Understorey structure and refuges from predators influence habitat use by a small ungulate, the Indian chevrotain (Moschiola indica) in Western Ghats, India

Abstract

The availability of refuges from predators and high quality food are thought to determine habitat use in small ungulates. We tested this hypothesis on habitat use by the Indian chevrotain in a tropical rainforest in the Western Ghats, using pellet-groups to infer habitat use. Between December 2009 and April 2010, we sampled 204 grids of 50 m × 50 m with four spatial replicates in each, using occupancy framework. We quantified refuges such as fallen logs and boulders, understorey complexity, and noted the presence of fruiting trees in the grid. Detection probability, p, of pellet-groups was estimated at 0.61. The naive estimate of occupancy was 0.52, which increased to 0.73 when p was accounted for. Out of eight candidate models compared using AICC, the one with the number of refuges and understorey complexity was the best model. Both covariates had non-zero positive slopes. Fruiting trees occurred very infrequently and did not influence habitat use, perhaps because the chevrotain fed more on buds, shoots and young leaves during the dry period covered by this study. The strong influence of understorey complexity on habitat use was perhaps also due to the abundance of these food items. These results highlight the need to control human use that adversely impacts the availability of refuges and understorey complexity such as removal of fallen logs and rattans. This study also demonstrates the application of the occupancy approach in the study of small and elusive ungulates.

Full text

Ecological Research
Understorey structure and refuges from predators influences habitat-use by a small
ungulate, the Indian chevrotain (Moschiola indica)
--Manuscript Draft--
Manuscript Number:
Full Title: Understorey structure and refuges from predators influences habitat-use by a small
ungulate, the Indian chevrotain (Moschiola indica)
Article Type: Original Article
Keywords: cover from predators; habitat use; Moschiola indica; occupancy models;
Someshwara Wildlife Sanctuary
Corresponding Author: Sachin Sridhara
National Centre for Biological Sciences
Bangalore, INDIA
Corresponding Author Secondary
Information:
Corresponding Author's Institution: National Centre for Biological Sciences
Corresponding Author's Secondary
Institution:
First Author: Sachin Sridhara
First Author Secondary Information:
Order of Authors: Sachin Sridhara
Advait Edgaonkar
Ajith Kumar
Order of Authors Secondary Information:
Abstract: The body size of small ungulates is thought to have significant influence on their
ecology. It is assumed to impose constraints on anti-predatory behavior and selectivity
in feeding, consequently influencing their habitat requirements. We tested whether
availability of cover from predators and high-quality resources like fruits influence
habitat use by the Indian chevrotain (Moschiola indica), a small ungulate widely
distributed in peninsular India. Being highly cryptic with low detectability, we used
pellet-groups to infer habitat-use by the Indian chevrotain. A grid based occupancy
approach was used to account for imperfect detections. We sampled 204 grids of size
50m x 50m, with four spatial replicates in each to build detection histories. We
recorded presence of pellet-groups in the replicate. Further, we quantified habitat
features that could potentially provide cover from predators and noted the presence of
fruits in the grid. Detection probability of pellet-groups was 0.61 and comparable to
direct evidence. Our results indicate that habitat use was positively influenced by
understorey structure and the presence potential refuges. However, the presence of
trees in fruit had no influence on habitat use, perhaps due to their scarcity in the study
period. Our study is among the first to establish habitat relationships for small
ungulates using a robust analytical framework. Given that small ungulates are largely
threatened by hunting and habitat fragmentation, this study suggests that retaining
understorey complexity and potential refuges are important for the management and
conservation of these species.
Powered by Editorial Manager® and Preprint Manager® from Aries Systems Corporation

1
Understorey structure and refuges from predators influences habitat-use 1
by a small ungulate, the Indian chevrotain (Moschiola indica) 2
3
Sachin Sridhara1, *, Advait Edgaonkar2 & Ajith Kumar3 4
5
1 Post-graduate Program in Wildlife Biology and Conservation, WCS-India Program, 6
National Centre for Biological Sciences, Bangalore, India. 7
2 Indian Institute of Forest Management, Bhopal, India. 8
3 Wildlife Conservation Society - India Program, Bangalore, India. 9
* Present address: Centre for Ecological Sciences, Indian Institute of Science, Bangalore. 10
11
Corresponding author: 12
Sachin Sridhara 13
Post-graduate Program in Wildlife Biology and Conservation, WCS India Program 14
National Centre for Biological Sciences, GKVK Campus, Bangalore, India. 15
sachin.sridhara@gmail.clom 16
17
18
Manuscript
Click here to download Manuscript: MS.doc
Click here to view linked References

2
Abstract 19
The body size of small ungulates is thought to have significant influence on their 20
ecology. It is assumed to impose constraints on anti-predatory behavior and selectivity 21
in feeding, consequently influencing their habitat requirements. We tested whether 22
availability of cover from predators and high-quality resources like fruits influence 23
habitat use by the Indian chevrotain (Moschiola indica), a small ungulate widely 24
distributed in peninsular India. Being highly cryptic with low detectability, we used 25
pellet-groups to infer habitat-use by the Indian chevrotain. A grid based occupancy 26
approach was used to account for imperfect detections. We sampled 204 grids of size 27
50m x 50m, with four spatial replicates in each to build detection histories. We recorded 28
presence of pellet-groups in the replicate. Further, we quantified habitat features that 29
could potentially provide cover from predators and noted the presence of fruits in the 30
grid. Detection probability of pellet-groups was 0.61 and comparable to direct evidence. 31
Our results indicate that habitat use was positively influenced by understorey structure 32
and the presence potential refuges. However, the presence of trees in fruit had no 33
influence on habitat use, perhaps due to their scarcity in the study period. Our study is 34
among the first to establish habitat relationships for small ungulates using a robust 35
analytical framework. Given that small ungulates are largely threatened by hunting and 36
habitat fragmentation, this study suggests that retaining understorey complexity and 37
potential refuges are important for the management and conservation of these species. 38
39
40

3
Keywords 41
cover from predators; habitat use; Moschiola indica; occupancy models; Someshwara 42
Wildlife Sanctuary 43
44
Introduction 45
Small ungulates (2-15 kg weight) are widely distributed through the tropics in 46
taxonomically distinct genera (Wilson & Reeder, 2005). Although they are threatened 47
across the world, they are amongst the least studied ungulates (Corlett 2007, Sodhi et al 48
2004). Consequently, many aspects of their ecology are still poorly understood, 49
hindering the effective conservation of these species (Baillie et al. 2004). Specifically, 50
understanding the habitat relationships of these small ungulates is important to their 51
conservation. Their cryptic nature (Geist 1998; Jarman 1974) poses a challenge in 52
quantitatively assessing their habitat needs. However, due to the remarkable degree of 53
convergence in morphological and behavioral adaptations of small ungulates (Barette 54
1987; Bodmer 1990), a broad understanding of their habitat requirements has 55
considerable applicability across species. 56
57
Habitat requirements of small ungulates are thought to be driven by the constraints 58
imposed by their body size (Geist 1998, Jarman 1974). It is hypothesized that the 59
‘slinker’ body form, found consistently across small ungulate species, is adapted for 60
quick ‘saltation’, predisposing them to remain close to thick cover. Anecdotal reports 61
suggest that small ungulates are predominantly solitary and use stealth to avoid 62

4
predators, often taking cover in buttresses, boulders and dense undergrowth, avoiding 63
open areas (Geist 1998; Macdonald 2001). Additionally, small ungulates are predicted to 64
show selection for high quality forage due to a combination of body size and digestive 65
physiology (Demment and van Soest 1985). The fact that metabolic rates increase with 66
declining body size, whereas digestive capacity decreases, may constrain small 67
ungulates to consume food items like fruits, known to be low in fiber content, and highly 68
digestible (Demment and van Soest 1985; Hofmann 1989; Prins et al 2006). Nearly 50-80 69
% of the diet of small ungulates is fruits (Bodmer 1990; Branan 1985; Dubost 1984; 70
Gautier-Hion et al. 1980), a much larger proportion compared to their larger 71
counterparts (Gagnon and Chew 2000). In this study we used the Indian chevrotain 72
(Moshiola indica) as a model species to test whether the availability of cover from 73
predators, and presence of fruit resources positively influenced habitat-use of small 74
ungulates. 75
76
The Indian chevrotain is a small ungulate (1 - 3kg) widely distributed in peninsular India 77
(Groves and Meijaard 2005). Although its conservation status is categorized as ‘Least 78
Concern’ (Duckworth et al. 2008), it is among the most frequently hunted animals in its 79
habitat (Madhusudhan and Karanth 2002; Kumara and Singh 2004). Yet, apart from 80
anecdotal accounts of distribution and behaviour (Eisenberg and Lockhart 1972; 81
Krishnan 1972; Ramachandran et al. 1986), and their seed dispersal efficacy (Prasad and 82
Sukumar 2010) no study has quantitatively assessed factors influencing habitat use, to 83
the best of our knowledge. Specifically, in this study we examined whether the amount 84

5
of available cover in an area positively influences habitat use. Further, we recorded the 85
presence of fruit in potential food tree species and assessed its influence on habitat use. 86
Since the detection probability of the Indian chevrotain is not very high (Nag 2008), we 87
used an occupancy framework to account for imperfect detections (MacKenzie et al. 88
2002) using the presence of dung pellets as an indicator of use. 89
90
Materials and Methods 91
Study area 92
We conducted the study in Someshwara Wildlife Sanctuary (88.40 km2), located in the 93
Western Ghats mountain range in the Karnataka, India (Fig 1(b)). The altitude in the 94
Sanctuary ranges from 75 m to 870 m and the annual rainfall is about 4000 mm, most of 95
it from the southwest monsoon during June to September (Pascal, 1988). Tropical 96
rainforest is the major vegetation type, with Dipterocarpus indicus - Diospyros 97
candolleana - Diospyros oocarpa association below 850 m and Dipterocarpus indicus - 98
Humboldtia brunonis - Poeciloneuron indicum associations above 850 m elevation 99
(Pascal 1988). Large wild mammals in the Sanctuary include barking deer (Muntiacus 100
muntjac), sambar (Rusa unicolor), gaur (Bos gaurus), lion-tail macaque (Macaca silenus), 101
tiger (Panthera tigris) and leopard (Panthera pardus). 102
103
Field Methods 104
We conducted the study between December 2009 and April 2010, the dry period in the 105
Western Ghats (Pascal 1988). Due to the cryptic and elusive nature of the species, we 106

6
used pellet-groups as an indicator of habitat-use. Although previous studies on the 107
species have used direct evidence through camera traps (Nag 2008; Prasad et al. 2009), 108
pellets-groups have been successfully used to study habitat relationships of ungulates 109
(Gopalaswamy et al. 2012; Krishna et al. 2008). Surveys that use pellet-groups have the 110
additional benefit of significant reduction in cost and complexity compared to those 111
using camera traps. 112
113
The study area was overlaid with grids of size 50 m x 50 m using QGIS 1.6.0. Each grid 114
was further sub-divided into four equal square sub-grids, the diagonals of which formed 115
the spatial replicates (Fig 1 (a)). The grids were realized on ground using Garmin eTrex 116
Vista GPS. Two observers walked along the four spatial replicates and carefully scanned 117
the forest floor for pellet-groups. All sighted fresh pellet-groups were recorded. Pellet-118
groups could be unambiguously identified as that of Indian chevrotain since they were 119
much smaller in size and differed in shape compared to that of the immediately large 120
ungulate, Indian muntjac (Muntiacus muntjac) (Fig 2) and considerably smaller than all 121
other ungulates found in the Sanctuary. 122
123
In each grid we measured the following five habitat covariates expected to influence the 124
detection of pellet-groups, and habitat-use by the chevrotain: (a) Potential food tree 125
species (henceforth FRUIT) was recorded as ‘1’ if present within the grid and in fruit. A 126
‘0’ was recorded if no fruit tree was present or if they were present but not in fruit. (b) 127
Refuges from predators (henceforth REFUGE) were measured as the total number of 128

7
buttresses, fallen logs, thick breaks of canes or lianas, crevices in rocks, and boulders in 129
the grid. (c) Since the understorey structure (henceforth USI) itself could provide cover 130
for the Indian chevrotain, an index of understorey complexity/density was computed by 131
combining the visual estimation of height and extent of the undergrowth. These were 132
estimated from circular plots of 3 m radius, laid at two points along each spatial 133
replicate of the grid. Height of the undergrowth was categorized as 0-50cm, 51-100 cm, 134
or >100 cm, while the extent of undergrowth in each plot was visually estimated in 135
percentage. The index USI was computed as the product of undergrowth height and its 136
extent. (d) As a surrogate for the thickness of undergrowth, visibility in the understorey 137
(henceforth VISIB) was estimated. At two locations in each spatial replicate a rod of 138
length 1 m marked at 10 cm intervals was placed vertically at a distance of two meters 139
from the observer, thus totaling … points per grid. The observer recorded the number of 140
marks visible at each location. Finally, an average was computed for each grid. (e) 141
Canopy cover was measured using a standard spherical densitometer (Forestry Supplies) 142
at two points in every spatial replicate and an average computed for a grid. 143
144
Analysis 145
Since it is impossible to perfectly detect all the pellet-groups, we explicitly accounted for 146
missed detections in our modeling approach. Habitat use and detection probability was 147
modeled as a function of covariates in an occupancy framework (MacKenzie et al. 2002). 148
We constructed detection histories for each grid based on whether pellet-groups were 149
detected in the four spatial replicates. For example, a detection history of “0110” means 150

8
that the pellet-groups were detected in second and third spatial replicates but not in the 151
first and fourth spatial replicates of the grid. Using likelihood functions (MacKenzie et al. 152
2002), detection probability ‘p’ of pellet-groups was modeled in software PRESENCE 153
(Hines 2006). Visibility in the understorey (henceforth VISIB), and the understorey 154
structure (USI) were used since we expected detection to drop in areas of thick 155
undergrowth or poor visibility. We also modeled ‘p’ independent of any covariates to 156
understand whether factors that were not measured by us influenced detectability 157
(MacKenzie et al. 2006; Williams et al. 2002). 158
159
The home range of species closely related to the Indian chevrotain, yet smaller, the 160
lesser mouse-deer (Tragulus javanicus), varies between 3 and 6 hectares (Matsubayashi 161
et al. 2003), while for the larger Natal duiker (Cephalophus natalensis), home ranges 162
varies between 7 to 11 hectares (Bowland and Perrin 1995) suggesting that the area of 163
each sampled grid in our study, 0.25 hectares (50 m x 50 m), was very likely to be 164
smaller than the home range of the Indian chevrotain. Since the area of a single 165
sampling unit is smaller than the home range of the target species the parameter ‘ψ’ 166
estimated using software PRESENCE has to be interpreted as probability of use rather 167
than occupancy (MacKenzie et al. 2006). While ‘ψ’ for a specific site indicates the 168
probability that that site is being used, an overall value indicates the proportion of sites 169
being used. 170
171

9
We modeled habitat-use ‘ψ’, as a function of the measured covariates. Based on our a 172
priori expectation that the presence of cover from predators (measured as REFUGES, 173
USI and VISIB) and fruiting trees (covariate FRUIT) would influence habitat-use we 174
constructed eight candidate models. Our expectations are summarized in table 2. 175
176
In order to check whether detection in one spatial replicate was independent of the 177
other (MacKenzie and Bailey 2004; Williams et al. 2002) we estimated the over-178
dispersion parameter ‘
c
ˆ
’ for our most complicated model using the method outlined in 179
MacKenzie and Bailey (2004). Parameter ‘
c
ˆ
’ was estimated for the global model 180
incorporating all measured covariates. AICC, which corrects the Akiake Information 181
Criteria (AIC) value for small sample size was used to rank the models (Burnham and 182
Anderson 1998). Model selection was based on the lowest AICC score and Akiake 183
weights. 184
185
Results 186
In all, 204 grids were surveyed. The naive occupancy (i.e. the proportion of grids in 187
which pellet-groups were found, without accounting for imperfect detection) was 0.52. 188
The overall estimated detection probability ‘
p
ˆ
’ was 0.61. After explicitly accounting for 189
detection probability the estimated parameter ‘

ˆ
’ or habitat-use was 0.73 ± 0.06. 190
191
Among the predictors measured, canopy cover was very high with very little variation 192
(mean = 98.64, SE ± 0.79%) and was therefore dropped from further analyses. 193

10
Understorey structure (USI) influenced detection probability while visibility (VISIB) did 194
not (Table 1). The constant ‘p’ model (Table 1: ψ(.),p(.)) performed poorly in 195
comparison. Akaike weights, ‘wi’, of the models (Table 1) suggest good support for USI 196
in comparison to VISIB. Therefore, for subsequent modeling of habitat-use, detection 197
probability was always modeled as a function of USI. 198
199
We evaluated 8 a priori models (Table 2). Based on AICC values, the model incorporating 200
REFUGES and USI were ranked at the top. Akaike weight, wi, for this model was close to 201
1 (0.94) and an evidence ratio of 31.3 with the next best model, indicated a high level of 202
certainty for this model (Burnham and Anderson 1998). However, support for the model 203
that included fruiting trees was poor with a delta AICC of 23.4 compared to the top 204
model. Summary of the estimated parameters for the candidate models (Table x) 205
indicates that covariates used have reasonable errors and exclude zero. 206
207
The over-dispersion parameter ‘
c
ˆ
’ was estimated as 0.69 for our global model 208
incorporating all measured covariates. 209
210
Discussion 211
Small-bodied ungulates are socially, morphologically and physiologically dissimilar to 212
their larger counter-parts. Their small body size is thought to result in stealth behavior 213
selectivity of food items (Barrette 1987). Additionally, their mostly solitary behavior 214
forces them to remain concealed in habitat features that provide cover from predators 215

11
(Geist 1998). Our study examined the influence of cover from predators and fruiting 216
trees on use of habitat by the small ungulate Indian chevrotain (Moshiola indica) using 217
the occupancy based approach. 218
219
Detection probability ‘
p
ˆ
’ for Indian chevrotain pellet-groups was moderate at 0.61, in 220
our study and higher than detecting individuals (0.48) of Indian chevrotain through 221
camera traps, (Nag, 2008). Since Nag (2008) did not model ‘
p
ˆ
’ as a function of sampled 222
covariates, estimates of ‘
p
ˆ
’ based on camera traps may have been higher. 223
Nevertheless, our study demonstrates that pellet-groups are a good indicator of Indian 224
chevrotain presence. A ‘
p
ˆ
’ of 0.61 is also within the range of 0.3 to 0.7 for the 225
occupancy framework to be useful for such a study (McKenzie et al. 2006). Modeling ‘
p
ˆ
’ 226
as a function of understorey structure (USI) significantly altered the parameter ‘

ˆ
’. 227
Understorey structure was also found to strongly influence the detection probability of 228
pellets of a forest dwelling bovid, four-horned antelope Tetracerus quadricornis (Krishna 229
et al. 2008). 230
231
By explicitly accounting for detection probability in our models, ‘

ˆ
’ (habitat-use) 232
increased by 40% (0.52 to 0.73) underscoring its importance. Our results largely match 233
estimates of occupancy of Indian chevrotain (0.69 to 0.78) estimated using camera traps 234
(Nag, 2008), a more intensive and costly method. In spite of the small size of the 235
adjacently placed grids in our study, there was perhaps little spatial correlation in 236
detecting signs between and within grids as indicated by an over-dispersion parameter 237

12
of less than one (MacKenzie et al., 2006, Williams et al., 2002), further implying that this 238
can be used successfully to model habitat-use in small ungulates. However, ‘

ˆ
’ may be 239
positively biased since we used spatial replicates in our study. Hence, for studies 240
conducted over longer periods of time, temporal replicates are recommended (Kendall 241
and White 2009). 242
243
Our results indicate that availability of cover from predators provided by understorey 244
structure, and refuges like buttresses and boulders, positively influenced the use of a 245
habitat by the Indian chevrotain. This is consistent with predictions of anti-predatory 246
behavior by small ungulates (Jarman 1974). Small ungulates are known to spring into 247
thick cover when closely approached, but otherwise remain concealed using buttresses, 248
boulders, snags as cover (Eisenberg and Lockhart 1972; Geist 1998). In south-east Asia 249
the lesser mouse-deer, (Tragulus javanicus), similarly used areas with cover during the 250
day (Matsubayashi et al. 2003). African chevrotains (Hyemoschus aquaticus), also use 251
areas that provide cover (Dubost, 2001). Consistent with this behavior from other small 252
ungulates, there was strong support for our expectation that, cover from predators 253
positively influences habitat-use by the Indian chevrotain. 254
255
Contrary to our a priori expectation, we did not find evidence in support of fruiting trees 256
influencing habitat-use by the Indian chevrotain. Fruit availability shows marked 257
seasonal fluctuation in various forest types including rainforests (van Schaik et al. 1993). 258
During such periods of resource scarcity, small ungulates like the steenbok (Raphicerus 259

13
campestris) and brocket deer (Mazama gouzoubira, M. americana), are known to 260
broadened their dietary width to include young leaves and shoots (Richard and Julia 261
2001; Toit 1993). Since the period of this study coincides with reduced fruit availability 262
in the Western Ghats (Sundarapandian et al. 2005), the Indian chevrotain may have 263
included food items like young shoots and leaves in its diet. Recent evidence also seems 264
to suggest that there may be no constraint on the quality of food intake by small 265
ungulates (Kerley et al. 2010). However, it is necessary to examine the diet of this 266
species in relation to inter and intra-annual phenological changes in the forest to help 267
confirm our results. As of now, our results indicate a greater role of habitat complexity 268
in the form of cover elements and understorey thickness, than fruit availability, in 269
influencing habitat-use by the Indian chevrotain. 270
271
The occupancy approach has been successfully used for a wide range of species recently 272
from (e.g. Baldwin and Bender 2008; Martin et al. 2006; Welsh et al. 2008). However, 273
this study to the best of our knowledge is novel in successfully interpreting the 274
parameter ‘

ˆ
’ as habitat-use rather than occupancy. Our study also demonstrates the 275
utility of the occupancy approach at very small scales, a useful method for many, 276
endangered but cryptic, elusive, and poorly studied animals. Recent studies have 277
recognized the role of small ungulates in seed dispersal of tropical trees (Brodie et al. 278
2009; Prasad and Sukumar 2010). Our study has provided a first step towards 279
integrating information on their habitat requirements to gain better insights into the 280
ecological processes mediated by these animals. 281

14
282
Conservation implications 283
284
Although the Indian chevrotain has been categorized as ‘Least Concern’, it suffers 285
moderate to heavy hunting pressures in many parts of its range (Madhusudhan and 286
Karanth 2002; Kumara and Singh 2004). Further, many parts of its forest habitat suffer 287
from habitat conversion to plantations and agriculture, livestock grazing, and extraction 288
of timber, fuel wood, leaf litter, and non-timber forest produce (Daniel 1991). In 289
addition to controlling hunting, to ensure long term conservation of the species it is 290
essential to retain elements of habitat complexity. Preventing drastic changes in 291
understorey structure and features providing cover, induced by increased use and 292
modification of forests by humans, may be very important for the management of small 293
ungulates throughout the world. 294
295
Acknowledgements 296
The project was funded by Department of Science and Technology, Government of 297
India. We thank National Centre for Biological Sciences, Wildlife Conservation Society-298
India Program and Centre for Wildlife Studies for logistic and financial support; and the 299
Karnataka Forest Department for permission and logistic support. Discussions with Ullas 300
Karanth, Samba Kumar, Arjun Gopalaswamy, Devcharan Jathanna, and my batchmates 301
helped conceptualize the study. We are grateful to Srinivasa for his invaluable help in 302

15
the field. Meghna Krishnadas is deeply acknowledged for the support provided at all 303
stages of the project. 304
305
References 306
Baillie JEM, Hilton-Taylor C, Stuart S (2004) 2004 IUCN Red List of threatened species: A 307
global assessment. IUCN, Switzerland. 308
Baldwin RA, Bender LC (2008) Distribution, occupancy and habitat correlates of 309
American martens (Martes americana) in Rocky Mountain National Park, Colorado. J 310
Mammal 89:419–427 311
Barrette C (1987) The comparative behavior and ecology of chevrotains, musk deer, and 312
morphologically conservative deer. In: Wemmer CM (ed) Biology and management of 313
the Cervidae. Smithsonian Institution Press, Washington DC, pp 200-213. 314
Bodmer RE (1990) Ungulate frugivores and the browser-grazer continuum. Oikos 315
57:319-325 316
Bowland AE, Perrin MR (1995) Temporal and spatial patterns in blue duikers Philatomba 317
monticola and red duikers Cephalophus natalensis. J Zoo 237:487-498 318
Branan WV, Marchinton RL (1985) Food habits of brocket and white-tailed deer in 319
Suriname. J Wild Manage 49:972-976 320
Brodie JF, Helmy OE, Brockelman WY, Maron JL (2009) Functional differences within a 321
guild of tropical mammalian frugivores. Ecology 90:688-698 322
Burnham KP, Anderson DR (1998) Model Selection and Multi-Model Inference. Springer-323
Verlag, New York. 324

16
Corlett RT (2007) The impact of hunting on the mammalian fauna of tropical Asian 325
forests. Biotropica 39:292–303 326
Daniel JC (1991) Ungulate conservation in India - Problems and prospects. Appl Anim 327
Behav Sci 29:349-356 328
Demment MW, van Soest PJ (1985) A nutritional explanation for body-size patterns of 329
ruminant and non-ruminant herbivores. Am Nat 125:641-672 330
Eisenberg JF, Lockhart M (1972) An ecological reconnaissance of Wilpattu National Park, 331
Ceylon. Smithsonian Contrib Zool 101:1-118 332
Dubost G (2001) Chevrotains. In: MacDonald D and Norris S (eds) The new encyclopedia 333
of mammals. Oxford University Press, Oxford pp 500-501. 334
Dubost G (1984) Comparison of the diets of frugivorous forest ruminants of Gabon. J 335
Mammal 65:298-316 336
Duckworth JW, Hemsagar Baral, Timmins RJ, (2008) Moschiola indica. IUCN red list of 337
threatened species. http://www.iucnredlist.org/apps/redlist/details/136585/0 338
Gagnon M, Chew AE (2000) Dietary preferences in extant African Bovidae. J Mammal 339
81:490–511 340
Gautier-Hion A, Emmons LH, Dubost G (1980) A comparison of the diets of three major 341
groups of primary consumers of Gabon primates, squirrels and ruminants. Oecologia 342
45:182-189 343
Geist V (1998) Deer of the world: their evolution, behaviour, and ecology. Stackpole 344
Books, Mechanicsburg. 345

17
Gopalaswamy AM, Karanth KU, Kumar NS, Macdonald DW (2012) Estimating tropical 346
forest ungulate densities from sign surveys using abundance models of occupancy. DOI 347
10.1111/j.1469-1795.2012.00565.x 348
Groves CP, Meijaard E (2005) Interspecific variation in Moschiola, the Indian chevrotain. 349
The Raffles Bulletin of Zoology 12:413-421 350
Hines JE (2006) PRESENCE2 - Software to estimate patch occupancy and related 351
parameters. 352
Hofmann RR (1989) Evolutionary steps of ecophysiological adaptation and 353
diversification of ruminants: A comparative view of their digestive system. Oecologia 354
78:443-457 355
Jarman, PJ (1974) The social organisation of antelope in relation to their ecology. 356
Behaviour. 48:215-267 357
Kendall WL, White GC (2009) A cautionary note on substituting spatial subunits for 358
repeated temporal sampling in studies of site occupancy. J App Ecol 46:1182-1188 359
Kerley GIH, Landman M, de Beer S (2010) How do small browsers respond to resource 360
changes? Dietary response of the Cape grysbok to clearing alien Acacias. Func Ecol 361
24:670–675 362
Krishna YC, Krishnaswamy J, Kumar NS (2008) Habitat factors affecting site occupancy 363
and relative abundance of four-horned antelope. J Zoo 276:63-70 364
Krishnan M (1972) An ecological survey of the larger mammals of Peninsular India. J 365
Bombay Nat Hist Soc 69:469-501 366

18
Kumara HN, Singh M (2004) The influence of differing hunting practices on the relative 367
abundance of mammals in two rainforest areas of the Western Ghats, India. Oryx 368
38:321-327 369
Macdonald DW (2001) The New Encyclopedia of Mammals. Oxford University Press, 370
Oxford. 371
MacKenzie DI, Nichols JD, Lachman GB, Droege S, Royle AJ, Langtimm CA (2002) 372
Estimating site occupancy rates when detection probabilities are less than one. Ecology 373
83:2248-2255 374
MacKenzie DI, Nichols JD, Royle JA, Pollock KH, Bailey LA, Hines JE (2006) Occupancy 375
modeling and estimation. Elsevier, San Diego. 376
Madhusudan MD and Karanth KU (2002) Local hunting and the conservation of large 377
mammals in India. Ambio 31:49-54 378
Matsubayashi H, Bosi E, Kohshima S (2003) Activity and habitat use of lesser mouse-deer 379
(Tragulus javanicus). J Mammal 84:234-242 380
Martin TG, Mcintyre S, Catterall CP, Possingham HP (2006) Is landscape context 381
important for riparian conservation? Birds in grassy woodland. Biol Conserv 127:201–382
214 383
Nag K (2008) Assessing animal abundance from photographic capture data using an 384
occupancy approach. MSc Thesis, Manipal University 385
Pascal JP (1988) Wet evergreen forests of the Western Ghats of India: ecology, 386
structure, floristic composition and succession. French Institute of Pondicherry, 387
Pondicherry. 388

19
Prasad S, Pittet A, Sukumar R (2009) Who really ate the fruit? A novel approach to 389
camera trapping for quantifying frugivory by ruminants. Ecol Res 25:225-231 390
Prasad S, Sukumar R (2010) Context-dependency of a complex fruit-frugivore 391
mutualism: temporal variation in crop size and neighborhood effects. Oikos 119:514-523 392
Prins HH, de Boer WF, van Oeveren H, Correia A, Mafuca J, Olff H (2006) Co-existence 393
and niche segregation of three small bovid species in southern Mozambique. Afr J Ecol 394
44:186-198 395
Ramachandran KK, Nair PV, Easa PS (1986) Ecology of larger mammals of Periyar Wildlife 396
Sanctuary. J Bombay Nat Hist Soc 83:505-524 397
Richard E, Julia JP (2001) Dieta de Mazama gouazoubira (Mammalia, Cervidae) en un 398
ambiente secundario de Yungas, Argentina. Iheringia, Serie Zoologia 90:147-156 399
Sodhi NS, Koh LP, Brook BW, Ng PKL (2004) Southeast Asian biodiversity: an impending 400
disaster. Trends Ecol Evol 19:654–660 401
Sundarapandian SM, Chandrasekaran S, Swamy PS (2005) Phenological behaviour of 402
selected tree species in tropical forests at Kodayar in the Western Ghats, Tamil Nadu, 403
India. Curr Sci 88:805-810 404
Toit JT (1993) The feeding ecology of a very small ruminant, the steenbok (Raphicerus 405
campestris). Afr J Ecol. 31:35-48 406
van Schaik, Terborgh J, Wright JS (1993) The phenology of tropical forests: adaptive 407
significance and consequences for primary consumers. Ann Rev Eco Syst 24:353-377 408

20
Welsh HH Jr, Pope KL, Wheeler CA (2008) Using multiple metrics to assess the effects of 409
forest succession on population status: a comparative study of two terrestrial 410
salamanders in the US Pacific Northwest. Biol Conserv 141:1149–1160 411
Wilson DE, Reeder DM (2005) Mammal Species of the World. 3rd edition. Johns Hopkins 412
University Press, USA. 413
414

21
Fig 1 Map of study area (Someshwara Wildlife Sanctuary). 415
416
(a) A schematic representation of each sampling unit (grid of dimension 50m x 50m), 417
indicating four spatial replicates and the locations at which various predictors were 418
measured in each replicate. 419
420

22
Fig 2 Comparison of pellet groups between the Indian chevrotain (a) and muntjac (b) 421
422
423

23
Table 1 Model selection results for covariates influencing detection probability of Indian 424
chevrotain pellet-groups 425
Model
AICC
ΔAICC
wi
K
ψ(.), p(USI)
771.84
0.00
0.99
3
ψ(.), p(.)
784.16
12.32
0.00
3
ψ(.), p(VISIB)
785.61
13.77
0.00
5
426
AICc is the small sample Akaike Information Criterion. ΔAICc is the difference between 427
the least AICc value and the AICc value of a model. wi is the Akaike model weight. K is the 428
number of model parameters. USI is an index of understorey structure, while VISIB is a 429
measure of understorey thickness. 430
431

24
Table 2 Summary of model selection procedure for parameter

ˆ
, interpreted as 432
habitat-use 433
434
Model

ˆ
ES ˆ
AICC
ΔAICC
wi
K
ψ(REFUGE+USI), p(USI)
0.72
0.02
749.89
0.00
0.94
5
ψ(REFUGE), p(USI)
0.81
0.01
756.89
7.00
0.03
4
ψ(REFUGE+VISIB), p(USI)
0.80
0.01
757.79
7.90
0.02
5
ψ(USI), p(USI)
0.79
0.02
758.31
8.42
0.01
4
ψ(VISIB), p(USI)
0.98
0.01
765.52
15.63
0.00
4
ψ(.), p(USI)
0.93
0.09
771.84
21.95
0.00
3
ψ(FRUIT), p(USI)
0.93
0.01
773.29
23.40
0.00
4
ψ(.), p(.)
0.82
0.08
782.30
32.41
0.00
2
435

ˆ
is the estimated probability of use (habitat-use).
ES ˆ
is estimated standard error 436
around

ˆ
. AICC is Akaike Information criterion corrected for small sample size. ΔAICC is 437
the difference between the least AICC value and the AICC value of a model. wi is the 438
Akiake model weight. K is the number of model parameters. REFUGE is the number of 439
potential features that provide cover, USI is an index of understorey structure, VISIB is a 440
measure of understorey thickness and FRUIT is a measure of availability of trees in fruit 441
in the grid. 442
443

25
Table 3 Expected response of the Indian chevrotain for the measured variables 444
Measured variable
ψ
p
USI
+
-
VISIB
+
-
REFUGE
+
NA
FRUIT
+
NA
445
Ψ is the parameter of interest, habitat-use and p the detection probability of Indian 446
chevrotain pellet-groups. REFUGE is the number of potential features that provide 447
cover, USI is an index of understorey structure, VISIB is a measure of understorey 448
thickness and FRUIT is a measure of availability of trees in fruit in the grid.449

26
Table 4 Estimates of slope parameters and standard errors from the candidate set of 450
models. 451
Model
Slope estimate
Standard Error
ψ(REFUGE+USI), p(USI)
REFUGE: 0.46; USI: 0.18
REFUGE: 0.16; USI: 0.09
ψ(REFUGE), p(USI)
REFUGE: 0.66
REFUGE: 0.22
ψ(REFUGE+VISIB), p(USI)
REFUGE: 0.68; VISIB: 0.37
REFUGE: 0.22; VISIB: 0.41
ψ(USI), p(USI)
USI: 0.25
USI: 0.09
452

Figure
Click here to download high resolution image

Figure
Click here to download high resolution image