Home range and use of diurnal shelters by the Etendeka round-eared sengi, a newly discovered Namibian endemic desert mammal

Capture and radio-tracking

We accumulated 2,742 trap-nights, capturing 3 rodents (one each of Gerbillurus, Petromyscus, Petromus; 0.11% trap success) and 7 M. micus individuals (0.26%). To try to capture all the sengi individuals at the study site, we often set trap transects across areas where we had already captured sengis, in addition to adjacent areas. Remarkably, we only once recaptured one of our tagged sengis (#4612F). We captured an eighth sengi by hand at night (#4585M), but only collared seven (Table 1); a single young female was only ear-tagged. Both #4427F and #4585M disappeared soon after collaring, and provided no data. For any particular analysis, a subset of only relevant data were used, thus sample sizes did not always conform to the overall totals shown in Table 1.

Home range

The average home range sizes of the five radio-collared sengis, as determined by the different methods of analyses, were highly variable (Table 2), spanning from 7.2 to 22.8 ha. The average maximum length of the home ranges, calculated using the 100% convex polygon method, was 705 m. However, this was greatly influenced by #4254M that had a remarkably large oblong-shaped home range (Figs. 3 and 4). The average distance between the arithmetic mean centers of overlapping 100% convex home ranges was 425 m, range 256–608 m, which is a useful comparative measure of sengi dispersion (see interpretation of M. flavicaudatus home range estimates in Discussion). Because of our small sample size and uncertainty that all resident sengis in the study area were captured, we have not presented detailed overlap analyses.

The home range size (Table 2) and shape (Figs. 3 and 4) of #4254M was odd compared to the other four sengis. We located this male mostly at each end of his oblong-shaped home range, which spanned over 1.5 km (Table 2). He moved from end-to-end of his home range 11 times, making the journey so quickly that we were only able to roughly track his path once, when he traveled the length of his home range (about 1.5 km and over 80 m in elevation) within 60 min, presumably in a relatively straight course with few pauses. The area between the ends of his home range was atypical habitat for M. micus, being a slightly sloping alluvial fan composed of softer and lighter-colored gravels and fewer rocks than on the surrounding higher slopes (Fig. 3). The only other home range that was not completely located on rust-colored Etendeka volcanic substrates was that of #4020F, with about 0.73 ha at the southern edge falling on the lowest alluvial flats in the study area, which were composed of finer and lighter colored gravels with few rocks on the surface (Figs. 1 and 3). The home range areas of all the sengis tended to fall below the steeper areas of the Etendeka formation that had huge boulders and rock faces (Fig. 3).

We closely followed #4020F on her home range twice during the night of 1 October 2014 by keeping sight of her reflective ear-tag. Starting at 2152 h, she covered about 219 m in 10 min (1.3 km/h) and her route (based on the GPS-determined track of the observer) was a large circle that did not quite meet the starting point. The second track started at 2217 h, and covered 89 m in 3 min (1.6 km/h) in roughly a straight line. The sengi easily kept ahead of us as it bounded from rock to rock, obviously following a familiar route. During our study, we found no worn sengi paths across the substrate, but nevertheless they appeared to follow familiar routes, as demonstrated when we spotted a lone unmarked sengi (became #4585M) within the home range of #4856F. The sengi appeared unfamiliar with the area because he continually stumbled over and bumped into rocks as he clumsily fled, which allowed us to chase and hand-capture him while keeping him in the beam of our headlamp (no physical or visual impairments were noted when collared and released). It was impossible to similarly capture our tagged sengis because they were too agile and swift.

After we radio-collared #4585M, we only located him once the next day, even though we searched widely (several km) in areas adjacent to our study area on several days. Because our transmitters had a line-of-sight range of about 1 km, it seems unlikely that we lost the signal. It is possible that the transmitter failed, but we never spotted any male ear-tagged sengis without an associated radio signal. Sengi #4585M possibly became prey of a Cape fox (Vulpes chama A Smith, 1833) that we saw in our study site on several nights. This was probably also the fate of #4612F, given that we found her shed and functioning transmitter with tooth damage (Table 1).

The areas (100% convex polygons) encompassing all day shelters for each of the five radio-tagged sengis averaged 36.8% of their respective home range (Table 2). The distribution of day shelters within a home range showed no obvious pattern, other than the sengis used locations with suitable rock shelters and tended to be well inside the home range boundaries (Fig. 4).

Shelter characteristics

We examined a sample of day shelters used by the five collared sengis (#4020F n = 13, #4254M n = 9, #4612F n = 5, #4856F n = 11, #4947M n = 11). The ground surrounding the shelters was always rock and boulder strewn, averaging 52% coverage (range 40–95%). Aspect and slope varied by animal, but showed no overall trend that differed from the surrounding habitat in each home range. Shelters were typically under a single rock with a horizontal crevice (Fig. 2B) with an average height of 6.6 cm (range 3–12 cm). No shelters showed any obvious signs of alteration, such as excavation, digging, or collected bedding. Three of 49 shelters had some windblown grasses or plant matter, but it was never noticeably arranged or manipulated, and seemed typical of the surrounding boulder fields. Interior substrates of dust, sand, or gravel more or less matched the surrounding substrate. Only one shelter of 49 contained feces (3 pellets), and none of the shelters had partially eaten food or scraps. The entrances to shelters showed significant directionality (Raleigh’s Z test, n = 41, z = 3.66, p < 0.05) with an average compass direction of 193° south by southwest, despite the fact that slope aspect varied among individuals and showed no overall directionality (Raleigh’s Z test, n = 49, z = 0.35, p > 0.2).

We regressed shelter temperature against the shelter rock thickness and recovered a significant negative relationship (n = 47, R² = 0.396, p < 0.01, Fig. 5A), thus confirming that thicker rocks may provide more stable temperature environments and protection from wide temperature fluctuations. Because we measured shelters on different days, we additionally sought to control for differences in midday temperature by subtracting shelter temperature from local ambient air temperature. We again found a significant positive relationship, suggesting that thicker rocks were cooler relative to air temperature (general linear regression, n = 47, R² = 0.4117, p < 0.01, Fig. 5B). Despite confirming the potential benefit of thicker shelter rocks to protect from extreme temperatures or wide fluctuations, we cannot confirm whether sengis are actually choosing shelters to take advantage of these benefits. In fact, most shelters were under rocks with smaller thicknesses (Fig. 5), but it is not clear whether this is due to an active choice on the part of sengis, or whether they are constrained by availability.

Shelter use

The collared sengis were strictly nocturnal. Once sheltered at night, usually near dawn, they normally remained in the same shelters throughout the day, and were very reluctant to leave. For example, we checked 33 occupied shelters twice during the day between 13 and 5.5 h prior to sunset, and in only two cases did a sengi change shelters. In one case (#4947M) the distance between shelters was about 3 m, and in the second case (#4612F) it was about 30 m. On three days we checked #4020F four different times during daylight, and on one day three times, and #4856F at four different times on one day. Neither of these sengis shifted shelters during the day. When we recaptured the four remaining radio-tagged sengis at the end of the study on 26 October 2014, between 1000 h and 1145 h, we had to dislodge or remove the shelter boulders to get the animals to flee into the capture nets, which further demonstrated their reluctance to leave their day shelters.

We never observed or radio-tracked any diurnal sengi movements, and they all were active on every night with one exception. During the night of 8 October, #4020F did not leave her day shelter, and when we checked her after dawn she was torpid in her shelter. We thought she might have entangled a forefoot in her collar, but upon capture we found no problems. She quickly came out of torpor and after her release she resumed her typical nocturnal activity pattern.

We determined whether collared sengis used different shelters during late night (usually just before dawn) compared to the following day. In 26 of the 31 cases, switching did not occur, indicating that the sengis often sheltered for the day well before first light. Related to this pattern, we extracted location data for four sengis (those with the most robust overall data sets: #4020F, #4254M, #4856F, #4947M) and determined whether we found them in a night shelter or not during two periods: between 2100 and 0100 h (early night), and between 0200 and 0600 h (late night). In the early period, there were 130 pooled observations, with 14 in night shelters (10.8%). During the late period, we had 117 observations with 37 (31.6%) in night shelters. These data support our subjective assessment that the animals were more active early in the night compared to late at night. This pattern made it nearly impossible for us to determine when animals retreated to shelters for the day, compared to when they left their day shelters for a night of activity. We monitored 40 day shelters starting at about sunset and the average departure time was 1938 h, with a range of 1913–1959 h. In two additional cases, a sengi (#4254M) had not left the day shelter by 2010 and 2015 h, when we terminated observations.

Even though the sengis rarely switched shelters within a day, they readily switched shelters from day to day, rarely using a site more than once (Table 3). Pooling individuals, we monitored 85 day shelters and 93% were used once, 5% twice, and 1% each for three and four times. The average interval between using the same shelter was 3.2 days, with a range of 1–9 days. We found no evidence that different individuals used the same shelter, nor that more than one sengi occupied a shelter at the same time, although it is remotely possible that untagged sengis paired with our collared animals.

Twice we located sengis under low bushes at night—a 1 m high Commiphora bush and a 2 m high Boscia bush (Fig. 1). Bushes in this size range only numbered 2 or 3 individuals in each home range. While under the canopy of these bushes, the sengis were “nervous” and easily disturbed by our movement, running to the opposite side of the bush from the observer on several occasions, but they did not flush into the open nor did they foot-drum. While we were about 5 m from the animals, we observed them for about 15 min while they groomed and rested (Fig. 2A) on the surface of the gravel substrate. They were always alert with their eyes open and ready to flee. These observations were terminated after they bounded off into the night.

Home ranges

The estimated home range sizes of the five sengis we collared are dependent on the method of analysis. Both convex polygon and kernel methods incorporate areas that were rarely if ever used, but we have included both metrics to allow comparison with published data. Our limited data suggest that a more conservative representation of the home ranges of our tagged sengis is obtained with the relatively new OREP method, especially for #4254M (Table 2, Figs. 3 and 4). Unfortunately, no previous sengi studies have used this method, as is the case with the concave polygon technique. We nevertheless have included both with the hope that future studies will also find that they are a useful alternative to convex polygon estimates because they may provide better insights into space use by sengis.

The three species of Macroscelides occupy very arid habitats (Dumbacher et al., 2012; Dumbacher et al., 2014) compared to other sengis, thus ecological insights may be gained from these and other sengi species. Schubert et al. (2009) provides quantitative home range data for the Karoo round-eared sengi (Macroscelides proboscideus Shaw 1800) near Springbok, South Africa, based on radio-tracking methods. Using direct observations, Franz Sauer (1973) with his wife Elinore report home ranges of about 1 sq km for the Namib round-eared sengi (Macroscelides flavicaudatus Lundholm 1955) in the Namib Desert southeast of Walvis Bay, Namibia, which is over 80 times larger than what Schubert et al. (2009) found (Table 4). As additional home range data for other species were published (Table 4), sengi home ranges of a square kilometer seemed almost unbelievable.

To better understand the Sauer (1973) home range estimate, we closely examined his definitions of space use, which are different than what is typically used. For example, Sauer (1973, pages 74 and 94 among others) states that M. flavicaudatus had an average home range of a square kilometer, but he also indicates that this was in fact a crude calculation of density based on his 20 sq km study area. On the other hand, based on his descriptions, it is almost certain that his sengis had home ranges that were larger than other small sengis, which are less than 2 ha (Table 4). To obtain a crude home range estimate that more closely conforms to the more widely accepted definition of Burt (1943), we used the 300 m mode of the average distances between the main shelters used by two closest neighbors on adjacent home ranges (Sauer, 1973, pg 71, Table 1, Fig. 7, pg 95). If we assume that the length of each of two adjacent sides of a hypothetical home range is thus 300 m, and adjoining home ranges were in relatively homogeneous habitats (Ibid.), we obtain an estimated home range area of about 9 ha, which is over an order of magnitude smaller than the density of 100 ha/sengi that he called a “home range”. This re-estimation of home range size (sensu Burt, 1943) reconciles the seemingly small modal average distances between shelters on adjacent home ranges of 300 m and an estimate of home range areas of 1 sq km. Lastly, 300 m is not very different from our similar metric of 425 m for M. micus (Table 4). Even through our home range estimate for M. flavicaudatus is not strictly comparable with other sengi home range estimates, it probably is sufficient for our discussion.

The literature related to mammalian home ranges is large, including attempts to relate the sizes of home ranges with physiological factors such as trophic level (calorie sources), body size (calories needs), metabolic rate (rate that calories are used), social structure (group versus individual needs), and phylogeny (McNab, 2002, pages 335–336). These factors are highly variable across a wide range of mammals, making comparisons difficult, except for the sengis, which share a very tightly defined adaptive syndrome with very similar phylogeny, metabolic rate, morphology, diet, reproduction, locomotion, social structure, etc. (Rathbun, 1979; Rathbun, 2009). In contrast, the variation in the body size and habitats occupied by sengis stands out (Rathbun, 2009, see ‘Introduction’).

Although body weight data for sengis are available (Table 4), there are several metrics that might be used to quantify the habitats used by sengis. Given the life history traits of sengis, we believe that prey abundance is particularly important. Unfortunately, prey abundance is not easily measured or available from most sengi study sites (but see Rathbun, 1979; FitzGibbon, 1995). However, rainfall is probably a reasonable proxy, and these data are available (Table 4). When sengi home range sizes are plotted against rainfall, the points for M. flavicaudatus and M. micus are far removed from the rest of the sengis in the plot (Fig. 6A), despite our using conservative estimates for these two species (see discussion of Sauer above, and Table 2). Although M. proboscideus occupies a low-rainfall habitat similar to its congeners, it clusters with the other smaller sengis (Fig. 6A), which suggests that low rainfall habitats do not fully explain home range size for similarly sized sengis (keeping in mind their very similar adaptive syndrome). The Succulent Karoo, where the data for M. proboscideus were gathered (Schubert et al., 2009; Schubert, 2011), is a relatively small area between the very low and concentrated winter rainfall regime of the Namib Desert to the north, and the low summer rainfall regime of the Mediterranean climate to the south. Although the Succulent Karoo is arid, the rainfall is spread across both winter and summer months (Desmet & Cowling, 1999). This rainfall pattern results in a richer vegetation (Cowling & Hilton-Taylor, 1999) and invertebrate fauna (Vernon, 1999) than might be expected based only on total average rainfall. Thus, the home range area of M. proboscideus clusters closer to the other small sengis than with M. flavicaudatus and M. micus (Fig. 6A). The general positive relationship of mammalian body weight and home range size (McNab, 2002) is supported by syntopic Petrodromus and Rhynchocyon in a coastal forest in Kenya (FitzGibbon, 1995; Table 4; Fig. 6B), but is overshadowed by the Macroscelides species, especially the two in the Namib Desert (Fig. 6B). Based on our analysis, we hypothesize that prey availability may have the greatest influence on the home range sizes of sengis, but unfortunately data on prey availability are lacking for most sengis.

In most sengi home range studies, a male will occasionally attempt to overlap with more than one female, often resulting in an exceptionally large and oblong home range. However, this configuration is not stable due to the would-be polygamous male retreating when a new male appears to associate with (and mate-guard) one of the females (Komer & Brotherton, 1997; references in Table 4). We speculate that the large hour-glass-shaped home range of #4254M represented a similar attempt at polygamy (although we did not trap the northwestern end of his home range to determine if there was a female in that area).

Male–female sengi pairs exhibit few pair-bond behaviors and spend relatively little time together (except during brief periods of estrus), yet some species have home ranges that are virtually congruent (Rathbun, 1979), while in others the ranges only partially overlap (all other references in Table 4 and Fig. 3). One hypothesis to explain this variation is that the degree of overlap is density dependent. In habitats where sengis essentially occupy all suitable space and thus are dense, male and female home ranges are nearly congruent and intra-sex overlaps are rare because the areas are defended sex-specifically, whereas when sengis are more dispersed, the home range overlap within male–female pairs is reduced (Rathbun & Rathbun, 2006). At our Namibia study site, we were unable to capture and radio-track as many sengis as we had hoped, and we could not document that all resident sengis were radio-tagged. However, based on our intensive trapping effort and extensive radio-tracking and observation activities on the study site, it is unlikely that many if any resident sengis escaped our notice. We speculate that the home range configurations that we documented (Fig. 3) are consistent with the density dependent home range overlap model. We again hypothesize that prey availability may be the most important underlying factor in determining sengi density, and thus many home range characteristics. However, multiple factors may be involved (Di Stefano et al., 2011), including mate availability. Sauer (1973) suggested that shelters may have been limiting for M. flavicaudatus, although we doubt this was the case at our study site (Fig. 1).

Sheltering

There were two noteworthy findings regarding shelters. First, was the lack of a central or home burrow or shelter, as found in many other small mammals. Second, was how unremarkable the shelters were; there was no sign of bedding, excavation or alteration, and every shelter seemed to simply be a small space or crevice under a rock where sengis hid during the day. Both findings are similar to the sheltering habits of other Macroscelidinae (Rathbun, 2009). It is difficult to determine which factors were motivating the use of rock shelters by M. micus because of the large number of possible factors, including sengi behavior, predation threat, weather, environmental conditions, and shelter availability. We believe the most important two factors were the thermal traits of the shelters and predation threat.

We found that midday temperatures of shelter rocks were inversely related to rock thickness (confirming the ability of thicker rocks to resist temperature fluctuations), and we found that shelter openings were significantly orientated toward 193° south. We suspect that these two features have related consequences for sheltering sengis. Like many deserts, the Namib is characterized by frequent high and low temperature extremes (Seely, 2004). Thus, the size and orientation of a shelter rock may allow M. micus to passively (behaviorally) avoid temperature extremes and thus reduce energy needed for thermoregulation (McNab, 2002), which is likely important for such a small-bodied desert dweller. For example, the sengis might choose shelters in order to take advantage of the thermal inertia of rock to buffer day and night temperature extremes. In western Namibia, the prevailing winds come from the south (Mendelsohn et al., 2002). Winds often blew hard (we measured up to 13.5 m/s) during the midday and afternoon. Thus, south-facing shelter entrances may be more exposed to cooling breezes during the heat of the day. In addition, the south side of a shelter corresponds with the shady and thus cooler side of the rock during the heat of midday because the sun is slightly angled toward the north during this time of year.

Lovegrove, Lawe & Roxburgh (1999) documented daily torpor in M. proboscideus, which is likely a physiological strategy that sengis use under conditions of limited food availability and low temperatures to conserve energy (Mzilikazi, Lovegrove & Ribble, 2002). It is possible that the torpid sengi we encountered (#4020F) was implementing this strategy in this hyper-arid study site with a hypothesized low abundance of prey. However, more research is needed to further explore the relationships between shelter traits, shelter choice, and sengi behavior.

Sauer (1973) also believed that thermoregulation was an important feature of the shelters that were used by M. flavicaudatus. However, potential shelters were less abundant at Sauer’s study site compared to our site, as clearly illustrated by his numerous figures (Ibid.). Low shelter availability may also partially explain why M. flavicaudatus either uses abandoned rodent burrows, or excavates shallow shelters in the gravel substrates (Sauer, 1973). The only hint that M. micus might excavate shallow shelters was the use of two shallow holes (9 and 22 cm deep) by the two young sengis that we captured. We had no direct evidence that sengis fashioned these sites, so they may have been abandoned rodent burrows, although rodents were even less common than sengis at our study site.

Predation is often difficult to document, but one of our collared sengis was preyed upon, possibly by a Cape Fox, suggesting that avoiding predation may be challenging. Thus, one important feature of shelters is likely the availability of refuges that provide adequate protection from predators. Perhaps just as important is the use of multiple shelters with a very low rate of return to any single shelter, and the lack of feces accumulation in the shelters. These behaviors may be related to reducing visual or olfactory cues that predators use to find sengis. Related explanations were proposed for the similar spatial and temporal traits of sheltering sites of Elephantulus intufi (Rathbun & Rathbun, 2006), and also the nesting traits of Rhynchocyon (Rathbun, 1979). Additionally, at our study site there was little cover other than under rocks. This may explain the strictly nocturnal behavior of M. micus, which effectively would avoid predation by the numerous diurnal predators, including several raptors and bustards.

Sengi adaptive syndrome

We found that M. micus largely conformed to the life history features characteristic of other sengi species, especially the Macroscelidinae, including swift and agile cursorial locomotion, relatively exposed multiple sheltering sites, and possibly spatial organization. Additionally, M. micus has small litters of precocial young (Dumbacher et al., 2014). However, we were unable to confirm whether M. micus has a female absentee maternal care system, and whether its diet was composed mainly of small invertebrates, as it almost surely is based on the near absence of any other visible food at our study site and its morphology, which is very similar to other sengis that are known to prey on invertebrates (Rathbun, 2009).

There are several behavioral features that are worth discussing for future comparative studies, but may only be peripheral to the adaptive syndrome. We failed to find any indication that M. micus created trails on the substrate, as do other sengis (Rathbun, 1979), including M. flavicaudatus (Sauer, 1973). We suspect that M. micus used familiar travel routes, but paths did not form because the substrate was dominated by rock. If sengis traveled mainly on familiar routes, we probably only trapped them when a route coincided with a trap location, which might explain our low capture rate. We also failed to see or hear foot-drumming during stressful situations, including while in live traps, which is also characteristic of other sengis (Rathbun, 1979; Faurie, Dempster & Perrin, 1996). Neither distinctive latrines of dung pellets (Rathbun, 1979), nor scent-marking behaviors (Rathbun, 1979; Faurie & Perrin, 1995) were observed during our study, despite M. micus having a very large subcaudal scent gland (Dumbacher et al., 2014). Daily torpor, which is an energy conservation strategy in M. proboscideus (Lovegrove, Lawe & Roxburgh, 1999), may be used by M. micus, based on our observation of a torpid individual.