R: Format EBD data for occupancy modeling with 'unmarked'

format_unmarked_occu {auk}

R Documentation

Format EBD data for occupancy modeling with `unmarked`

Description

Prepare a data frame of species observations for ingestion into the package unmarked for hierarchical modeling of abundance and occurrence. The function unmarked::formatWide() takes a data frame and converts it to one of several unmarked objects, which can then be used for modeling. This function converts data from a format in which each row is an observation (e.g. as in the eBird Basic Dataset) to the esoteric format required by unmarked::formatWide() in which each row is a site.

Usage

format_unmarked_occu(
  x,
  site_id = "site",
  response = "species_observed",
  site_covs,
  obs_covs
)

Arguments

`x`	`data.frame`; observation data, e.g. from the eBird Basic Dataset (EBD), for a single species, that has been filtered to those with repeat visits by `filter_repeat_visits()`.
`site_id`	character; a unique idenitifer for each "site", typically identifying observations from a unique location by the same observer within a period of temporal closure. Data output from `filter_repeat_visits()` will have a `.site_id` variable that meets these requirements.
`response`	character; the variable that will act as the response in modeling efforts, typically a binary variable indicating presence or absence or a count of individuals seen.
`site_covs`	character; the variables that will act as site-level covariates, i.e. covariates that vary at the site level, for example, latitude/longitude or habitat predictors. If this parameter is missing, it will be assumed that any variable that is not an observation-level covariate (`obs_covs`) or the `site_id`, is a site-level covariate.
`obs_covs`	character; the variables that will act as observation-level covariates, i.e. covariates that vary within sites, at the level of observations, for example, time or length of observation.

Details

Hierarchical modeling requires repeat observations at each "site" to estimate detectability. A "site" is typically defined as a geographic location visited by the same observer within a period of temporal closure. To define these sites and filter out observations that do not correspond to repeat visits, users should use filter_repeat_visits(), then pass the output to this function.

format_unmarked_occu() is designed to prepare data to be converted into an unmarkedFrameOccu object for occupancy modeling with unmarked::occu(); however, it can also be used to prepare data for conversion to an unmarkedFramePCount object for abundance modeling with unmarked::pcount().

Value

A data frame that can be processed by unmarked::formatWide(). Each row will correspond to a unqiue site and, assuming there are a maximum of N observations per site, columns will be as follows:

The unique site identifier, named "site".
N response columns, one for each observation, named "y.1", ..., "y.N".
Columns for each of the site-level covariates.
Groups of N columns of observation-level covariates, one column per covariate per observation, names "covariate_name.1", ..., "covariate_name.N".

Examples

# read and zero-fill the ebd data
f_ebd <- system.file("extdata/zerofill-ex_ebd.txt", package = "auk")
f_smpl <- system.file("extdata/zerofill-ex_sampling.txt", package = "auk")
# data must be for a single species
ebd_zf <- auk_zerofill(x = f_ebd, sampling_events = f_smpl,
                       species = "Collared Kingfisher",
                       collapse = TRUE)
occ <- filter_repeat_visits(ebd_zf, n_days = 30)
# format for unmarked
# typically one would join in habitat covariates prior to this step
occ_wide <- format_unmarked_occu(occ,
                                 response = "species_observed",
                                 site_covs = c("latitude", "longitude"),
                                 obs_covs = c("effort_distance_km", 
                                              "duration_minutes"))
# create an unmarked object
if (requireNamespace("unmarked", quietly = TRUE)) {
  occ_um <- unmarked::formatWide(occ_wide, type = "unmarkedFrameOccu")
  unmarked::summary(occ_um)
}

# this function can also be used for abundance modeling
abd <- ebd_zf %>% 
  # convert count to integer, drop records with no count
  dplyr::mutate(observation_count = as.integer(observation_count)) %>% 
  dplyr::filter(!is.na(observation_count)) %>% 
  # filter to repeated visits
  filter_repeat_visits(n_days = 30)
# prepare for conversion to unmarkedFramePCount object
abd_wide <- format_unmarked_occu(abd,
                                 response = "observation_count",
                                 site_covs = c("latitude", "longitude"),
                                 obs_covs = c("effort_distance_km", 
                                              "duration_minutes"))
# create an unmarked object
if (requireNamespace("unmarked", quietly = TRUE)) {
  abd_um <- unmarked::formatWide(abd_wide, type = "unmarkedFrameOccu")
  unmarked::summary(abd_um)
}

[Package auk version 0.7.0 Index]

Format EBD data for occupancy modeling with unmarked