Data from: Release from captivity allows African savannah elephant movement patterns to converge with those of wild and rehabilitated conspecifics

The release of captive animals through rewilding or conservation translocation is an important strategy for the rehabilitation of individuals and ecosystems. Assuming that wild animals are better adapted to their environment than captive ones, comparisons between the behaviour of released animals and their wild counterparts would allow the evaluation of rewilding success. We compared the movement patterns of six captive African savannah elephants, subjected to a three-year soft release in the western Okavango Delta, Botswana, with those of two elephants released over a decade previously and of four wild elephants. GPS fixes at 1800hrs were used to calculate daily displacement, 30-minute diurnal and nocturnal distances, cumulative daily distances, and home ranges for all study animals; the effects of elephant group, phase, year, and season on these movement metrics were analysed. Captive elephants were most affected by the phase and changed their movements after release. After release, the movement patterns of captive elephants were less different from those of rehabilitated elephants than wild elephants, possibly due to sample size, which could indicate that rewilded animals may not be able to fully approximate the behaviour of wild-born individuals. However, the captive elephants should exhibit seasonality in movement patterns just like rehabilitated and wild elephants. These results highlight the critical importance of long-term monitoring of animals to allow the evaluation of release and translocation, which is recommended for other similar studies. The data collection was approved by the UB ethics committee (UBR/RES/IRB/SOC/132). # Data from: Release from captivity allows African savannah elephant movement patterns to converge with those of wild and rehabilitated conspecifics ### Analysis Dataset and Code **General Introduction** The project "Release from captivity allows African savannah elephant movement patterns to converge with those of wild and rehabilitated conspecifics" investigates the movement of rewilded African savannah elephants (Loxodonta africana) in the western Okavango Delta, Botswana. This study compares movement patterns of captive, previously released (rehabilitated), and wild elephants using GPS data from collared elephants to evaluate rewilding success. Four R scripts—30_minute_script.R, Cumulative_Daily_Distance_script.R, Displacement_script.R, and Home_range_script.R—analyze are saved in the zip file called R_Scripts.zip. Four key movement metrics: 30-minute movement distances, Cumulative daily distances, daily displacement (net displacement between consecutive days at 18:00), and monthly 95% Kernel Density Estimation (KDE) home range areas were analysed using these R_scripts as stated below. These scripts fit generalized linear mixed-effects models (GLMMs) to assess the effects of factors such as phase, season, year, day/night (for 30-minute data), and elephant origin on movement metrics. They also generate visualizations to compare results across groups. This work highlights the importance of long-term monitoring in evaluating rewilding outcomes and understanding how captive elephants adapt to wild movement patterns. The Botswana Ministry of Environment, Natural Resources Conservation and Tourism granted the research permission (Permit ENT8/36/4L(43)) and the Botswana Department of Wildlife and National Parks issued the release permit (WP SAF 16/4/1 II (1)), while the University of Botswana Office of Research and Development provided the ethics approval (UBR/RES/IRB/SOC/132). The R scripts were developed and tested on R version 4.5.1 Instructions for Use 1. Place the data files (30_min_dist.csv, daily_range.csv, last_coords_local.csv, kde_95_monthly.csv) in your working directory. Download the Data.zip and extract the csv files into your working directory. 2. Open the R scripts 30_minute_script.R, Cumulative_Daily_Distance_script.R, Displacement_script.R, and Home_range_script.R in R or RStudio. 3. Update the setwd() path in each script to match the location of your working directory. For example: setwd("/path/to/your/directory") 4. Install the required R packages listed below for each script. 5. Run the scripts to preprocess data, generate visualizations, and fit GLMMs. Three datasets are provided: * 30_min_dist.csv: 30-minute movement data\ Columns/Variables: * Column A - row number for the data. No column heading * Column B - "CollarID", is the column for the names given to each of the elephants in the study. Each elephant had a distinct name * Column C - "date_timestamp_local", is the timestamp (month/day/year hour:minutes) for the Botswana local time (+ 2 GMT) * Column D - "seasons", is the column for the seasons of the Okavango Delta flooding patterns. The seasons are 4 months long each * Column E - "release_phase", is the column for the names of different phases of the soft release. * Column F - "ele_origin", is the column that shows the category of the elephant in this study.  * Column G - "dist_moved", is the distances measured between 2 consecutives GPS fixes in meters. * Column H - "Year", is the column that shows the years of the data was collected with 1st December 2021 marking the beginning of year 1 of the data used in this study. * Column I - "day_night", is the column that shows the time of day when the data was collected. * last_coords_local.csv: Daily displacement data\ Columns/Variables: * Column A - row number for the data. No column heading * Column B - "CollarID", is the column for the names given to each of the elephants in the study. Each elephant had a distinct name * Column C - "date_timestamp_local", is the timestamp (month/day/year hour:minutes) for the Botswana local time (+ 2 GMT) * Column D - "seasons", is the column for the seasons of the Okavango Delta flooding patterns. The seasons are 4 months long each * Column E - "release_phase", is the column for the names of different phases of the soft release. * Column F - "ele_origin", is the column that shows the category of the elephant in this study.  * Column G - "dist_to_next_day", is the distance between GPS fixes of consecutive 18:00 hours in meters * Column H - "Year", is the column that shows the years of the data was collected with 1st December 2021 marking the beginning of year 1 of the data used in this study. * Column I - "season_number", is the number allocated to each number starting with the first season as the season when the captive elephants were collared * Column J - "year_regular", is the column of the Calendar year in Gregorian calendar * Column K - "log_dist_to_next_day", is the log of the dist_to_next_day (Column G) * kde_95_monthly.csv: Monthly 95% KDE home range data\ Columns/Variables * Column A - row number for the data written as the 95...[row number], where 95 shows the kernel density estimate 95% contour. No column heading * Column B - "CollarID", is the column for the names given to each of the elephants in the study. Each elephant had a distinct name * Column C - "year_month", is the column of  the calendar year (Gregorian calendar) and month for which the kernel density estimate (95 % contour) was generated. Month is given as a three-letter English abbreviation * Column D - "year", is the column of the Calendar year in Gregorian calendar * Column E - "month", is the column of the month as a three-letter English abbreviation * Column F - "period", is the column of the period when the elephants were controlled (before) and when they were officially released with their movements not controlled (after) * Column G - "ele_origin", is the column showing the category of the elephants, whether it was captive (elephants that were long-term captive and released during the course of this study), rehabilitated (elephants that were previously captive and released before this study commenced, over ten years ago), or wild (wild elephants with no known history of captivity) * Column H - "seasons", "seasons", is the column for the seasons of the Okavango Delta flooding patterns. The seasons are 4 months long each * Column I - "kde_95_area", is the column of the calculated area of the 95% kernel density estimate (KDE) contour calculated for each elephant in each year-month in square kilometres (km²) Analysis Scripts 30-Minute Movement Analysis Introduction script processes 30-minute movement distances from 30_min_dist.csv to analyze fine-scale movement patterns. It generates visualizations (histograms, bar plots) to compare movement distances by elephant origin, period (Before vs. After), season, and day/night. It also fits GLMMs to assess the effects of season, phase, year, day/night, and elephant origin on movement distances, including model diagnostics and post-hoc comparisons. CSV File 30_min_dist.csv Requirements R packages:  MuMIn (for model selection with dredge) lme4 (for generalized linear mixed-effects models),  DHARMa (for model diagnostics),  dplyr (for data manipulation),  lubridate (for date-time handling),  ggplot2 (for plotting),  patchwork (for combining plots),  e1071 (for skewness calculation),  emmeans (for post-hoc comparisons),  geosphere (for Haversine distances in preprocessing) All packages can be installed using the code below:  install.packages(c("MuMIn", "lme4", "DHARMa", "dplyr", "lubridate", "ggplot2", "patchwork", "e1071", "emmeans", "geosphere")) When running the script 30_minute_script.R, it should: * Generate histograms of 30-minute movement distances by period, overall, and for Captive, Before, and After subsets, with density curves. * Generate bar plots of average movement distances by elephant origin, period, season, and day/night, with standard error bars. * Fit generalized linear mixed-effects models to assess the effects of season, release phase, year, day/night, and elephant origin on movement distances. * Provide summaries of the fitted mixed-effects models and dredge results. * Provide post-hoc comparisons using emmeans for Captive, Before, and After periods. * Generate residual diagnostic plots using DHARMa to assess model fit. * Calculate skewness for movement distances by period. Cumulative daily distance. This script aggregates 30-minute movement data from 30_min_dist.csv into cumulative daily distances, representing the total distance moved per day. It generates visualizations (histograms, bar plots) to compare cumulative daily distances by elephant origin, period, and season, and fits GLMMs to assess the effects of season, release phase, and elephant origin on daily movement distances, including model diagnostics and post-hoc comparisons. CSV Files 30_min_dist.csv Requirements R packages:  MuMIn (for model selection with dredge) lme4 (for generalized linear mixed-effects models) DHARMa (for model diagnostics),  dplyr (for data manipulation),  lubridate (for date-time handling),  ggplot2 (for plotting),  patchwork (for combining plots),  e1071 (for skewness calculation),  emmeans (for post-hoc comparisons),  geosphere (for Haversine distances in preprocessing) All packages can be installed using the code below:  install.packages(c("MuMIn", "lme4", "DHARMa", "dplyr", "lubridate", "ggplot2", "patchwork", "e1071", "emmeans", "geosphere")) When running the script 30_minute_script.R, it should: * Generate histograms of cumulative daily distances by period, overall, and for Captive, Before, and After subsets, with density curves. * Generate bar plots of average cumulative daily distances by elephant origin, period, and season, with standard error bars. * Fit generalized linear mixed-effects models to assess the effects of season, release phase, and elephant origin on daily movement distances. * Provide summaries of the fitted mixed-effects models and dredge results. * Provide post-hoc comparisons using emmeans for Captive, Before, and After periods. * Generate residual diagnostic plots using DHARMa to assess model fit. * Calculate skewness for cumulative daily distances by period. Displacement This script processes daily displacement data at 18:00 from last_coords_local.csv, capturing net movement between consecutive days. It generates visualizations (histograms, bar plots, boxplots) to compare displacement by elephant origin, period, and season, and fits GLMMs to assess the effects of season, year, release phase, and elephant origin on displacement, including model diagnostics and post-hoc comparisons. CSV File last_coords_local.csv Requirements R packages:  MuMIn (for model selection with dredge) lme4 (for generalized linear mixed-effects models) DHARMa (for model diagnostics, s),  dplyr (for data manipulation),  lubridate (for date-time handling),  ggplot2 (for plotting),  patchwork (for combining plots),  e1071 (for skewness calculation),  emmeans (for post-hoc comparisons) All packages can be installed using the code below:  install.packages(c("MuMIn", "lme4", "DHARMa", "dplyr", "lubridate", "ggplot2", "patchwork", "e1071", "emmeans")) When running the script Displacement_script.R, it should: * Generate histograms of daily displacement distances by period, overall, and for Captive, Before, and After subsets, with density curves. * Generate bar plots of average displacement by elephant origin, period, and season, with standard error bars. * Fit generalized linear mixed-effects models to assess the effects of season, year, release phase, and elephant origin on displacement. * Provide summaries of the fitted mixed-effects models and dredge results. * Provide post-hoc comparisons using emmeans for Captive, Before, and After periods. * Generate residual diagnostic plots using DHARMa to assess model fit. * Calculate skewness for displacement distances by period. Home RangeThiss This script processes precomputed monthly 95% KDE home range areas from kde_95_monthly.csv, representing the spatial extent of elephant movement. It generates visualizations (histograms, bar plots) to compare home range sizes by elephant origin, period, and season, and fits GLMMs to assess the effects of release phase, season, and elephant origin on home range sizes, including model diagnostics and post-hoc comparisons. CSV File kde_95_monthly.csv Requirements R packages:  MuMIn (for model selection with dredge) lme4 (for generalized linear mixed-effects models),  DHARMa (for model diagnostics),  dplyr (for data manipulation),  ggplot2 (for plotting),  e1071 (for skewness calculation),  emmeans (for post-hoc comparisons) All packages can be installed using the code below:  install.packages(c("MuMIn", "lme4", "DHARMa", "dplyr", "ggplot2", "e1071", "emmeans")) When running the script Home_range_script.R, it should: * Generate histograms of monthly 95% KDE home range areas by period, overall, and for Captive, Before, and After subsets, with density curves. * Generate bar plots of average 95% KDE areas by elephant origin, period, and season, with standard error bars. * Fit generalized linear mixed-effects models to assess the effects of release phase, season, and elephant origin on home range sizes. * Provide summaries of the fitted GLMMs and dredge results. * Provide post-hoc comparisons using emmeans for Captive, Before, and After periods. * Generate residual diagnostic plots using DHARMa to assess model fit. * Calculate skewness for home range areas by period. Reproducibility * Ensure all required R packages are installed using the provided install.packages() commands. * Verify that 30_min_dist.csv, daily_range.csv, last_coords_local.csv, and kde_95_monthly.csv are in the working directory. * Update the setwd() path in each script to point to your working directory. * Run the scripts in R (version 4.5.1 or higher) to reproduce the analyses. * Check for missing values or data inconsistencies using summary() and table() functions, as outlined in the scripts. * For home range analysis, note that kde_95_monthly.csv contains precomputed KDE areas. To recompute these areas from raw GPS data, users would need additional scripts and the All_elies_regular_30min.csv dataset (not included). For questions about this dataset or scripts, please contact:  Murphy Tladi ([201000042@ub.ac.bw](mailto:201000042@ub.ac.bw) or [murphytladi@ymail.com](mailto:murphytladi@ymail.com))