plotunmarked The goal of plotunmarked is to provide convenient plotting functions for visualising predictions from unmarked models, particularly occupancy estimates – a quick and dirty approach! Currently, functions are available for static occu occupancy models. Future updates will incorporate dynamic colext and multispecies interaction occuMulti models.

Installation

You can install the development version of plotunmarked from GitHub with:

# install.packages("pak")
pak::pak("ugyenpenjor1/plotunmarked")
# OR using devtools:
devtools::install_github("ugyenpenjor1/plotunmarked")

Example

This is a basic example which shows you how to use the functions of the package.

Plot occupancy vs covariate

Below is an example of how to plot occupancy against a covariate using the plot_occupancy() function

library(plotunmarked)
if (requireNamespace("unmarked", quietly = TRUE)) {
  # Plot occupancy vs covariate 
  plot_occupancy(
    model,
    "droa", # change this to plot for other variables
    xlab = "Distance to road (standardised)"
  )
}

A similar plot can be made for the model-averaged coefficient using the plot_modavg_response() function.

plot_modavg_response(
  model_selection_table = ms,   # model selection table
  model_list = mod_ls_2,        # list of models within some delta AIC threshold (delta AIC < 2)
  covariate = "droa",           # covariate name as it appears in your model
  fixed_vals = list(driv = 0, elev = 0, tree = 0), # other covariate values fixed to 0 if standardised
  response_type = "state",      # occupancy; "det" for detection probability
  xlab = "Distance to road"     # intuitive label
)

Plot predicted occupancy map from a single model (this can be your top model or best model)

library(terra)
riv <- rast("C/yourdirectory/dist_river.tif")

# Or
library(raster)
riv <- raster("C/yourdirectory/dist_river.tif")

Standardise raster using mean and SD of covariates

Use standardise_raster() function to standardise raster.

driv <- standardise_raster(
  raster_layer = riv,    # raster layer
  cov_data = cov,        # covariate data frame
  cov_name = "DRiver"    # covariate name in cov 
)

Likewise, for other raster layers…

Create a raster list and make a map

# Create a raster list
raster_list <- list(driv = driv, droa = droa, elev = ele)

# Make a map
psi_map <- occupancy_map(
  model = model,                       # top or best model
  covariate_rasters = raster_list,     # standardised raster list
  return_class = "terra",              # if input was loaded using the raster package, choose "raster"
  plot_map = TRUE,                     # display map
  plot_ci = TRUE,                      # plot lower and upper prediction maps
  ci_level = 0.95,                     # confidence threshold
  save_plot = TRUE,                    # saves the plot in your working directory
  plot_filename = "occupancy_map_top"  # if save_plot = TRUE
)

Plot predicted occupancy map using model-averaged coefficients

# Create model list (models within some delta AIC threshold)
mod_list <- list("occ.1" = model1, "occ.2" = model2)

# Create raster stack
raster_stk <- stack(driv_s, droa_s, ele_s, tree_s)         # using raster package
#raster_stk <- c(driv_s, droa_s, ele_s, tree_s)            # using terra package
names(raster_stk) <- c("driv", "droa", "elev", "tree")     # rename to match covariates names in models

psi_map_ma <- modavg_occupancy_map(
  d2_mod_list = mod_list,                  # model list
  pred_cov_stack = raster_stk,             # stacked rasters
  agg_factor = 4,                          # raster resolution, higher number produces coarser map but runs quicker
  plot_ci = TRUE,                          # plot lower and upper prediction maps
  save_plot = TRUE,                        # saves the plot
  plot_filename = "plot_name",             # plot name
  plot_format = "png"                      # plot format (also available as "jpeg" and "tif"
)

ROC curve and AUC to validate prediction

# AUC and ROC for a single model
auc_plot(model, use_detection = FALSE)   

# Model-averaged AUC and ROC
auc_modavg_plot(
  model_list = list(model1, model2),  # list of models
  use_detection = TRUE,               # whether raw detection to use
  detection_history = detn            # if use_detection = TRUE, supply detection matrix
) 

This work is in progress. If you have any questions or discover bugs, please report to [email protected].

To cite the package plotunmarked in publications, use:

Penjor, U. (2025). plotunmarked: Utility Tools for Visualising unmarked Package Models. R package version 0.1.0. Available at: https://github.com/ugyenpenjor1/plotunmarked

Or

citation("plotunmarked")