Mobility, behaviour and land-use data from GPS tracking and hidden Markov model training on cattle herds in Sahel agropastoral systems

Contains positional, behavioural and environmental data for cattle herds in West Africa. (Illustration from the related publication) Extent of the data The dataset contains tracking data from 15 collars, fitted on 21 Zebus of the Gobra breed in 19 transhumant and resident herds. Observations were recorded every 30 minutes, between May 2021 and December 2023. Three seasons of transhumance from the home-villages in the Niakhar area, to the outskirts of Tambacounda and the Ferlo region were recorded. Data description Each observation contains recorded (R) and computed (C) information. - Identification of the collar and animal Trajectory data: - Positional information (R): longitude, latitude, altitude and timestamp - Dilution of precision of the point (R) - Distance to the previous point (C) - Angle between the trajectories from the previous point and to the next (C) Behavioural data: - Activity state the herd is in, as estimated by the HMM (C) Environmental data: - Temperature of the air, as recorded by the collar (R) - Ongoing season (cold dry season: November – February, warm dry season: March – May, rainy season: June – October) (C) - Whether it is daytime or nighttime, based on the sunset and sunrise time (C) - Soil occupation data from the coarse, Globcover dataset (C) Data collection Raw GPS and accelerometer data were collected using Vertex Lite GPS Collars (see below). Behavioural activity was inferred using a hidden Markov model, the Viterbi algorithm, a non-linear minimisation algorithm and parameter sampling, as described in the related publication (DOI). Data preparation and analysis was conducted using the HAMAC routine (https://gitlab.cirad.fr/selmet/hamac) and the moveHMM R package (https://doi.org/10.1111%2F2041-210X.12578).