How Big Is It Really? Assessing the Efficacy of Indirect Estimates of Body Size in Asian Elephants

Abstract

The purpose of the demographic analyses in this study was to calculate the basic life tables to determine the effects of the long-term captivity of Asian elephants (Elephas maximus), which are utilized extensively as draught animals, on survival, fecundity and viability. The studbook data were collected from the elephant log books and the annual reports of the Extraction Department, Myanma Timber Enterprise of the Union of Myanmar. We had access to a near-total of the records (n (9600) of elephants captured or born after the year 1875, including 3 070 calving records. It was documented that 32.5 percent of calves born in captivity failed to reach the age of five years. Life table analysis revealed that most mortality occurred before the age of five. Survivorship analysis of adults and sub-adults (more than five years) showed that wild caught elephants and female elephants had significantly higher survival rates (P <0.001) than captive born and male elephants, respectively. A similar analysis was conducted for calves (under five years) and comparisons were made between dam origins and sex. It was revealed that calves born from wild caught (WC) dams had higher survival rates than those born from captive born (CB) dams (P <0.001), while survivorship and sex showed no correlation.

Citation: Chapman SN, Mumby HS, Crawley JAH, Mar KU, Htut W, Thura Soe A, et al. (2016) How Big Is It Really? Assessing the Efficacy of Indirect Estimates of Body Size in Asian Elephants. PLoS ONE 11(3): e0150533. doi:10.1371/journal.pone.0150533

Editor: David B. Allison, University of Alabama at Birmingham, UNITED STATES

Received: September 21, 2015; Accepted: February 15, 2016; Published: March 3, 2016

Copyright: © 2016 Chapman et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are available from Dryad (doi:10.5061/dryad.p9r4k).

Funding: VL received funding from the European Research Council and Leverhulme Trust for this study. HSM received a Natural Environmental Research Council PhD studentship, and post-doctoral funding from the Leverhulme Trust, Society in Science—Branco Weiss Fellowship and a Drapers' Company Junior Research Fellowship. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The size of an organism has a major role in its life-history, influencing aspects such as mating behaviour [1], reproduction [2,3], offspring mass [4] and dominance [5,6]. Knowledge of an individual’s size can also be used to tailor medical treatment or intervention, to aid conservation efforts [2,7,8], detect the extent of population declines [9,10], and for studies on the behaviour, ecology and evolution of animals [7,11,12]. For wild, large or dangerous animals, however, this data may be difficult to obtain [1317], especially as immobilisation may negatively affect the animal [18].

In these cases, size can be measured indirectly. There is already a diverse array of methods through which linear size measurements can be estimated, including photogrammetric methods [5,8,15,16,1921], use of acoustic signals [2224], using footprint size [25] or brood mass [2], and very basically by observer estimations [9,13,26]. Weight is an especially important measure of body size; not only does it play a role in the traits already listed, but it also is a marker of body condition [27]. It is, however, often even less feasible to obtain a measure of an animal’s weight than its height or length, especially for larger species [17]. Whilst body parameters such as height, girth and length can quickly and easily be measured on immobilised animals if required [8,15,28], weighing large animals, particularly mammals, requires equipment that may not be practical to use in the field [15]. Measurements of these 

other, more easily obtained, body parameters are frequently used in equations to predict weight [5,14,2830], such as using chest girth to estimate weight of polar bears Ursus maritimus [28]. Other techniques, though less commonly used, include estimation from faecal bolus weight [31,32] or the size of teeth [33].

Though these techniques are frequently used in lieu of direct measurements, their accuracy cannot always be properly assessed, often because of the difficulty in obtaining these measurements directly. For instance, in sperm whales Physeter macrocephalus, acoustically ascertained lengths are validated with photogrammetric lengths [22,23], but this has shortcomings, most notably the difficulty in linking acoustically obtained size to a specific individual when more than one whale is present [10]. Photogrammetric techniques rely on calibration to a scale of known size, and this is often used to validate the accuracy of the method in place of direct comparison between photographic and actual measurements [16,1921]. A range of prediction equations based on measurements have been developed to estimate animal weight, though to what extent these prediction equations for body weight are suited to a specific population, rather than the species as a whole, is often unknown–their applicability to populations beyond the individuals included in the original studies has rarely been tested.

 The Asian elephant Elephas maximus is an ideal species with which to test the accuracy of predictive tools for a number of reasons: size may be linked to traits such as male dominance [34], conservation is of vital importance in this species (the Asian elephant is classified as endangered), and elephants can be dangerous and immobilisation undesirable [35], so lack of body size data is often an issue. To counter the latter problem, numerous equations exist for the species that predict weight based on other body parameters [3640], and photogrammetry has recently been used to obtain length measurements [41]. Though photogrammetry is often used to obtain measures of height in other elephant species [32,42,43], it has not been validated with directly obtained measures in Asian elephants and a study of semi-captive elephants provides an opportunity to do so for the first time. We also estimate both weight and height from photographs and test their accuracy at predicting real body size, as these parameters are important in distinguishing between growth and increasing weight alone [34].

In this study, we aim to assess the efficacy of existing prediction equations for estimating weight by using a large semi-captive population of known age Asian timber elephants in Myanmar, for which cross-sectional and semi-longitudinal weight and morphometric measurements (height, chest girth, neck girth, foot circumference and back length) are available, allowing us to directly compare predicted to actual values. In addition, we develop sex-specific equations to allow individually-tailored veterinary care for elephants of all ages, for example to enable the application of more accurate doses of medication when dosage is weight-dependent. We also test how body height and chest girth from photographs of individual elephants correlate to directly measured height and chest girth. In combination, our study forms an extensive and detailed appraisal of the use of direct and photogrammetric methods to accurately estimate the height and weight of a larger-bodied mammal.

Methods

Study Population

Asian elephants are an endangered species [44], present in discontinuous populations across South East Asia and the Indian sub-continent. The second largest wild population of Asian elephants is found in the Union of Myanmar, with as many as 5,000 individuals [45,46]. Myanmar also has the largest captive population, of around 5,000 elephants [46], 2,700 of which are government-owned through the Myanma Timber Enterprise and are used for sustainable logging [47]. Historically the captive-born population has been supplemented with elephants captured from the wild [48], to the extent where almost half of the living elephants were born in the wild. The logging elephants are released into the forest at night for up to fourteen hours to forage at will and mate unsupervised [48]; both wild and captive bulls have access to oestrus females. As there is no additional supplementation or selective breeding, they are classed as a semi-captive population.

Every elephant is marked with a unique identification (ID) number, and has important life-history information recorded in log books [49], including information on health, mother ID, body measurements taken by vets at each health check, and dates of birth and death. For wild-born individuals, age is not accurately known, so year of birth is visually estimated upon capture based on numerous body condition markers: facial concavity, skin pigmentation, ear folding, and size (though this only applies for younger individuals) [50]. This population is therefore ideal for testing prediction tools.

Infants stay at the mother’s heel until the age of 5, at which point they are separated and trained. The work season lasts from mid-June to mid-February, with a rest period during October. This working season is timed to coincide with the monsoon (July-October) and cool (November-February) seasons, so that no work is done during the dry season (March-June) [51], when temperature-related mortality is highest [52]. Elephants have well-defined workload limits each year, which cannot be exceeded. As well as an annual maximum tonnage of logs each elephant can move, there are strict limits to daily and weekly work–in 2010 these limits were set to a daily maximum of eight hours, with a break at noon, and no more than five days of work in a week [53]. Elephants are only used for light work from age 5 years until 17, becoming part of the true working population after this. The main working population are able to drag logs and engage in heavier work, continuing up until retirement at around 55 years old [54]. Pregnant females are given time off work from mid-pregnancy through to a year after birth [48], and are then given lighter work up until weaning, at which point the mother begins heavier work again.

Data Selection

The weight of elephants and five morphometric measures—back length (from the base of the neck to the fold of the tail), chest girth (measured behind the forelegs with care taken to ensure it was not affected by inhalation by the elephant), foot circumference (right forefoot), height (from scapula to the ground) and neck circumference (taken close to the body, with care taken that the measurement was not affected by breathing or any ropes or bells worn by the animal)—were recorded in camps in five locations in Myanmar: Pyinmana, Monywa, East Katha, West Katha and Kawlin. Weight was measured to the nearest 1kg using EziWeigh 3000 scales, whilst the other measurements were taken with tape measures, and were recorded either in inches or centimetres, depending on location. A single trainer trained a measurer at each camp and observed them taking at least one measurement of each individual herself; repeated monthly measures were then taken by the measurer independently. For consistency, all elephants were measured in the morning on non-work days. Individuals in the final year of pregnancy were not included in the analyses. All measurements were converted into centimetres during collation of data for consistency, though this means measurements are only correct to the nearest multiple of 2.54cm (i.e. 1 inch). A total of 404 elephants aged between 9 months and 71 years were included in this study, with each individual having one or more measures of weight and at least one other body parameter measured. Of these, 224 were female and 180 were male.

In Pyinmana (n = 72), individuals were measured on a monthly basis between December 2011 and October 2012. Elephants were recorded cross-sectionally just once in Monywa in 2012 (n= 69). In East Katha (n = 71), data were collected monthly from December 2012 until the end of 2013. In West Katha (n = 45), monthly measurements have been taken since June 2012, with data up until March 2015 available. Kawlin (n = 147) data have been collected since November/December 2012, with records up until April 2015 available. However, for camps where monthly measurements had been taken, there are some months with missing data. Multiple measurements of the same individuals, which can be averaged, are likely to reduce variation in the measurements due to feeding patterns, illness and work schedules.

In addition to the body measurement data, photographs of 151 elephants in profile were taken in Monywa, East Katha, West Katha and Kawlin, with a measuring stick of known length held up directly against them. One photo was selected for each elephant, based on three criteria: whether the elephant was perpendicular to the camera, whether the ends of the scale were visible, and whether the foreleg was straight. Using ImageJ software version 1.48 [55], the 'photographic' heights of elephants were ascertained. This was done by using the visible measuring stick to set the scale for each photo, and then measuring the shoulder height from scapula to the base of the foot. If the leg was straight but not completely vertically orientated, the measurement would be taken from the scapula to the area of ground level to the base of the foot. A proxy for chest girth was also measured on the photographs: from the underside of an elephant behind the leg in a vertical line to the corresponding point on the elephant’s back. Measurements were independently taken by four people, to test if the method can consistently be applied. This research has ethics approval from the University of Sheffield, Department of Animal and Plant Sciences, Department Ethics Committee. Fieldwork permission was granted from the Ministry of Environmental Conservation and Forestry in Myanmar.

Statistical Analyses

All analyses were conducted using R version 3.2.1 [56].

Predicting weight.

To test the ability of previously published growth equations (Table 1) to accurately estimate the weight of elephants using different available body parameters, the equations were first applied to subsets of the data which included only those elephants that had weight and the other body parameters specified in the equation recorded for them. When there were many equations in a published study, only those reported to fit well were selected. Furthermore, only elephants within the original age range of each equation were used, with those outside the range excluded from testing for that particular equation. The predicted values of weight obtained from this procedure were averaged for each individual, and the observed monthly weights were averaged for each individual, giving one predicted and one actual value for each elephant. This allowed us to approximate the cross-sectional nature of the original studies as closely as possible. There were two exceptions to this: the two equations from Sukumar et al. [38], which were created from measurements taken over a period of four years. These predicted values were then run in linear regression equations against the actual weights of the elephants, with a separate regression for each equation.

To select the best fitting equations, we assessed values of R2 and the mean and standard deviation of predicted weights relative to actual weights, represented here as a percentage of actual weight, with a value of 100% showing predicted and actual measures to be the same.R2 by itself is insufficient to inform on the accuracy of these equations [57]–it is a measure of relation, rather than agreement–whereas relative mean is directly linked to how close predicted weights are to actual weights–a representation of agreement.

New equations for predicting weight.

In order to create new equations that best predict weight from the available body parameters (height, chest girth, neck circumference, foot circumference and length) and in this specific population, we first explored the correlations separately between each of the different body parameters and weight for each sex. The individual body parameter correlates most closely related to weight were used in linear regression mixed-effects models using the function lmerfrom the package lme4 version1.1–9 [58]. Separate models were created for each sex, as growth curves show large differences in the pattern of weight gain and values of weight between the sexes [34]. Different body parameters were not used in combination in any of the equations; as they are closely correlated with each other they should not be included in a single model—they add little in the way of accuracy, whilst adding a degree of complexity that would make them less applicable in the field [17].

Initially, a series of full models were created using a training set of 70% of the dataset (n = 188 females and 155 males). The data were split by observation rather than individual to ensure a range of ages were present in the training dataset and that a few young elephants with only a single row of data each would not affect the results. The fixed effects in these full mixed-effects models were as follows: the body parameter (height, chest girth, neck circumference, foot circumference and length), with the closest correlates being height and weight; age in years, to account for the difference between age-related increases in the body parameter and weight; age squared, to allow non-linear increases in weight; and season of measurement, which was a factor with three levels (Dry, Cool, Monsoon) to account for seasonal variation in body condition and weight [27]. There were two random effects: elephant ID, a factor with nlevels, with n being the number of elephants in that particular subset, which was included to account for between individual variation, and location, a factor with five levels (West Katha, East Katha, Pyinmana, Monywa, Kawlin), to account for variation by region. As the variance in the age of wild-born elephants is likely to be greater than a year, they were not used in the construction of the models in which age was present (in these models, n = 131 captive-born males and 132 captive-born females—subset from the training data). We also constructed linear models without random effects, but with the same fixed factors as above—to be used as prediction equations in the field—and tested their fit to the data in comparison to the models with random terms to determine how much individual-level variation and location-level variation alters the model. We then produced reduced linear models, which excluded both age and the random effects, to allow application to all elephants, regardless of birth origin. Finally, prediction equations were extracted from the models and tested on a testing subset of the data, which comprised the remaining 30% of the dataset (n = 139 females and 121 males). This training and testing procedure was appropriate for this dataset, with repeated measures for individuals being treated as independent and, in the mixed-effects models, the random ID term statistically controlling for between-individual variation.

Prediction equations that fitted well with the testing subset, based on regression outputs (S1 Table), were then used to obtain predictions of weight from the entire dataset. These predicted weights were compared to the actual measured weights of elephants through calculating the mean and standard deviation of predicted weights relative to actual weights and from the outputs of linear regressions—regressing observed weight against predicted weight—to test for applicability in the field. The fits of all these equations were then compared with the fits of the previously published equations (which were also compared to the whole dataset), to allow direct comparison. No values were averaged for this analysis. If an existing equation was applicable to both sexes, it was run separately on both.

Predicting height and chest girth.

In order to test the photographic method of estimating height and chest girth, estimates of these parameters obtained from analysis of photographs were compared to the actual measurements of the animals taken in the field. To test for differences in values obtained through the ImageJ method by different people, Pearson’s product moment correlation coefficients were calculated. Linear regressions were then run between actual measurements and the heights and proxy measurements for girth obtained using ImageJ software.

Results

Measurements in elephants, in fitting with standard mammalian growth patterns, differ with age and sex (Fig 1), and appear in this population to represent determinate growth [34]. Males are larger than females of the same age in every body parameter within a few years of birth, and continue to gain weight throughout life.

Existing Prediction Equations

We applied twenty-one prediction equations from five previous studies in the literature to data on the timber elephants in order to test whether they are applicable in the field for a different population than the one from which they were created. 2 from these models ranged from 0.080 to 0.956 for these prediction equations, highest from Kurt and Garai’s [36] male equation (R2 = 0.956), which uses chest girth and height, and lowest from Sreekumar and Nirmalan’s [39] length-based equation (R2 = 0.080). The Kurt and Garai [36] equation for infants provides the most accurate fit (R2 = 0.814, mean±SD (%) = 99.3±7.7), though this only applies for elephants under the age of 6 (Fig 2Table 2).

New Prediction Equations

We found that weight was highly correlated in both sexes with chest girth (male r = 0.949, n = 177, p <0.001; female r = 0.916, n = 221, p <0.001) and height (male r = 0.954, n = 178, p<0.001; female r = 0.913, n = 223, p <0.001); these parameters were selected for use in the models. Neck circumference was also highly correlated, but less so in females than males (male r = 0.938, n = 65, p <0.001; female r = 0.837, n = 82, p <0.001), and fewer elephants had available data on neck circumference. In contrast, foot circumference (male r = 0.888, n = 162, p <0.001; female r = 0.641, n = 189, p <0.001) and length (male r = 0.702, n = 162, p<0.001; female r = 0.678, n = 197, p <0.001) predicted weight considerably more poorly, especially for females. Model comparison showed chest girth to be the best single predictor of weight in both sexes (Table 3; see S2 and S3 Tables for model parameters). The ΔAIC values indicate that the model including height and chest girth is better-supported than either chest girth or height alone. However, as discussed in the methods, any additional accuracy is counteracted by an increase in complexity that would make field use less practical, and the correlation between the body parameters makes it difficult to interpret models including more than one parameter. Therefore, chest girth is the preferable single measurement for use in the equations. The most parsimonious model has the terms chest girth, age and age2. Season of measurement had no significant effect on predicted weight in either males (χ22 = 1.13, p = 0.569) or females (χ22 = 2.06, p = 0.357) in this sample.

The comparison of training and testing results is provided in S1 Table. All prediction equations below are based upon models created with the training dataset.

The predictions of weight obtained from the mixed-effects models are accurate, predicting values very close to the actual weights of the individuals in the model (male R2 = 0.986, mean±SD (%) = 100.3±5.8; female R2 = 0.963, mean±SD (%) = 100.3±5.5). As random effects cannot be included in an equation, field-applicable equations were derived from the basic linear regression models (S4 Table). The equations for chest girth are as follows, where W = weight,CG = chest girth and t = age:

When random terms accounting for ID and camp were removed, the fit was slightly poorer than for the mixed-effects models, as expected (male R2 = 0.944, mean±SD (%) = 101±10.2; female R2 = 0.873, mean±SD (%) = 101.6±10.9). However, there was little difference between this test of known-age captive-born elephants only and that for all elephants, including age-estimated wild-born individuals (male R2 = 0.935, mean±SD (%) = 101±10.6; female R2 = 0.884, mean±SD (%) = 101.4±10.2). The reduced, chest girth only model with no age term, to be used in populations of elephants of unknown ages, has a larger deviation (male R2 = 0.903, mean±SD (%) = 100.3±14.8; female R2 = 0.839, mean±SD (%) = 101.5±13.2), though it is better at prediction at higher weights than the equations derived from the mixed-effects model (Fig 3). The new equations for the Myanmar elephants with age terms were more effective for predicting weight than any of the previously published equations (Fig 4S5 Table). The reduced equations with chest girth as the sole predictor were also preferable for predicting weight. However, for young elephants, the Kurt and Garai [36] <6 years old equation (female R2 = 0.815, mean±SD (%) = 97.8±6.0; male R2 = 0.616, mean±SD (%) = 99.1±9.5) was of a similar efficacy to the female equation (Eq 3) applied to only those under 6 years (R2 = 0.745, mean±SD (%) = 102.9±7.5), and more accurate than the male equation (Eq 1) applied to the same age group (R2 = 0.553, mean±SD (%) = 99.4±10.2). The reduced equations were also less accurate when applied to young elephants (female R2 = 0.691, mean±SD (%) = 106±9.6; maleR2 = 0.542, mean±SD (%) = 96.1±18.2).

Predicting Height and Chest Girth

The accuracy of using photographs to estimate an elephant’s height and chest girth is high (Figs 5 and 6), with the values predicted from the photographs being highly correlated with the measured heights (r = 0.946) and chest girths (r = 0.940). There was little difference in values obtained by different people measuring height and chest girth from photographs: r was greater than 0.97 in all cases (S6 Table). Most height values are underestimated by a similar amount—between ~5 and 20cm—meaning a simple equation can be used to obtain heights closer to the actual value (Fig 7):

Similarly, the proxy photo measurements of chest girth can be adjusted for estimated values close to the actual chest girth of elephants (Fig 7):(6)

Discussion

Weight is a particularly important measure of size, serving as an indicator of body condition [27] as well as influencing other traits [35], whilst height and/or length is more indicative of skeletal size [34]. It is, however, more difficult to directly measure weight than other body parameters, so prediction equations are frequently used, though often without validation using known values of weight. We found chest girth, sometimes referred to as heart girth or thorax circumference in the literature, to be the best predictor of weight in Asian elephants, in line with previous studies on large mammals [14,17,28,5961], and some of the other weight prediction equations for Asian elephants [36,37,39]. Whether the relationship between chest girth and body weight, and thus the equation used for predicting weight from chest girth, differs between distinct populations of a species is not known outside a few species, and there is no consensus: white-tailed deer Odocoileus virginianus populations differ in this relationship [59], though populations of the grizzly bear Ursus arctos may not [62]. Similar to Weckerly et al. [59], we found that equations from other populations of Asian elephants did not fit the Myanmar timber elephant dataset particularly accurately; only the Kurt and Garai [36] equation for elephants specifically under age 6 was more accurate than the new equations we propose in this paper, possibly due to lower variation in body weight in early life [38]. Future work on animal species in which weight is predicted from equations should therefore exercise caution, rather than using established, but untested, equations for predicting weight. We have also shown that although mixed-effects models provide the best predictions of weight by taking individual variation into account, their actual application in field studies are limited: random effects (in this case ID and location) cannot, by their very nature, be included in a single, simple equation.

For indirectly measuring linear body measurements, such as height or length, photogrammetry is used, especially for large animals [8,10,41]. As with weight, measures obtained in this manner have rarely been tested against actual values. By comparing actual heights with heights estimated through photogrammetry, we have shown that a basic photogrammetric method is sufficiently accurate for estimation of height, and is likely to be useful for estimation of other linear morphometric measurements. We have also shown that a proxy measurement for chest girth is highly correlated with actual chest girth and that these photogrammetric measurements could therefore be used for prediction of weight. This may allow estimation of the weight of wild elephants without sedation. The difference in measurement values obtained by different measurers was negligible; basic photogrammetry of this kind can therefore be easily implemented, with little training required. We found that the accuracy of this method is high, without the need for high-end equipment such as lasers, which have been used in recent photogrammetric studies [19,20]. The main challenge of this basic technique, however, is trying to include a scale directly next to an animal with which measurements can be calibrated–for this population of semi-captive elephants the scale is held directly against an individual, but for wild or dangerous species other solutions will have to be found. Furthermore, though a scale is required for calibration, it is not sufficient for validation, as the technique does not guarantee accuracy; a simple equation, calculated from validation with actual values, may be required for increasing the accuracy of further measurements collected with photogrammetry [32].

For the Myanmar timber elephants, the prediction equation most commonly used in the field by local veterinarians is the Hile et al. [37] equation for all elephants, though this tends towards overestimation (R2 = 0.869, mean±SD (%) = 105.8±11.6), and as such is ill-suited for accurate prediction of weight in this population. Of the equations proposed in this paper, the new equations we have produced including age provide very good fits, even with wild-born elephants included. The inclusion of age terms does, however, limit their applicability as management tools to elephants of known age, as wild-born elephants cannot be guaranteed to have ages estimated to within a year of their actual age. For wild-born elephants captured later in life where age cannot be accurately estimated, the new reduced equations (Eqs 2 and 4) we have produced should be used instead of previous equations. This could be because wild-caught elephants have growth patterns that are different to captive-born elephants [38], even if they are aged accurately. If the equations with age are too complex for field application, the reduced equations are suited for all elephants over age 6. Even though these new equations are more accurate over the entire population, the Kurt and Garai [36] young elephant equation (Table 1) provided the best accuracy for calves aged under 6.

For the purposes of conservation, knowing body size is critical. Size is of great ecological importance: at a species level, body size can dictate niche differentiation [63], population density [64] and range size [65]; at an individual level, body size can influence resource choice and trophic level [66], metabolic rate [67] and intraspecific variation in range size [11]. Accurate measurements of body weight are vital for practical management: dosages of both medicinal and immobilising drugs often depend on weight [35,40,6871], which can be used for veterinary care, collection of other body data, and for fitting GPS collars [68]. Ideally animals will be weighed on scales, though this is not always possible, either because of logistical issues or the difficulty of encouraging wild animals to stand on them; prediction equations, therefore, are often the only viable alternative. It is especially important to know the weight of Asian elephants as they are an endangered species, and this particular population is already unsustainable [47].

In conclusion, we have demonstrated that weight prediction equations provide a high enough level of accuracy that they can be used in the absence of scales, though they may be too specific to a certain population for use on other, distinct, populations of the same species without validation first. Basic photogrammetry can easily be implemented without specialist training and can also be highly precise, so long as a scale of known size is present and wholly visible, and that at least a small number of individuals are of known size, so that an appropriate equation can be used to obtain highly accurate measures of both weight and height.

Supporting Information S1 Table. Fit of predictions on training and testing dataset

From linear models regressing observed weight against predicted weight.

doi:10.1371/journal.pone.0150533.s001

S2 Table. Mixed-effects model estimates.

Female n = 132 (720 observations); male n = 131 (689 observations). Estimates are for full models with all parameters included. The season estimates are in comparison to the cool season.

doi:10.1371/journal.pone.0150533.s002

S3 Table. Model estimates for predicting body weight.

Models included the following as explanatory variables: i) chest girth, age and season of measurement (female n = 132; male n = 131); ii) chest girth and age (female n = 132; male n = 131); iii) chest girth only (female n = 188; male n = 155). Intercepts are for the ‘Cool’ season, and are in kg. All models use a Gaussian distribution.

doi:10.1371/journal.pone.0150533.s003

S4 Table. Regression equations for estimating weight from other body parameters

Where P is the body parameter.

doi:10.1371/journal.pone.0150533.s004

S5 Table. Summary statistics of prediction equations.

R2, intercept and slope were obtained from running linear models of actual weight against predicted weight. Relative mean and standard deviation are the mean and standard deviation of predicted weights as a percentage of actual weights.

doi:10.1371/journal.pone.0150533.s005

S6 Table. Correlation coefficients between photo-measured heights and girths from different measurers.

doi:10.1371/journal.pone.0150533.s006

We thank the Leverhulme Trust (VL and HSM), European Research Council (VL), Drapers’ Company Fellowship at Pembroke College and a Society in Science, Branco Weiss Fellowship, administered by the ETH Zürich (HSM) for funding. We thank the Ministry (MOECAF) for granting permission for our group to access the timber elephants, and the MTE vets and mahouts for arranging elephant measurements, as well as the field and administration staff in Myanmar who helped with data collection and collation. We also thank Alysia Dyke, Alex Pryke and Eloise Wells for measuring heights with imageJ.

Author Contributions

Conceived and designed the experiments: HSM VL SNC. Performed the experiments: KUM WH ATS HHA HSM. Analyzed the data: SNC JAHC HSM VL. Contributed reagents/materials/analysis tools: KUM WH ATS HHA. Wrote the paper: SNC HSM JAHC KUM WH HHA ATS VL.

References

  1. 1.Aubin-Horth N, Dodson JJ. Influence of individual body size and variable thresholds on the incidence of a sneaker male reproductive tactic in Atlantic salmon. Evolution. 2004;58: 136–144. doi: 10.1111/j.0014-3820.2004.tb01580.x. pmid:15058726
  2. 2.Hall PM. Estimation of nesting female crocodilian size from clutch characteristics: Correlates of reproductive mode, and harvest implications. J Herpetol. 1991;25: 133–141. doi: 10.2307/1564639.
  3. 3.McElligott AG, Gammell MP, Harty HC, Paini DR, Murphy DT, Walsh JT, et al. Sexual size dimorphism in fallow deer (Dama dama): do larger, heavier males gain greater mating success? 2001;49: 266–272.
    • 4.Blueweiss L, Fox H, Kudzma V, Nakashima D, Peters R, Sams S. Relationships between body size and some life history parameters. Oecologia. 1978;37: 257–272. doi: 10.1007/bf00344996
    • 5.Haley M, Deutsch C, Le Boeuf B. Size, dominance and copulatory success in male Northern elephant seals, Mirounga angustirostris. Anim Behav. 1994;48: 1249–1260. doi: 10.1006/anbe.1994.1361.
    • 6.Schuett G. Body size and agonistic experience affect dominance and mating success in male copperheads. Anim Behav. 1997;54: 213–24. doi: 10.1006/anbe.1996.0417. pmid:9268451
    • 7.Rayadin Y, Spehar SN. Body mass of wild bornean orangutans living in human-dominated landscapes: Implications for understanding their ecology and conservation. Am J Phys Anthropol. 2015;346: n/a–n/a. doi: 10.1002/ajpa.22709.
    • 8.Berger J. Estimation of body-size traits by photogrammetry in large mammals to inform conservation. Conserv Biol. 2012;26: 769–777. doi: 10.1111/j.1523-1739.2012.01896.x. pmid:22834930
    • 9.Bradshaw CJA, Fitzpatrick BM, Steinberg CC, Brook BW, Meekan MG. Decline in whale shark size and abundance at Ningaloo Reef over the past decade: The world’s largest fish is getting smaller. Biol Conserv. 2008;141: 1894–1905. doi: 10.1016/j.biocon.2008.05.007.
    • 10.Jaquet N. A simple photogrammetric technique to measure sperm whales at sea. Mar Mammal Sci. 2006;22: 862–879. doi: 10.1111/j.1748-7692.2006.00060.x.
    • 11.Lindstedt SL, Miller BJ, Buskirk SW. Home range, time, and body size in mammals. Ecology. 1986;67: 413–418. doi: 10.2307/1938584
    • 12.Ma X, Lu X, Merilä J. Altitudinal decline of body size in a Tibetan frog. J Zool. 2009;279: 364–371. doi: 10.1111/j.1469-7998.2009.00627.x.
    • 13.Woodward AR, Moore CT. Use of crocodilian night count data for population trend estimation. Second Regional Conference of the Crocodile Specialist Group, Species Survival Commission. 1993.
      • 14.De Esteban-Trivigno S, Köhler M. New equations for body mass estimation in bovids: Testing some procedures when constructing regression functions. Mamm Biol. Elsevier GmbH; 2011;76: 755–761. doi: 10.1016/j.mambio.2011.07.004.
      • 15.Bell CM. Growth of the southern elephant seal Mirounga leonina (Linnaues 1758) at Macquarie Island. MSc. Thesis. University of Tasmania. 1997.
        • 16.Breuer T, Robbins MM, Boesch C. Using photogrammetry and color scoring to assess sexual dimorphism in wild Western gorillas (Gorilla gorilla). Am J Phys Anthropol. 2007;134: 369–382. doi: 10.1002/ajpa. pmid:17657788
        • 17.Talbot LM, McCulloch JSG. Weight estimations for East African mammals from body measurements. J Wildl Manage. 1965;29: 84–89. Available: http://0-www.jstor.org.innopac.up.ac.za/stable/10.2307/3798635\nhttp://www.jstor.org/stable/10.2307/3798635
        • 18.Alibhai SK, Jewell ZC, Towindo SS. Effects of immobilization on fertility in female black rhino (Diceros bicornis). J Zool. 2001;253: 333–345. doi: 10.1017/S0952836901000309.
        • 19.Deakos MH. Paired-laser photogrammetry as a simple and accurate system for measuring the body size of free-ranging manta rays Manta alfredi. Aquat Biol. 2010;10: 1–10. doi: 10.3354/ab00258.
        • 20.Webster T, Dawson S, Slooten E. A simple laser photogrammetry technique for measuring Hector’s dolphins (Cephalorhynchus hectori) in the field. Mar Mammal Sci. 2010;26: 296–308. doi: 10.1111/j.1748-7692.2009.00326.x.
        • 21.Jeffreys GL, Rowat D, Marshall H, Brooks K. The development of robust morphometric indices from accurate and precise measurements of free-swimming whale sharks using laser photogrammetry. J Mar Biol Assoc United Kingdom. 2013;93: 309–320. doi: 10.1017/S0025315412001312.
        • 22.Gordon JCD. Evaluation of a method for determining the length of sperm whales (Physeter catodon) from their vocalizations. J Zool. 1991;224: 301–314. doi: 10.1111/j.1469-7998.1991.tb04807.x.
        • 23.Rhinelander MQ, Dawson SM. Measuring sperm whales from their clicks: stability of interpulse intervals and validation that they indicate whale length. J Acoust Soc Am. 2004;115: 1826–1831. doi: 10.1121/1.1689346. pmid:15101660
        • 24.Drouot V, Gannier A, Goold JC. Diving and feeding behaviour of sperm whales (Physeter macrocephalus) in the Northwestern Mediterranean Sea. Aquat Mamm. 2004;30: 419–426. doi: 10.1578/AM.30.3.2004.419.
        • 25.Western D, Moss C, Georgiadis N. Age estimation and population age structure of elephants from footprint dimensions. J Wildl Manage. 1983;47: 1192–1197. doi: 10.2307/3808191
        • 26.Norman BM, Stevens JD. Size and maturity status of the whale shark (Rhincodon typus) at Ningaloo Reef in Western Australia. Fish Res. 2007;84: 81–86. doi: 10.1016/j.fishres.2006.11.015.
        • 27.Mumby HS, Mar KU, Thitaram C, Courtiol A, Towiboon P, Min-oo Z, et al. Stress and body condition are associated with climate and demography in Asian elephants. Conserv Physiol. 2015;3: 1–14. doi: 10.1093/conphys/cov030.Introduction.
        • 28.Kolenosky GB, Lunn NJ, Greenwood CJ, Abraham KF. Estimating the weight of polar bears from body measurements. J Wildl Manage. 1989;53: 188–190. doi: 10.2307/3801329
        • 29.Smith RJ. Estimation of body mass in paleontology. J Hum Evol. 2002;43: 271–287. doi: 10.1006/jhev.2002.0573.
        • 30.Rising JD, Somers KM. The measurment of overall body size in birds. Auk. 1989;106: 666–674. Available: http://www.jstor.org/stable/4087673
        • 31.Coe MJ, Carr RD. The relationship between large ungulate body weight and faecal pellet weight. Afr J Ecol. 1983;21: 165–174. doi: 10.1111/j.1365-2028.1983.tb01178.x.
        • 32.Morgan BJ, Lee PC. Forest elephant stature in the Reserve de Faune du Petit Loango, Gabon. Afr J Ecol. 2007;45: 519–526. doi: 10.1111/j.1365-2028.2007.00762.x.
        • 33.Calzada J, Haydon DT, Palmares F. Estimating the size of European rabbits consumed by predators: Relationship between body mass and tooth dimensions. Acta Theriol (Warsz). 2003;48: 101–111. doi: 10.1007/BF03194270.
        • 34.Mumby HS, Chapman SN, Crawley JAH, Mar KU, Htut W, Thura Soe A, et al. Distinguishing between determinate and indeterminate growth in a long-lived mammal. BMC Evol Biol. 2015;15: 214. doi: 10.1186/s12862-015-0487-x. pmid:26464339
        • 35.Isaza R, Hunter RP. Drug delivery to captive Asian elephants—treating Goliath. Curr Drug Deliv. 2004;1: 291–298. doi: 10.2174/1567201043334641. pmid:16305391
        • 36.Kurt F, Garai ME. The Asian elephant in captivity: a field study. Cambridge: Cambridge University Press; 2007.
          • 37.Hile ME, Hintz HF, Erb HN, Hintz F. Predicting body weight from body measurements in Asian elephants (Elephas maximus). J Zoo Wildl Med. 1997;28: 424–427. pmid:9523637
          • 38.Sukumar R, Joshi N V., Krishnamurthy V. Growth in the Asian elephant. Proc Indian Acad Sci (Animal Sci.) 1988;97: 561–571. doi: 10.1007/BF03179558.
          • 39.Sreekumar KP, Nirmalan G. Estimation of body weight in Indian elephants (Elephas maximus indicus). Vet Res Commun. 1989;13: 3–9. doi: 10.1007/BF00366845. pmid:2773303
          • 40.Kanchanapangka S, Supawong S, Koedlab K, Kaewpannarai J, Khawnual P, Tummaruk P, et al. Body weight formulation in Asian elephant. Thai J Vet Med. 2003;37: 49–58.
          • 41.Wijeyamohan S, Sivakumar V, Read B, Schmitt D, Krishnakumar S, Santiapillai C. A simple technique to estimate linear body measurements of elephants. Curr Sci. 2012;102: 26–28.
          • 42.Shrader AM, Ferreira SM, McElveen ME, Lee PC, Moss CJ, van Aarde RJ. Growth and age determination of African savanna elephants. J Zool. 2006;270: 40–48. doi: 10.1111/j.1469-7998.2006.00108.x.
          • 43.Lee PC, Moss CJ. Statural growth in known age African elephants (Loxodonta africana). J Zool. 1995;236: 29–41. doi: 10.1111/j.1469-7998.1995.tb01782.x
          • 44.Choudhury A, Lahiri Choudhury DK, Desai A, Duckworth JW, Easa PS, Johnsingh AJT, et al. Elephas maximus. In: The IUCN Red List of Threatened Species. 2008 [cited 10 Jun 2015]. Available: www.iucnredlist.org
            • 45.Leimgruber P, Min Oo Z, Aung M, Kelly DS, Wemmer C, Senior B, et al. Current status of Asian elephants in Myanmar. Gajah 35. 2011;35: 76–86.
            • 46.Sukumar R. A brief review of the status, distribution and biology of wild Asian elephants Elephas maximus. Int Zoo Yearb. 2006;40: 1–8. doi: 10.1111/j.1748-1090.2006.00001.x.
            • 47.Leimgruber P, Senior B, Uga , Aung M, Songer MA, Mueller T, et al. Modeling population viability of captive elephants in Myanmar (Burma): Implications for wild populations. Anim Conserv. 2008;11: 198–205. doi: 10.1111/j.1469-1795.2008.00172.x.
            • 48.Gale Toke. Burmese timber elephants. Yangon, Burma: Trade Corporation; 1974.
              • 49.Zaw K. Utilization of Elephants in Timber Harvesting in Myanmar. Gajah. 1997;17: 9–22.
              • 50.Arivazhagan C, Sukumar R. Constructing age structures of Asian elephant populations: A comparison of two field methods of age estimation. Gajah. 2008;29: 11–16. Available: http://www.asesg.org/PDFfiles/2009/33121-Gajah-29.indd.pdf#page=15
              • 51.Mumby HS, Courtiol A, Mar KU, Lummaa V. Birth seasonality and calf mortality in a large population of Asian elephants. Ecol Evol. 2013;3: 3794–3803. doi: 10.1002/ece3.746. pmid:24198940
              • 52.Mumby HS, Courtiol A, Mar KU, Lummaa V. Climatic variation and age-specific survival in Asian elephants from Myanmar. Ecology. 2013;94: 1131–1141. doi: 10.1890/12-0834.1. pmid:23858653
              • 53.Hayward AD, Mar KU, Lahdenperä M, Lummaa V. Early reproductive investment, senescence and lifetime reproductive success in female Asian elephants. J Evol Biol. 2014;27: 772–783. doi: 10.1111/jeb.12350. pmid:24580655
              • 54.Mar KU. The demography and life history startegies of timber elephants in Myanmar. Changes. 2007.
              • 55.Rasband WS. ImageJ. Bethesda, Maryland, USA: U.S. National Institutes of Health; Available: http://imagej.nih.gov/ij/
                • 56.R Core Team. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2015. Available: http://www.r-project.org/
                  • 57.Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1: 307–310. doi: 10.1016/S0140-6736(86)90837-8. pmid:2868172
                  • 58.Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed-effects models using Eigen and S4. 2015. Available: https://cran.r-project.org/package=lme4
                    • 59.Weckerly FW, Leberg PL, Van Den Bussche RA. Variation of weight and chest girth in white-tailed deer. J Wildl Manage. 1987;51: 334–337. Available:http://www.jstor.org/stable/3801012
                    • 60.Kelsall JP, Telfer ES, Kingsley M. Relationship of bison weight to chest girth. J Wildl Manage. 1978;42: 659–661. doi: 10.2307/3800837
                    • 61.Payne N. Estimating live weight of black bears from chest girth measurements. J Wildl Manage. 1976;40: 167–169. Available: http://www.jstor.org/stable/3800173
                    • 62.Nagy JA, Kingsley MC, Russell RH, Pearson AM, Goski BC. Relationship of weight to chest girth in the grizzly bear. J Wildl Manage. 1984;48: 1439–1440. doi: 10.1002/jwmg.ll6.
                    • 63.Ramesh T, Kalle R, Sankar K, Qureshi Q. Role of body size in activity budgets of mammals in the Western Ghats of India. J Trop Ecol. 2015;31: 315–323. doi: 10.1017/S0266467415000188.
                    • 64.LaBarbera M. Analyzing body size as a factor in ecology and evolution. Annu Rev Ecol Syst. 1989;20: 97–117. doi: 10.1146/annurev.es.20.110189.000525.
                    • 65.Fernández MH, Vrba ES. Body size, biomic specialization and range size of African large mammals. J Biogeogr. 2005;32: 1243–1256. doi: 10.1111/j.1365-2699.2005.01270.x.
                    • 66.Heupel MR, Knip DM, Simpfendorfer CA, Dulvy NK. Sizing up the ecological role of sharks as predators. Mar Ecol Prog Ser. 2014;495: 291–298. doi: 10.3354/meps10597.
                    • 67.Gillooly JF, Brown JH, West GB, Savage VM, Eric L. Effects of size and temperature on metabolic rate. 2014;293: 2248–2251.
                      • 68.Johansson Ö, Malmsten J, Mishra C, Lkhagvajav P, McCarthy T. Reversible immobilization of free-ranging snow leopards (Panthera uncia) with a combination of Medetomidine and Tiletamine-Zolazepam. J Wildl Dis. 2013;49: 338–346. doi: 10.7589/2012-02-049. pmid:23568909
                      • 69.Hebert DM, Lay DW, Turnbull WG. Immobilization of coastal grizzly bears with etorphine hydrochloride. J Wildl Dis. 1980;16. doi: 10.7589/0090-3558-16.3.339
                      • 70.Renecker LA, Olsen CD. Antagonism of xylazine hydrochloride with yohimbine hydrochloride and 4-aminopyridine in captive wapiti. J Wildl Dis. 1986;22: 91–96. pmid:3005668 doi: 10.7589/0090-3558-22.1.91
                      • 71.Millspaugh JJ, Brundige GC, Jenks JA, Tyner CL, Hustead DR. Immobilization of Rocky Mountain elk with Telazol and xylazine hydrochloride, and antagonsim by yohimbine hydrochloride. J Wildl Dis. 1995;31. doi: 10.7589/0090-3558-31.2.259
                      Back to Top