Study Area
We carried out this study in the buffer area of Panna Tiger Reserve,
Madhya Pradesh, India in April 2019. The dominant vegetation is dry
deciduous forests interspersed with grassy meadows. The region has a
tropical climate with mean summer temperatures between 24-40 °C. Summers
are dry with mean rainfall of 6 mm in April. The rugged land features of
the Park – plateaus and gorges – along with tall large trees provide
suitable nesting and roosting sites for several vulture species
(Red-Headed Vulture Sarcogyps calvus , White-rumped VultureGyps bengalensis , Long-billed Vulture Gyps indicus ,
Egyptian Vulture Neophron percnopterus , Griffon VultureGyps fulvus , Himalayan Griffon Gyps himalayensis , and
Cinereous Vulture Aegypius monachus ) that along with the striped
hyena (Hyaena hyaena ) are the obligate scavengers of this system.
The study area supports other meat-eating animals such as the tiger
(Panthera tigris tigris ), leopard (Panthera pardus ), sloth
bear (Melursus ursinus ), jungle cat (Felis chaus ), golden
jackal (Canis aureus ), wild pig (Sus scrofa ) and several
smaller mammals which also often scavenge. Herbivores which constitute
natural prey and carrion in the Park include spotted deer (Axis
axis ), Sambar (Rusa unicolor ), Nilgai (Boselaphus
tragocamelus ) Indian gazelle (Gazella bennettii ) and four-horned
antelope (Tetracerus quadricornis ), in addition to domestic
cattle (Bos taurus ).
Experimental Method
To observe the effects of different scavenging guilds (microbial,
invertebrate, and vertebrate) on carcass consumption and decay, we
conducted an experiment using fresh chicken carcasses (1-2 kg) (Figure
1). To control for other confounding factors that can influence carcass
persistence rates, we placed all carcasses in similar conditions –
under shade of dry deciduous vegetation with similar canopy cover and
horizontal cover (scrub/herbaceous layer density). The above conditions
ensured consistency in scavenger composition and similarity in carrion
discovery rates. The experimental control and treatments were as
follows:
- Control (for estimating moisture loss): Carcasses with guts removed to
prevent decay due to gut bacteria (n=3) were treated with
antimicrobial agent (Neosporin in powder form) on the outside and
inside of the body cavity, placed in a cloth sack and hung in a cage
to exclude invertebrate and vertebrate scavengers. These were weighed
daily to record moisture loss in the absence of microbial
decomposition.
- Microbial decomposition: Whole carcasses (n=3) were placed in cloth
sacks to exclude invertebrate scavengers, and placed in a wire cage to
exclude vertebrate scavengers. These were weighed daily to record loss
of biomass due to microbial decomposition as well as by desiccation.
- Invertebrate scavenging: Carcasses were placed inside wire mesh cage
on the ground with no cloth covering, to allow invertebrate scavengers
but to exclude vertebrate scavengers (n=6). These were weighed daily
to record invertebrate scavenging, microbe decomposition, as well as
moisture loss.
- Vertebrate scavenging: Carcasses were weighed and staked in the open
and monitored with camera traps to observe scavenging by vertebrate
scavengers (n=5). All carcasses were placed in the day time, before
noon, to give equal opportunity to avian and mammalian scavengers, as
mammals may also be active at night but avian scavengers were active
only in the day time. All carcasses were weighed at the time of
placement and then daily, to measure biomass loss due to scavenging by
vertebrates, invertebrates, microbes and by desiccation (figure 2).
Analytical Methods
Proportion of biomass remaining was calculated daily for each carcass.
To observe the effects of each ‘treatment’ (control, microbe,
invertebrate, vertebrate) on biomass removal, Linear Mixed Effect Models
(LMEs) were used to account for the correlated nature of the data,
arising from observing the same carcass over time. We modelled
logit-transformed (log-odds of) proportional biomass remaining over days
against treatment types as fixed effects, and the carcass replicate as
random effect. We modelled both additive (time + treatment) and
multiplicative (time x treatment) models to find the best fit. An
additive model implies that both time and treatment types have
independent effects on the loss of carrion biomass, and the resulting
graph would have parallel lines for each treatment type. However, a
multiplicative model points to an interactive effect between time and
treatment on the loss of carrion biomass, which would more closely
approach a natural system.
Logit-transformation allowed biomass persistence to be modelled as a
logistic function and constrained the predicted values between 0 and 1.
Proportion of carcass remaining on day 0 was considered as 0.9999, to
avoid the transformed value from becoming infinity. All analysis was
carried out in R (R Core Team, 2021). Inferences were based on the least
Akaike Information Criterion (AIC) model, which was used to predict
carcass persistence over time for different treatments.
We examined the effects (slopes) of each treatment on the decay rate of
carcass over time using ‘lstrends’ and ‘pairs’ functions in ‘emmeans’
package of R (Lenth, 2021). However, the effects thus estimated werecompounded effects as the treatments were nested. Vertebrate
scavenging included some carcass weight loss due to invertebrate
scavenging, microbial decomposition, and water loss; whereas,
invertebrate scavenging included some microbial decomposition and water
loss. Hence, to tease apart the individual effects of scavenger guilds
on carrion removal rates, we subtracted a) the daily mean carcass weight
loss due to evaporation (estimated from control experiment) from the
daily proportional carcass weight in the microbes treatment, b) the
daily mean weight loss due to microbial decomposition from the daily
proportional carcass weight in the invertebrate treatment, and c) daily
mean carcass weight loss in invertebrate treatment from the daily
proportional carcass weight in vertebrate treatments. The above
modelling exercise was repeated on these adjusted measurements, to
estimate individual effects of each treatment. Graphs were
generated using ‘ggplot2’ package in R (Wickham, 2009).