Instantaneous energetics of puma kills reveal advantage of felid sneak attacks

See allHide authors and affiliations

Science  03 Oct 2014:
Vol. 346, Issue 6205, pp. 81-85
DOI: 10.1126/science.1254885

The costs and benefits of stalking and chasing

Organisms live under a constant balance between getting and using energy. Large carnivores may feel this balance more acutely because of the large amounts of energy needed to capture and subdue their prey. Williams et al. and Scantlebury et al. used remote measures of physiology and behavior to identify the hunting strategies of the stalking North American puma and the speedy African cheetah (see the Perspective by Laundré). In both cases the cats' hunting strategies are well matched to produce a balance between the energy they spend on the hunt and the energy they acquire from their prey, despite their very different strategies and levels of competition.

Science, this issue p. 81, p. 79; see also p. 33


Pumas (Puma concolor) live in diverse, often rugged, complex habitats. The energy they expend for hunting must account for this complexity but is difficult to measure for this and other large, cryptic carnivores. We developed and deployed a physiological SMART (species movement, acceleration, and radio tracking) collar that used accelerometry to continuously monitor energetics, movements, and behavior of free-ranging pumas. This felid species displayed marked individuality in predatory activities, ranging from low-cost sit-and-wait behaviors to constant movements with energetic costs averaging 2.3 times those predicted for running mammals. Pumas reduce these costs by remaining cryptic and precisely matching maximum pouncing force (overall dynamic body acceleration = 5.3 to 16.1g) to prey size. Such instantaneous energetics help to explain why most felids stalk and pounce, and their analysis represents a powerful approach for accurately forecasting resource demands required for survival by large, mobile predators.

A central tenet of foraging theory is that animals manage energetic costs and benefits when feeding (1). Yet measuring these costs for large, highly active predators that hunt, chase, and kill mobile prey has been exceedingly difficult, resulting in a poor understanding of how physiological capacities and environmental factors affect foraging success (2). This is especially apparent for species within the family Felidae. Among terrestrial carnivores, felids show a large range in body size, prey preferences, and predatory movements, each of which are linked to the landscape in which they live (36). The lankiest cat, the African cheetah (Acinonyx jubatus), engages in astonishing high-speed pursuits in open habitats to outmaneuver and overtake smaller, swift prey (7, 8). In contrast, heavy-bodied species including African lions [Panthera leo (9)], leopards (Panthera pardus), tigers (Panthera tigris), and pumas residing in forested or grassland habitats tend to stalk, ambush, and pounce to overpower prey up to several times their size (6).

Of the 36 extant wild felid species, the majority are considered cryptic ambush hunters (5, 6), which would suggest an energetic advantage for this predatory tactic (1). However, because of the covert nature of these activities, it is difficult to observe felid ambush behaviors or measure energetic efficiency. By necessity, field energetic studies of most terrestrial carnivores have relied on broad approaches, including doubly labeled water methods and field metabolic rate (FMR) modeling that integrate energetics across days or seasons [e.g., (1012)].

Here, we monitored behavior-specific energetics of a large cryptic felid, the puma, to evaluate the cost of discrete physiological states involved with ambush hunting. Using a laboratory-to-field approach, we developed and calibrated a new SMART (species movement, acceleration, and radio tracking) wildlife collar on captive, adult pumas (n = 3), with subsequent deployment on wild pumas (n = 4 pumas with SMART collars, 1 puma with GPS only; tables S1 and S2). We created a library of behavior-specific collar acceleration signatures by filming instrumented, trained pumas as they traversed a 20-m level course at different speeds, performed routine activities (resting, eating, grooming, etc.), and ran on a motorized treadmill (13). Accelerometer signatures were then correlated to energetic costs by simultaneously measuring oxygen consumption (Embedded Image), kinematics [stride frequency, stride length (14)], and overall dynamic body acceleration [ODBA (2, 15)] of the pumas during steady-state resting and treadmill running. Cost of transport (COT, the energy expended per meter) and cost of acceleration (COA, the energy expended per g) were calculated by dividing Embedded Image by speed and ODBA, respectively.

In the field, we captured and instrumented wild pumas (n = 2 females, 2 males) with the calibrated SMART collar. An additional female instrumented with a GPS collar allowed for a detailed analysis of hunting movements and longer deployment due to the conservation of collar batteries. All pumas were then released into a 1700-km2 study area within the Santa Cruz Mountains of California. The collar integrated a three-axis accelerometer and magnetometer that continuously monitored activity level and body position at 64 Hz, with a GPS/VHF unit with remote download capabilities that sampled every 4 hours to provide movement and kill locations (13). Potential kill sites were initially identified by the clustering of GPS positions indicating a feeding event, using a custom program integrated in ArcGIS (v.10; ESRI, 2010). Carcass visitation based on the GPS data allowed positive prey identification (13, 29). The timing of the GPS cluster was then correlated to time-synched accelerometer signals and behavioral signatures that indicated a pouncing event. Together, the suite of data from the collar allowed us to map both physiological and behavioral responses of wild pumas onto the physical landscape in which they hunted.

Our results indicate that the biomechanics and energetics of running by pumas are typical of most quadrupedal mammals including other adult felids (16). Both stride frequency and stride length of pumas increased linearly with running speed (Fig. 1A). Pumas also switched among walking, trotting, and running gaits at transition speeds predicted for their body mass (17). Routine preferred speed on the outdoor course and treadmill was 1.1 m s−1; maximum speed (4.9 m s−1) was only one-fifth of that measured for cheetahs chasing prey (7, 8).

Fig. 1 Stride mechanics and energetic costs of running pumas.

(A) Stride frequency and stride length in relation to outdoor speed (n = 3 pumas). Points represent individual walk (cyan), trot (blue), or run (red) trials. Dashed lines show predicted gait transition speeds (17). (B) Oxygen consumption in relation to speed for pumas on a treadmill. Points represent individual steady-state measurements. The least-squares regression (solid line, Eq. 1) is compared to basal metabolic rate [BMR, dashed line; (25)]. (C) Minimum cost of transport (COT) for canids [green circles (16)], pumas [black circles, present study and (30)], and other felids [domestic cat, black square; African lion, upward triangle; African cheetah, downward triangle (16)]. The green line is the allometric regression for COTMIN of quadrupeds (16).

The oxygen consumption (ml O2 kg−1 min−1) of pumas increased linearly with running speed (m s−1; Fig. 1B) according toEmbedded Image (1)Minimum cost of transport (COTMIN) was 0.17 ml O2 kg−1 m−1 (3.42 J kg−1 m−1), as predicted for canids and other felids on the basis of body mass (Fig. 1C) (16). This includes large and small felids that stalk, pounce, and perform high-speed chases. Immature African lions, however, are outliers with a COTMIN that is 2.4 times that predicted for running mammals (18) and 2.1 times the value for pumas. If adults follow the trend for immature lions, then comparatively high locomotory costs of African lions may help to explain the tendency of this species to rely on cooperative hunting, which is unique among felids (6).

Continuous monitoring of acceleration in wild pumas wearing SMART collars allowed us to determine how these costs vary with hunting across time and space for individuals. Figure 2A shows a typical GPS track on the day of a kill for a 42-kg female puma hunting black-tailed deer (Odocoileus hemionus). The cat moved downhill through a residential area, stalked and killed a deer, and stayed within 200 m of the carcass until moving into the mountains the next day. What is not apparent in such GPS tracks is the considerable variation in behaviors and energy expended by a puma when pursuing different-sized prey. Here the continuous, high-resolution time-stamped accelerometer traces from the SMART collar provide remarkable insight. First, unique accelerometer signatures within the traces reveal individual behaviors that can be localized via GPS on the terrain where they occur (Fig. 2A). Second, raw triaxial acceleration traces (Fig. 2A, bottom) can be collapsed into overall dynamic body acceleration (ODBA) values and then transformed into activity-specific metabolic costs in terms of Embedded Image (ml O2 kg−1 min−1) using the regression Embedded Image (2)from the instrumented pumas walking on a treadmill (Fig. 2B) (13).

Fig. 2 Field movements, acceleration, and energetic costs for pumas.

(A) GPS track for female 13F preying on deer (Google Earth, Minimum distances moved, time of day, and kill time are indicated. Locations of behavior-specific acceleration signatures (n = 3840) (X-surge, red; Y-sway, cyan; Z-heave, black; from collar orientation, inset) are superimposed on the GPS track with corresponding color-coded ODBA (g) and Embedded Image (ml O2 kg−1 min−1). (B) Relationship between Embedded Image and ODBA for instrumented pumas on a treadmill. Points are mean steady-state measurements ± 1 SE for two pumas. See text for regression (black line) and COA (inset) statistics. (C) Energetic cost of a pounce and kill by pumas in relation to deer mass. Points represent separate kills defined by pounce peak and immediate prey handling (Fig. 3 and table S3). The linear regression (black line, Eq. 4) for five known kills (black symbols) is extrapolated (red dashed line) to predict the size of an unidentified kill according to pounce energetics (red symbol).

Equation 2 allows the acceleration signatures of wild pumas to be assigned to the energetic cost associated with each behavior. As reported for smaller animals and humans (15, 19), pumas show a significant, linear correlation between oxygen consumption and ODBA. Like COT (16), COA declines asymptotically with ODBA (Fig. 2B, inset), showing a decline in the variation of acceleration costs as activity level increases (i.e., <10% change in COA with ODBA from 0.5 to 3.0 g). The lowest COA (~58 ml O2 kg−1 min−1/g ODBA) occurs at the highest ODBA and is equivalent to the slope of the relationship in Eq. 2. Together, these relationships enable energetic costs to be ascribed to a wide variety of aerobic activities as long as ODBA accurately reflects the puma’s movements (2, 15, 19).

Using these methods, we assigned energetic costs for pumas during three characteristic periods: (i) pre-kill activity (locating prey), (ii) pounce and kill, and (iii) post-kill prey handling and eating. These periods constitute a characteristic 2-hour interval we term the “hunt” (Fig. 2 and Fig. 3). Activity levels during the pre-kill phase varied markedly among individuals, ranging from activities with low energetic cost [sit-and-wait (male puma 5M) and slow-walking (female 2F) behaviors] to those with moderate cost [coursing-type movements (16M) and stalk-and-ambush activities (7F)] (Fig. 3). Total energy expended during the pre-kill period (Hunting CostPRE-KILL in kJ) decreased with increased use of cryptic behaviors according to Embedded Image (3)(Fig. 4A). Here, percent time cryptic represents the proportion of the pre-kill period comprising all low-acceleration, low-cost activities [including rest, slow-walking (e.g., stalking), and sit-and-wait behaviors]. Overall, pre-kill hunting costs represented 10 to 20% of the estimated FMR of pumas (13).

Fig. 3 Activity levels and energetic costs determined from ODBA of wild pumas.

Each panel represents a separate kill identified by puma ID and by size and age of the deer. Two-hour hunt segments, including pre-kill, pounce spike, and post-kill prey handling periods, are compared for pumas 2F, 5M, 7F, and 16M. The red dashed line indicates predicted Embedded Image (20). Colored bars denote speed and behavior according to the inset scale. Low-cost periods (blue bars) were used to calculate percentage of time spent in cryptic activities for each puma.

Fig. 4 Effects of behavior and body mass on energetic costs of hunting carnivores.

(A) Total energy expended during the pre-kill periods in Fig. 3 in relation to percentage of time pumas spent performing low–activity level (cryptic) behaviors. See text for regression statistics. (B) Minimum (green line from Fig. 1), predicted [black line (14)], and measured (red line, Eq. 5) transport costs for hunting canids (squares) and felids (male and female pumas in this study, circles; male and female tigers, triangles). See supplementary materials for data sources.

Pumas generally remained aerobic during the hunt, only briefly exceeding their predicted Embedded Image [49.2 to 54.7 ml O2 kg−1 min−1 (20)] when pouncing. This high-energy activity, along with killing and prey handling, resulted in different acceleration signatures for pumas preying on fawns or bucks (Fig. 3). Predator costs (kJ) are related to prey mass (kg) asEmbedded Image(4)(Fig. 2C). Extrapolating from this relationship, we estimated the size of an animal that had been attacked, despite being unable to view the event or carcass. Although some prey may be harder to kill and may require more handling than others, pumas appear to precisely gauge the magnitude of the initial pounce to account for the size of the animal to be overtaken (table S3).

We find that the investment of energy for traveling by large terrestrial carnivores is considerable and often underestimated. Comparing pre-kill transport costs for hunting pumas (COTHUNT in J kg−1 m−1) to those calculated for foxes (11, 12), wild dogs (10), and tigers (21) results in the allometric regressionEmbedded Image (5)(Fig. 4B) (13). Energetic demands for pre-kill movements by these large mammalian carnivores average 2.3 times the levels routinely used to model energetic costs from distances moved or GPS tracks (2123) and 3.8 times the predicted COTMIN (16, 24). These elevated costs undoubtedly reflect the metabolic demands of carnivory (16, 25) as well as the SMART collar’s ability to account for previously overlooked energetic expenditures associated with intermittent locomotion, turning maneuvers, and kinematic changes due to uneven or variable substrates (2628) that are often imperceptible in most GPS tracks (13). Pumas can mitigate high hunting costs by matching maximum pouncing force to prey size and using cryptic tactics. With such an energetic advantage, it is not surprising that felid morphology and physiology have been shaped over evolutionary time for the stalk-and-pounce sneak attack (5, 6).

Although our study highlights one energetically expensive activity (hunting) for a single carnivore, there are more than 245 mammalian carnivore species ( and numerous omnivores that pursue and kill mobile, elusive prey. Most have specialized hunting styles and would be expected to display highly variable instantaneous rates of energy expenditure that differ widely among predator and prey, habitats, and even individuals. The laboratory-to-field approach used here, which pairs basic physiological attributes of the animal with standard wildlife monitoring, provides a powerful way of quantifying such species-specific energetic demands. Some activities, including hunting and locomotion across complex habitats, are energetically costly; most are woefully underestimated by many predictive algorithms in common use. Correcting this will become increasingly important for the preservation of large carnivores as foraging becomes progressively affected by continued habitat degradation and loss of vegetative cover essential for cryptic movements (4, 29).

Supplementary Materials

Materials and Methods

Tables S1 to S3

References (3138)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: Supported by NSF grants DBI-0963022 and 1255913 (T.M.W., C.C.W., and G.H.E.) with in-kind support from the California Department of Fish and Wildlife. We thank P. Houghtaling, Y. Shakeri, C. Wylie, and D. Tichenor for assistance in catching wild pumas and finding kill sites, and J. A. Estes and the anonymous referees for critical review of this manuscript. Animal procedures were approved by the UCSC IACUC. Statistical data are tabulated in the supplementary materials, with electronic versions available upon request from the senior author through the UCSC Mammal Physiology Project database.
View Abstract

Stay Connected to Science

Navigate This Article