Skip to main content
Dryad logo

Bushmeat yields, species extinction rates and ecosystem-level impacts of bushmeat harvesting as predicted by the Madingley General Ecosystem Model


Barychka, Tatsiana; Mace, Georgina; Purves, Drew (2021), Bushmeat yields, species extinction rates and ecosystem-level impacts of bushmeat harvesting as predicted by the Madingley General Ecosystem Model, Dryad, Dataset,


The datasets contain data generated using the Madingley General Ecosystem Model for experiments decribed in the paper: "T. Barychka, G.M.Mace and D.W.Purves (2021) The Madingley General Ecosystem Model predicts bushmeat yields, species extinction rates and ecosystem-level impacts of bushmeat harvesting. Oikos."

The Madingley General Ecosystem Model was used to generate predictions of bushmeat yields, extinction rates and broader ecosystem impacts for a range of harvesting intensities of duiker-sized endothermic herbivores. Duiker antelope (such as Cephalophus callipygus and Cephalophus dorsalis) are the most heavily hunted species in sub-Saharan Africa, contributing 34%-95% of all bushmeat in the Congo Basin. In the Madingley, the harvested group was described as "Heterotroph – Herbivore – Terrestrial – Mobile – Iteroparous– Endotherm", with adult bodymasses of 13-21 kg and juvenile bodymasses of >100 g. Harvesting period was set at 30 years (=30).

In the first experiment, we used the Madingley model to predict bushmeat yields ("Harvested Biomasses") and extinction rates ("Density") from harvesting duiker-sized herbivores using proportional harvesting strategy, with harvest rates ranging from 0 to 0.90. These were compared to the estimates of bushmeat yields and survival probabilities for two duiker antelope species (Cephalophus callipygus and Cephalophus dorsalis) from conventional single-species Beverton-Holt model. 

In the second experiment, we used the Madingley model to generate data on the state ("State") of the harvested ecosystem. The datasets contain estimates of biomasses, abundances, adult and juvenile bodymasses, etc. of the harvested duiker-sized herbivores as well as unharvested herbivores, omnivores and carnivores present in the simulated ecosystem. In the paper we focused on abundances; however, other estimates e.g., adult bodymasses can be used to conduct further studies on the effects of harvesting in tropical ecosystems.

Main results of the experiments are that: 1) the Madingley model gave estimates for optimal harvesting rate, and extinction rate, that were qualitatively and quantitatively similar to the estimates from conventional single-species Beverton-Holt model; 2) the Madingley model predicted a background local extinction probability for the target species of at least 10%; 3) at medium and high levels of harvesting of duiker-sized herbivores, the Madingley model predicted  statistically significant, but moderate, reductions in the densities of the targeted functional group; increases in small-bodied herbivores; decreases in large-bodied carnivores; and minimal ecosystem-level impacts overall.



In the Madingley model, a 100-year 'burn-in' (no harvesting, 30 replicates) followed by a 30-year harvesting period was run to produce estimates of the following: for each functional group (carnivore/omnivore/herbivore), the number of surviving animal cohorts, abundances, biomass and adult body masses. 
We used a constant proportional harvesting policy, where each year a proportion of animals were targeted. This harvest rate remained constant for the duration of harvesting period. Experiments were replicated 30 times at each harvest rate, with harvest rate ranging from 0 (no harvest) to 0.90 in discrete steps of 0.05 (smaller steps for survival probability). Harvesting was applied from year 1 onwards (no harvesting took place in year 0).

Target Group

We simulated harvesting strategies for animals similar to duiker antelope Cephalophus callipygus and Cephalophus dorsalis. We set up harvesting in the Madingley model to target terrestrial herbivorous endotherms, described using the following categorical traits: ‘Heterotroph – Herbivore – Terrestrial – Mobile – Iteroparous– Endotherm’. This definition was further narrowed using two continuous traits: adult body mass (13-21 kg) and juvenile body mass (>100 g).


Our experimental site was simulated in the Madingley model on a 1° × 1° geographic grid cell (111.32 × 110.57 km) centred on 1°S, 15°E; the coordinates were selected to fall within the known duiker range in the tropical forests of the Republic of Congo. For the purposes of this study, no inter-cell migration was modelled, i.e. no animals were allowed from outside the experimental area.


The rate of extinction of duiker-like animals in the Madingley model was estimated at each time step. Extinction was defined when the total density of animals that matched the definition of duiker-like fell below 0.1 animals km−2 during  a simulation run.

Ecosystem response

The ecosystem-level information was recorded in the Madingley at each time step, such as, for each functional group, adult body masses, animal biomasses and
abundances. Overall, the ecosystem-level response to harvesting was analysed as follows. First, each cohort’s functional group was identified as being a herbivore, omnivore or carnivore. Individuals were also allocated into a body mass bin ranging from the smallest body mass (10^−3 to 10^−2 g) to the largest bin (10^6–10^7 g). Total abundances were then calculated for each functional group in each body mass bin and logged (on log10 scale). Changes in abundances were calculated by subtracting total animal abundances under one of the four harvesting regimes (20%, 50%, 70% and 90% of duiker-like population harvested) to total animal abundances without harvesting.

All data processing and analysis were carried out in R ver. 3.6.3.

Usage Notes

For Yields use 'HarvestedBiomasses' data.

For Survival Probability use 'Density' data.

For Ecosystem Impacts use 'State' files.

In each file, 'Set' refers to order in which files were created (3 sets, 10 replicates each); 6-digits number gives the date of the simulation; 'DL' stands for duiker-like harvesting; non-integer (e.g. 0.30) represents harvest rate (between 0 and 0.90) and final integer (e.g. 4) stands for the replicate number (0 to 9).


Natural Environment Research Council, Award: NE/L002485/1