases. Oikos 41: 402-410. Jeeva, V., Ramakrishnan, P.S. 1997. Studies on imp act of plantation forestry in Nilgiri Hills of the Ghats on soil quality and nutrient cycling. Trop ical Ecology 38: 215-235. Johnston, V.R. 1947. Breeding birds of the fo rest edge in Illino is. Condor 2: 45-53. Jose, S. Sreepathy, A. Mohan Kumar, B., Venugopa l, V.K. 1994. Structural, floristic and edaphic attributes of the shola-grassland forests of Er avikulam in peninsular India. Forest Ecology and Management 65: 279-291.

PAGE 116

116 Jose, S. Gillespie, A.R., Goerge, S.J., Ku mar, M.K. 1996. Vegetation response along edgeinterior gradients in a high alt itude tropical forest in penins ular 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 conservati on of the grasslands of Eravikulam National Park a nd the Western Ghats. Wildlife In stitute 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 a nd fuzzy mapping in diverse tropical deciduous ecosystems using multi-date NDVI. International Jour nal of Remote Sensing 25: 1185-1205. Lpez-Barrera, F. Newton, A., Manson, R. 2005. Edge effects in a tropical montane forest mosaic: experimental tests of post-dispersa l acorn removal. Ecological Research. 20: 3140. 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 ha bitat fragmentation research has transcended island biogeographic theory. Biological Conservation 141: 1731-1744. Laurance,W.F., Ferreira,L.V. Rankin-De Mer ona, J.M., Laurance, S.G., Hutchings, R.W. Lovejoy, T.E. 1998. Effects of forest fragme ntation on recruitment patterns in Amazonian tree communities. Conserva tion 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. Cons ervation Biology 16: 605-618. Laurance, W.F., Yensen, 1991. Predicting the impact s of edge effects in fragmented habitats. Biological Conservation 69: 23-32. Leopold,A. 1993.Game management. Scribner Sons, New York, USA. Lpez-Barrera, F., Manson, R.H. Gonzlez-Esp inosa, 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.

PAGE 117

117 Lpez-Barerra, F. Newton, A., Manson, R. 2005. Edge effects in a tropical montane forest mosaic: experimental tests of post-dispersa l acorn removal. Ecological Research. 20: 3140. 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 a nd its measurement. Cr oom Helm, London, UK. Malcolm, J.R. 1994. Edge effects in central Amazonian forest fragments. Ecology 75: 24382445. Mancke, R.G.,Gavin, T.A.2000. Breeding bird de nsity in woodlots: Effects of depth and buildings at the edges. Ecol ogical 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 vege tation zonation in the Cordillera Central, Dominican Republic. Journal of Biogeography 34: 1792-1806. Mather, P.M. 1976. Computational methods of mu ltivariate analysis in physical geography. J Wiley and sons, London, UK. May, R.M. 1977. Thresholds and breakpoints in eco systems with multiplicity of states. Nature 269: 471-477. McCune, B., Mefford, M.J. 2006. PC-ORD. Multivaria te analysis of ecological data. Version 5. MjM Software, Glenden Beach, Oregon, USA. McNab, W.H. 1989. Terrain Shape Index: Quantify ing effect of minor la ndforms on tree height. Forest Sciance 35: 91-104. Meher-Homji, V.M. 1965. Ecological status of the montane grassla nds of the sout h Indian hills: A phytogeographic reassessment. I ndian Forester 91: 210-215. Meher-Homji, V.M. 1967. Phytogeography of the s outh Indian hill sta tions. 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 Kera la: Environment and Biodiversit y. Kerala Forest Research Institute, Peechi, India.

PAGE 118

118 Menon, S., Bawa, K.S. 1997. Applications of ge ographic information systems, remote-sensing and a ladscape ecology approach to biodive rsity 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., Delamnica, P., Laurance, W.F. 1999. Effect of surrounding vegetation on edge-related mortality in Amazonian forest fragments. Biological Conservation 91: 129134. Mishra, C., Johnsingh, A.J.T. 1998. Population a nd conservation status of the Nilgiri tahr Hemitragus hylocrius in Anamalai hill, south India. Biological Conserva tion 86: 199-206. Moore, I.D., Gessler, P.E., Nielse n, 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 Environm ental 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 priori ties. Nature 403: 853-858. Nadkarni, N.M. Wheelwright, N.T. 2000. Montev erde: Ecology and Conservation of a Tropical Montane Cloud Forest. Oxford Univ ersity 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 Conserva tion of a Tropical Montane Cloud Forest. Oxford University Press, New York, USA. Nair, K.K.N., Menon, A.R.R. 2001. Endemic arbores cent 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 In stitute, Peechi, India. Nair, K.K.N., Khanduri, S.K., Balasubraman ayam, K. 2001. Shola forests of Kerala: Environment and Biodiversity. Kerala Fore st Research Institute, Peechi, India. Nameer, P.O. 2001. Richness of avifauna in th e sholas of Munnar, Idukki District. Pages 365389 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., Na rendra Babu, K. 2001. Characteriza tion 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 In stitute, Peechi, India.

PAGE 119

119 NATMO.2009. Soil map of India (National Atlas and Thematic Mapping Organization). Available online at www.indiabiodive rsity.org. Last accessed on May 04th 2009. Noble, W.A. 1967. The shifting balance of grassl ands, shola forests and planted trees on the upper Nilgiris, southern India. Indian Forester 93: 691-693. Oosterhoorn, M. Kappelle, M. 2000. Vegetation st ructure and composition along an edge-interior gradient in a Costa Rican montane cloud fo rest. 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 Eu calyptus. Vol II. Available online at http://www.fao.org/docrep/005/ ac772e/ac772e06.htm #bm06 Pavlacky, D.C., Anderson, S.H. 2007. Does avia n species richness in natural patch mosaics follow the forest fragmentation para digm? Animal Conservation 10: 57-68. Qian, H. 2008. A latitudinal gradient of beta dive rsity 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 stat istical computing. R Foundation for Statistical Co mputing, 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., Samarasin ghe, 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 gra sslands of the Nilgir is: Past and present scenario. Range Management a nd 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. Ecol ogical responses to hab itat edges: Mechanisms, models and variability explained. Annual Review of Ec ology Evolution and Systematics 35: 491-522. Ries, L., Sisk, T.D. 2004. A predictive m odel of edge effects. Ecology 85: 2917-2926.

PAGE 120

120 Sahoo, D.C., Sharda, V.N., Jayakumar, M., Trip athi, 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 micr oflora 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 Ke rala: Environment and Biodiversit y. 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 gr owth of different Shol a species under a Shola afforestation programme in N ilgiris 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 Conser vation 91: 135-142. Smith, E.P., van Belle, G. 1984. Nonparametric es timation of species richness. Biometrics 40: 119-129. Somanathan, H., Borges, R.M. 2000. Influence of exploitation on populatio n structure, spatial distribution and reproductive success if dioeci ous 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: 195207. 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. Clim ate change and its impact on tropical montane ecosystems in southern India. J ournal 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 trop ical lower montane forest in northern Thailand. Ecological Reseach 18:114.

PAGE 121

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 gras sland ecosystem of upper Palnis. Indian Journal of Forestry 24: 324-327. Still, C.J., Foster, P.N. Schneider, S.H. 1998. Simu lating the effects of clim ate 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., Picke tt, S.T.A., Belnap, J. 2003. A classification of ecologi al boundaries Bioscience 53: 723-729. Sudhakara, K. 2001. Inventory and computerized herbarium of highe r plants in the sholas of Munnar, Idukki dist rict. Pages 179-208 in Nair, K.K.N., Khanduri, S.K., Balasubramanayam, K., editors. Shola forests of Kerala: Environment and Biodiversity. Kerala Forest Research In stitute, Peechi, India. Svenning, J-C. 2001. Environmental heterogene ity, recruitment limitation and the mesoscale distribution of palms in a tropical montan e forest (Maquipucuna, Ecuador). Journal of Tropical Ecology 17: 97-113. Swarupanandan, K., Sashidharan, N., Chac ko, 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 Fore st Research Institute, Peechi, India. Thomas, S.M. Palmer, M.W. 2007. The montane grasslands of the We stern Ghats, India: Community ecology and conservatio n. 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 Kera la: Environment and Biodiversit y. 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 In stitute, Peechi, India. Tabarelli, M., Mantovani, W., Peres, C.A. 1999. Effects of habitat frag mentation on plant guild structure in the montane Atlantic forest of southeastern Braz il. Biological C onservation 91: 119-127. Troll, C. 1973. The upper timberlines in different climatic zones. Arctic and Alpine Research 5: A3-A18.

PAGE 122

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 a nd environmental conditions of forest edges in Panama. Journal of Ecology 78: 356-373. Williams-Linera, G., Domnguez, Garcia-Zurita, M. E. 1998. Microenvironment and floristics of different edges in a fragmented tropical rainforest. Conservati on Biology 12: 1092-1102. Williams-Linera, G. 2002. Tree species ric hness, complementarity, disturbance and fragmentation in a Mexican tr opical 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 th e 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 Montan e Cloud Forests. Springer-Verlag, New York, USA. Yahner, R.H.1988. Changes in communities near edges. Conserva tion Biology 2: 333-339. Young, K.R., Keating, P.L. 2001. Remnant fore sts of Volcn Cotacachi, Northern Ecuador. Arctic, Antarctic and Alpine Research 33: 165-172. Young, K.R., Leon, B. 1995. Distribution and conservation of Perus montane forests: Interactions between the biota and human society. Pages 362-276 in Hamilton, L.S., Juvik, J.O., Scatena, F.N., editors. Tropical Mont ane 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.

PAGE 123

123 BIOGRAPHICAL SKETCH Milind Christopher Bunyan was born in Banga lore, India in 1979. The pursuit of an undergraduate degree in Environm en tal Science at St. Josephs Co llege 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 worl d. He went on to receive a Masters 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 Re search 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 Shi bu Jose, he earned his PhD from the School of Forest Resources and Conservation, at the Univer sity of Florida in Gain esville, Florida studying edge effects in tropical montane fo rest fragments in southern India.