GPS-identified, low-level nocturnal activity of vervets (Chlorocebus pygerythrus) and olive baboons (Papio anubis) in Laikipia, Kenya

2.1 Study site and subjects

We conducted this research at the Mpala Research Centre on the Laikipia Plateau of central Kenya (0.29°N, 33.90°E), a semi-arid environment (total rainfall in 2014: 443.2 mm) that includes riverine habitat along the Ewaso Nyiro River, dominated by Acacia xanthophloea, and bushlands away from the river, dominated by A. etbaica, A. mellifera, A. brevispica, and Boscia angustifolia. The study area supports a nearly full complement of wild mammals (Isbell & Bidner, 2016; Young, Okello, Kinyua, & Palmer, 1998).

We monitored five groups of vervets and four groups of olive baboons with GPS and VHF/GPS radio-collars, respectively, all with tri-axial accelerometer data loggers (Savannah Tracking, Inc., Nairobi, Kenya). The goal was to place collars on two adult females in each study group. Females were targeted because they are philopatric and served as representatives of their groups. We trapped vervets in drop-traps modified from Grobler and Turner (2010) and baboons in wire cage traps placed near regular sleeping sites (Jolly, Phillips-Conroy, & Mueller, 2003). The traps were prebaited with dried maize prior to trapping for 4–24 days. When an adult female entered a trap, we caught vervets by pulling the prop away using a rope tied to it, and baboons, by releasing a taut rope attached to the wire cage door, both from inside a vehicle at least 20 m away. Trapped primates were immobilized with 10 mg/kg ketamine administered via hand injection or blow dart. After weighing and measuring the animals, and fitting the collars on them, we returned them to cage traps or to two drop-traps stacked together to recover from immobilization. We fitted nine vervets and six olive baboons with collars in January 2014, and fitted three more vervets with collars in March 2014.

2.2 Data collection

The data collection period reported here spanned 7.5 months from 16 January to 31 August 2014. We programmed the collars to take GPS fixes every 15 min, and accelerometer recordings for 3 s/min at 32 Hz, continuously throughout the study. When we were within Ultra High Frequency (UHF) range of each collar, we downloaded the collar's GPS and accelerometer data using a base station (E-obs GmbH, Gruenwald, Germany) fitted with either an omnidirectional marine antenna (cxl 900–3LW: Procom, Frederikssund, Denmark) or a nine-element yagi antenna (YAGI-869A: Low Power Radio Solutions, Witney, UK).

We also monitored terrestrial activities of primates at or near seven sleeping sites with camera traps (Reconyx Hyperfire PC900 and Rapidfire RM45; Reconyx, Inc., Holmen, WI) that ran continuously day and night. Body heat or movement triggered the cameras to take bursts of three photographs at the rate of 1/sec. All cameras used infrared illumination for nocturnal photos.

2.3 Data analysis

“Travel” and “activity” are measures that both involve movement. Travel was identified via GPS data and always involved spatial displacement of the collared individual greater than the precision of our GPS units. Activity was identified based on accelerometer data and involved bodily movement (McFarland, Hetem, Fuller, Mitchell, Henzi, & Barrett, 2013) that could, but did not have to, involve spatial displacement. For example, a primate engaged in grooming, local foraging, scratching, or body shifting would be scored as active but not traveling in our analysis. To quantify individual activity patterns, we calculated the overall dynamic body acceleration (ODBA), an aggregate measure that combines data from all three accelerometer axes. We first generated a histogram of ODBA values during the daytime, from which we identified a threshold ODBA value for distinguishing active from inactive states. We then applied this threshold to the night-time ODBA values to estimate the proportion of ODBA measures that were above this threshold (and thus the proportion of time spent active).

2.4 Travel

We extracted GPS data from 3,138 primate-nights between 20:00 hr and 04:00 hr, inclusive (local time). We limited our analyses to these hours to avoid confounding effects from animals arriving late at their sleeping trees or leaving particularly early. We first explored the characteristics of GPS units left in situ where vegetation was either sparse or wooded for 1–4 days to examine the accuracy of the units. We found that when the GPS unit's internal horizontal inaccuracy estimate was >23 m for a given point, the error of the point became unpredictable. In contrast, when the value was <23 m, the location error had a 95% error range of 11 m (but up to about 20 m error). Thus, we expect about 95% of the location estimates for a stationary GPS collar to lie within a 22 m-diameter circle of the collar's true location.

To determine whether animals traveled during the night, we developed and applied a spatiotemporal clustering algorithm to identify changes in the centroid of consecutive observations. This algorithm involved first calculating the distance between consecutive points (step length) and multiplying this by the time between points (in minutes). We then applied a hierarchical clustering algorithm (using the complete linkage algorithm) to these distances to generate a hierarchical tree in which the branch length is equal to the spatiotemporal distance between clusters of observations. We found that the distribution of between-cluster distances was bimodally distributed, and applying a cut-off of 45 m (which is twice the maximum GPS error) as the minimum spatiotemporal distance between clusters enabled us to generate an a priori set of candidate travel nights (Figure 1). The two modes in the distribution of travel nights are shown in Figure 1, with most nights having very little travel (gray) and fewer nights potentially having some travel (red). These candidate travel nights (n = 577) included many false positives, so we visually examined the nightly travel tracks for each animal-night found in the right-hand half of that distribution to manually extract the ones that showed clear evidence for travel (i.e., two clusters with separated centroids) in the movement plots.

When trying to characterize nightly travel of these diurnal primates, we found that there was no simple way of classifying distance traveled into movement versus nonmovement due to GPS noise. For example, we found the log of the distance between each consecutive GPS point to be normally distributed. This suggests that there were no cases where an individual moved a distance in a 15-min period that was larger than GPS error (or at least they were very rare). Had individuals made significantly larger movements than the size of the GPS error during the course of a night, for example, hundreds of meters, these points would have appeared as a second “hump” in the distribution of between-point distances (step lengths) and thus would have been clearly identifiable.

2.5 Activity

The total time spent active, pooled across individuals for each species, was estimated using data taken from 3D accelerometers integrated into the GPS collars. For each 3-s measurement of activity, we calculated ODBA by summing the total acceleration across all three axes of the accelerometer (Wilson et al., 2006). We then calculated the frequency distribution of log(ODBA) values and found that these were clearly bimodal. Comparing the nighttime distribution of log(ODBA) values to those from the daytime suggested that “active” and “non-active” were clearly different modes, and that an ODBA threshold value of log(992) = 6.9 would separate out observations of active individuals from those of nonactive individuals. This number represents the mean minima between distributions across all individuals. Thus, to calculate total time spent active for each night, we calculated the proportion of log(ODBA) values above 6.9. We calculated the moon phase for each night using the R package lunar (Lazaridis, 2014) to examine rates of activity relative to moon phase.

3.1 Travel

We analyzed 3,138 primate-nights (vervets: n = 2,029; baboons: n = 1,109) and found evidence of nighttime travel on only 21 of these nights (vervets: n = 9; 0.4% of vervet-nights; baboons: n = 12; 1.1% of baboon-nights). Only 3 of 12 vervets, all in different groups, ever traveled at night. In contrast, five out of six baboons, representing all four study groups, engaged in nighttime travel at least once. Baboons traveled on significantly more nights than vervets (χ² = 4.33, p = .04). On travel nights, vervets moved an average of 29.7 m (range: 11.−66.5 m, SD: 21.5 m) whereas baboons moved an average of 22.8 m (range: 11.6–54.8 m, SD: 17.2 m). Figures 2 and 3 provide examples of nights with and without travel in each species. None of the camera traps ever revealed primates on the ground at night. Inspection of GPS locations on those nights using Google Earth indicates that vervets remained in the trees on six nights, but on three nights the same vervet did not move into her sleeping trees until 21:00–03:00. On these nights, before ascending she was in the same bushland area, suggesting that there might have been a unique refuge at that particular location, but we cannot identify it. One night of arboreal travel may have been in response to a leopard, which was caught on camera at the sleeping site that night. Baboons were located on kopjes (rocky outcrops) on nine of the 12 baboon-nights in which they traveled, indicating terrestrial travel. On one night, two females from the same group crossed the river together, suggesting the possibility of a predator attack. On another night, a female appears to have descended to the ground before returning to the trees.

3.2 Activity

Figure 4 contrasts high activity during the daytime with low activity during the nighttime for vervets and baboons. Separating out the active state (the right-hand peak) from the “inactive” state (the left-hand peak) on the daytime plot, and then applying this same threshold to the nighttime reveals that vervets were active 13% of the night, and baboons, 15%. Activity varied little throughout the night (Figure 5) and was also not obviously affected by moon phase (Figure 6).