A lot of these data cleaning scripts were written during my dissertation and use outdated file paths, naming conventions, etc. I haven't gone back and changed them, since I haven't done additional data cleaning but rather used the cleaned data in subsequent analyses.
01-timeshift: Change date and time metadata associated with image files; to do before generating record table in camtrapR. Input is a spreadsheet with the time corrections for all camera folders that must be edited. (Also associated timeshift files for WildCam data cleaning in December 2019)
02-create-record-table-year1 and 02a-create-record-table-year2: Generate record table from raw photos that have been sorted into folders by species (used in many of the scripts below); generate species-specific detection matrices for later use in Occupancy_matrix_wrangling.Rmd
03-remove-duplicate-records: written by Lynn, gets record table into format with one row per burst (to account for differences in sorting method between seasons). Input and output are both record tables
04-sunrisesunset-season.R: assigns time of day and season to each record, based on date and time. Script written for elephant activity pattern paper. Input a record table, and information on daily sunrise and sunset time, which I downloaded from a NOAA (?) spreadsheet - sunrise_sunset.csv. Compares the time to sunrise/sunset on that date to assign day or night. Output is Record table with columns for time period and season (ex. Record_table_all_1hr_timeperiods.csv)
05-recordtable-all-to-15min.R and 05a-recordtable-all-to-10min.R: Cleans master record table, eliminating records that are outside of camera operation and reducing to “independent” records. Input is a complete record table with all detections (generated by camtrapR, columns for elapsed time) (Gaynor_ALLrecords_cleaned_June16_to_June17.csv). Also need camera metadata with columns with dates for Start, End, Problem1_From, Problem1_To (optional) (cam_metadata_all.csv). Output is a record table that only includes detections separated by a given interval (here, I did 15 min, but could be easily changed), during period of camera operation (Gaynor_15min_records_June2016_to_June2017.csv). Can be brought into rai-calculation.R to calculate RAI
06-make-master-occupancy-matrix.R: Create a master occupancy matrix for later analysis, using record tables and camtrapR functions.
07-gorongosa-gis.RMd: Compile and clean spatial vector and raster layers, and extract values at camera locations. Crop, change resolution/extent, convert vector to raster, create raster stack, calculate distance to features, calculate neighborhood raster values. Output is a spreadsheet with extracted raster values at camera locations (ex. cam_metadata_102518.csv)
08-occupancy-matrix-wrangling.Rmd: Generate summarized occupancy records for different time periods in the format necessary for analysis using Lindsey’s code. Input is master file with a row for each camera-species, and column for each date, with 0, 1, and NA for non-detection, detection, and inoperability. Subsets by dates of interest. Sum “1” columns to calculate total number of sampling periods with detections, spread data so there is one column per species and one row per camera; Count “0” and “1” columns to calculate total number of nights that each camera was operational. Output = detection counts for each species at each camera (ex. Detections_7_1_16_9_30_16.csv), camera operation for each camera (ex. Camoperation_7_1_16_9_30_16.csv)
08a-occupancy-matrix-wrangling-monthly-loop.R: Does the same as above, but allows the creation of multiple matrices for multiple periods using a for loop
01-covariate-correlations.R: calculates correlations among variables. Input camera metadata file (cam_metadata_master.csv). Determines Pearson’s r coefficients and p-values for relationships between each of covariates
02-normalization-comparison.R: Was curious to see how similar the metadata values were at each camera site if the covariate rasters were standardized before extraction, versus standardized among the 60 values at the camera sites Lindsey suggested they should be similar, or the camera locations don't accurately represent the study area.
03-aerial-count-exploration.R: Basic exploration of aerial count data, to determine percentage/biomass of ungulates on floodplain. (Doesn't actually integrate camera trap data; that's another project)
01-rai-calculation.R: Calculates RAI for each species at each camera from the record table for a given time period. Input cleaned record table with only independent records. Removes species that are not to be included in analysis (incl. Ghost), subsets record table to time period of interest (after you input start and end dates), calculates number of records of each species at each camera, calculates RAI based on number of records and operation date. Outputs table for each season, with columns for species, camera, operation, count, and RAI.
01a-rai-calculation-10min.R: Does the same thing as above, but with 10 min interval rather than 15
01b-rai-calculation-monthly-for-loop.R: Does the same thing as above, but with a loop to allow for calculation of RAI every month
02-rai-analysis.Rmd: Variety of analyses related to RAI.
01-occupancy-models.Rmd: Lindsey’s occupancy models. Input detection histories for each species at each camera, plus camera operation days for each camera (from occupancy-matrix-wrangling.Rmd). Runs multi-species hierarchical occupancy models with specified covariates for occupancy and detection. Outputs various files with occupancy, detection means and CIs for different groups/species, modeled species richness, etc.
01a-occupancy-models-aerial-period.Rmd: Run occupancy models for the 2016 aerial count. For Integrative Zoology / Animal Conservation manuscript
01b-occupancy-models-combinedyearseason.Rmd: Run occupancy models for both years and seasons in a single model, where year and season are included as predictors (technically not an appropriate approach)
02-occupancy-covariate-comparison.R: for exploring the effects of individual covariates on overall occupancy (Lindsey’s suggested method for Bayesian "model selection"). Essentially does what occupancy-models does, but with just one detection covariate, and only exports community-level hyperparameters. Output = Community-level hyperparameter - coefficient corresponding to effect of the predictor variable on occupancy (mean, SD, upper and lower CIs)
02-occupancy-covariate-comparison-2.R: for exploring the effects of individual covariates on overall occupancy (Lindsey’s suggested method for Bayesian "model selection"). Input individual spreadsheets for the community coefficient model output (outputs from occupancy-covariate-comparison.R
03-occupancy-results-exploration.Rmd: Explore the results of the occupancy models
01-temporal-overlap.R: Calculates pairwise temporal (24-hr) overlap between all pairs of species, using overlap package. Input is record table, and outputs long data matrix with pairwise overlap values for all 703 species pairs, with columns for overlap in dry, wet3mo, and wet4mo seasons
01-format-data-for-sif.R: Reformats detection and operation matrices for pairwise species interaction factor modeling (from Lindsey) Input is detection and operation matrices (as generated by occupancy-matrix-wrangling.Rmd). Output is spreadsheet of species pair co-occurrence data ready for SIF analysis using run-sif-model.R (ex. Dry.peak.SppPairs.csv)
02-run-sif-model.R: Runs species co-occurrence models. Code from Lindsey Rich. Input species pair detection information - from Sformat-data-for-sif.R (ex. Dry.peak.SppPairs.csv). Output SIF results, with columns for psiA (probability A occurred in study area), psiBA (probability of B given A), psiBa (probability of B given A), psiB (derived quantity from the other values), pA (A detection probability), pB (B detection probability) - ex. SIF_2018Dry.csv
03-explore-sif-results.Rmd: Analysis of SIF results, in conjunction with temporal overlap, RAI overlap, dietary overlap, etc.