Long-term demography of spotted hyena (Crocuta crocuta) in a lion-depleted but prey-rich ecosystem
Data files
Mar 03, 2025 version files 1.02 MB
-
CH_HY_GLE_knowndeadexcl.csv
44.37 KB
-
CH_Monthly_HY_fordensity_allindividuals.csv
163.58 KB
-
Query_for_locs_for_KUD.csv
801.41 KB
-
README.md
6.56 KB
Abstract
Interspecific competition has strongly shaped the evolution of large carnivore guilds. In Africa, the lion (Panthera leo) and spotted hyena (Crocuta crocuta, hereafter hyena), exert direct and indirect competitive impacts on each other and on subordinate guild members. The impacts of competition on demography are complex and not well-understood. With carnivore guilds now ubiquitously impacted by humans, disentangling effects of interspecific competition and other drivers of hyena demography is important.
Western Zambia’s Greater Liuwa Ecosystem (GLE) provides a unique natural experiment where lions were functionally eliminated from the system. Hyenas are the apex predator, with an abundant prey base, and low levels of human-hyena conflict. We measured GLE hyena survival and density, using mark-recapture models fit to ten years of data from 663 known individuals in 11 clans.
GLE hyena densities were high, though slightly lower than other wildebeest-dominated systems, and stable over 10 years. Survival rates were high for all age-sex classes, and higher than other systems with high lion density, suggesting the possibility of competitive release from lion competition.
These findings provide insights into long-term hyena demography in the absence of their top competitor, but with an abundant prey base. As humans continue to alter ecosystems and fundamental ecological relationships such as interspecific competition, altered dynamics such as competitive release are likely to be widespread and should be a focus of future research.
https://doi.org/10.5061/dryad.612jm64d5
Description of the data and file structure
These data were collected to fit Cormack-Jolly-Seber (CJS) models for Bayesian statistical analysis for the estimation of age-sex-specific survival rates as well as time-trends in density of a Spotted hyena (Crocuta crocuta)* *population.
Data were collected year-round between June 2010 and November 2019 through a combination of stratified random sampling and opportunistic sightings. Sightings of individuals were recorded and, when possible, sex of the individuals was determined. Upon first sighting of a new individual, a Date of Birth (DOB) was estimated and an error-range relative to the certainty of the estimated DOB was assigned.
For the survival analysis, individual detection histories (excluding those that were confirmed to have died during the study period; n=13) were binned into 2-month occasions, generating a 57-occasion detection history for each individual. All individuals were divided into 3 age-classes (age = 0-0.99; age = 1-2.99; age = >3) and 3 sex classes (Female, Male, Unknown sex) and any changes of age-class, e.g. from cub to subadult, were included into the detection history.
CJS models were constructed to provide an estimated survival conditional on first detection, allowing for individuals to enter in a staggered way, with fixed effects of sex and age on survival rates.
For density estimates we fit closed mark-recapture models of hyena abundance in wet season (December - April) and dry season (May - November) per year. Each season's detection history was aggregated into 1-month time bins to create capture histories with 5 (wet season) and 7 (dry season) occasion encounter histories per individual per year. A model with individual heterogeneity effects on the detection probability (p) was fitted to each season's detection history to get a population size estimate per season. Density was estimated by dividing this population size estimate by the study area covered in that season. Study area was estimated as the 90th percentile isopleth of a kernel utilization distribution fit to all sighting locations of all individuals included in the density estimate for that season.
Files and variables
File: CH_HY_GLE_knowndeadexcl.csv
Description: Detection history per individual, plus relevant information about each individual (age and sex), used for age-sex dependent survival analysis.
Variables
- Nr: Number of individual in file
- ch: Capture history
- AnimalID: ID number of the individual
- Sex: Sex classes; Male, Female of Unknown
- DOB: Date of Birth, to assign age-class at each bin
- error DOB: Given error relative to how certain the DOB was upon assignment.
- CurrentGroup: Clan
- yearatfirst: Year of first sighting, For assignment of age-class in each time-bin
- monthatfirst: Month of first sighting, For assignment of age-class in each time-bin
- ageatfirst: Age of the individual when sighted first in years
File: CH_Monthly_HY_fordensity_allindividuals.csv
Description: Detection history per month for density estimate. Each row is one individual, each column defines data about that individual. X1 - X114 being 0's (non-detection) and 1's (detection).
Variables
- Nr: Number of individual in the file.
- X1 - X114: Months of the detection history, X1 being month 1 (= June 2010), X2 being month 2 (= July 2010), etc.
- clan: Clan the individual belongs to.
- sex: Sex of the individual.
File: Query for locs for KUD.csv
Description: Provides the exact location (coordinates) of each sighting of each individual and dates, used to generate 90th percentile isopleth Kernel Utilization Distribution per season per year to get study area size. The population size estimate per season per year is then divided by the study area size estimate to get density estimates per season per year.
Variables
- AnimalID: ID number of the individual
- SightingID: ID number of the sighting record
- SightingDate: Date at which the animal was sighted.
- Latitude of coordinate in WGS84, unprojected: Latitude of the coordinate of sighting
- Longitude of coordinate in WGS84, unprojected: Longitude of the coordinate of sighting
Code/software
All analyses were done in R version 4.2.2 - "Innocent and Trusting" but may be done in later versions as well.
The analyses and models are based on the book Bayesian Population Analysis Using WinBUGS by Marc Kéry and Michael Schaub (https://doi.org/10.1016/C2010-0-68368-4). We strongly suggest keeping this book at hand when going through these scripts to provide methodological clarification.
**Sex-age dependent survival: **
R-script "SA - CJS_models_survival-2.R" was used to fit the CJS models for age-sex dependent survival. The script contains all comments to be able to build the models. Included in the script are models to test for the effects of age or sex only.
To generate these models, installment of JAGS-4.3.0 (or later version) is required.
Packages that are required:
Lattice
coda
R2WinBUGS
R2jags
TeachingDemos
dplyr
lubridate
**Population size estimate per season **
R-script "SA - CJS_models_densityperseason.R" was used to fit the CJS models for population size estimate per season and per year. The script contains all comments to be able to build the models. Remove the comments on the relevant sections to get the estimates per season (e.g. go through the steps, then comment them out for the next season and remove the comments on the next season to generate the next population size estimate).
To generate these models, installment of JAGS-4.3.0 (or later version) is required.
Packages that are required:
Lattice
coda
R2WinBUGS
R2jags
ggplot2
ggjoy
dplyr
tidyverse
**90th percentile isopleth Kernel Utilization Distribution **
R-script "SA - KUD_period.wholepop.R" was used to generate 90th percentile isopleth Kernel Utilization Distribution to estimate study area size per season per year, based on each individual's locations. The KUDs were then visualized QGIS, but may also be displayed in any other GIS software or in R.
Packages that are required:
adehabitatHR
dplyr
purrr
reshape
readr
tidyr
plyr
tibble
lattice
coda
ggplot2
ggjoy
tidyverse
raster
rgdal
maptools
Over a time period of 9 years (2010 - 2019) of year-round monitoring, a combination of stratified random sampling and opportunistic sightings was used to observe 663 individually known hyenas. We binned sightings of each individual into 2-month bins and created capture histories for each individual.
We then used Bayesian statistical methods to fit Cormack-Jolly-Seber and closed-capture models of age-sex specific survival, reproduction and population densities per season.