Description of methods used for collection/generation of data: 20 horses housed in a paddock-trail system were monitored under four different shelter layouts using a Latin square design. Accelerometers were mounted on the left front leg of each horse and recorded activity continuously. Lying behaviour was classified from sensor orientation. Rhythmicity metrics (DFC, HP) were computed from 7-day windows using Fourier decomposition via the digiRhythm R package. Weather data were retrieved from the nearest MeteoSwiss station and aggregated to each 7-day window. Methods for processing the data: Raw accelerometer files → orientation → binary lying classification → daily & weekly metrics. Fourier spectral analysis of activity (24 h & harmonics) → DFC & HP. Weather variables aggregated using daily means or sums depending on parameter. All analyses performed in R 4.x using lme4, MuMIn, DHARMa, and digiRhythm. Instrument- or software-specific information needed to interpret the data: R package digiRhythm (SoftX, 2025; doi:10.1016/j.softx.2025.102184) R packages: lme4, dplyr, lubridate, boot, ggplot2, rcompanion, MuMIn, DHARMa