Dataset DOI: https://doi.org/10.5061/dryad.q573n5tv4

Description of the data and file structure

The dataset provided can be used to reproduce results associated with the manuscript titled "Habitat selection and outdoor recreation explain human-mountain lion conflict" by Morgan et al. 2026 (Current Biology).

These data consist of .csv files and spatial raster files originally analyzed using R (version 3.6.0). All GPS movement data were collected by the Santa Cruz Puma Project within California's Santa Cruz Mountains, USA, between 2018 and 2023. GPS data have been stripped of location information due to the sensitive nature of mountain lion space use within the study area, but are available upon request to the corresponding author. Human outdoor recreation data were derived from Strava Metro data collected from 2018–2023. Recreation data represent on-trail recreation intensity at multiple unique spatial and temporal scales around each used and available GPS location. Further details on how recreation intensity was calculated can be found in the manuscript. Human-mountain lion conflict data are based on Wildlife Incident Reports (WIR) compiled by the California Department of Fish and Wildlife from 2018–2023, supplemented by two confirmed attacks on people. Raw conflict event locations are not shared to protect location privacy; conflict indicators have been pre-assigned to raster grid cells. Together these data can be used to model mountain lion behavioral response to outdoor recreation activity and to better understand the drivers of human-mountain lion conflict.

---

Files and variables

File: issa_steps_top_model.csv

Description: Step dataset for the top iSSA model (avg01hr_365da temporal scale, 30m spatial scale). Each row represents one used or available step. This file contains movement characteristics, standardized habitat covariates, and the standardized recreation covariate at the top-performing spatiotemporal scale. Used to fit the top iSSA model and behavioral subgroup models. One used step and 20 available steps per stratum.

Variables

• pumaID — individual mountain lion identifier (coded; e.g., "102F")
• start_gpsID — unique stratum identifier; each stratum consists of one used step and its 20 available alternatives
• case — used locations = 1, available locations = 0
• sex — individual sex: "M" (male) or "F" (female)
• diel2 — diel period at step end time: "day" or "night", classified by civil sunrise/sunset
• wkend — TRUE if step end time falls on a weekend (Saturday or Sunday), FALSE otherwise
• sl_km — step length in kilometres
• sl_log_km — natural log of step length in kilometers
• ta_rad — turning angle in radians. Contains NA for steps without a valid prior relocation.
• slope_end_std — standardized terrain slope at step end location. Derived from Bay Area Open Space Council DEM.
• hd_150_end_std — standardized housing density within 150m kernel density estimator at step end location. Derived from Microsoft Building Footprints. Units before standardization: buildings per km².
• cover_end_std — standardized canopy cover at step end location. Derived from NLCD 2019. Units before standardization: percent (0–100).
• dist_urban_pos_log_end_std — standardized log-transformed distance to urban edge at step end location. Positive values only (distance from urban boundary into wildland). Derived from county and city limit boundaries.
• dist_trail_end_std — standardized distance to nearest trail at step end location. Units before standardization: meters.
• avg_rec_all_30_end_log1p_std — standardized log(x+1)-transformed recreation intensity within 30m buffer at step end location. Recreation intensity = average hourly trail use × trail length within buffer. Average hour based on the previous 365 days. Derived from Strava Metro trip count data.
• trail_dens_30_end_std — standardized trail density within 30m buffer at step end location. Units before standardization: meters of trail.
• trail_dens_30_start_std — standardized trail density within 30m buffer at step start location.

Note: All covariates are standardized using the mean and SD of the full dataset (used + available steps combined). Unstandardized covariate values and raw GPS coordinates are not included.

---

File: issa_steps_scale_optimization_example.csv

Description: Step dataset used to demonstrate the AICc-based spatiotemporal scale optimization procedure. Contains the same base columns as issa_steps_top_model.csv plus log(x+1)-transformed and standardized recreation intensity and trail density columns at all 8 spatial scales (30–500m) for the avg01hr_365da temporal scale only. The full optimization tested 109 models across all temporal and spatial scale combinations; complete AICc results are in Table S2 of the manuscript.

All columns from issa_steps_top_model.csv are present, plus:

• avg_rec_all_Xm_end_log1p_std — standardized log(x+1)-transformed recreation intensity at spatial scale X meters, where X =c(30, 60, 100, 150, 200, 250, 350, 500)
• trail_dens_Xm_end_std — standardized trail density at spatial scale X meters (end location)
• trail_dens_Xm_start_std — standardized trail density at spatial scale X meters (start location)

---

File: indiv_exposure.csv

Description: Individual-level recreation exposure values. One row per individual (n = 36). Used to fit functional response models examining whether individual variation in recreation avoidance scales with exposure to recreation.

Variables

• pumaID — individual mountain lion identifier. Matches pumaID in step datasets.
• avg_rec_30_used — mean log(x+1)-transformed recreation intensity at used GPS locations within the individual's home range. Derived from the 30m buffer recreation covariate.

---

File: sim_surface_aligned_strava.tif

Description: 31-layer GeoTIFF raster stack defining the simulation landscape used in movement simulation. All layers are standardized using the mean and SD from the fitted iSSA dataset so that raster values are on the same scale as the model coefficients. Used by 04_movement_simulation.R.

• CRS: UTM Zone 10N (EPSG:32610)
• Resolution: 30m
• NA values: Cells outside the study area boundary and open water

---

Files: valid_start_cells_30m.rds, params_move.rda, fixef_m5b_day.rds, vcov_m5b_day.rds, fixef_m5b_night_noRE.rds, vcov_m5b_night_noRE.rds

Description: Model objects and spatial index required to run the movement simulation (04_movement_simulation.R). All files are R binary format; load with readRDS() or load() as appropriate.

• valid_start_cells_30m.rds — integer vector of raster cell indices eligible as simulation start locations. Excludes urban areas, agriculture, open water, and cells within 500m of the study area edge.
• params_move.rda — data frame containing gamma distribution parameters for the step-length movement kernel (columns: sl_km_shape, sl_km_rate, sl_km_scale, i_strata).
• fixef_m5b_day.rds — named numeric vector of fixed-effect coefficients from the daytime iSSA model.
• vcov_m5b_day.rds — full variance-covariance matrix for the daytime iSSA model (fixed effects + random-effect SD parameters). Used to draw correlated coefficient sets via MASS::mvrnorm().
• fixef_m5b_night_noRE.rds — fixed-effect coefficients from the nighttime iSSA model. The recreation random effect is excluded due to insufficient individual variation at night.
• vcov_m5b_night_noRE.rds — variance-covariance matrix for the nighttime iSSA model.

---

Files: ud_all_all.tif, ud_shy_all.tif, ud_median_all.tif, ud_bold_all.tif

Description: Normalized utilization distribution (UD) rasters produced by the movement simulation. Each cell value gives the predicted probability of mountain lion space use; values sum to 1.0 within each raster. Used by 05_conflict_models.R.

• CRS: UTM Zone 10N (EPSG:32610)
• Resolution: 30m
• Data type: Float (FLT8S)

---

File: conflict_grid_30m.csv

Description: Pre-classified conflict analysis dataset at 30m resolution. One row per cell within the study area that contains at least one Strava-defined trail segment and falls within the mountain lion UD extent. Conflict events have been assigned to raster cells; raw event locations are not provided to protect location privacy. Used by 05_conflict_models.R.

Variables

• rec_intensity_log1p_std — standardized log(x+1)-transformed yearly-average recreation intensity. Recreation intensity = average yearly trip count × trail length within cell, averaged across 2018–2023.
• prob_lion_all_std — standardized population-level UD value
• prob_lion_shy_std — standardized shy behavioral class UD value
• prob_lion_med_std — standardized median behavioral class UD value
• prob_lion_bold_std — standardized bold behavioral class UD value
• case_conflict_all — 1 if cell contains at least one WIR event or confirmed attack (2018–2023), 0 otherwise
• case_conflict_nosight — 1 if cell contains at least one WIR event excluding sightings, or a confirmed attack, 0 otherwise

---

File: conflict_grid_1km.csv

Description: Same structure as conflict_grid_30m.csv at 1km resolution. UD values were derived by resampling the 30m UD rasters to 1km using bilinear averaging. Recreation intensity was independently computed at 1km resolution using the same Strava Metro trip count data.

Same column definitions as conflict_grid_30m.csv.

---

Code/software

All models were run using R (version 3.6.0). iSSA models were fit using the function fitTMB in the glmmTMB R package. Generalized linear models were fit using glm in the stats R package. Movement simulation used the terra, data.table, and doParallel packages. Full code and package version information are available on Zenodo (https://doi.org/10.5281/zenodo.15660336).

---

Access information

Data related to mountain lion attacks on people were derived from California Department of Fish and Wildlife:

• https://wildlife.ca.gov/Conservation/Mammals/Mountain-Lion/Attacks

Recreation intensity estimates were derived from Strava Metro trip count data licensed from Strava Inc.:

• https://metro.strava.com