Estimating the Body Weight of Florida Panthers From Standard Morphometric Measurements

Body weight is an important variable in studies of mammalian reproductive biology (Blueweis et al. 1978; Clutton-Brock 1988; Weckerly 1998) and ecology (Clutton-Brock and Harvey 1983; Peters 1983; Lindstedt et al. 1986). In pumas Puma concolor Linnaeus, also regionally known as cougar, mountain lion and panther, body weight is a particularly important measure of an individual's health and serves as an indicator of spatial variation in food habits (Iriarte et al. 1990), growth (Maehr and Moore 1992; Laundré and Hernández 2002; Bartareau et al. 2013), prey requirements (Ackerman et al. 1986; Laundré 2005), and has a pivotal role in reproduction and survival. For example, female reproductive maturity depends on when an individual attains a critical minimum body weight (Robinette et al. 1961), and an individual's reproductive success is correlated with energetic investment in gestation and rearing offspring that would benefit from increased energy stores in amassed adipose tissue (Gittleman and Thompson 1988). In addition, larger females may be better able to protect their offspring and deterring infanticide is an important part of early puma survival. Male reproductive success is dependent on the ability to thwart competitors (Maehr et al. 1991), so heavier males should be more proficient in fighting ability than lower-weight males and would be able to obtain and defend a territory, allowing access to females in estrus. Furthermore, intraspecific aggression is the foremost natural cause of mortality (Taylor et al. 2002; Hornocker et al. 2010), and larger body weight most likely reduces casualty risk, thereby conferring an advantage to survivorship and reproduction. Collecting body weight measurement is therefore recommended during handling because demographic and reproductive variables are functionally dependent on weight rather than age, which has important implications for assessing trends in population health.

The puma is highly vagile and measuring body weight of free-ranging individuals is difficult. The ability to directly measure body weight of pumas in the field using calibrated scales can be impractical when equipment needed to weigh individuals is inadequate or unavailable (e.g., in large-bodied animals, in remote localities). As an alternative method, Jansen and Jenks (2011) created a body-weight estimation model for pumas in South Dakota and Wyoming based on the relationship between known measures of body weight and body length, head girth, and chest girth morphometric measurement predictor variables that are obtainable in the field. Such body-weight estimation models with multiple predictor variables are generally more accurate than any single-variable model because multiple morphometric measurements describe body shape better than a single measurement alone (Cattet 1990; Bartareau 2017). However, a limitation of the current model for retrospectively estimating the body weight of a puma in other studies or subpopulations is that head girth has not been a standard morphometric measurement from pumas (Jansen and Jenks 2011). In addition, sexual and interpopulation variation in puma growth rates (Maehr and Moore 1992; Laundré and Hernández 2002; Bartareau et al. 2013) and adult size (Kurtén 1973; Iriarte et al. 1990) might affect the body-weight-shape relationship and accuracy of body-weight estimation model. Thus, the body-weight–morphometric measurement relationship of pumas in a given subpopulation cannot be assumed to be the same as another and it may be necessary to construct and validate an explicit body-weight estimation model from available predictor variables for different populations.

My objective here was to develop a model to accurately estimate the body weight of Florida panthers Puma concolor coryi Bang based on the relationship between scale weight, sex, and standard morphometric measurement predictor variables easily obtainable in the field. This ability will be useful to personnel conducting field research when scale weight is impractical to obtain. It will also enable retrospective estimates of the body weight of panthers in cases where sex and standard morphometric measurements are available but the individuals were not weighed.

Personnel from the Florida Fish and Wildlife Conservation Commission and National Park Service captured and handled panthers of all sex and age classes spanning the full geographic range of the species in Florida from 1982 to 2012. They used trained hounds to locate individuals and captured the panthers using immobilizing drugs delivered via remote injection. The capture and handling procedures followed a standard data collection protocol that was approved by U.S. Fish and Wildlife Service (i.e., Endangered Species Collection Permit TE051553-4 and TE146761-2) and is consistent with that of Kreeger (1996) and Gannon and Sikes (2007). For a detailed description of the study area, and capture and handling method, see Florida Fish and Wildlife Conservation Commission (2012) and Jansen et al. (2013).

While immobilized, the field personnel measured the scale weight, body length, tail length, chest girth, and neck girth on 46 females and 68 males (Table S1, Supplemental Material). They measured body weight to the nearest ounce or pound using a calibrated hanging spring scale, and I converted weights to kilograms. As the panther was lying on its side, they measured body length to the nearest centimeter using a nonstretchable tape measure pulled tight along the contour of the spine from the distal hairline on the nose to the junction of the sacral and caudal vertebrae. They measured tail length to the nearest centimeter from the junction of the sacral and caudal vertebrae to the distal end of the final caudal vertebra. They measured total length as the sum of body length and tail length. They measured chest girth to the nearest centimeter at the largest circumference of the thorax, and measured neck girth to the nearest centimeter at the smallest circumference in the area closest to the head. I did not include age or age classes in the analyses because estimates of age as measured in the field by tooth eruption and wear (Ashman and Greer 1976) or counts of cementum annuli in an extracted tooth (Matson 1981) were too imprecise and unreliable. Furthermore, Jansen and Jenks (2011) found that the predictor variable of age class did not contribute significantly to their predictive model.

To assess how accurately the best-fitted model in the set would perform in practice to the population the sample that was chosen from, I created candidate models using a model data set with 70% of the data (n = 32 females and 47 males) and a validation data set that included the remaining 30% of the data (Snee 1977). I split the data randomly by weight so that a wide range of weights was present in each data set and the few small and large panthers would not bias the models. I compared the observed body weights to the estimated values and used the residual for descriptive statistics.

I used t-tests to compare differences between sexes in the means of morphometric measurements (Zar 1999). I assessed the degree of linear association between body weight and the morphometric predictor variables with Pearson correlation (r) matrix. I fitted body weights and predictor variables to simple and multiple linear regression body-weight estimation models using the Marquardt-Levenberg least squares method (Statistix 9.0, Analytical Software, Tallahassee, FL). I used corrected Akaike's information criterion (AIC_c) and Akaike weight (w_i) to evaluate the suitability of candidate models in the set based on a balance of model fit and accuracy of estimates (Burnham and Anderson 2002; Whittingham et al 2006; Gergely and Garamszegi 2011). I used Shapiro-Wilk (W) tests to determine whether the residuals conformed to a normal distribution and that they met this assumption for linear regression (Zar 1999). I used the coefficient of determination (R²) value to evaluate the amount of variation explained by the models (Zar 1999). I assessed the accuracy of models as applied to the validation data set through the mean residuals (i.e., difference between estimated and observed body weights), standard deviation of the residuals (SD), and 95% confidence interval on the bias of the estimated body weight. I used paired t-tests to compare differences between observed weights and estimated weights for the best model (Zar 1999). All values are presented as mean ± standard error (SE).

Total length and body length were significantly (P < 0.001) correlated in both sexes (female: r = 0.977, n = 32; male: r = 0.945, n = 47). Because the tail in some individuals was kinked or injured and represented little mass relative to body weight, I excluded tail length from the analyses (sensu Jansen and Jenks 2011). I fitted the body weights and remaining predictor variables to linear regression models of the form Y = B₀ + B₁X₁ + B₂X₂ + … + B_nX_n + e, where Y is the dependent variable body weight (kg), B₀ is the intercept, B_1…n is the coefficient of independent variable, X_1…n is the independent variable, and e is the random error. The predictor variables for model building included sex (S = 0 female and 1 male), body length (L = cm), chest girth (G = cm), chest girth² (G² = cm²), and neck girth (N = cm). Body length reflected an individual's skeletal growth and relative maturity (Bartareau et al. 2013). Chest and neck girth indicated tissue growth and nutritional condition (Marta et al. 2014). I added the chest girth² predictor variable to more accurately model the potential effect of chest girth, which may have a nonlinear relationship with body weight. I initially created sex-specific models because growth curves show large differences between the sexes in the rate of body size gain and mature weight (Bartareau et al. 2013). I tested sex differences in model intercept and coefficient of predictor variables and combined the models if they did not differ (Zar 1999).

The sample of panthers included dependent kittens ≤ 0.25 y (n = 4 females and 2 males), juveniles 0.25–1 y (n = 10 females and 16 males), and adults 1–8 y (n = 32 females and 50 males) that fully represented the population in Florida by sex and standard morphometric measurements (Table 1). Morphometric measurements from the females were more variable than those from males. Body weight was the most variable morphometric measurement between sexes followed by body length, chest girth, and neck girth. Body weight, chest girth, and neck girth of males was significantly (P < 0.001) larger than in females. Body length was homogenous between sexes (P = 0.064). Pearson correlation coefficients indicate that all morphometric measurements were significantly (P < 0.001) correlated with each other (Table 2). Body weight showed the strongest correlation with chest girth², followed by chest girth, and neck girth. Body weight was also correlated with body length, but more so in females than males.

Confidence limit tests revealed that none of the coefficients of the female or male subsample models differed significantly (P > 0.05) from its combined sample counterpart. For this reason, the single combined sample model can stand as the optimum estimator for its sample. All top models (Table 3) explained a similar amount of variation (R² = 0.936–0.937) as well as degree of dispersion (SD = 4.11–4.15). The R² value indicated that chest girth² explained the greatest amount of variation in body weight (R² = 0.908). The three top models included body length, chest girth², and neck girth measurements. I chose the model with body length, chest girth², and neck girth measurements as the best because it was better supported than any other multiple predictor variable modeled relationships with lowest AIC_c (AIC_c = 231.2), highest Akaike weight (w_i = 0.36) and R² (R² = 0.937), and was the most parsimonious.

The regression analysis results indicate that all morphometric measurement betas were significantly different (P ≤ 0.003) from zero if all other variables are already in the model (Table 4). My best model to estimate the body weight of panthers based on the relationship with standard morphometric measurement predictor variables was as follows:

The measured and estimated body weights were highly correlated in the validation model set (r = 0.962, P < 0.001). Residuals conformed to a normal distribution in both the model data set (W = 0.989, n = 79, P = 0.721) and validation data set (W = 0.992, n = 35, P = 0.994), indicating constancy of variance. Chest girth² had the strongest relationship to body weight according to the t value. The differences between the observed and estimated body weight was not significant in the model (t₇₈ = 0.2, P = 0.877) and validation data set (t₃₄ = −0.7, P = 0.497). The mean residual was 0.1 ± 0.5 kg in model dataset and −0.5 ± 0.7 kg in the validation data set. The 95% confidence interval on the bias of the estimated body weight ranged from −0.8 to 0.9 kg in model data set and from −1.1 to 0.9 kg in the validation data set.

It is more difficult to directly measure a puma's body weight than other standard morphometric measurements so researchers have undertaken estimation with scale weight-morphometric measurement predictor variable relationship using a multiple regression model (Jansen and Jenks 2011). An important question is whether the results of a modeled relationship on the model data set can be extended to the population the sample has been chosen from. Previously, researchers overlooked model validation and accuracy consisted of quoting the R² and 95% confidence interval on the bias of the estimated body weight in the model data set. The application of a modeled relationship without independent validation of the obtained model could result in biased parameter estimates and inaccurate inference on new subjects (Snee 1977; Whittingham et al 2006; Gergely and Garamszegi 2011). By using an information-theoretic approach to evaluate competing models with a model data set and then comparing observed and estimated body weight measurements through a validation data set, I have shown that body weight of Florida panthers can be accurately estimated based on the relationship between scale weight and standard morphometric measurement predictor variables obtained in the field. I found that a model including body length, chest girth², and neck girth predictor variables was better supported than any other modeled relationships in the model data set, and the relationship was consistent between the sexes.

Jansen and Jenks (2011) concluded that sex information did not contribute significantly to their best body-weight estimation model for pumas in a South Dakota and Wyoming subpopulation. Likewise, my findings indicated that the body-weight predictor variable modeled relationship for Florida panthers was consistent over the range of the standard morphometric measurements using one multiple regression model for both sexes. In my sample, panthers of a given body weight differed in body length, chest girth², and neck girth measurements, yet the modeled relationship indicates that the relative proportions of these morphological variables varied predictably as body weight increased or decreased in both sexes. Body length was the worst single predictor of body weight, indicating that it was the least sensitive model parameter. Chest girth² was the best single predictor of body weight followed by neck girth. This finding suggests that body weights were heavier with increasing body length, neck girth, and chest girth, though at larger chest girths weights increased at a greater rate. The thickness of the subcutaneous muscle and adipose layers should increase and decrease as a panther's body weight is gained or lost and these multilayered tissue deposits appear to be tightly correlated with both neck girth and chest girth.

My body-weight estimation model explained 94% of the variation in body weight of panthers in the model data set and correctly predicted weights to within −0.5 ± 0.7 kg of the scale weight in the validation data set. Pumas are known to consume up to 10 kg of prey in a single feeding period (Ackerman et al. 1986) and researchers have documented that they contain food material in their stomachs 30–79% of the time (Robinette et al. 1959; Thompson et al. 2009). Thus, the body weight of pumas may vary temporally independently of body-weight–morphological measurement relationships in relation to the time since feeding and how much food was consumed. Although further studies are needed to determine to what degree panther morphometric measurements may be affected by stomach fullness, my body-weight estimation model estimates were very close to scale weight. The 95% confidence interval on the bias of estimated body weight in my best model ranged from −1.1 to 0.9 kg and this is well below the potential temporal variation in stomach contents expected when weighing Florida panthers at random capture events.

My results differed from those of Jansen and Jenks (2011), whose predictive model for estimated body weight included body length, head girth, and chest girth predictor variables. Their model accounted for 89% of the variation in body weight of pumas and the 95% confidence interval on the bias of the estimated body weight in the model data set was −6.3 to 6.3 kg. I was unable to evaluate the efficacy of head girth predictor variable because it has not been a standard morphometric measurement from Florida panthers. It is possible incorporating head girth morphometric measurement might have resulted in a body-weight estimation model more similar to that of Jansen and Jenks (2011). However, the R² of my model was slightly higher and the predictive accuracy much better without utilizing a head girth measurement and instead incorporating the chest girth² predictor variable. Similarly, accurate body-weight estimation has been practiced through chest girth² measurements using simple and multiple regression models in many black bear (Ursus americanus) populations and researchers found it to be the best single predictor of weights in both sexes (Swenson et al. 1987; Bartareau 2017). Thus, differences in body-weight estimation model performance may be attributable to variation between models in constituent morphological measurement predictor variables.

Whether the relationship between body weight and standard morphometric measurement predictor variables, and thus the resultant body-weight estimation model, differs between distinct puma populations is unknown. Because of differences in genotype (Culver et al. 2000), food habit (Iriarte et al. 1990; Maehr et al. 1990; Thompson et al. 2009), and climate (Gay and Best 1996) regional variations may exist in the physical proportions of bodies of pumas that contribute to the differences in body-weight estimation models. The broad similarity of body shape and growth patterns (Maehr and Moore 1992; Laundré and Hernández 2002; Bartareau et al. 2013) among different puma subpopulations suggests that, while the coefficients in the model for different geographic regions may vary (i.e., the rate of change in allometric relationships), body length, chest girth², and neck girth morphometric measurements are most likely the best predictor variables of body weight in each case. These three morphometric predictor variables of body weight are standard morphological measurements, easy to measure in the field, and replicable among individual personnel.

It might be possible to develop a reliable body-weight estimation model applicable to the puma species as a whole incorporating standard morphological measurement predictor variable data from many different subpopulations. However, the body-weight–morphological measurement relationships may change over time (Cattet and Obbard 2005). Ideally, body-weight–morphometric measurement relationships should be validated periodically to ensure that the bias has not changed substantially and, if so, to modify the current body-weight estimation model. I recommend that personnel handling pumas measure the scale weight whenever possible and record the body length, chest girth, and neck girth for all individuals handled. The availability of observed and estimated body-weight measurements from the same pumas would allow routine assessment of the accuracy of the current body-weight estimation model. Furthermore, though a scale weight is required for body-weight estimation model calibration, it is not sufficient for model validation. Validation of the obtained model with an independent data set is required to evaluate the accuracy of a putative body-weight estimation model.

Many panthers are handled annually in Florida, but due to their large size they are not always weighed because the personnel often do not have the equipment needed to weigh individuals in the field. Without the need for specialized equipment, my body-weight estimation model can be used by personnel involved in Florida panther research and management. Although direct measurement using a calibrated scale is preferable to estimates using a body-weight–morphometric measurement predictor variable relationship, the validated model presented should prove useful to provide more body-weight data from panthers handled in Florida that are not weighed with a scale and the estimate will be more accurate than visual estimates.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the author.

Table S1. Data file (.xls) of date, sex, body weight (kg), chest girth (cm), chest girth² (cm²), neck girth (cm), body length (cm), tail length (cm), total length (cm), and data set of 46 female and 68 male Florida panthers (Puma concolor coryi) captured 1982–2012.

Found at DOI: http://dx.doi.org/10.3996/042017-JFWM-036.S1 (38 KB XLS).

Reference S1. Florida Fish and Wildlife Conservation Commission. 2012. Annual report on the research and management of Florida panthers: 2011–2012. Fish and Wildlife Research Institute and Division of Habitat and Species Conservation. Florida Fish and Wildlife Conservation Commission.

Found at DOI: http://dx.doi.org/10.3996/042017-JFWM-036.S2 (1636 KB PDF).

Reference S2. Jansen D, Kellam J, Johnson A. 2013. Florida panther (Puma concolor coryi) research and monitoring in Big Cypress National Preserve 2011–2012 annual report. Ochopee: Florida. National Park Service.

Found at DOI: http://dx.doi.org/10.3996/042017-JFWM-036.S3 (4570 KB PDF).

This study was made possible thanks to many people associated with the Florida Fish and Wildlife Conservation Commission and the National Park Service who assisted with data collection. Thanks to Dr Walter Meshaka, the Editor-In-Chief, the Associate Editor, and two anonymous reviewers for constructive comments that improved the manuscript, and to Linda Pulliam for diligence in sourcing literature.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.