Ecological drivers of female lion (Panthera leo) reproduction in the Kruger National Park

Abstract The role of social cues in the reproduction of social mammals, particularly carnivores, has been thoroughly studied and documented in literature. However, environmental cues such as resources of water, food, and shelter have been identified to a lesser extent. Pregnant lions (Panthera leo) are notoriously secretive during the final stages of pregnancy and postpartum. Behavioral indicators depicted by movement patterns obtained by remote detection of collared female lions in the Kruger National Park were necessary for the monitoring of birth timing. Over the study period, eight plus a potential three parturition incidences of collared females were recorded. Of the variables measured (step length, range size, duration, prey biomass, and rainfall), range size during the month of parturition was the most indicative movement pattern of a successful birth. By backdating the potential birth month of the litters, date of conception was calculated and our results revealed a correlation between the birthing peaks of preferred prey during the month of conception. Birth timing in conjunction with remote sensing and ecological factors were thus identified behaviors associated with denning.

Both these studies, however, only focused on birthing incidences and ecological conditions. Although ecological and environmental cues such as rainfall and prey biomass are identified as factors influencing reproductive schedules of lions, research into these cues is limited (Bertram, 1975;Ogutu & Dublin, 2002).
Rainfall influences resource availability, such as food, water, and shelter, which are critical for the successful rearing of cubs Van Orsdol, Hanby, & Bygott, 1985). The overriding driver that influences interactions and survivorship of African predators is resource availability, mainly rainfall (Maruping-Mzileni, Funston, & Ferreira, 2017). Seasonality of rainfall determines the amount of available surface water; this in turn influences the distribution and abundance of lion prey species. This is because energy is required for gestation, lactation, and eventual protection of offspring. Changes in energy requirements suggest an alteration of physiological demands that manifest in behavioral changes. Literature suggests that a behavioral change independent of environmental changes indicated by altered or reduced movement patterns could indicate potential physiological stress (Hinton & Chamberlain, 2010) such as pregnancy.
The spatial ecology of lions is based on their need to fulfill physiological, ecological, and social requirements (Ogutu & Dublin, 2004).
These requirements are met based on habitat suitability (Ogutu & Dublin, 2002), resource availability , and social dynamics (Loveridge et al., 2009). Home ranges are large enough to ensure access to key resources such as food, water, and shelter and in some cases are defended either fully or in part as territories. Lions will adjust their location in space until the requirements have been met and in so doing define a home range (Beyer et al., 2012).
Prey movement and abundance are significant determinants of home range size and use by lions (Loveridge et al., 2009;Morales et al., 2012). Lion home ranges are smaller in areas of high prey availability compared with those of lions living in areas of low prey availability . De Boer et al. (2010) found that animal distribution is nonrandom and tend to congregate around water where lion hunts occur most frequently. These findings suggest that resource availability in the form of prey and water are key drivers of lion home range dynamics. Consequently, home range dynamics may thus influence reproductive activity of lions.
The ability to survive and access all necessary resources can be compromised by physiological stress. Disease, injury, chronic food restriction, or pregnancy can impose physical stressors on the body such that limit an individual's ability to obtain resources. For example, these stressors have been found to exacerbate and influence behavioral patterns and movement patterns (Wu, Bazer, Cudd, Meininger, & Spencer, 2004). Chronic food restriction experienced or injuries are clear examples of physiological stress. Physical stress resulting from chronic food restriction altered reproductive development and estrous cycles of spotted hyenas and lions in the Maasai Mara National Park (Holekamp et al., 1999) lions in Nairobi National Park (Rudnai, 1973) and sheep in New Zealand (Rae et al., 2002). Similarly, when seasonal rainfall and ecological conditions favored the success and thus an increase in wildebeest numbers of the Serengeti, lion cub survival increased because of an increase in resource availability . Disease and pregnancy, however, are less understood as physical stressors that limit the ability to access resources. Bovine tuberculosis (bTB) is an emergent infectious chronic disease that was opportunistically identified in buffalo (Syncerus caffer) of the Kruger National Park (KNP) in the 1960s and in lions in the 1990s (Renwick, White, & Bengis, 2007). Symptoms include swollen joints, weight loss, and lameness (Renwick et al., 2007).
Such physical stress can reduce a lion's ability to obtain resources, influence social interactions, and reduce energetic expenditure for reproduction.
Here, we hypothesized that for adult female lions, energetic constraints such as the ability to hunt or obtain resources are imposed when the body is stressed during pregnancy as a result of factors such as disease. The inability to access resources can potentially result in reduced conception rates, reduced ability to maintain pregnancy, and changes in movement patterns to establish a denning site where cubs are born and reared. Therefore, we asked ( To test these hypotheses, we first determined the potential month when a birth occurred through several indicator variables collected through GPS radio-telemetry and GPS satellite. The indicator variables measured were the number of fixes at a localized site, the elapsed time of consecutive fixes at a localized site, and the distance between the fixes. We then verified whether a successful birth occurred according to the following: (a) the duration of stay at a particular denning site using GPS fix clusters; (b) core and total home range size during the potential birth month, and then (c) variation in daily net displacement (km) during the wet and dry seasons; and (d) variation in daily net displacement (km) during the potential birth month.
These variables were chosen because they are the best indicators of changes in prenatal and postnatal behavioral patterns. To test for the season of reproduction, we measured movement patterns by calculating net displacement as a function of energy expenditure and range size as a function of resource availability. To test for the association of ecological factors on birth timing, we calculated prey biomass and rainfall as a function of resource availability. To test the association of fecundity with ecological factors, we compared the KNP with other game reserves as benchmarks.
Mean annual rainfall varies from the north to the south between 450 to 750 mm per annum, respectively. Precipitation generally decreases from south to north and increases from east to west.
The wet season is from October to March and receives approximately 80% of the precipitation, with the driest months being April to August (Du Toit, Biggs, & Rogers, 2003). Relative temperature ranges from 0 to 40°C, with high peaks from December to February.
June and July are the coldest months when frost occurs, particularly in the low-lying areas. We divided KNP into three zones according to prey abundance, and rainfall, with all factors being highest in the south and lowest in the north. Given the vastness of the KNP landscape, to best classify the environmental and ecological factors, we used Acocks (1988) vegetation classification. The ecological factors used to classify the study area into regions were prey biomass and average rainfall. The regional categories were the southern region which had high prey biomass, and high rainfall and the northern region which had low prey biomass, and low rainfall (Figure 1 (Ferreira et al., 2013). Once immobilized, assessment of captured lions by veterinarians from the Veterinary Wildlife Services (SANParks-VWS) included general health and body condition such as muscle tone, coat condition, dehydration and reproductive state, age assessment, and tissue sampling for assessing disease status.
During subsequent monitoring, demographic information including total pride size and structure was identified and recorded.
In each pride captured, one adult female was fitted with a collar.
In the southern region, where prey biomass and rainfall is high, captures (N = 10) took place between February and March 2010 while in the northern region, where prey biomass and rainfall is low, captures (N = 13) were conducted from August to September 2010. We also captured lions throughout the study period when collars needed to be changed. Of the collared females in the northern region, only 10 were used for this analysis because of inconsistent data when collars failed or females were not seen for more than three months.
An adult female from each of the pride was fitted with a standard mammalian collar. The collar includes a battery pack and a built-in tracking unit and antennae sewn between the belt layers. The width and weight is species-specific (www.awt.co.za) [Accessed: 2015, 22 July]. All collars had Global Positioning System (GPS) and Very High awete lemet ry.com). Variations in vegetation structure, terrain, climatic factors, and power lines may have affected collar upload frequency, accuracy, and precision (Di Orio, Calla, & Schaefer, 2003). Of the total number of fixes obtained, the fix percentage was approximately 90% (SD ± 20). Satellite collars were the most reliable with low failure-to-download incidences.
Duration at a site was calculated by measuring the amount of time spent at a denning site or at a locality. The duration of stay was calculated by identifying clusters defined as the number of consecutive fixes within a 200 m radius. The information included longitude, latitude, altitude, temperature, date, and time. We used the data from the GPS fixes to measure daily movement patterns of the collared lionesses.

| Data analysis
For each collared female in the north and south, we used generalized linear models (GLMs) to identify which of the explanatory variables (range size, daily net displacement, and rainfall) were most indicative of the likelihood of a birth. These models do not force data into predefined artificial scales, thus allowing for nonlinearity and nonconstant variance structures in the data (Hastie & Tibshirani, 1990). We compared Akaike information criterion (AIC) among candidate models with a single and combination explanatory variables (Johnson & Omland, 2004) for the southern and northern region. From the AIC, corrected AICc was calculated. This adjusts the AIC for small sample sizes. The AIC∆ was then calculated, which is the difference between the AIC value of all candidate models and the candidate model with the lowest AIC score. The Akaike weight (wj) was then derived from the candidate model averages to determine the relative likelihood of the model (Johnson & Omland, 2004). This method was used because it independently evaluates the model discriminatory power (Wisz et al., 2008). A candidate model for the duration of stay was not run due to the subjective nature of the data.
Range size (core 50% and total 90%) and daily net displacement were measured in ArcGIS 9 ® ArcMap v9.3 (Environmental Systems Research Institute, Midrand, Gauteng, South Africa) using LoCoh (Getz et al., 2007) (Tambling, Cameron, Toit, & Getz, 2010). Fixes were taken at four-hour intervals with time associated with when the fix was uploaded and could thus be calculated into days spent at a location.
The southern and northern regions were analyzed separately and then compared to distinguish the differences in birth timing (season F I G U R E 1 Ecological zones of Kruger National Park according to available prey biomass and the home ranges of lions collared between 2010 and 2012 (Ferreira & Funston, 2010)   To answer the first hypothesis, we used the movement pattern outcomes from the potential birth month calculations to identify reproductive activity. We calculated potential month when a birth occurred by the following metrics collected through GPS radio-telemetry: (a) the duration of stay at a particular denning site by identifying GPS fix clusters; (b) core and total home range size during the potential birth month; (c) variation in daily net displacement (km) during the wet and dry season; and (d) and variation in daily net displacement (km) during potential birth month.
To address the second hypothesis, we calculated the association between the birth timing and ecological factors by measuring known birth and conception with (a) prey biomass, (b) lambing and calving of preferred prey, and (c) rainfall. We used verified cub births to identify whether there was an association between the occurrence of births and conception incidences and ecological factors.
To address the third hypothesis, we measured the fecundity of female lions in the KNP; (a) age at first birth, (b) litter size, and (c) birthing intervals. We compared the fecundity of lionesses in the KNP with other lion populations.

| Seasonality of reproduction
To address our first hypothesis, do lions show seasonality of reproduction, we analyzed movement patterns through net displacement, range size, and duration of stay at a site.

| Birth timing
To test the second hypothesis, of whether lion birth timing associate with ecological factors, we compared the ecological conditions in the form of monthly rainfall and prey biomass. We did not take into account landscape features or habitat quality because we used prey biomass as an indication of landscape and habitat quality. Prey biomass was associated with game counts and relative to lambing and calving of preferred prey. We used monthly rainfall because it indicated the variation between seasons in more detail than annual rainfall. Previous studies indicated that buffalo (Pienaar, 1969), wildebeest (Smuts, 1976), kudu (Smuts, Hanks, & Whyte, 1978), and zebra (Mason, 1990) have peak births during the wet season in southern Africa, usually in late October and January to March (Ogutu & Dublin, 2002). We included rainfall because previous research has indicated a strong relationship between lion behavior and prey activity and movement patterns (Tambling et al., 2010).
To investigate the relationship between lion birth timing and prey biomass, 15 prey species included in game counts were taken into account (Fairall, 1968 Toit, 1995;Klingel, 1975;Pienaar, 1969;Skinner, Zyl, & Heerden, 1973). Game counts in the KNP have been recorded through annual censuses since 1970 between July and October; however, the method was changed in 1998 from total counts to sample surveys with an average of 22% coverage (Kruger et al., 2008). The change in methodology was found to be sufficiently precise in estimating herbivore populations (Kruger et al., 2008). The prey biomass was scaled to the total home range size of the pregnant female to adjust for prey availability. Essentially, the prey biomass data from the region where the female occurred were used. We then used GLM's to identify which of the explanatory variables, prey biomass and rainfall, were most contributing to the birth timing.

| Fecundity
To verify whether fecundity alters because of ecological factors, we analyzed the average age at first birth, birthing intervals, and litter size of collared females in the KNP.
Fecundity data were analyzed using one-way ANOVA (p < .05) to identify whether there was variance in the fecundity of age at first birth, litter size, and birth intervals in the southern and northern March] using the Rcmdr package (Fox, 2005).

| RE SULTS
Of the 20 collared females, 12 were recorded as having had cubs, 6 in the southern region and 6 in the northern region. From these collared females only, a total of 37 cubs were observed and recorded.
When we included all pride females, both with and without collars, a total of 20 lionesses had cubs, nine lionesses in the southern region and 11 lionesses in the northern region. In total, 97 cubs were observed during the study period, 62 in the south and 35 in the north regions.
Of the 12 recorded births by collared females, 10 birth events were identified using net displacement and GPS clusters of the collared females to locate denning sites. Of these denning sites, eight were ground proofed by direct observations with litters belonging to the radio-collared lionesses. For two lionesses, cubs were seen during captures in excess of five months after birth, which made it difficult to prove whether the collared female was the mother. For the remaining two lionesses, no ground proofing was possible. Of the radio-collared lionesses with confirmed denning sites with litters, three in the southern and two in the northern region had synchronized births with other lionesses within the pride.
When compared, the pregnant females in the southern region had a shorter daily net displacement (mean = 4.3 km, SE = 1.3) than the pregnant females in the northern region (mean = 5.4 km, SE = 1.2); (F 2, 32 = 0.42, p = .66). When the southern and northern pregnant females were compared with nonpregnant females during the birth month, the pregnant females had a shorter step length (distance between consecutive GPS fixes), than the nonpregnant females (F 2, 26 = 3.87, p = .03). There was a 51%, reduction in net displacement in the northern (from 9.48 to 5.81 km) and southern (from 7.40 to 3.08 km) regions during the birth month (SE = 4.3, 95% CI; Figure 3).
Median total range size of the collared pregnant females differed between birth month and a seasonally similar month when the female was not pregnant for both core home range (R 2 = 0.608, p = .004) and total home range (R 2 = 0.655, p = .002; Table 1). During the birth month, collared pregnant females operated over a smaller core home range (mean = 0.2 km 2 , SE = 0.1) compared with the core range (mean 3.4 km 2 , SE = 1.7) of nonpregnant females (F 1, 12 = 24.55, p < .05).
During the birth month, total home range of collared pregnant females (mean = 6.3 km 2 , SE = 3.4) was smaller compared with the total monthly home range (mean = 19.5 km 2 , SE = 12.2) of the nonpregnant females (F 1, 12 = 7.55, p < .05). The best candidate GLM models to predict a successful birth incident and denning site comprised core range and the combination of core and total range (Table 1).

| Birth timing
One-way ANOVA suggested no relationship between rainfall and con-  (Table 2).

| Fecundity
From the data collected in this study, there was no significant difference in the average age at first birth for lionesses in the southern (mean = 3.7 years, SE = 0.3) and northern (mean = 3.9 years, SE = 0.1) regions (F 2, 7 = 0.3, p = .8). Nor was there a significant difference for litter size (south: mean = 2.6 cubs, SE = 0.3, and north: mean cubs = 2.7, SE = 0.3,); (F 2, 7 = 0.24, p = .62). There was insufficient data from the pregnant females to compare birthing intervals (south N = 3; and north N = 0), which averaged 2.9 years (SE = 1.1) in the south.
Our findings for fecundity were compared with KNP historical data and data from other large reserves. No significant difference between the current data collected during this study, historical KNP data from Serengeti National Park, Nairobi National Park, and Makali Nature Reserve were found using the Kruskal-Wallis test (X 2 6 = 6, p = .423). Nor was there a significant difference found in the fecundity of KNP between this study and 1978 (Smuts et al., 1978; X 2 3 = 3, p = .416; Table 3).

| D ISCUSS I ON
Energetic and metabolic needs vary at different stages of reproduction, from conception, through gestation to birth (Woodroffe, Chapman, & Lemusana, 2009;Wu et al., 2004). Our findings indicated that core range size was a key indicator of a denning site and birth event. Although there was no correlation between rainfall and conception, there was a correlation between seasonality of lion conception and the birth peak of key preferred prey namely buffalo, kudu, zebra, and wildebeest. Prey biomass peaked during the early  Note: Assessed through candidate general linear models (GLMs) with the number of parameters (k) using the Akaike information criterion (AIC). Included are the corrected (AICc) for small sample size, the difference between the AIC candidate models and the smallest AIC recorded for the GLM (AICc∆I), and the weight (wi) of the model. Although both pregnant and nonpregnant lionesses in the southern region had shorter daily net displacements than lionesses in the northern region, this was likely due to higher frequency of prey encounters, water distribution, and the general quality of the home ranges (De Boer et al., 2010;Loveridge et al., 2009;Ogutu & Dublin, 2004) in the southern region (Funston et al., 2003). Our results confirmed that pregnant females alter their range use at various phases of pregnancy. Reproduction is a physiological state that places the female body under stress because it requires higher amounts of energy

Number of liƩers
Wet Dry input and output (Wu et al., 2004). Energy from food prepares the body for conception and sustains a pregnancy through the gestation period (Woodroffe et al., 2009). Accordingly, pregnant females need energy to sustain the pregnancy and, therefore, cannot expend it by walking great distances to find resources such as water, prey, and shelter (Du Toit, 1995;Wu et al., 2004). As expected, in our study the net displacement of all pregnant lionesses was substantially less (56%) than nonpregnant lionesses leading to significantly smaller core and total range of pregnant females compared with nonpregnant females in both the southern and northern regions during the birth month. This is consistent with the reduction in red wolf female and breeding pairs range size during birth and pup-rearing months. Further investigation into the landscape and resource features of these core ranges would provide insight into the key causal determinants of core range selection and denning sites by female lions (Abade et al., 2014).
Our results revealed a correlation between the birthing peaks of preferred prey during the month of conception. Interestingly, while there was no correlation between conception and rainfall, there was evidence that peak conception occurred during the wet season when buffalo, wildebeest, kudu, and zebra have birth peaks-usually in late October and January to March (Mason, 1990;Pienaar, 1969;Smuts, 1976;Smuts et al., 1978). Coinciding conception with calving and lambing of preferred prey strategically allows for pregnant females to easily access resources needed to sustain a pregnancy (Holekamp et al., 1999;Ogutu & Dublin, 2002). The wet season conception peaks identified in our study most likely indicate lionesses accessing food and water resources to sustain body conditions that facilitated conception. Lion cubs start eating meat at three months of age and are completely weaned by six months. Prior to birth and for the first three months after birth, a lioness is often accompanied by another female that assists with cub rearing and hunting. This companionship alleviates energy output that would otherwise be associated with cub rearing . It is anticipated that once the cubs start eating meat exclusively at six months and onwards the energy output of a female increases. However, this would be balanced by cooperative group hunting.
In Kenya, spotted hyena births coincide with response to season variation and resource availability (Holekamp et al., 1999), although spotted hyenas can breed year-round. Likewise, lions in the Maasai

Mara in Kenya displayed birth peaks during the wet season between
March and June, coinciding with the increase in migratory prey (Ogutu & Dublin, 2002 Our third hypothesis, fecundity alters as a result of ecological factors, was not supported by our results. Firstly, litter size, age at first reproduction, and birthing interval were not significantly different between the resource-rich southern region and poorer resourced northern regions. Secondly, when compared with historical data (Smuts, 1976;Smuts et al., 1978), and data from lions in the Nairobi National Park (Rudnai, 1973) and Makalali Nature Reserve (Druce et al., 2004) all representing differing resource conditions, the fecundity of lions in our study fell within the same range. Accordingly, our study suggests that ecological conditions are important, but they are not drivers of the fecundity of lions in the KNP. There may be other factors unaccounted for that drive fecundity. This study, however, identified ecological mechanisms that may influence conception.
Using movement patterns as behavioral indicators, we were able to detect the denning sites remotely and identify reproductive activities of lions. In existing protected areas, conservation practitioners can be more selective in their management of denning sites to avoid burning and minimize disturbance during denning. In the event of poaching activity on a reserve, conservation mangers will be better suited to direct law enforcement to the denning sites. Although this study did not focus or take into account poaching of wild lions, the lessons learned could be applicable to addressing such incidences.
Kruger National Park has not been identified as a source of lion poaching. Lions that have been killed outside the park are ad hoc animals that have left the park and entered neighboring communal property. Throughout Africa, retaliation killing for livestock lose is one of the primary human-wildlife interactions leading to death and population declines of lions (Tumenta et al., 2010). Although the trade in lion bones has put a spotlight on lion parts for trade, this is most evident in captive-bred lions more so than wild lions (Department of Environmental Affairs South Africa, 2018). The ability for conservation managers to detect births and denning activities of lionesses allows for better preparedness in the event of a poaching incident.
This study provided insight into areas of study not previously questioned such as the importance of conception rather than just the birthing. Population dynamics could consequently be a function of poor conditions during conception and gestation. Similar to the importance of understanding the role of landscape features and habitat quality in lion denning site selection, so is understanding the triggers and drivers of conception. This study suggests that space use in the form of landscape features and habitat quality have an impact on reproduction.
Although this study was conducted in an area with known bTB disease prevalence, disease did not appear to be a significant driver.
It is important to note that this study did not explore social cues of synchronized births of pride members or the effect of male takeovers. To strengthen our analysis, social cues in the form of male takeovers, estrus synchrony, and crèche formations could be incorporated. Nevertheless, ecological and environmental cues that influenced the conception period were identified, particularly regarding the influence of births of preferred prey. Identifying the resources and landscape features associated with denning sites is an important next step to explore the effect of seasonality and elucidate denning site characteristics.

ACK N OWLED G M ENTS
The opportunity to conduct this research in the Kruger National Park

CO N FLI C T O F I NTE R E S T
None declared.

AUTH O R CO NTR I B UTI O N S
All authors contributed equally to this work including the conception, data analysis, and manuscript synthesis.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data will be stored on the SANParks Data Repository and made accessible from http://datak np.sanpa rks.org/sanpa rks/metac at/ judit hk.112037.1/sanparks, on request from SANParks or the corresponding author.