Abstract:
Savannas once constituted the range of many species that human encroachment has now reduced to a fraction of their former distribution. Many survive only in protected areas. Poaching reduces the savanna elephant, even where protected, likely to the detriment of savanna ecosystems. While resources go into estimating elephant populations, an ecological benchmark by which to assess counts is lacking. Knowing how many elephants there are and how many poachers kill is important, but on their own, such data lack context. We collated savanna elephant count data from 73 protected areas across the continent estimated to hold ~50% of Africa’s elephants and extracted densities from 18 broadly stable population time series. We modeled these densities using primary productivity, water availability, and an index of poaching as predictors. We then used the model to predict stable densities given current conditions and poaching for all 73 populations. Next, to generate ecological benchmarks, we predicted such densities for a scenario of zero poaching. Where historical data are available, they corroborate or exceed benchmarks. According to recent counts, collectively, the 73 savanna elephant populations are at 75% of the size predicted based on current conditions and poaching levels. However, populations are at <25% of ecological benchmarks given a scenario of zero poaching (~967,000)—a total deficit of ~730,000 elephants. Populations in 30% of the 73 protected areas were <5% of their benchmarks, and the median current density as a percentage of ecological benchmark across protected areas was just 13%. The ecological context provided by these benchmark values, in conjunction with ongoing census projects, allow efficient targeting of conservation efforts.
Description:
S1 File. Additional methods and supporting results and references. This file contains Supporting
Information Materials and Methods; Figure A (study schematic), Figure B (23 time
series and best-fit population models); Figure C (partial residual plots of components of model
averaged GLM to explain PIKE values from MIKE sites); Figure D (comparison of GAMs to
explain extracted stable densities with and without influential points); Figure E (histogram of
ecological benchmark estimates for each of 73 protected areas illustrating results of Monte
Carlo simulation to incorporate uncertainty); Figure F (histogram of cumulative ecological
benchmark across 73 protected areas recalculated for each of 1x10^6 runs from the Monte Carlo simulation to incorporate uncertainty); Table A (alternative candidate models to
describe the population dynamics of 23 time-series populations); Table B (summary information
on time series and extracted stable density and SE); Table C (summary information for 43
MIKE sites and the explanatory variables used to explain PIKE); Table D (Summary statistics
and variables included in most likely quasi-binomial generalized linear models to explain
PIKE for 43 MIKE sites across Africa and final predictive average model used to generate
PIKE estimates for non-MIKE sites); Table E (selection parameters of candidate generalized
additive models explaining variation in extracted stable population size for 18 populations);
Table F (summary information, predicted stable density (give current PIKE), and ecological
benchmark density (given zero PIKE), and comparisons between most recent density and population
size estimates and ecological benchmark density and population size for 73 protected
areas across 21 countries); and Supporting Information References.