The African wild dog Lycaon pictus is endangered, with anthropogenic impacts, pack size dynamics and competing predators explaining its decline. Relative to solar and lunar events, analysis of diel activity in two parapatric Zimbabwean populations revealed behavioural plasticity in response to human activity. In Hwange, human presence was low; in Nyamandlovu, human presence and persecution were high. In both populations, Lycaon frequently hunted by moonlight, with 3–4 lux of light restricting nocturnal hunting to 13 days/lunar month. With diurnal hunts commencing at ‘civil twilight begin’ and ending at ‘astronomical twilight end’, light intensity was confirmed as a limiting factor.
Nyamandlovu dogs exhibited behavioural plasticity, demonstrated by scattered rather than clumped organization when at rest, and masked the zeitgeber by utilizing evenings and moonlight for more days under suboptimal light conditions than did Hwange dogs. Significantly, different allocation of morning, evening and moonlight hunts between Hwange (47%, 36%, 15%) and Nyamandlovu (28%, 31%, 41%), reduced the temporal potential for human encounter by 64%, but increased this potential for hyaena and lion encounters by 70% and 37%, thus highlighting the trade-off of this switch. Finally, we tentatively conclude that the cue masking the ‘zeitgeber’ is risk, rather than gain related, and could be seen as an evolutionary ‘emergency exit’, the understanding of which is important to conservation in landscapes that are increasingly dominated by people.
Daily rhythms are controlled by a circadian clock, entrained to the overriding cue of light intensity (a ‘zeitgeber’ in the terms of Lorenz & Kickert, 1981), and in evolutionary terms, responding to a zeitgeber facilitates efficient use of the environment (Kronfeld-Schor et al., 2001). Here, it triggers appropriately timed, physiological and behavioural responses (Heldmaier et al., 1989; Refinetti, Nelson & Menaker, 1992; Aronson et al., 1993), and facilitates interspecific coexistence (Schoener, 1974; Richards, 2002). Though temporal partitioning in communities has never been a strong focus of ecology (Kronfeld-Schor & Dayan, 2003) and biologists are aware that there is a degree of rigidity in the response to light, there are few field data to reveal the plasticity of this endogenous rhythmicity. In particular, little is known of what triggers are likely to mask the zeitgeber, although there are examples where one species causes another to adopt an opposite activity pattern [e.g. mink Neovison vison : otter Lutra lutra and fox Vulpes vulpes : rat Rattus norvegicus interactions (Fenn & Macdonald, 1995; Harrington et al., 2009)].
Furthermore, in the context of landscapes increasingly dominated by people, behavioural plasticity may reduce the threats to a species but will incur a cost [e.g. hyaenas Crocuta crocuta (Boydston et al., 2003)]. With the African wild dog or painted hunting dog (Courchamp, Rasmussen & Macdonald, 2002) Lycaon pictus (hereinafter referred to as Lycaon) representing a monotypic genus and listed as endangered by the International Union for Conservation of Nature/Species Survival Commission (Woodroffe, Ginsberg & Macdonald, 1997), the aim of this article is therefore to explore (1) the relationship between activity patterns of Lycaon and sympatric competition under ‘natural’ conditions of coexistence; (2) plasticity in response to high anthropogenic activity; (3) potential costs of sub-optimization and masking behaviours. We present data from two parapatric Lycaon populations in Zimbabwe, and their competitors. As circadian entrainment is essentially light driven, we make our measurements relative to solar and lunar light cues.
Lycaon are eusocial (Sherman et al., 1994; Rasmussen et al., 2008) kin-selected, obligate cooperative breeding canids (Courchamp et al., 2002), living in packs of up to 20 adults. Usually, only the alpha pair breeds, with the remaining adults being reproductively suppressed. It is among the most endangered large carnivores in Africa, with most of the remaining packs being in populations too small to be viable (Woodroffe et al., 1997) and the question is raised as to why Lycaon populations not only declined faster than those of sympatric carnivores, but also have failed to recover (Creel & Creel, 1998). Though causes of Lycaon mortality have been well documented, with anthropogenic mortality being recorded as a significant factor depressing populations in some systems (Woodroffe et al., 2007), the relationship between diel activity, how it could increase their conspicuousness and hence vulnerability to anthropogenic impact (Rasmussen, 1999), has not. This article investigates the hypothesis that the optimal foraging conditions for Lycaon impose high temporal niche overlap with humans, thereby putting them at greater risk than some other sympatric carnivores.
To date, on the assumption that moonlight hunting does not occur, the activity of Lycaon has been described as crepuscular to diurnal (Saleni et al., 2007). Lycaon hunt small to large ungulates (Childes, 1988; Creel & Creel, 2002; Rasmussen et al., 2008), and occasionally livestock (Rasmussen, 1999). Lycaon select for sick and weak individuals (Pole, Gordon & Gorman, 2003), which considering the extreme energetic cost of chasing may be a crucial life strategy (Rasmussen et al., 2008). Hyaenas Crocuta crocuta and lions Panthera leo kleptoparasitize Lycaon, with this impact being particularly significant in packs of less than six individuals (2009), so with lions also killing adults and pups, any changes in encounter with these predators is likely to have major implications.
It is plausible that changing pack dynamics will also affect diel activity, time windows utilized and encounters with competitors to include humans, which as a consequence of shooting, cars and snares, contribute to 93% of all Lycaon mortality in Zimbabwean ranch land (Rasmussen, 1997). This high figure is not unusual for canids, for which anthropogenic mortality is often the greatest threat. For example, human-induced mortality in wolves ranges from 80% in America (Ballard, Whitman & Gardner, 1987; Fuller, 1989) to 92% in parts of Europe (Smietana & Wajda, 1997). Similarly, humans are responsible for most coyote, Canis latrans mortalities (Windberg, Anderson & Engeman, 1985; Gese, Rongstad & Mytton, 1989). As predators are known to respond behaviourally to levels of anthropogenic disturbance (Vila, Urios & Castroviejo, 1995; Ciucci et al., 1997; Sillero-Zubiri & Macdonald, 1997; Kitchen, Gese & Schauster, 2000; Boydston et al., 2003), it is likely that Lycaon will too. In such cases, while behavioural plasticity can facilitate survival, it will come at energetic cost.
Fieldwork was conducted in two parapatric study sites separated by 150 km: Hwange National Park in the north-west of Zimbabwe, and adjacent areas, totalling 5500 km2 (April 1994 and December 2002); and the Nyamandlovu farming region totalling 1000 km2 (April 1994 until June 1997), with Lycaon densities being 0.93/100 km2 and 0.84/100 km2, respectively (Rasmussen, 1997). Data for the Nyamandlovu study came from three packs coexisting with spotted hyaenas at a density of 6/100 km2, with lions being occasional vagrants (this study). Land use was 93% cattle ranching, 7% hunting safari area with suitable prey species including kudu (Tragelaphus strepsiceros), duiker (Sylvicapra grimmea) and impala (Aepyceros melampus; Childes, 1988; Rasmussen, 1997). Cattle stocking rates (including calves) averaged 5.5/km2 in winter to 13.2/km2 in summer (Rasmussen, 1999) with trophy hunting of ungulates occurring from May to October. As the Nyamandlovu ranching region was 60 km from the nearest town, light sources for both study areas were the same.
Hwange focal packs were those that either resided entirely in areas contiguous with the park or occupied home ranges within 60 km of the park border, lions [2.7/100 km2 (Loveridge et al., 2007)], hyaenas [10.2/100 km2 (Salnicki, 2002)] leopards and suitable prey (Bougarel, 2004) being present throughout the study area. Land use comprised 35% national park, 25% photographic safari area, 35% hunting safari area and 5% cattle ranching. Data came from 22 known packs, 13 of which were radio collared for all or part their study, and by using foot tracking, a small number of unidentified units. Study time, in months, for the known packs, ranged from 3.9 to 73.3, , sd = 20.11.
For this study, 18 dogs (11 males, 7 females) were chemically immobilized with a ketamine : xylazine (Pfizer, Kent, UK) dose of 180 mg : 33 mg. Only adults over the age of 14 months were collared, with the individual being selected on the basis of the safety of the shot. Alpha females were never collared even if not suspected to be pregnant because ketamine is known to cross the placental barrier. Darting was only undertaken in the mornings in order to reduce the predation risk from interguild competitors, and restricted to open habitats to reduce the likelihood of losing an anaesthetized animal. Administration of the drug was intramuscular in the rear quarter using 1.8 mm Dan-Inject syringes (Dan-Inject ApS., Copenhagen, Denmark) with 2.0 mm, side-ported needles and a Dan-Inject 1M rifle. Uncollared dropout needles were used as a precaution against either an incomplete injection leaving an uncaptured animal with a needle left in (that it is not believed would fall out), or an inter-os misplacement that if caused by a collared or barbed needle would create excessive site trauma both on entry and removal. Rectal body temperature, breathing, pulse rates, oxygen saturation and capillary refill time were monitored throughout the anaesthesia. Dogs were regularly turned to reduce oedema, with this procedure being executed sternally to avoid stomach torsion. Once recumbent, 1 mL vitamin B complex (Alphasan, Woerden, Netherlands) was given as a compensator for stress-induced losses, along with Effortil (Boehringer Irgelheim Vetmedica GmbH, Irgelheim, Germany) to improve cardiac output, mean systemic blood pressure and aorto-coronary bypass flow were administered. All procedures on anaesthetized animals were carried out in situ with the dogs being laid on the non-thermal side of a heavy duty ‘All weather blanket’ (http://www.forestry-suppliers.com), which could be reversed if a hypothermic condition arose. Conversely, water from a knapsack sprayer was used to counter any hyperthermic condition. As the depth of anaesthesia could not be measured, precautions were taken to reduce possible stress from awareness of close proximity with humans. These measures involved the dogs being blindfolded and fitted with earmuffs specially designed to allow easy removal by the study animal in case of an unexpected recovery. As frequently, other members of the pack were waiting as close as 10 m away, no erect postures were adopted by assisting personnel and communications were kept silent by using predetermined hand signals. If extended anaesthesia was needed, top-up ketamine : xylazine doses were 100 mg : 10 mg concomitant with the fact that xylazine has a longer half-life than ketamine. When vital reflex signs indicated that the ketamine (whose half-life is shorter than xylazine) was nearly metabolized, the immobilizations were reversed with 4–6 mg of atipamezole (Pfizer) intramuscularly.
In order to reduce the need to re-anaesthetize an animal, the collars (mass 425 g, 1.70% mean body weight mass, n = 18, range 1.89–1.49%) from Sirtrack (http://www.sirtrack.com), were designed to have a battery life of 6 years at the expense of lower output. In order to spread the weight, reduce the likelihood of chafing and inhibit dorsolateral movement, belting width was increases from the standard 35 mm to 50 mm. The lower frontal section of the collar was pre-moulded to the neck of the dogs, with the batteries spread from the transmitting unit so that the weight of the batteries was evenly distributed over the whole lower section of the collar. Finally, the antenna was re-routed to exit at right angles to the collar and run along the shoulder to minimize irritation or interference with the dog's movement. When a dog was no longer being monitored, the collar was removed. All immobilized dogs were monitored for 24 h post-anaesthesia to ensure safe return and integration into their pack with no adverse effects being seen from either procedures or the collar itself.
Once packs were located, they were followed for as many days as possible. For the period of the study, a ceasefire agreement was negotiated with farmers in both study areas, but as some land owners’ attitudes were hostile to both Lycaon and the researchers, compounded by difficult terrain, poor road network, dense habitat, lack of landowner compliance, vehicle breakdown and punctures, some hunt follows were only partially completed. The collars included activity sensors such that 15 beats per minute (bpm) = mortality, 30 bpm = rest, 45 bpm = active, with individual collared dogs having separate frequencies. Once packs were located, using telemetry, they were monitored by a field observer (G. R.) and national park scout continuously doing shifts during the hours the dogs were resting. This entailed remaining with the dogs throughout their activity and rest periods at a distance of ≥50 m for up to a maximum of 28 days. To determine hunt period time and time windows utilized, data were recorded at 5 min scan intervals. Time event data collected were as follows: (1) commenced hunt, defined as leaving the resting site; (2) end hunt, defined as the commencement of the first rest period greater than 30 min; (3) hunt period (HP), denoted as the time interval in minutes between consecutive rests for a morning, afternoon, moonlight or middle of the day activity period. Any short periods of rest >10 min were subtracted from the time interval of the rest-to-rest period, hunt period time (HPT) was the duration of this interval, (4) number of HP per day (nHP) was defined as the sum of all HP recorded during the 24-h period between 00:00 and 23:59 h.
Almanac data to equate the time of dog events in relation to solar and lunar phases were compiled for all years and obtained from http://aa.usno.navy.mil/ for the relevant latitudes and longitudes (Hwange: 18-30S 27-00E; Nyamandlovu 19-30S 28-30E). These event data were then related in minutes to the pertinent solar and lunar events and denoted (−) = before (+) = after. Definitions of the solar and lunar events from http://aa.usno.navy.mil/ are as follow:
Civil twilight is defined to begin in the morning, and to end in the evening when the centre of the sun is geometrically 6 degrees below the horizon. This is the limit at which twilight illumination is sufficient, under good weather conditions, for terrestrial objects to be clearly distinguished.
Nautical twilight is defined to begin in the morning, and to end in the evening, when the centre of the sun is geometrically 12 degrees below the horizon. At the beginning or end of nautical twilight, under good atmospheric conditions and in the absence of other illumination, general outlines of ground objects may be distinguishable.
Astronomical twilight is defined to ‘begin’ in the morning, and to ‘end’ in the evening when the centre of the sun is geometrically 18 degrees below the horizon. Before the beginning of astronomical twilight in the morning and after the end of astronomical twilight in the evening, the sun does not contribute to sky illumination. At the beginning or end of astronomical twilight, under good atmospheric conditions and in the absence of other illumination, general outlines of ground objects are not distinguishable.
Moon transit time refers to the instant that its centre crosses an imaginary line in the sky, the observer's meridian, running from north to south. For observers in low to middle latitudes, transit is approximately midway between rise and set, and represents the time at which the body is highest in the sky on any given day.
Twilight to sunrise and civil to astronomical twilight time intervals were calculated from the almanac data compiled using the mean value of all the study years.
Percent of the moon illuminated was denoted as the fraction of the moon illuminated.
Relationship to the nearest full moon in days before full moon at transit time was denoted by ‘minus’ for days before and plus for days after.
As there is no morphological evidence to imply that Lycaon have any specialized night vision adaptation, the official definitions are believed appropriate for this canid.
Finally, in order to incorporate interspecific competition into modelled time niche overlaps, activity data were collated from the literature for lions and hyaenas, with human activity being known from the local area.
SPSS v.11 (SPSS Inc., Chicago, IL, USA) was used for all statistical analyses. For the data pertinent to the utilization percentage of the moon visible, non-parametric Kolmogorov–Smirnov tests were used. All tests were two tailed with significance = P < 0.05. Pearson's correlations were used to test for relationships between variables.
In Hwange, 571 HP follows were attempted, with activity being almost exclusive to three periods, morning (AM), evening (PM) and when there was sufficient moonlight (ML). Hunts close to midday (MD) were rare. Number of complete hunts followed were AM = 206, PM = 185, ML = 90, MD = 3. Partial hunts (p) followed were AMp = 38, PMp = 23, MLp = 24, MDp = 3. Total activity pattern was complete for 316 days resulting in the following HP allocation: 244 AM hunts (47%), 186 PM hunts (36%), 79 ML hunts (15%) and 5 MD hunts (1%). In the Nyamandlovu study, though one dog was collared, farmland fences made hunt follows impossible and though total activity time was not deduced, the HP allocation was obtained as follows n = 99, AM hunts (28%), 186 PM hunts (31%), 79 ML hunts (41%).
Activity in relation to time of day
AM, PM and ML activity times in minutes differed significantly from each other F2,479 = 22.69, P < 0.0001 with mean times as follows:
AM hunts commenced closer to, and just before civil twilight (n = 227, , sd = 33.1, min = −116, max = 137) than to nautical twilight end (, sd = 33.8, min = −98, max = 166) thus indicating that because civil twilight is the limit at which a terrestrial object can be clearly distinguished, light may be deemed a limiting factor (see definitions). Hunts occurring considerably earlier than civil twilight were facilitated by the light of a setting moon. AM hunts ended 2 h after sunrise (n = 219, , sd = 59.5, min = 10, max = 283). Kills (n = 350) occurred on average 54 min after sunrise (, sd = 61.1, min = −95, max = 280). Overall AM activity period time was (sd = 53.3, min = 40, max = 320).
PM hunts commenced 1 h before sunset (n = 199, , sd = 30.8, min = −136, max = 27), and ended (n = 195) 5 min before astronomical twilight end (, sd = 47.2, min = −118, max = 158) when by definition there is no utilizable light from the sun, again suggesting light as a limiting factor. Extended hunts, resulting in positive outliers, were concurrent with a rising moon. Kills occurred on average 7 min after sunset (n = 258, , sd = 58.8, min = −174, max = 236).
ML hunts maximized on moonlight by starting 1 h before the moon was at its apex at moon meridian time (n = 119, MMT) (, sd = 62.3, min = −204, max = 190), and ended (n = 95) 100 min after MMT (, sd = 73.4, min = −74, max = 350). No hunts were recorded on nights when available moonlight was obscured by cloud cover. In both study areas during each lunar month, dogs hunted by moonlight a maximum of 13 days (7 days before the full moon to 6 days after). Kills (n = 63) occurred 36 min after MMT (, sd = 81.7, min = 139, max = 269). ML activity period time was (, sd = 55.82, min = 55, max = 320).
In relation to the percentage of the moon visible, Lycaon hunted only with ≥49% of the moon visible on a rising moon and ≥58% on a setting moon. Nyamandlovu dogs however utilized lower light conditions more frequently (Fig. 1). Testing for both percentage of hunts undertaken in relationship to the available moon visible, and days before/after the full moon, showed these population differences to be significant (Kolmogorov–Smirnov z = 1.839 P = 0.002 and z = 1.567 P = 0.015). Use of a light meter (Extech Foot Candle/Lux Meter, Extech Instruments Corp., Waltham, MA, USA) during the moonlit hunts indicated that the limiting light condition was between 3 and 4 lux. Attempts to use the meter for the solar twilights failed to detect the breakpoint as the light conditions changed so rapidly that the meter (designed for lower light levels) would, in a time span too fast for the observer eye, go from reading nothing to a light level off the scale.
There were only three MD hunts (, sd = 147.4, min = 60, max = 340) so no inferences could be drawn and they were excluded from analyses.
Spatial organization and pup provisioning
The two populations showed different behavioural patterns by exhibiting different spatial organization when at rest and different pup provisioning patterns. In accordance with other studies (Scott, 1991; McNutt et al., 1997; Creel & Creel, 2002), the Hwange study packs rested as a group or at least in close proximity to one another (<50 m); however, the Nyamandlovu dogs were never detected at rest as a group and on all encounters following foot tracking (n = 43), were scattered, often resting >200 m apart as singletons or as pairs. This was evidenced from the multidirectional alarm calls of the dogs upon being detected, as well as trackers pointing out where individual dogs had been resting. With respect to pup provisioning in the Hwange study, in only five cases out of 155 AM hunts did the dogs not return to the den after killing and feeding successfully. By contrast in Nyamandlovu from 38 AM hunts, on no occasion, did they return to the den until either sunset (n = 2) or after astronomical twilight end (n = 36).
In spite of no dogs being shot during the period of study, mean adult, yearling (AY) and adult, yearling, pup (AYP) pack sizes were significantly lower in the Nyamandlovu region (F(AY)1,2143 = 8.67, P = 0.003) (F(AYP)1,2143 = 43.77, P ≤ 0.001). Means for the two regions follow: Hwange [n = 2045 (, sd = 2.7) (, sd = 5.0)]; Nyamandlovu [n = 99 (, sd = 2.7) (, sd = 4.3)].
Activity in relation to pack size
Previously published data from the same population over the same time period in Hwange during the denning season and nomadic phase (Rasmussen et al., 2008) showed no significant relationship between pack size and HPT, but significant differences between pack size and nHP. Based on the number of weeks for which packs of given sizes were denned or nomadic (Rasmussen et al., 2008), these data were used to calculate: An annual mean HPT of 138 min; a relationship between nHP and pack size. Multiplying the number of hunt periods per day by the hunt period time gave the relationship between daily HPT and pack size (Fig. 2)
Using nHP data, relative percentages of AM, PM and ML hunts per day relative to pack size were determined. This gave the results (Fig. 3), that as pack size increased, moonlight hunts increased (r2 = 0.59, P = 0.05), AM hunts decreased (r2 = 0.60, P < 0.001) and PM hunts showed no significant change (r2 = 0.121, P = 0.643). Part of the decrease in AM hunts is explained by the fact that post-ML hunts, the dogs were sated, as reinforced by the data revealing that on only 27% of occasions did AM hunts follow ML hunts (n = 116). The rest of the resultant decrease in AM hunts may therefore indicate either a preference, or need, for larger packs to undertake ML hunts, which also coincidentally reduces their likelihood of encounter with humans.
Combining regressions from Figs 2 and 3, and the differences between activity pattern in Hwange and Nyamandlovu thus enabled a time window utilization to be calculated for AM, PM and ML hunts in both study areas (Fig. 4). It also highlights both the contribution of the hitherto unstudied ML time niche and the altered time dynamics of the behavioural shift.
Lions are primarily nocturnal (Kruuk & Turner, 1967; Schaller, 1972; Van Orsdol, 1984; Prins & Iason, 1989; Stander, 1991), with main activity (activity > 20 min) commencing 1–2 h after sunset (astronomical twilight), peaking between midnight and 04:00 h and ceasing at sunrise (Van Orsdol, 1984). The same activity pattern was found in a study of buffalo and their vigilance response to lion activity (Prins & Iason, 1989). Studies also show that lions adjust their nocturnal hunting period to coincide with either moonless hours or periods of cloud cover (Schaller, 1972; Van Orsdol, 1984). In this study, out of 520 kills (AM = 281, PM = 198, ML = 39, MD = 2), lions were present on only eight occasions (1.5%; AM = 4, PM = 3, ML = 1).
Hyaenas are predominately nocturnal, commencing activity after sunset and operating through the night from moonless to full moon nights (Cooper, 1990), though in cooler weather, they do hunt in daylight (Cooper, 1990). Therefore, it is more likely that Lycaon will encounter hyaenas than lions. In this study, hyaenas were present at 41 kills (7.9%) of which the greater percent were AM and PM hunts (AM = 19, PM = 18, ML = 4), and at one kill both hyaenas and lions were present.
Humans utilize very closely the time niche used by Lycaon with ranchers and rural communities commencing work as soon as there is available light, which by definition would begin and end at civil twilight, with a slowing down of activities close to midday due to heat. This being the case, with the exception of moonlight hunting, in terms of time overlap and the 53-min interval between the end of civil and astronomical twilight Lycaon mirrors the time niche of humans.
Modelling of time overlaps
Using the aforementioned data, time niche overlaps were determined to be as follows:
AM = Time sympatry for whole HP
PM = Time sympatry for whole HP minus 53 min
ML = Total time allopatric
AM = Time sympatry from civil twilight to sunrise
PM = Time sympatry from civil twilight to astronomical twilight end
ML = Total time sympatric
AM = Time sympatry from civil twilight to sunrise
PM = Time sympatry from civil twilight to astronomical twilight end
ML = (Hwange = Time sympatry for 18% of ML activity; Nyamandlovu = Time sympatry for 49% total ML activity)
Note well that these differences arise because Nyamandlovu dogs utilized days further from the full moon (Fig. 2) and thus overlap more with lions.
These overlaps, shown in time sympatry (Fig. 5), demonstrate how by changing allocation of AM, PM and ML hunts, Nyamandlovu dogs shifted their activities to reduce the probability of encounter with humans by 64%, but increase those of encounters with hyaenas and lions by 70% and 37%, respectively. By introducing niche overlap factors, defined as the time active when the interacting competitor was also active/total activity time (Fig. 6), these changing dynamics further highlight the consequence of switching to more nocturnal activity, whereby encounters with humans decreases at the cost of increased probability of hyaena encounters.
This study of diel activity of Lycaon in relation to solar and lunar events has not only revealed light as a limiting ecological factor, but also demonstrates behavioural plasticity, and temporal activity that changes with pack size and anthropogenic activity. It also highlights the importance of interpreting events in the context of solar/lunar patterns rather than using the arbitrary 24-hour clock. In theory, with the lunar month not being synchronous with the solar month, only studies on the equator where organisms respond exclusively to solar cues and not lunar ones, are unlikely to fall foul of noise generated using clock time. Even in latitudes as close to the equator as 5 degrees, the time differential over the year is 45 min. Furthermore, with some events being before twilight and some after, the bias could be double this.
Previous Lycaon studies have not noted the utilization of the moonlight niche (Mills, 1993; McNutt et al., 1997; Creel & Creel, 2002); however, this phenomenon is not exclusive to the Hwange population. During the course of this study, this behaviour was also opportunistically observed in the Lowveldt region of south-east Zimbabwe, Mana pools and Kanyemba in north-west Zimbabwe. Our results should be considered in the light of four aspects of Lycaon biology: (1) larger packs are more able to defend kills against hyaenas (Carbone, du Toit & Gordon, 1997); (2) larger packs require more food, with larger prey rather than an increased number of kills providing a more economical option; (3) kudu, which form the most significant part of Lycaon diet (Rasmussen et al., 2008), are nocturnal to crepuscular and thus are more available in this time window, and indeed in Rasmussen et al. 2008, it is also demonstrated that larger packs select larger prey commensurate with pack size; consequently with packs also commanding territories that are parapatric in time (Rasmussen, 1997), intraspecific competition is deemed an unlikely cause for this finding; (4) due to low light conditions, flight distances are smaller, and therefore, due to the extreme cost of chasing (Rasmussen et al., 2008), this time window is energetically more profitable. These points apply equally to both the Hwange population and the Nyamandlovu one, between which we could detect no difference in the hunting conditions. So as pack sizes in the Nyamandlovu study were smaller, one would have expected less moonlight activity, not more. The only observable difference between the two areas lay in the extent of anthropogenic disturbance, so we conclude that this explains the contrasting behaviour of the two Lycaon populations.
We now turn to the potential costs of sub-optimization and masking behaviours. Firstly, increased foraging time associated with moonlit hunts, which due to the hypercursorial nature of this species, will represent appreciable energetic cost (Rasmussen et al., 2008). Furthermore, with light being a limiting factor, the costs of this suboptimal strategy is likely to reduce hunting success and cause even longer hunting hours than the model predicts. Secondly, a twofold increase in the probability of hyaena encounters will significantly increase kleptoparasitic cost (Carbone et al., 1997). While it may be hard to quantify the ‘cost’ of the behavioural adaptations, either due to social deficit incurred by spreading out when resting up (rather than sleeping in physical contact or close proximity as they usually do) perhaps resulting in reduced pack cohesion, or the lowered security for the pups, it is likely that such costs exist. Equally, rich literature on human shift workers demonstrates that diametric utilization of the diel cycle entails severe costs in health and performance (Van Reeth, 1998), so similar costs could affect the dogs. The accumulated impacts of these factors on the Nyamandlovu population may be reflected in the pack sizes AY and AYP being significantly less by 0.8 and 3.4, respectively. Nonetheless, in the short term at least, it appears that the tactic has some success insofar as the population persisted and produced dispersers. One of which that had originated from the source Hwange population, survived for 4 years, successfully dispersing a total of 570 km Hwange, Nyamandlovu, West Nicholson, Beitbridge, ending up in Messina South Africa, where it was shot. This is to date the longest distance a dog has been recorded to disperse, and perhaps facilitated by masking the zeitgeber.
In terms of likelihood of encounter, hyaenas represent a greater threat to Lycaon than do lions, and it is unlikely that in Nyamandlovu, Lycaon could switch diel activity so radically if other predators were at the higher density typical of Hwange. In that event, Lycaon would be trapped between an anthropogenic rock and a kleptoparasitic hard place. The latter is particularly pertinent, for had the other predators been present to the same degree, they too probably would have become even become more nocturnal as documented for hyaenas elsewhere (Boydston et al., 2003). Under these circumstances, it seems likely that niche overlaps would be even higher than predicted by our model.
Turning to the significance of triggers that mask the zeitgeber, while behavioural plasticity is not unique, the few published field studies on temporal shifts suggest that the masking behaviour represents a response to extreme risk rather than gain. This study, and those of (Fenn & Macdonald, 1995) and (Boydston et al., 2003) now add weight to the argument that ‘zeitgeber masking’ functions to ameliorate direct mortality risk rather than accrue foraging gain. Furthermore, we argue that the zeitgeber is only masked in extremis and this is also supported by a study on spiny mice (Acomys cahirinus and A. russatus), where (Kronfeld-Schor & Dayan, 2003) it was proposed that while the rigidity of endogenous rhythmicity ensures that species are not misled by minor environmental disturbances, this may also be an evolutionary constraint. Therefore, any signal that is capable of masking it might, by inference, be of major survival significance. Consequently, we suggest that the mechanism that masks the zeitgeber can be thought of as an evolutionary ‘emergency exit’ which, if successful, might evolve into a mechanism of entrainment over time.
Overall, these results highlight the value of understanding not only temporal activity and interspecific interactions, but also the role of the zeitgeber. Furthermore, insofar as there are good reasons that circadian entrainment represents an adaptation of organisms to their environment, it should be recognized that when it is masked, the accrued energetic losses may become the ‘last straw on the camels back’. Consequently, when the zeitgebers ‘emergency exit’ is taken, it should sound a warning of conservation risk.
We are grateful to the Director General of the Parks and Wildlife Management Authority for permission to conduct research and logistical support in Zimbabwe. To national parks scout Felix Banda, Peter Blinston and Jealous Mpofu for fieldwork; Greg Gibbard and Esther van der Meer for assisting with data cleaning and almanac work, Paul Johnson; Phil Riordon, Jorgelina Marino and to the referees for their insightful comments.