Hunting data collected in the Ebo forest, Cameroon from 2008 to 2023
Data files
Nov 26, 2024 version files 352.99 KB
Abstract
The long-term survival of many mammal populations relies on how effectively we mitigate the threat from unsustainable hunting. Yet, hunting activities are often cryptic, especially in unprotected forests. Here, we investigate whether hunting signs can help understand the spatiotemporal dynamics of hunting activities in an unprotected African rainforest and examine how landscape characteristics predict various indicators of hunting. We recorded hunting signs (e.g., shotgun cartridges, wire snares, direct sightings) systematically on 23 parallel recce lines across the Ebo forest, Cameroon from 2008 to 2023. We assigned hunting data and spatial covariates (e.g., elevation, distance to the village) to 1×1 km grid cells and applied generalized linear mixed models to predict the effects of these covariates on hunting. We found that hunting was commonplace across the entire Ebo forest. The best-fitting models for each hunting sign differed considerably. Shotgun cartridges and all hunting signs combined increased significantly from 2016 to 2023 and varied non-linearly along the village-distance gradient. We found a progressive inversion of hunting trends along the anthropogenic gradient; between 2016 and 2018, wire snares declined with the distance to the road but from 2021, they increased along the road distance gradient. Wire snares showed a similar pattern along the river distance gradient. Our results also revealed differences between shotgun hunting and snaring along the altitudinal gradient; the effect of elevation was positive on shotgun cartridges and negative on wire snares. Hunting signs and trails decreased significantly with increasing terrain ruggedness. Using long-term monitoring data, we show how hunting patterns change dynamically with respect to human and landscape-related features. We also demonstrate complex hunting patterns along the gradient of human influence, therefore questioning the use of proxies such as the distance to human settlements and even topography to account for hunting pressure. Overall, we show that hunting sign data can reveal the spatiotemporal patterns of hunting, crucial in evaluating the effectiveness of conservation interventions and guiding the prioritisation of limited conservation resources.
https://doi.org/10.5061/dryad.8931zcrzk
This dataset contains hunting data from the Ebo landscape, as well as other landscape features derived from GIS sources as indicated below. For the sake of the analyses, the data were assigned to 1 x 1 km grid cells, resulting in 346 cells.
Description of the data and file structure
The data contained in the CSV file is organized in columns as follows:
Site: this represents the ID of the 1 x 1 km grid cells. There are a total of 346 grid cells, therefore, the column "site" varies from 1 to 346.
Year: represents the year during which the survey took place. Each year is coded as a number as follows: 1 = 2008, 2 = 2012, 3 = 2016, 4 = 2017, 5 = 2018, 6 = 2019, 7 = 2021, 8 = 2022, 9 = 2023. In 2019, most grid cells were not surveyed; this year should therefore not be used in analyses.
Effort: represents the distance in km walked in each grid cell every year.
Elev: is the mean elevation (altitude asl) of each 1*1 km grid cell obtained from a Shuttle Radar Topography Mission (SRTM) layer of 30 m resolution obtained from https://search.earthdata.nasa.gov (download: May 2023)
Rugg: is the terrain ruggedness index obtained from QGIS using the above-mentioned STRM layer.
MovCost: is the mean anisotropic cumulative cost of moving from villages to different grid cells. It was obtained using the r.walk.points function in QGIS, taking into consideration the STRM layer and the location of the villages.
VillDist: The Euclidean distance (km) from the centre of grid cells to the nearest village
RoadDist: The Euclidean distance (km) from the centre of grid cells to the nearest road
RivDist: Euclidian distance (km) from the centre of grid cells to the nearest river
HFI: Mean Human Footprint Index within each grid cell
Data collection
We established twenty-three parallel recce lines placed 4 km apart and totalling c. 345 km across the entire Ebo forest, Cameroon in 2008 (see Fig. 1 in the publication). The recce lines were oriented perpendicularly to the main rivers and were surveyed in 2008, 2012, 2016, 2017, 2018, 2019 (partially due to the COVID-19 pandemic), 2021, 2022, and 2023. The surveys generally took place from October to April in the subsequent year, with the initial year regarded as the year in which the survey occurred. The survey team consisted of 2-3 trained observers followed by 3 porters, all moving at about 1 km per hour, using a handheld GPS (Garmin) and a manual compass to navigate along the predefined recce lines. The team used only pruning shears (secateurs) to ease movement, minimise damage to vegetation, minimise the risk of hunters using the recces, and distinguish their cuts from hunters’ machete cuts. We recorded different types of signs indicating human activities, including locations of wire snares (both active and inactive), hunting camps, trails regularly used by humans, cuttings (machete cuts and shrub breakings made by hunters while moving across the forest), human footprints, audible gunshots, and shotgun cartridges encountered on both sides of the recce lines. Perennial Permanent signs such as hunting camps and hunting trails were recorded when it was evident that they were used after the previous survey. All detected shotgun cartridges were collected and later destroyed to avoid double counting during subsequent surveys. We associated covariates related to the configuration of the landscape and anthropogenic pressures (distance to village, distance to road, distance to river, elevation, terrain ruggedness index, movement cost, human footprint index) to analyse the spatiotemporal patterns of hunting activities across the landscape.
Data Preparation
We placed 1 x 1 km grid cells (hereafter referred to as the site) across the entire study area using QGIS version 3.28.5 (QGIS Development Team, 2022; Whytock et al., 2021), and extracted a total of 346 sites which intersected with recce lines. We considered each site as a sampling unit (Wessling et al., 2020; Whytock et al., 2021) and assigned the data collected on recce lines across the years to their respective sites using their coordinates. We also calculated the survey effort (i.e., the kilometres walked) in each site. We only used hunting signs (shotgun cartridge, wire snare, hunters’ sign of passage) that were assumed to be less than one year old. The age of hunting signs was not documented in 2008 and 2012; we therefore excluded them from in-depth analyses. Since there were no law enforcement activities undertaken in the area during the period of surveys, we assumed that the number of shotgun cartridges displaced from the original location where they were produced was negligible, as hunters would not attempt to hide their traces.
For each site, following Soofi et al. (2018), we created secondary sampling units by dividing the recce line into segments of approximately 200 m each. On these segments, for each survey year, we recorded each hunting sign only once per segment to obtain presence and absence data (Karanth et al., 2011; Laurance et al., 2008; Soofi et al., 2018). Then, we summed the results from the segments within each site and obtained the number of events for each hunting sign which could vary from 0 to 5 given that each site had a maximum of five 200 m segments. This allowed us to reduce potential pseudoreplication and increase the independence of the signs (Soofi et al., 2018) which may arise from multiple signs created by the same hunter at the same place (e.g. group of snares deployed by the same hunter, multiple shotgun cartridges produced by a hunter when attempting to kill one animal).