1

LRH: Loveridge, Wearn, Vieira, Bernard and Ewers 2

RRH: Logging Facilitates Invasion of Small Mammals 3

4

5

6

Movement behaviour of native and invasive small mammals shows logging may 7

facilitate invasion in a tropical rainforest 8

9

Authors: Robin Loveridge* (1) (2), Oliver R. Wearn (1)(3), Marcus Vieira (4), Henry Bernard 10

(5) and Robert M. Ewers (1). 11

12

(1) Department of Life Sciences, Imperial College London, Silwood Park Campus, Buckhurst 13

Road,
 Ascot, Berkshire, SL5 7PY, UK 14

(2) BirdLife International Cambodia Programme, #2, Street 476, Toul Tompung 1, 15

Chamkarmon, P.O. Box 2686, Phnom Penh, Cambodia 16

(3) Institute of Zoology, Zoological Society of London, Regent’s Park, London, NW1 4RY, 17

UK 18

(4) Departamento de Ecologia, Instituto de Biologia, Universidade Federal do Rio de Janeiro, 19

Caixa Postal 68020, 21941-902, Brazil 20

(5) Universiti Malaysia Sabah, Institute for Tropical Biology and Conservation, Sabah, 21

Malaysia 22

23

24

* Corresponding author: Robin Loveridge (robin.loveridge@gmail.com)

2

Received ____; revision accepted ___ . 1

Abstract 2

Invasive species pose one of the greatest threats to biodiversity. This study investigates the 3

extent to which human disturbance to natural ecosystems facilitates the spread of non-native 4

species, focusing on a small mammal community in selectively logged rainforest, Sabah, 5

Borneo. The microhabitat preferences of the invasive Rattus rattus and three native species 6

of small mammal were examined in three-dimensional space by combining the spool-and-line 7

technique with a novel method for quantifying fine-scale habitat selection. These methods 8

allowed the detection of significant differences for each species between the microhabitats 9

used compared with alternative, available microhabitats that were avoided. Rattus rattus 10

showed the greatest preference for heavily disturbed habitats and, in contrast to two native 11

small mammals of the genus Maxomys, R. rattus showed high levels of arboreal behaviour, 12

frequently leaving the forest floor and travelling through the under and mid-storey forest 13

strata. This behaviour may enable R. rattus to effectively utilize the complex three-14

dimensional space of the lower strata in degraded forests, which is characterized by dense 15

vegetation. The behavioural flexibility of R. rattus to operate in both terrestrial and arboreal 16

space may facilitate its invasion into degraded forests. Human activities that generate heavily 17

disturbed habitats preferred by R. rattus may promote the establishment of this invasive 18

species in tropical forests in Borneo, and possibly elsewhere. We present this as an example 19

of a synergistic effect, whereby forest disturbance directly threatens biodiversity, and 20

indirectly increases the threat posed by invasive species, creating habitat conditions that 21

facilitate the establishment of non-native fauna. 22

23

Key words: Invasion, synergistic effects, Rattus rattus, selective logging, habitat preference, 24

arboreality. 25

3

1

VAST AREAS OF THE REMAINING NATURAL FOREST IN SE ASIA EXIST IN A LOGGED 2

AND DISTURBED STATE. In Borneo and Sumatra, for example, 46 percent and 72 percent of 3

natural forest is degraded, respectively (Margono et al. 2012, Gaveau et al. 2014). Within the 4

Malaysian state of Sabah, Borneo, less than 2 percent of the remaining lowland forests 5

remain undisturbed, mostly in just a few sites making up a total of 700 km2 (Reynolds et al. 6

2011). With the vast majority of forests existing in an altered state, it is essential for 7

conservation to understand the processes that drive community change in modified 8

landscapes (Hansen et al. 2001, Meijard & Sheil 2007). Despite this we still have only a 9

rudimentary mechanistic and theoretical understanding of why these changes in biological 10

communities occur, and the possible role that invasive species may play (Stokes et al. 2009). 11

In particular, synergistic effects between threat processes, such as between logging and 12

biological invasion, or biological invasion and the spread of disease (Wells et al. 2014a, 13

Wells et al. 2014b), have been poorly explored in tropical forests thus far, but have the 14

potential to increase the extinction risk of native species (Brook 2006). 15

16

Small mammals play a key role in biological communities, acting as both seed 17

predators and dispersers (Asquith et al. 1997, Wells & Bagchi 2005, Wells et al. 2009). 18

Small mammals are also important prey items for large avian and mammalian predators 19

(Puan et al. 2011, Wilting et al. 2006). Changes in the abundance of this functionally 20

important group in altered forests may cause cascading effects on other trophic levels 21

(Grassman et al. 2005). Native rodent species may be particularly threatened by the invasion 22

of the black rat, Rattus rattus (Stokes et al. 2009; Gibson et al. 2013). Rattus rattus is now 23

spread over most of the world (Amori & Clout 2003), and when it has invaded native habitats 24

on islands it has led to well-documented extinctions of native bird fauna (Atkinson 1985, 25

4

Barnett 2001), disruption to nutrient transport and soil fertility (Fukami et al. 2006) and 1

carbon sequestration (Wardle et al. 2007). The multiple ecological effects that cascade from 2

the invasion of R. rattus places a premium on understanding how those invasions occur in 3

order to limit the future spread of this species. 4

5

In Borneo, R. rattus was traditionally thought to be restricted to urban areas (Payne 6

and Francis 1985). However, R. rattus has been recently observed in a small number of 7

primary and secondary tropical forests of northern Borneo (Wells et al. 2006b, Cusack et al. 8

2014) and detected in the oil palm plantation matrix surrounding the forests examined in this 9

study (Cusack 2011). Although this species has not been systematically surveyed across 10

Borneo, the limited data available suggest it is not yet ubiquitous in natural forest areas 11

(Nakagawa et al. 2007, Wells et al. 2007, Bernard et al. 2009, Charles & Ang 2010). This 12

provides a unique opportunity to examine the niche overlap of R. rattus with native species at 13

a relatively early stage in the invasion process. 14

15

Previous survey work has demonstrated that the occurrence of R. rattus increases 16

along a gradient from undisturbed to heavily disturbed forest (Cusack et al. 2014). Here, we 17

build on this work by examining whether logging facilitates the invasion of R. rattus into 18

tropical forests by investigating their movement behaviour and microhabitat selection at 19

small spatial scales. Our hypothesis is that the invasive R. rattus has a stronger preference for 20

heavily disturbed forest microhabitat compared to native small mammals, hence logging 21

disturbance creates favorable microhabitats for the establishment of R. rattus in tropical 22

forests. 23

24

25

5

We compare the habitat preferences of the invasive R. rattus with three native species 1

of small mammals regarding (1) vertical space use, (2) preference for more (or less) disturbed 2

forest microhabitats, and (3) preference for moving along and through particular substrates 3

associated with forest disturbance. To effectively quantify a species’ habitat preference, 4

habitat measurements must be made at the spatial scale at which the species makes habitat 5

selection choices (Manly et al. 2002). We were able to study fine-scale microhabitat 6

preference using the spool-and-line technique (Miles et al. 1981, Boonstra & Crane 1986), as 7

it allows even very subtle route choices to be quantified (Wells et al. 2006, Harris et al. 8

2006). In addition, we employed a novel matching methodology to establish actual preference 9

for microhabitats, comparing used vs. control microhabitats. 10

11

METHODS 12

13

STUDY SITE AND SPECIES.- The study was conducted within the Stability of Altered Forest 14

Ecosystems (SAFE) project site in Sabah, Malaysian Borneo (Ewers et al. 2011). Small 15

mammal trapping was carried out in a 7200 ha area of repeatedly-logged forest, with high 16

spatial variation in levels of forest disturbance. The continuous forest of the SAFE project 17

area is connected to a large (> 1 million ha) area of similar forest to the north, and is 18

otherwise surrounded by an oil palm plantation matrix. 19

20

The four study species belong to the Muridae family (rats and mice). Native species 21

included the red spiny rat (Maxomys surifer), a widespread generalist species with a 22

characteristic red pelage, Whitehead’s rat (Maxomys whiteheadi), classified as Vulnerable on 23

the IUCN red list (IUCN 2014), and the long-tailed giant rat (Leopoldamys sabanus). We 24

compared these species to the newly-invading black rat (Rattus rattus). Ear tissue samples 25

6

were taken from all individuals and stored in 95% alcohol. MtDNA sequencing of the Rattus 1

species confirmed that it was a member of the Rattus rattus species complex (sensu Aplin et 2

al. 2011, Pagès et al. 2010) based on both cytochrome b and CO1 gene sequences (M. Pagès, 3

pers. comm.). 4

TRAPPING AND TRACKING INDIVIDUALS.- Three trapping grids were established at locations 5

that encompass a gradient of logging intensity and forest modification. Grids were located 6

more than a kilometre from the forest edge and were all connected by continuous forest. 7

Trapping was conducted for three months between May and July of 2012. Traps were 8

checked in the morning and captured individuals belonging to the four study species were 9

anaesthetized using diethyl ether and a small fur clip was made dorsally between the shoulder 10

blades. A spool was then attached to the under-fur using acrylamide gel (Loctite). Spools 11

were made of nylon quilting cocoons and weighed between 1 and 2.5 g, extending over 12

distances of 100 to 200 m (Danfield Ltd). Spools were wrapped in a thin layer of cling-film, 13

followed by a layer of electrical tape to prevent snagging on vegetation. The weight of 14

spools was adjusted to less than five percent of the individual’s body weight following the 15

ratio used in radio-tracking studies (Kenwood 2001). In addition, all trapped target species 16

were injected with a unique passive induced transponder tag (Francis Scientific Instruments, 17

Cambridge) to enable individual identification. Some individuals that were re-captured were 18

spool-tracked multiple times. 19

20

HABITAT VARIABLES.- Habitat use was examined at two spatial scales. The largest spatial 21

scale was the scale at which the availability of microhabitats changed in the forest 22

environment and was standardized to intervals of 10 m along the animal’s path. The smallest 23

spatial scale was the scale at which substrates (individual fallen branches, leaf litter patches, 24

7

open ground) changed along the animal’s path. This spatial scale varied from approximately 1

0.5 to 5 m. 2

3

QUANTIFYING MICROHABITAT PREFERENCES.- The first ten meters of each spool track were 4

discarded as a flight response (Harris et al. 2006), and the first microhabitat measurements 5

were recorded a further ten meters from the starting point of normal behaviour. Preliminary 6

trials were conducted to estimate the separation distance at which repeated measures on the 7

spool track became independent from each other. The autocorrelation between consecutive 8

10 m observations was assessed using the autocorrelation function (ACF) in R (R 9

Development Core Team, 2014). Negligible correlation was found at 10 m intervals 10

(correlation = 0.046), so this was subsequently used as the spacing between repeated 11

microhabitat measurements. 12

13

Microhabitat was defined as the area within a 1 m radius of the spool and nine 14

microhabitat variables were measured at 10 m intervals along the animal’s track (a detailed 15

description of microhabitat variables is provided in Table 1 of online supporting 16

information). All microhabitat variables were recorded by the same researcher (R.L). The 17

density of vegetation in four layers of forest strata was estimated visually for each 10 m 18

interval along the spool track following Puttker et al. (2008) (near the ground 0-0.5 m, 19

understory 0.5-3.0 m, mid-story 3.0-20 m, and canopy > 20 m). We assigned a score of 20

between one and four to each stratum to give an index of vegetation density. Four further 21

variables indicating potential resource availability and predation risk were also recorded 22

(total canopy closure, forest quality, tree basal area, habitat concealment score). As covariates 23

that may influence the space use of individuals, we recorded rainfall during the night and 24

lunar phase. Rainfall may influence small mammal movement by masking the noise of travel 25

8

across complex substrates (Vickery & Bider 1981; Browman et al, 2000), and variation in 1

moonshine may influence the visibility of the microhabitats, making some locations less 2

desirable when well lit by a full moon (Kotler et al. 1991). 3

4

An assessment of microhabitat preference requires a comparison between the 5

microhabitats that are selected by a species with other microhabitats that are available, but 6

not used. To compare observed vs. control microhabitat, measurements taken every 10 m on 7

the spool track were paired with measurements at a spatially matched location off the track. 8

The locations of these control measurements were determined by walking 10 m in a straight 9

line on a bearing relative to the spool track of 45 °, 135 °, 225 °, or 315 °, with bearings 10

selected sequentially (Fig. 1). 11

12

QUANTIFYING SUBSTRATE PREFERENCES.- We broke spool tracks into a series of discrete 13

steps, where a step was defined as a straight-line section of track with no change of direction 14

greater than 20 °. For each step, we characterized the dominant substrate into one of eight 15

categories following Wells et al. (2006), and the length of the step (a detailed description of 16

the different substrate categories is provided in Table 1 of online supporting information). 17

Species-specific step lengths were similar; M. surifer had the longest mean step length at 2.0 18

m, and M. whiteheadi the shortest at 1.8 m. 19

20

As with the microhabitat variables, we recorded paired control observations for each 21

step by noting the dominant substrate along a control step located off the spool track. Control 22

steps were determined by walking the same distance as the observed step on a bearing of 45 23

°, 135 °, 225 ° or 315 ° relative to the direction of the spool step, with bearings selected 24

sequentially. 25

9

1

STATISTICAL ANALYSIS.- Linear mixed effects models were used to account for the nested 2

repeated measures taken on the same spool track from individually-identified small 3

mammals. The R packages nlme and lme4 (Bates et al. 2011, Pinheiro et al. 2011) were used 4

for these analyses. The categorical rain and lunar phase variables were included as covariates 5

in all models but are not reported as they do not impinge on our hypotheses. The significance 6

of fixed effects, including interaction terms, was determined from likelihood ratio tests by 7

comparing models with and without the fixed effect or interaction. 8

9

To test for vertical spatial segregation between species, we used a binomial general 10

linear mixed effects model to compare repeated measures of the height above ground at 10 m 11

intervals along the spool track across species. Height was categorized into ground (< 0.5 m) 12

or above-ground (> 0.5 m) for analysis, and species, rain and lunar phase were fitted as fixed 13

effects. The nested data structure was explicitly modelled by nesting repeated measures taken 14

on the same spool track within individually-identified small mammals. 15

The microhabitat variables used in this study examined different layers of forest strata and 16

together provided a holistic measure of forest disturbance. The values were highly 17

intercorrelated so we reduced the number of variables to a smaller set of principal component 18

axes using the vegan package in R (Oksanen et al. 2011). Variable standardization was 19

carried out by dividing the value of each variable by the maximum values for that variable. 20

Much of the variation (35 %) was captured within the first principle component (PC1), and 21

we therefore used this as the only variable in subsequent modelling of species’ microhabitat 22

preference. This single composite variable provided a robust and easy to interpret axis from 23

low forest disturbance (negative PC1 values) to high forest disturbance (positive PC1 values). 24

Negative PC1 values were related to high forest quality scores, high canopy and mid-storey 25

10

density and high tree basal area. Positive PC1 scores were associated with high understory 1

and high ground vegetation density. 2

3

To compare the strength of microhabitat preference between species, a single 4

“microhabitat preference” response variable was generated for each 10 m observation along 5

the spool track. This response variable was defined as the difference in PC1 scores between 6

observed and paired control observations. This provided a measure of the strength of 7

preference for more or less disturbed microhabitats in relation to the immediate, surrounding 8

environment. A linear mixed effects model was used to compare microhabitat preference 9

among species, retaining rain and lunar phase as additional fixed effects and nesting repeated 10

measures of tracks within individually-identified small mammals as random effects. 11

12

To compare the use of substrates across species, we first calculated the ratio between 13

the number of times a substrate category was recorded in observed versus matched control 14

steps per spool track. This ratio was then used to standardize the observed frequency of 15

substrate use by the availability of the substrate in the immediate vicinity of each spool track. 16

We log-transformed the ratio so that positive and negative values scaled equally. Positive 17

values of this “substrate preference” score indicate preference for that particular substrate and 18

negative values indicate avoidance of that particular substrate. A linear mixed effects model 19

was used to compare substrate preference among species and substrate type, retaining rain 20

and lunar phase as additional fixed effects and nesting repeated measures of tracks within 21

individually-identified small mammals as random effects. 22

23

RESULTS 24

25

11

SPECIES’ SPATIAL BEHAVIOUR.- We attached 53 spools to 41 individuals, from which a total of 1

293 movement steps were recorded: 15 M. surifer individuals with 139 steps, 13 M. 2

whiteheadi individuals with 81 steps, nine L. sabanus individuals with 41 steps and four R. 3

rattus individuals with 32 steps. The small number of R. rattus reflects their relatively low 4

abundance in this habitat at what we believe is an early stage in the invasion process. The 5

total spool track followed was 3.4 km with a mean spool length of 64 m per individual (SE = 6

44.3 m). All species were captured on each of the three trapping grids and individuals of 7

different species displayed a large amount of spatial overlap in their terrestrial foraging 8

behaviour. Spool tracks of different species released on the same night were occasionally 9

found to cross over each other, and individuals of different species were recorded using the 10

same fallen log or dry riverbed. 11

12

Vertical segregation of space was observed between species (χ2 = 26.43, df = 3, p < 13

0.001). All observations of both Maxomys species were recorded on the forest floor (< 0.5m 14

above the ground). However spool tracks of both R. rattus and L. Sabanus were frequently 15

observed to pass along lianas (woody vines) that emanated from the forest floor and 16

penetrated up into the under and mid-storey forest strata. 28 percent of R. rattus observations 17

(n = 32) and 27 percent of L. Sabanus observations (n = 41) were recorded in the understory 18

(0.5 to 3 m) and a further nine percent of R. rattus observations and 15 percent of L. sabanus 19

observations were recorded in the mid-storey of the forest (3 to 10 m). 20

21

MICROHABITAT AND SUBSTRATE PREFERENCE.- There were significant differences among 22

species in their microhabitat preferences (Fig. 2) (χ2 = 17.91, df = 3, p < 0.001). Across all 23

species, R. rattus had the strongest preference for the most heavily disturbed forest 24

(preference = 0.31, SE = 0.06), followed by the two Maxomys species that had weaker 25

12

preferences for disturbed forest (M. whiteheadi: preference = 0.13, SE = 0.01; M. surifer: 1

preference = 0.10, SE = 0.01). Pairwise comparison between the two Maxomys species 2

confirmed that M. whiteheadi showed the marginally stronger, though non-significantly 3

different preference for more disturbed habitats (difference in preference score =0.03, 4

SE = 0.04, df = 27, p = 0.45). Of the four species, only L. sabanus had a preference for less 5

disturbed habitats (preference = -0.089, SE = 0.01). 6

7

The four species showed similar patterns of preference across the substrate types (Fig. 8

3), with no significant differences among species (χ2 = 4.33, df = 3, p = 0.228) and neither 9

was there a significant interaction effect between species and substrate preference (χ2 = 10

24.60, df = 17, p = 0.104), despite an apparent preference for R. rattus to prefer moving 11

through suspended leaf litter that the three native species showed no preference for (Fig. 3). 12

Across species, however, there were strong preferences for some substrates over others (χ2 = 13

88.97, df = 6, p < 0.001). Overall, small mammals showed the strongest preference for fallen 14

wood (preference value = 0.85, SE = 0.17) and dry stream beds (preference value = 0.83, SE 15

= 0.38), followed by rocky substrates (preference value = 0.53, SE = 0.18). Species tended to 16

actively avoid leaf litter (preference value = -0.47, SE = 0.07), but were not strongly 17

influenced by suspended leaf litter (preference value = 0.32, SE = 0.06), vines (preference 18

value = 0.25, SE = 0.06) or bare ground (preference value = -0.06, SE = 0.01). Comparing the 19

two Maxomys species; M. surifer had the slightly stronger preference for travelling along 20

vines, while M. whiteheadi had the stronger preference for travelling over rocky terrain (Fig. 21

3). 22

23

DISCUSSION 24

25

13

Our results emphasise how species make habitat selection choices at multiple spatial scales 1

(Manly et al. 2002), and that small-scale processes can be used to help explain large-scale 2

phenomena such as the invasion of R. rattus into modified forests. We have confirmed 3

earlier survey work that demonstrated R. rattus is more likely to be present in heavily 4

modified than unmodified tropical rainforest (Cusack et al. 2014), and confirmed our 5

hypothesis that this pattern appears to arise from specific, small-scale alterations to 6

microhabitat structure that R. rattus actively prefers to move in. 7

8

VERTICAL SPATIAL SEGREGATION.- This study represents one of the only efforts to quantify 9

the extent of arboreal space use by R. rattus (Key and Woods 1996), and how this ability may 10

facilitate invasion in logged tropical forests. Neither Maxomys species showed arboreal 11

behaviour, however both L. sabanus and R. rattus showed significantly different vertical 12

space use compared to these terrestrial species, utilizing higher layers of the forest strata. 13

Indeed, the foot morphology of the two species is very similar, both having broad feet and 14

prominent plantar pads, adaptations known to indicate arboreality in murid rodents (Aplin et 15

al. 2003). Moreover, the long tails of both species, up to 120 percent of body length for R. 16

rattus and 135 percent of body length for L. sabanus, help with balance and may well be 17

adaptions for climbing (Boonsong et al. 1988). Therefore R. rattus is well-adapted to fully 18

exploit the three-dimensional complexity of forest habitat. This is of particular concern as 19

nest predation by this introduced omnivore may negatively impact the survival success of 20

native bird species, as has been the case on other tropical islands such as Hawaii 21

(Amarasekare 1993). 22

23

The behavioural flexibility of R. rattus to utilise both terrestrial and arboreal space 24

may significantly increase the species’ invasion capability into the complex three 25

14

dimensional forest environment. Indeed, invasive species are more frequently habitat 1

generalists than specialists (Olden et al. 2004, Whitney & Gabler 2008). When entering a 2

new forest habitat R. rattus can potentially access both terrestrial and arboreal foraging 3

resources and a greater extent of niche space than an obligately-terrestrial, or obligately-4

arboreal species. Additionally, in Sabah, arboreal murid rodent communities are relatively 5

species poor compared to terrestrial small mammals (Wells et al. 2004). Therefore the 6

arboreal adaptive zone could have more space for invasion. In addition, it has also been 7

suggested that arboreal behaviour may reduce exposure to terrestrial predators, and present 8

physical barriers for aerial predators (Montogomery & Gurnell 1985, Buesching et al. 2008). 9

10

MICROHABITAT PREFERENCE.- Both Maxomys species and R. rattus showed significant 11

preference for more disturbed microhabitats, corroborating the results of a previous study 12

using trapping data at the same study site (Cusack et al. 2014). Here, however, we have 13

demonstrated at a smaller spatial scale that these three species choose to move through more 14

disturbed habitats than less disturbed, alternative habitats in the immediate surrounding 15

environment. Disturbed habitats were characterized by a low presence of large trees, 16

allowing the growth of dense ground and under-storey vegetation layers. This provides 17

excellent cover from predators. Heavily-disturbed habitats may maximize both predator 18

avoidance and resource availability, making these habitats desirable for small mammals. The 19

greater abundance of these habitats after human disturbance may also explain why it has been 20

observed that small mammal populations tend to increase after logging activity (Adler and 21

Levins 1994, Pardini 2004). Among the three species, R. rattus showed the strongest 22

preference for more disturbed habitat. This suggests that heavily-disturbed forest habitat is a 23

more strongly preferred habitat for R. rattus than it is for the native Maxomys species. 24

15

Heavily disturbed forest habitat may therefore facilitate the invasion of this species into 1

logged forests in Borneo (Stokes et al. 2009, Cusack et al. 2014). 2

3

It is interesting to note that this same preference for disturbed habitat was not 4

replicated by L. sabanus, which showed a preference for less disturbed habitat. One 5

explanation for this reverse trend is that the semi-arboreal L. sabanus benefits from exploiting 6

resources present in the more intact canopy of less disturbed forest. This species also has a 7

relatively larger body size, making it more mobile. This, combined with semi-arboreal 8

behaviour, may make this species less susceptible to predation (Montogomery & Gurnell 9

1985) and therefore more able to exploit the microhabitats that lack the dense vegetation 10

layer associated with disturbed forest sites. 11

12

SUBSTRATE PREFERENCE.- All study species showed a clear preference for moving along 13

fallen wood (Fig. 3). Woody debris contains high concentrations of invertebrates, a key 14

resource for small mammals (Emmons 2000). Larger pieces of woody debris may also be 15

used as cover in order to reduce visual detection by predators (Browman et al. 2000). This is 16

supported by the observation that individuals in this study sometimes travelled parallel to 17

fallen wood, close against the side. Travelling on top of woody debris has also been 18

hypothesized to provide a simple substrate to allow for faster, more efficient travel and to 19

allow small mammals to scan more effectively for predators whilst moving (Shadbolt & 20

Ragai 2010). Therefore, in terms of foraging and predation risk, fallen wood may be an 21

important microhabitat feature for maintaining populations of small mammals. 22

23

Small mammals tended to avoid moving through leaf litter, a pattern that may be 24

explained by the complexity of this substrate inducing slow, noisy travel with greater risk of 25

16

predation (Shadbolt & Ragai 2010). By contrast, the preference for sunken, dry river beds 1

may be explained by the more enclosed nature of this substrate, reducing visibility to 2

predators. 3

4

Using trap-level habitat data, Cusack et al. (2014) found that the occurrence of L. 5

sabanus could be predicted by the presence of fallen wood and low levels of leaf litter. 6

However these were not significant predictors of occurrence for either Maxomys species or R. 7

rattus (Cusack et al. 2014). By contrast, we found that all study species showed a significant 8

preference for fallen wood and an avoidance of leaf litter. The slight contrast in the findings 9

between the two studies likely arises from differences in methodologies. Cusack et al. (2014) 10

used microhabitat variables associated with individual trap sites, representing point locations 11

where individual small mammals were captured, whereas we used more finely detailed 12

information on the movement tracks of individuals through the habitat. This apparent contrast 13

in results therefore reinforces the importance of measuring habitat preference at the spatial 14

scale at which the species makes habitat selection choices in order to accurately discern 15

patterns of behaviour (Manly et al. 2002). 16

17

IMPLICATIONS FOR IMPROVED MANAGEMENT OF LOGGED FOREST SYSTEMS.- The microhabitat 18

data presented demonstrates how one of the major threats to biodiversity in the tropics, 19

namely habitat degradation and modification (Brown & Brown 1992), may multiply the 20

likely impact of a second major threat, invasive species (Kot et al. 1996). Specifically, we 21

found that habitat degradation may generate habitat that is more suitable for R. rattus, so 22

promoting the spread of this invasive species, which may in turn threaten the persistence of 23

native fauna. Species potentially impacted by the invasion of R. rattus into tropical 24

rainforests include under and mid-storey nesting birds, which are often vulnerable to nest 25

17

predation by rodents, and other small mammals through direct competition. In addition, the 1

threat of R. rattus to native fauna may be further magnified by the species potentially acting 2

as a conduit for the introduction of novel diseases into native small mammal communities 3

(Wells et al. 2014b). 4

5

The findings of this study agree with those of Cusack et al. (2014) in demonstrating 6

that R. rattus shows strong preference for heavily disturbed habitats. We therefore 7

recommend that the best practical defence against the invasion of R. rattus into tropical 8

forests is to preserve forest habitats that are less suitable for the species (Bernard et al. 2009). 9

This may be achieved through reduced impact logging techniques that are designed to 10

minimise levels of habitat disturbance and degradation (Gerwing & Uhl 2002, Putz et al. 11

2008). 12

13

Our results also point towards specific new methods for minimising the likelihood of 14

R. rattus invasion based on the fine-scale habitat preferences of the species for particular 15

forest structural elements. It is a common logging practice to undertake some crude shaping 16

of felled wood extracted from the forest prior to transportation in order to reduce the weight 17

and cost of transporting the wood. This contributes to large quantities of wood fragments 18

being discarded within logged forest stands, at the sides of logging roads and along the forest 19

edge. Recent research by Pfeifer et al. (2015) has shown that the volume of deadwood within 20

a logged tropical forest increases with greater forest disturbance. We have demonstrated that 21

fallen wood is especially preferred by R. rattus. Therefore reducing the availability of this 22

favoured microhabitat type along the forest edge may be one means of reducing the 23

likelihood of R. rattus invading logged forest stands. This could be achieved by 24

implementing logging practices that promote extraction efficiency. For example improving 25

18

tree selection to only log the most suitable tall, straight trees would minimise the amount of 1

wood discarded. Improved disposal of waste wood could also be explored. 2

3

Furthermore, cutting back mid-storey vegetation, such as lianas and knocking 4

suspended leaf litter to the ground may reduce the ability of invasive small mammals to 5

penetrate the non-terrestrial layers of forest strata. Before committing to such an action 6

however, it will be important to ensure that this approach will not inadvertently compromise 7

the movement corridors of other native small mammal species, such as scansorial mice in the 8

genera Chiropodomys and Haeromys that may exploit lianas. Our data, nonetheless, provide 9

initial support for the use of liana cutting as an important component of enrichment planting 10

practices aimed at rehabilitating the structure and composition of degraded forests (Kettle 11

2010, Ansell et al. 2011). 12

13

14

15

ACKNOWLEDGEMENTS 16

We thank the Royal Society’s South East Asia Rainforest Research Programme for logistical 17

support, and Yayasan Sabah, Maliau Basin Management Committee, the State Secretary, 18

Sabah Chief Minister’s Departments and the Sabah Biodiversity Council for permission to 19

conduct research. This work was funded by the Sime Darby Foundation. RME was 20

supported by European Research Council Project number 281986. We are grateful for the 21

comments of T. Coulson, and A. M. Valenzuela. 22

23

References 24

25

ADLER, G., AND R. LEVINS. 1994. The Island Syndrome in rodent populations. Quarterly Review of 26

19

Biology 69: 473-490. 1

2

AMARASEKARE, P. 1993. Potential impact of mammalian nest predators on endemic forest birds of 3

western Mauna Kea, Hawaii. Conservation Biology 7: 316-324. 4

5

AMORI, G., AND M. N. CLOUT. 2003. Rodents on islands: a conservation challenge. In SINGLETON, G. 6

R., HINDS, C. A., KREBS C. J., AND D. M. SPRATT (Eds). Rats, Mice and People: Rodent Biology and 7

Management. ACIAR Monograph 96. pp 63–68. Australian Centre for International Agricultural 8

Research: Canberra. 9

10

ANSELL, F., EDWARDS, D., AND K. HAMER. 2011. Rehabilitation of Logged Rain Forests: Avifaunal 11

Composition, Habitat Structure, and Implications for Biodiversity-Friendly REDD+. Biotropica 43: 12

504–511. 13

14

APLIN, K. P., BROWN, P. R., JACOB, J., KREBS, C. J., AND G. R. SINGLETON. 2003. Field methods for 15

rodent studies in Asia and the Indo-Pacific. Monographs. 16

17

APLIN, K. P., SUZUKI, H., CHINEN, A. A., CHESSER, R. T., TEN HAVE, J., DONNELLAN, S. C., ... AND A. 18

COOPER. 2011. Multiple geographic origins of commensalism and complex dispersal history of black 19

rats. PLoS one 6: e26357. 20

21

ASQUITH, N. M., WRIGHT S. J., AND M. J. CLAUSS. 1997. Does mammal community composition 22

control recruitment in neotropical forests? Evidence from Panama. Ecology 78: 941-946. 23

24

ATKINSON, I. A. E. 1985. The spread of commensal species of Rattus to oceanic islands and their 25

20

effects on island avifaunas. In MOORS P, J. (Ed). Conservation of island birds. pp 35–81. 1

International Council for Bird Preservation: Bristol. 2

3

BARNETT, S., A. 2001. The Story of Rats. Their impact on us, and our impact on them. Allen & 4

Unwin: Crows Nest, Sydney. 5

6

BATES. D., MAECHLER. M., B. BOLKER. 2011. lme4:Linear mixed-effects models using S4 7

classes. R package version 0.999375-42. 8

available from http://CRAN.R-project.org/package=lme4 9

10

BOONSTRA, R. AND I. CRAINE. 1986. natal nest location and small mammal tracking with a spool and 11

line technique. Canadian Journal of Zoology-Revue Canadienne De Zoologie 64: 1034-1036. 12

13

BOONSONG, L., LEKAGUL, B, AND J. A. MCNEELY. 1988. Mammals of Thailand. Saha Karn Bhaet. 14

15

BROWN, K. S., AND G. G. BROWN. 1992. Habitat alteration and species loss in Brazilian 16

forests. Pages 119–142. In Whitmore, T.C. and J. A. Sayer editors. Tropical Deforestation 17

and Species Extinction. Chapman & Hall, New York City. 18

19

20

CUSACK, J. J. 2011. Characterising Small Mammal Responses to Tropical Forest Loss and 21

Degradation in Northern Borneo Using Capture-Mark-Recapture Methods. Unpublished 22

master’s thesis, Imperial College London. 23

24

CUSACK, J. J., WEARN, O. R., BERNARD, H., AND R. M EWERS. 2014. Influence of 25

21

microhabitat structure and disturbance on detection of native and non-native murids in logged 1

and unlogged forests of northern Borneo. Journal of Tropical Ecology, 31: 25-35. 2

3

EWERS, R. M., DIDHAM, R. K., FAHRIG, L., FERRAZ, G., HECTOR, A., HOLT, R. D., ... AND E.C 4

TURNER. 2011. A large-scale forest fragmentation experiment: the Stability of Altered Forest 5

Ecosystems Project. Philosophical Transactions of the Royal Society B-Biological Sciences 366: 6

3292-3302. 7

8

FUKAMI, T., WARDLE, D. A., BELLINGHAM, P. J., MULDER, C. P. H., TOWNS, D. R., YEATES, G. W., 9

BONNER, K. I., DURETT, M. S., GRANT-HOFFMAN, M. N., AND W. M. WILLIAMSON. 2006. Above- and 10

below-ground impacts of introduced predators in seabird-dominated island ecosystems. Ecology 11

Letters 9: 1299–1307. 12

13

GERWING, J. J. AND C. UHL. 2002. Pre-logging liana cutting reduces liana regeneration in logging 14

gaps in the eastern Brazilian Amazon. Ecological Applications 12:1642–1651. 15

16

GIBSON, L., LYNAM, A. J., BRADSHAW, C. J., HE, F., BICKFORD, D. P., WOODRUFF, D. S., 17

BUMRUNGSRI, S. AND W. F. LAURANCE. 2013. Near-complete extinction of native small mammal 18

fauna 25 years after forest fragmentation. Science 341:1508–1510. 19

20

GOVE, J, H., DUCEY, M, J., STAHL, G., RAINGVALL, A. (2001). Point relascope sampling - A new way 21

to assess downed coarse woody debris. Journal of Forestry 99(4): 4-11. 22

23

GRASSMAN, L., TEWES, M, E., SILVY, J, N., AND K. KREETIYUTANOUT. 2005. Ecology of three 24

sympatric felids in a mixed evergreen forest in north-central Thailand. Journal of Mammalogy 86: 25

22

29-38. 1

2

HANSEN, A. J., NEILSON, R. P., DALE, V. H., FLATHER, C. H., IVERSON, L. R., CURRIE, D. J., ... & P. J. 3

BARTLEIN. 2001. Global change in forests: Responses of species, communities, and biomes. 4

Bioscience 51: 765-779. 5

6

HARRIS, D. B., GREGORY S. D., AND D. W. MACDONALD. 2006. Space invaders? A search for patterns 7

underlying the coexistence of alien black rats and Galapagos rice rats. Oecologia 149: 276-288. 8

9

IUCN. 2014. The IUCN Red List of Threatened Species. Version 2014.1 10

available from http://www.iucnredlist.org 11

12

KEY, G. E. AND R. D. WOODS. 1996. Spool and line studies on the behavioural ecology of rats (Rattus 13

spp) in the Galapagos Islands. Canadian Journal of Zoology-Revue Canadienne De Zoologie 74: 14

733-737. 15

16

KOT, M., LEWIS M. A., AND P. VAN DEN DRIESSCHE. (1996). Dispersal data and the spread of 17

invading organisms. Ecology 77: 2027-2042. 18

19

LEMMON, P. E. 1956. A spherical densiometer for estimating forest overstory density. Forest Science 20

2: 314–320. 21

22

MANLY. B. J. F., MCDONALD. J. L., THOMAS. D., MCDONALD. T. L AND W. P. L. ERICKSON. 23

2002. Resource Selection by Animals: Statistical Design and Analysis for Field Studies. 24

Second Edition. Dordretch, The Netherlands, Kluwer Academic Publishers. 25

23

1

MEIJARD, E., AND D. SHEIL. 2007. A logged forest in Borneo is better than none at all. Nature 446: 2

974. 3

4

MONTGOMERY, W. I., AND J. GURNELL. 1985. The behaviour of Apodemus. Symposium of 5

the Zoological Society of London 55: 89-115. 6

7

NAKAGAWA, M., MIGUCHI, H., SATO, K., AND T. NAKASHIZUKA. 2007. A preliminary study of two 8

sympatric maxomys rats in Sarawak, Malaysia: Spacing patterns and population dynamics. Raffles 9

Bulletin of Zoology 55: 381-387. 10

11

OKSANEN. J., GUILLAUME. F., ROELAND. B., LEGENDRE. K. P., MINCHIN. P. R., O'HARA. R. B., 12

SIMPSON. G., SOLYMOS. STEVENS. S, AND H. WAGNER. 2011. vegan: Community Ecology 13

Package. R. package version 2.0-2. 14

Available from http://CRAN.R-project.org/package=vegan 15

16

OLDEN, J. D., LEROY POFF, N., DOUGLAS, M. R., DOUGLAS, M. E., AND K. D. FAUSCH. 2004. 17

Ecological and evolutionary consequences of biotic homogenization. Trends in Ecology & 18

Evolution 19: 18–24. 19

20

PAGÈS, M., CHAVAL, Y., HERBRETEAU, V., WAENGSOTHORN, S., COSSON, J. F., HUGOT, J. P., ... AND J. 21

MICHAUX. 2010. Revisiting the taxonomy of the Rattini tribe: a phylogeny-based delimitation of 22

species boundaries. BMC evolutionary Biology 10: 184. 23

24

PARDINI, R. 2004. Effects of forest fragmentation on small mammals in an Atlantic Forest landscape. 25

24

Biodiversity and Conservation 13: 2567-2586. 1

2

PAYNE, J. AND C. M FRANCIS. 1985. A field guide to the mammals of Borneo. Sabah Society, 3

Kota Kinabalu, Malaysia. 4

5

PINHEIRO J., BATES, D., DEBROY. S., SARKAR. D AND THE R DEVELOPMENT CORE TEAM. 6

2011. nlme: Linear and Nonlinear Mixed Effects Models. R package version 3.1-102. 7

Available from http://cran.r-project.org/web/packages/nlme/index.html 8

9

PFEIFER, M., LEFEBVRE, V., TURNER, E., CUSACK, J., KHOO, M., CHEY, V. K., PENI, M., AND 10

R. M EWERS. 2015. Deadwood biomass: an underestimated carbon stock in degraded tropical 11

forests?. Environmental Research Letters, 10: 044019. 12

13

PUAN, C., GOLDIZEN, A. W., ZAKARIA, M., HAFIDZI, M. N., AND G. S. BAXTER. 2011. Absence of 14

differential predation on rats by malaysian barn owls in oil palm plantations. Journal of Raptor 15

Research 45: 71-78. 16

17

PUTTKER, T., PARDINI, R., MEYER-LUCHT, Y. AND S. SOMMER. 2008. Responses of five small 18

mammal species to micro-scale variations in vegetation structure in secondary Atlantic Forest 19

remnants, Brazil. BMC Ecology 8:9. 20

21

PUTZ, F. E., SIST, P., FREDERICKSEN, T., AND D. DYKSTRA. 2008. Forest Ecology and 22

Management Reduced-impact logging : Challenges and opportunities. Forest Ecology and 23

Management, 256: 1427–1433. 24

http://cran.r-project.org/web/packages/nlme/index.html

25

1

R DEVELOPMENT CORE TEAM. (2014). R: A language and environment for statistical 2

computing. R Foundation for Statistical Computing, Vienna. 3

Available from: http://www.R.project.org/ 4

5

REYNOLDS, G., J. PAYNE, W. SINUN, G. MOSIGIL, AND R. P. D. WALSH. 2011. Changes in 6

forest land use and management in Sabah, Malaysian Borneo, 1990-2010, with a focus on the 7

Danum Valley region. Philosophical Transactions of the Royal Society B-Biological Sciences 8

366: 3168-3176. 9

STOKES, V. L., BANKS, P. B., PECH, R. P., AND R. L. WILLIAMS. 2009. Invasion by Rattus rattus into 10

native coastal forests of south-eastern Australia: are native small mammals at risk? Austral Ecology 11

34: 395– 408. 12

13

VICKERY, W., AND J. BIDER. 1981. The influence of weather on rodent activity. Journal of 14

Mammalogy 62: 140-145. 15

16

WARDLE, D. A., BELLINGHAM, P., FUKAMI, T., AND MULDER, C. P. H. 2007. Promotion of ecosystem 17

carbon sequestration by invasive predators. Biology Letters 3: 479–482. 18

19

WELLS, K., PFEIFFER, M., LAKIM M. B., AND K. E. LINSENMAIR. 2004. Use of arboreal and terrestrial 20

space by a small mammal community in a tropical rain forest in Borneo, Malaysia. Journal of 21

Biogeography 31: 641-652. 22

23

WELLS, K., AND R. BAGCHI. 2005. Eat in or take away - Seed predation and removal by rats 24

26

(muridae) during a fruiting event in a dipterocarp rainforest. Raffles Bulletin of Zoology 53: 281-1

286. 2

3

WELLS, K., KOCK, D. B., LAKIM, M. B., AND M. PFEIFFER. 2006. Is Rattus rattus invading the 4

primary rainforest on Borneo? Malayan Nature Journal 59: 73–79. 5

6

WELLS, K., KALKO, E. K. V., LAKIM, M. B., AND M. PFEIFFER. 2007. Effects of rain forest logging on 7

species richness and assemblage composition of small mammals in Southeast Asia. Journal of 8

Biogeography 34: 1087-1099. 9

10

WELLS, K., CORLETT, R, T., LAKIM, M, B., KALKO, E, K, V., AND M. PFEIFFER. 2009. Seed 11

consumption by small mammals from Borneo. Journal of Tropical Ecology 25: 555-558. 12

13

WELLS, K., LAKIM, M. B., AND R. B. O’HARA. 2014a. Shifts from native to invasive small mammals 14

across gradients from tropical forest to urban habitat in Borneo. Biodiversity and Conservation 23: 15

2289– 2303. 16

17

WELLS, K., O'HARA, R. B., MORAND, S., LESSARD, J. P., AND A. RIBAS. 2014b. The importance of 18

parasite geography and spillover effects for global patterns of host–parasite associations in two 19

invasive species. Diversity and Distributions 21: 477-486. 20

21

WHITNEY, K. D., AND C. A. GABLER. 2008. Rapid evolution in introduced species, “invasive traits” 22

and recipient communities: challenges for predicting invasive potential. Diversity and Distributions, 23

14: 569–580. 24

27

1

28

1

10m A

BCobs

CDobs

BCcont

CDcont

AB

DEobs EFobs

DEcont

EFcont

2

Fig. 1. An example spool track of a male M. surifer illustrating the spool track the animal 3

moved along (continuous black line) with observation locations (filled circles) and the 4

spatially matched control routes (dotted lines) with control locations (open circles). A= 5

release point; AB= first 10 m route of spool track that was discarded as a flight response; 6

BCobs= observed 10 m route of spool track; BCcont= spatially matched 10 m control route for 7

segment BC. 8

29

LS MS MW RR

-1.0

-0.5

0.0

0.5

1.0

Species

Mic

roh

ab

itat

pre

fere

nce

1

Fig. 2. Microhabitat preferences of three native (white) and one introduced species (grey) of 2

small mammals. Positive values of the habitat preference score indicate a preference for 3

selecting more disturbed habitats over less disturbed habitats, and negative values indicate the 4

reverse. LS = Leopoldamys sabanus, MS = Maxomys surifer, MW = Maxomys whiteheadi, 5

RR = Rattus rattus. 6

30

BG FW LL RK SLL ST V

-2-1

01

2

Substrate

Su

bstr

ate

pre

fere

nce

va

lue

a

BG FW LL RK SLL ST V

-2-1

01

2

Substrate

Su

bstr

ate

pre

fere

nce

va

lue

b

BG FW LL RK SLL ST V

-2-1

01

2

Substrate

Su

bstr

ate

pre

fere

nce

va

lue

c

BG FW LL RK SLL ST V

-2-1

01

2

Substrate

Su

bstr

ate

pre

fere

nce

va

lue

d

1

Fig. 3. Substrate preference values of Leopoldamys sabanus (A), Maxomys surifer (B), 2

Maxomys whiteheadi (C) and Rattus rattus (D) for bare ground (BG), fallen wood (FW), leaf 3

litter (LL), rocky substrates (RK), suspended leaf litter (SLL), dry stream beds (ST) and vines 4

(V). Positive preference values indicate substrates that are preferred and negative values 5

indicate substrates that are avoided. 6