Ethics Statement

This research was approved by the Institutional Animal Care and Use Committees of the University of Pennsylvania (protocol no. 801295) and Colorado State University (protocol no. 11–2816A). Personnel of the Department of Wildlife Conservation provided information on carcasses, and permission to conduct this research was granted by its Research Committee.

Study Site

Uda Walawe National Park (UWNP) is located between latitudes 6°25′–6°34′N and longitudes 80°46′–81°00′E in south-central Sri Lanka, very close to the equator. It features large and small man-made reservoirs that provide water for human and wildlife consumption. Other common herbivores include domestic and feral water buffalo (Bubalus bubalis) and spotted deer (Axis axis). The largest carnivores include the leopard (Panthera pardus cotiya), jackal (Canis aureus naria), and mugger crocodile (Crocodylus palustris), but it is unknown whether they pose any threat to elephants. UWNP has a seasonal, mesic environment, with nearly constant day length throughout the year. The region receives two annual monsoons, generally occurring from October–December and March-April. Rainfall averaged 1514±388 mm annually between 2007–2012 (measured by a standard US weather bureau rain gauge, and calibrated according to a national weather station in Rathnapura). Dynamic and persistent changes in vegetation due to fire and invasive Lantana camara (an introduced species of herbaceous flowering plant) during the course of the study converted the majority of tall grassland (the dominant vegetation cover on >50% of the study area at the outset) to mixed scrubland. Over 2/3 of the reserve perimeter is currently electric fenced, but wildlife movement in and out of the park remains possible.

Long-term Monitoring

UWNP hosts an estimated population of between 800 and 1160 elephants that remains relatively stable over multiple years, although individuals varied in the frequency with which they were seen at different times of year [17]. A pilot study spanning 20 days was conducted in August 2005 and the remaining data were collected between June 2006-December 2012 as part of an ongoing study. Observations inside the park were conducted between 0630–1800 h, three days per week on average by vehicle along the road network in approximately the middle third of the park (Figure 1). Effort was distributed evenly year-round and across years except January–April 2008 for which data were not available due to closure of the park during a period of political unrest [17].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Uda Walawe National Park.

The study was conducted along the road network in the middle portion of the park, represented by the hashed lines (<100 Km²). Images sources: NASA Earth Observatory (http://earthobservatory.nasa.gov), the Department of Wildlife Conservation, Sri Lanka, and the Center for Conservation Research, Sri Lanka.

https://doi.org/10.1371/journal.pone.0082788.g001

Photographic and demographic records for known individuals were maintained. Individuals were identified using natural features of the ears and other physical attributes [21] (see also Table S1 and Figures S1–S2). To date we have individually-identified 409 individuals in the subadult or adult male size classes, along with 465 individuals in the subadult or adult female size classes. Known adults for whom the interval between sightings was <1 year on average were recorded as dead either when they were observed to have died (N = 3) or when they were not seen for two or more years (N = 10). Mortalities of nursing calves were inferred if mothers were seen without them on three or more occasions and/or mothers exhibited reduction in breast size after the calf went missing (N = 9). Data on mortalities were additionally collected from opportunistic investigations of carcasses (N = 15) inside and within 5 km of the park boundaries and corridors whenever such deaths were reported. Causes of death were inferred whenever possible either through direct observation (such as gunshots and other physical injuries) or interviews with personnel from the Department of Wildlife Conservation (DWC) who conducted investigations.

Age Estimation

Estimated birth dates for individuals in the study population were assigned an accuracy rating of 1–5. Accuracy 1 indicates calves that were seen within one month of birth, 2 indicates those seen within two months, 3 indicates those for whom the year was known but month was not, 4 indicates those for whom the birth year was estimated by size relative to the height of an adult female [17], and accuracy 5 indicates those who were already adults at the time of the study and, as a result, for whom birth year was approximated according to physical attributes (Table 1). Over the entire monitored population (N = 419), 52 calves were seen within a month of birth (assigned an accuracy rating of 1), 32 within two months (rated 2), 38 within a year (rated 3), 49 were aged based on size at an older age (rated 4), and the remaining 248 were adults when first identified (rated 5). Adult females were defined as those having had at least one calf rather than according to their estimated age class [17]. Age classes 0–13 were known with a high degree of accuracy (1–4) based on the observation of growth in known-aged individuals who were five years old or less at the start of the study. In this paper we focus on demographic data for females and their dependent young, but provide information on injuries and mortalities for males as well.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Visual cues used for adult female age-class assignment.

https://doi.org/10.1371/journal.pone.0082788.t001

Among adults, age classification was established from (1) comparison with individuals aged post-death (Figure 2) and (2) age-specific physical attributes described elsewhere [21], [22] and developed in this study, including visual assessments of head morphology, body development, and size relative to one another as well as the height and length of our observation vehicle or trees when they stood next to it. Males may continue growing throughout their lives, but female gain little height once they are of reproductive age. A mature adult female in the oldest age class typically had a back length of ∼2.3 m and shoulder height of ∼2.1 m in our population, but high variability in body size limited the utility of using such characteristics without additional calibration. Females and the majority of males in this population are tusk-less, thus tusk growth could not be used as an indicator of age as in other populations. Molar assessments [23] were also impractical as most carcasses were never found. We anchored age classes relative to one of the oldest adult females (60+years) whose jaw was recovered after death (Figure 2). Because estimates based on these criteria are coarse, we grouped females aged 10–20 yrs into 5-year age classes and those over the age of 20 into 10-year age classes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Female no. 33 (‘Tailless’), reference female for oldest age class.

This unique female, who was distinctly identifiable even after death due to her tail which was broken at the base, was among the oldest individuals observed in this population. Top panels: Tailless in 2008 exhibits the sunken forehead and cheeks as well as reduction in breast mass characterizing the >60 yr age class. Bottom panels: both upper and lower molars exhibit severe wear with no signs of additional teeth to emerge.

https://doi.org/10.1371/journal.pone.0082788.g002

Fecundity, Primiparity, and Inter-birth Intervals

Because the duration of the study was short relative to the developmental trajectory of elephants, we offer multiple estimates of vital rates with varying degrees of accuracy. We calculated age-specific fecundity over all females aged 0 to over 60 for whom ages could be estimated (N = 280). We first estimated the age at primiparity (first birth) using only those females known with certainty to have had their first calves during the study period, for whom data are most accurate. We then used standard lifetable approaches to derive the hazard function of age-specific primiparity among all identified females born between 1995–2003 (N = 34).

We estimated the inter-birth intervals (IBI) using multiple cohorts representing decreasing levels of accuracy but increasing sample sizes: (a) females for whom two successive births were observed within two months of the event (N = 13) (b) those for whom two successive births were recorded any time during the study period (N = 27), and (b) those for whom one birth was estimated to have occurred within 2 years of the start of the study and the second birth was observed (N = 37). We then present the hazard function for IBI calculated using the cohort of females that were of reproductive age between 2005–2012 (N = 92), excluding individuals who died, were seen infrequently, were over the age of 60, or were not parous before 2010.

Results are reported as averages ± estimated standard errors except where noted, and samples sizes are as specified. All statistical analyses were conducted in R v. 2.15 [24].

Reproduction and Growth

Although the density of elephants in UWNP is very high compared to those reported in the literature (Table 2), the age at first reproduction and inter-birth intervals we observed indicate a productive population. The annual peak in births toward onset of the dry season (far in advance of the monsoon) is similar to patterns reported for captive elephants in Myanmar and India [16], [27], [30]. In the Myanmar timber camps, calves born during the peak birthing months had a lower probability of dying before the age of 5 [30]. However, as the same months also corresponded to resting rather than working periods for the mothers, seasonal and anthropogenic effects on births as well as mortalities were difficult to disentangle. Because this confound does not hold for the wild population in our study, the cycles are more likely to actually reflect ecological influences and it would be of interest to see in future whether there is a similar effect on calf survival. However, these findings contrast to what was observed in Gal Oya, Sri Lanka, during a 22 month study in which an apparent birth peak was observed toward the end of the dry season in the first year, but not the second [32]. There appears to be some variability across populations therefore in the degree of coupling between births and seasonality, which would benefit from further exploration.

The median age at first reproduction (as opposed to conception) was approximately 13 years. Fecundity remained constant from the age at first reproduction until around the age of 60, from which point no female in our study was observed to give birth. Note however that because there are fewer individuals in this age class (Table 5) and the relatively short duration of the study, this does not necessarily indicate that individuals stop reproducing beyond this age. Birth sex ratios did not show significant bias, although there was a slight tendency to observe more females than males compared to records from captive or semi-captive populations (Table 2). At least two (non-mutually exclusive) explanations are possible. Females in the wild may be in poorer condition than those in captivity and thus more likely to have female offspring, in conformity with the Trivers-Willard hypothesis [47]. When broken down by the age class of the mother (Table 4), there seems to be a tendency for primiparous females to have more males compared to females in their peak reproductive years. Alternatively, early male-biased mortality may go undetected in the wild, where there is a lag between birth and observation. Table 5, however, suggests this is not the case. Larger sample sizes are needed to evaluate these effects.

Estimates of average IBI derived from the most accurate data (calves seen within two months of birth, and births estimated to have occurred during the study period) are in close agreement despite doubling the sample size (Table 3). Both are slightly lower than other published estimates. These estimate may be biased low in that females with long birth intervals were not well-represented in this study and the subsets of known individuals who gave birth twice are small (Table 3). When we include calves born shortly before the study period, the estimated IBI rises by three months to the more typically reported range of 4–6.5 years (Table 2). The 50^th percentile IBI value (Figure 5b) falls at 6 years. It is interesting to observe that the average IBI also increased after excluding those females who lost their calf within the first year (despite a reduction in sample size). Taken together, these observations suggest that ignoring calf mortalities increases the apparent IBI. Published estimates of IBI for wild Asian elephant populations are typically inferences from cross-sections of the age structure of living offspring seen during short-term observations. Such estimates would miss calf mortality as well as individuals with longer birth intervals, where two consecutive offspring are not simultaneously visible. Additional years of data on the Uda Walawe population would enable the IBI to be estimated with greater confidence.

The individuals for whom the most complete and accurate data were collected also happened to be those occupying central areas of the park, largely buffered from human disturbance aside from tourism. These data potentially understate human impacts affecting the broader population. There is a high degree of seasonal turnover within the park [17]. Those occupying more peripheral areas and ranging outside the park, where observations are more difficult and where they are more exposed to human activity, may exhibit demographic differences.

Calves in this population were usually weaned as soon as a sibling was born, around the age of four. But a few may nurse up to the age of five or more in the absence of a younger sibling. Some females were observed nursing two calves simultaneously, and one exceptional female permitted all three of her calves to nurse in close succession. The older calf was approximately seven or eight years old. Our observations based on those rare cases in which calves appear to have been orphaned before the age of 4 suggest that milk is important for normal growth at least up to this age, even though they do consume herbaceous vegetation. This is similar to what was found using the largest demographic dataset on semi-captive elephants. A recent study compiling multiple decades of data on elephants from logging camps in Myanmar reported that while the leading cause of death among calves under the age of 5 was accidents (42.4%) the second most common cause was calf weakness due to maternal agalactica, or lack of milk, and general weakness of the neonate (26.3%) [26]. Maternal condition has been shown to affect the duration of lactation in several species of large bodied mammals, with young typically weaned when they attain four times their birth weight [33]. In the absence of predation, both in the wild and semi-captivity, maternal condition may play an important role in regulating generation time in Asian elephants via its effect on calf growth and survival.

Injuries and Mortalities

The majority of sightings involving injuries did not appear fatal though several adult males habitually appeared in Uda Walawe exhibiting gunshot wounds in the legs, side, or pelvic region. One adult male in the 31–50 age class appeared in 2012 with multiple swellings from gunshot wounds to the legs which subsided within a month. Three males including a tusker and one (pregnant) female were confirmed to have died from gunshot wounds (Figure 7 and S3). Anthropogenic activity was evident in 76% of the injuries and mortalities in males but only 29% in females. This tendency for more males than females to exhibit signs of conflict with humans is similar to what is observed in southern India where 49% of male mortality was attributed to poaching or conflict, compared to 12% for females [16]. The majority of shooting incidents are likely due to crop raiding events, and perhaps intended to deter rather kill the animal. Similarly, burn wounds are caused by defensively throwing or stabbing at animals with clay or cloth flame torches which then adhere to the skin. The tendency of males to be the primary crop raiders is typical of elephants [34], including the sister species L. africana which exhibits social differentiation among those that raid habitually and those that do not [35].

Carcasses exhibited two patterns of death other than by physical trauma. Four individuals, all adult females, died after several days of prolonged incapacitation due to unknown causes with no obvious external signs of injury. These were concluded to be predominantly natural, possibly due to malnourishment, old age, or disease etc. In contrast, two adult females and one sub-adult male exhibited signs of sudden death, also without any external indications of injury. Two of these were reported to be poisoned while the third cause was not determined. Our criteria for recording mortality were very conservative. Although the proportion of mortalities relative to births appears low, it is very likely deaths are being under-reported due to the difficulty of detecting carcasses in the dense vegetation as well as our inability distinguish death from dispersal in the intermediate age classes. We therefore present only raw counts of mortalities, rather than estimates of age-specific survivorship in this study. We hope that longitudinal monitoring of known individuals of all age classes will enable such estimates in the future.

Wider Context

This study demonstrates that wild populations can be productive despite range constriction and the presence of high human population densities. Sri Lanka is estimated to contain approximately 13.5% of the global Asian elephant population, the second largest of all range countries, yet it also contains the third highest density of people per unit area [36], [37]. It is generally thought that scrub and grassland environments can support higher densities of elephants than forest environments [16]. UWNP fits this perspective. In addition, the creation of man-made water sources in and around the park likely augment its carrying capacity. However, an alternate possibility is that the population is exhibiting demographic inertia. As with other slowly-reproducing mammals such as the great apes, the effects of habitat constriction and loss of gene flow may only be visible after substantial time lags [38]–[40]. UWNP is surrounded by agriculture and cultivation, but connected to other patches of scrub and deciduous forest in the north and east. While elephants within the park likely serve as a source population for the region, the ability of these remnant habitat patches to support dispersing individuals is not well known. This raises several important questions to be addressed by future research: how and where are individuals from this population dispersing each year? What other populations are linked to it and what is the extent of gene flow among them? How do dispersing individuals interact with human settlements?

Attention to the demographic dynamics of wild populations is also crucial because elephants influence ecosystem structure through their biotic and abiotic interactions [8]–[10]. Fernando & Pastorini (2011) note that distribution data for Asian elephants are still inadequate, and do not permit “accurate monitoring of distributional changes over time, other than the complete disappearance of an entire population” [36]. Extirpations could be forestalled if the parameters within which populations persist or decline locally could be identified. Failure to monitor them likewise weakens our ability to ensure the persistence of these remnant populations. Unfortunately, if data on distributions is scarce, data on demographics in the wild are almost nonexistent, but for the few studies referenced here. Maintaining longitudinal records on single populations is time-consuming, costly, labor-intensive, logistically challenging and vulnerable to political interference. The health of extant populations and the impacts of human activity on their viability therefore remain poorly known. Nevertheless, concerted efforts must be made to overcome these data gaps and challenges if we are to conserve both elephant populations, and the ecosystems they help to shape.