Noa Pinter-Wollman 1)
(Animal Behaviour Graduate Group, University of California, One Shields Avenue, Davis,
CA 95616, USA)
(Accepted: 15 January 2009)
When animals encounter a novel environment they can either reject it and leave or accept it and explore their new home. It is important to understand what governs animals’ response to a novel place because of the fitness consequences and wildlife management implications entailed. Here I examine the spatial behaviour of translocated African elephants (Loxodonta africana) upon arrival at a novel environment. I monitored the movement patterns of 12 radiocollared elephants for a year post-translocation. I documented the first account of both female and male African elephants homing back to their natal habitat. More males than expected left the release site, but female–calf units also homed to their natal habitat, demonstrating that homing is not confined to one sex or age.When examining the spatial behaviour of elephants that remained near the release site I did not find a relationship between habitat exploration and last distance from release site, elephant age, or social association. However, I did find a negative correlation between habitat exploration and distance from human activities. This work provides biological insights regarding individual variation in spatial activity of animals in a novel environment and offers recommendations for future management actions.
Keywords: African elephant, exploration, individual variation, novel environment, translocation, wildlife management.
1) Author’s current address: Department of Biology, Stanford University, USA, e-mail:
© Koninklijke Brill NV, Leiden, 2009 Behaviour 146, 1171-1192
DOI:10.1163/156853909X413105 Also available online – www.brill.nl/beh
Many species encounter novel environments both naturally, e.g., during dispersal
(Stenseth & Lidicker, 1992), and due to human activities, e.g., animal relocations (Stamps & Swaisgood, 2007). Upon arrival at a novel environment animals can either remain and assess their new home or reject it and leave.
Rejecting a release site and returning to the source site i.e., homing, is often observed in translocations. Most translocations aim to establish viable populations at the release site (Fischer & Lindenmayer, 2000) or permanently remove animals from the source site (Richard-Hansen et al., 2000). Therefore, it is important to understand the factors underlying homing e.g., life history traits (Tuberville et al., 2005), translocation timing (Belisle et al., 2001), and release procedure (Bangs & Fritts, 1996).
Remaining in a novel environment requires assessing it through habitat exploration (Stamps, 2001). Exploration can provide important information (Clark & Mangel, 1984; Eliassen et al., 2007) but can also entail costs due to predation (Larsen & Boutin, 1994; Yoder et al., 2004), or exhaustion (Baker & Rao, 2004 ). Despite the potential fitness consequences of exploration, studies of translocated animals often overlook the exploration process and report only the last distance from release site (Musil et al., 1993; Clarke & Schedvin, 1997; van Vuren et al., 1997; Armstrong et al., 1999; Cowan, 2001). Distance from natal nest or release site can affect survival and fitness (Byrom&Krebs, 1999; Hansson et al., 2004) but does not always correspond to the exploration exhibited (Moehrenschlager & Macdonald, 2003; Tweed et al., 2003, Selonen & Hanski, 2006). Thus, both distance and exploration should be addressed when studying spatial behaviour in a new environment.
Linking exploration with other biological variables can provide useful proxies for predicting individual variation in exploration and, thus, in fitness (Dingemanse et al., 2004). Such variables could be age (Mikheev & Andreev, 1993), social behaviour (Sunnucks, 1998; Stoewe et al., 2006) and other behavioural traits (Fraser et al., 2001; Watters & Meehan, 2007). Animals translocated for solving human-wildlife conflict can be easily assigned behavioural measures relating to their fear from humans such as, latency to approach a stationary observer i.e., the ‘human approach test’ (Hemsworth et al., 1989, 1996), and distance from roads (Theuerkauf et al., 2003; Whittington et al., 2005).
African elephants (Loxodonta africana) are a vulnerable species that is often translocated in wildlife management actions to reduce human-elephant conflict (Muir, 2000;Wambwa et al., 2001, Dublin & Niskanen, 2003). However, only little information has been thus far provided on the spatial behaviour of translocated male African elephants (Muir, 2000; Garai & Carr, 2001; Slotow & van Dyk, 2004) and none on females. Here I examine the spatial behaviour of both male and female translocated African elephants upon release
to a novel environment.
Some life history aspects of African elephants may influence their spatial behaviour in a novel place. Males are the dispersing sex and travel long distances in search for mates; females and their offspring live in matriarchal groups (Moss & Poole, 1983). Therefore, males would be more likely to leave the release site, or explore it more extensively if they remain, than females whose movements may be confined by the physical abilities of young calves. Furthermore, matriarch’s age governs both social (McComb et al., 2001) and spatial (Foley, 2002; Foley et al., 2008) knowledge and it has been suggested that old males are information repositories (Evans & Harris, 2008). Thus, older individuals may have prior knowledge about the new habitat, and may explore it less extensively than younger individuals. Finally, elephant social aggregations are hypothesized to serve as a platform for ecological information exchange (Foley, 2002). Therefore, I expect association with conspecifics to reduce self-exploration of the new environment.
In addition to the above life history aspects, I also examine the relationship between exploration and approach distance to human activities to provide an accessible proxy of exploration for future management actions. Previous research has shown a negative correlation between habitat exploration and latency to approach unfamiliar objects in birds (Verbeek et al., 1994). Therefore, I anticipate finding a negative correlation between exploration and approach to human activities.
Material and methods
Translocation and study site
During September 2005, 150 African elephants were translocated from Shimba Hills National Reserve and Mwaluganje Elephant Sanctuary on the coast of Kenya (4–4.3.S and 39.5–39.3.E) to Tsavo East National Park (2.00–3.70.S and 38.13–39.30.E), a distance of 160 km (Figure 1) as part of Kenya’s Wildlife Service (KWS) effort to decrease human-elephant con-flict around Shimba Hills. The translocation was conducted by the KWS according to the IUCN elephant translocation guidelines (Dublin & Niskanen, 2003) and funded by the Kenya Government. Elephant groups of fewer than
12 individuals were targeted and transported as intact units. Adultmales were targeted based on their location and accessibility by road during the translocation and were moved in pairs. The elephants were released in a ‘hard release’ method i.e., were allowed to walk directly into the park from the transportation trucks and not kept in an enclosure. Translocating the 150 elephants took 32 days during which 20 groups (average ± SD group size 6.8 ± 2.5) and 20 adult males were moved.
The release site, Tsavo East, differs greatly from the source site, Shimba Hills, in its climate, vegetation, size, and elephant density. Tsavo East is semi arid with an average annual rainfall ranging from 300 to 700 mm, while Shimba Hills is part of the coastal plateau with an average annual rainfall of 1500 mm and a humid equatorial climate. During the rains, vegetation growth in Tsavo East is spatially heterogeneous and unpredictable, in contrast to the spatially homogeneous and reliable vegetation growth in
Shimba Hills (van Wijngaarden, 1985). Tsavo East is the largest national park in Kenya (13 950 km2) and along with the adjacent Tsavo West National Park forms the largest protected area in the country (20 812 km2), which is home to approx. 9000 elephants (Blanc et al., 2007); density = 0.43 elephants/km2. The source site, Shimba Hills is a small (250 km2) reserve surrounded by human settlements containing approx. 600 elephants (Blanc et al., 2007); density = 2.4 elephants/km2. These ecological differences
between the release site and source site provide a unique opportunity to study the behaviour of elephants in a novel environment, the release site.
During the translocation all elephants were individually marked for posttranslocation
monitoring. All 150 elephants were tagged with yellow zip ties on their tails and painted with a unique white number on their backs for individual identification. Natural ear marks and tusk shapes were also used for individual identification (Moss 1996). The age of each translocated elephant was estimated, based on Moss (1996), according to body measurements (back length and shoulder height) taken during the translocation and observations later in the field. Of the translocated elephants, 12 adults moved on different days (3 independent males chosen haphazardly, and 9 females —the matriarchs of the first 9 groups moved) were fitted with GPS/VHF elephant collars (Sirtrack, New Zealand) to enable detailed post-release tracking of movement patterns. Due to a malfunction in the collars’ drop-off mechanism, only one collar was recovered and the GPS data from it retrieved and presented here (the collar of individual No. 89). Spatial data for all other collared individuals are based on radio-telemetry using the VHF signal only.
Post-translocation monitoring was conducted for 380 days after the release of the first group, providing at least a year of data for all translocated elephants. Collared elephants were tracked from the air and ground by locating the VHF signal of their radio collar using a TR-4 Tracking receiver (Telonics, USA). During ground surveys, a three element hand-held folding Yagi antenna (Sirtrack) was used to detect the collar signals. A compass was used to determine the bearing towards the signal and the location from which the bearing was taken recorded using a Geko 201 GPS unit (Garmin, USA).
The computer program Locate II (Nams, 2000) was later used to triangulate the elephants’ location. Aerial tracking was conducted with a light Super Cub aircraft fitted with wing-mounted antennae. Signal directionality was determined using a TAC-2-RLB Antenna Control Unit (Telonics). Elephants’ locations from air were recorded using a Geko 201 GPS unit (Garmin). Each collared elephant was sought at least 2–3 times a week, from air and ground, and located at least once a month. For additional information on the resolution of the spatial data, see Appendix A, Table A-1.
The location and identity of every translocated elephant spotted were recorded
providing data on the status of all translocated elephants, and not only collared individuals. Analysis of exploration patterns, however, was based only on data from collared individuals, due to the higher temporal resolution of collared individuals’ sightings.
Despite elephants living in matriarchal groups (Moss & Poole, 1983), the number of individuals associating with the collared elephants varied throughout the study due to their dynamic fission-fusion social behaviour (Moss & Poole, 1983; Wittemyer et al., 2005; Pinter-Wollman et al., 2009). Therefore, to examine the correlation between spatial and social behaviour posttranslocation, I used the average number of elephants associating with each collared elephant as a measure of sociality. Associating individuals were de-fined as elephants within 500 m of one another during a time window of 2 h,
based on previous work on African elephants’ social behaviour (McComb et al., 2000, 2001, 2003; Wittemyer et al., 2005). Additional information about the social behaviour of the translocated elephants can be found in Pinter-Wollman et al. (2009).
Minimal distance between the collared elephants and a stationary observer was estimated for each ground observation, similarly to obtaining minimal distance to observer in a ‘human approach test’ conducted in farm animals (Hemsworth et al., 1996):When collared elephants were seen during ground surveys I immediately stopped the vehicle and remained stationary. The minimal distance to which the collared elephant approached the stationary observer was estimated based on known distances to prominent topographical features in proximity to the elephants. Observations were carried out until the elephants could no longer be seen, allowing them enough time to sense the Observer’s presence. Minimal distance to observer was averaged across all sightings for each collared elephant. Statistical analysis was then conducted on the log of this average for normalization purposes.
Last distance from the release site and minimal distance from roads were calculated based on the collared individuals’ locations. Last distance from release site was calculated as the straight Euclidian distance between the release site and the location of the collared elephant at the end of the study, approximately a year after release. Distance from roads was calculated as the minimal distance between each collared elephant sighting and the nearest road, using GIS data from the Tsavo East research station. Average distance from roads was computed for each elephant for further statistical analysis. Distance calculations were implemented in ArcView 3.2 (ESRI, USA). Other spatial measures can be found in Appendix A, Table A-2.
A Moving Weighted Centroid (MWC) analysis was developed to accurately describe exploration patterns, taking into account the patchy manner in which elephants use their habitat. Elephants exhibit a heterogeneous usage of their habitat by moving great distances rapidly between areas of high use (Cushman et al., 2005), referred to as streaking (Douglas-Hamilton et al., 2005), also seen in other mammals (Sinclair, 1984; Sheppard et al., 2006). In the MWC analysis I calculated the distance of each collared elephant location (focal location) to the centroid of its locations from the previous 30 days. The centroid was calculated as a spatio-temporal weighted average:
Where X and Y are the x, y coordinates of the centroid; xi and yi are the x, y coordinates of a sighting (i) within the 30 days preceding the focal location; and ti is the number of days separating a sighting (i) from the focal location.
Weighing each location inversely proportional to the number of days separating it from the focal location assigned earlier locations a lower impact on the location of the centroid. Because the centroid for each new location was based on data from the 30 days preceding the focal location (average ± SD number of sightings available for each point was 7.4 ± 5.9), the centroid moved over time, and overall effectively created a moving average of the general movement patterns for each elephant.
The MWC is an extension of using a fixed time window, which creates discrete activity centers, as described in Waterman (1986). In expansion of the discrete activity centers, the MWC analysis creates a continuous activity center by employing principles from smoothing techniques, often used to analyze animal movements, such as moving windows (Pace, 2001), moving average (MA), and moving weighted average (MWA) (for a review of smoothing techniques see Hen et al. (2004)). However, in contrast to such smoothing techniques, whose goal is to average the movement pattern, in this study the deviation from the average smoothed movement was of interest here as an exploration measure. A similar approach was previously applied to describe elephants’ heterogeneous movement patterns by Cushman et al. (2005).
Calculating the distance (d) of each observation from the MWC provided information regarding the amount of localized movements each elephant exhibited during its exploration of the novel environment. The statistic used to describe the exploration value for each elephant was the median of d for all observations over a course of a year. The median of d and not its mean was used due to the skewed distribution (positive skew) of d. To test the robustness of d values median to sampling effort, I performed a 1000 run 95% cross-validation procedure (Efron & Tibshirani, 1993). The median of d showed small variation (SD = 0.52 km), indicating that d is robust to outliers and uneven sampling. For a comparison of d’smedian to other spatial measures, see Appendix A, Table A-2. All calculations were conducted in Matlab (MathWorks, USA).
To determine whether the probability of translocated independent males (older than 15 years: Poole, 1996) to leave the release site, Tsavo (East and West) National Parks, differed from that of female–calf social units, I used a two tailed Fisher’s exact test. Female–calf units were used and not larger social units because some social groups that were captured together broke up (Pinter-Wollman et al., 2009). Variability among elephants in last distance from release site and in exploration was expressed using a normalized measure of variability: coefficient of variation (Cv).
To study the relationship between exploration and other biological variables, linear regression was used when a dependent variable (exploration) and an independent variable (age) could be defined. Because I could not naturally define a dependent and an independent variable for all other cases, Pearson’s correlation coefficient was used when comparing exploration with social association, minimal distance to observer, and average distance to roads. Correction for multiple testing was conducted using the False Discovery Rate (FDR) method (Benjamini & Hochberg, 1995). Consequently, statistical significance was set at p-values less than 0.025 for testing the relationship between exploration and other biological variables. Statistical analyses were implemented in Matlab using its statistical toolbox (MathWorks) and in JMP (SAS Institute, USA).
Leaving Tsavo East and West National Parks
Eight of the 109 translocated elephants whose fate is known left the release site, Tsavo (East and West) National Parks, and either returned to Shimba Hills (N = 6) or ended up elsewhere on the coast, near Malindi (N = 2)