Hierarchical modeling of abundance and occurrence requires repeat visits to sites to estimate detectability. These visits should be all be within a period of closure, i.e. when the population can be assumed to be closed. eBird data, and many other data sources, do not explicitly follow this protocol; however, subsets of the data can be extracted to produce data suitable for hierarchical modeling. This function extracts a subset of observation data that have a desired number of repeat visits within a period of closure.

  min_obs = 2L,
  max_obs = 10L,
  annual_closure = TRUE,
  n_days = NULL,
  date_var = "observation_date",
  site_vars = c("locality_id", "observer_id"),
  ll_digits = 6L



data.frame; observation data, e.g. data from the eBird Basic Dataset (EBD) zero-filled with auk_zerofill(). This function will also work with an auk_zerofill object, in which case it will be converted to a data frame with collapse_zerofill(). Note that these data must for a single species.


integer; minimum number of observations required for each site.


integer; maximum number of observations allowed for each site.


logical; whether the entire year should be treated as the period of closure (the default). This can be useful, for example, if the data have been subset to a period of closure prior to calling filter_repeat_visits().


integer; number of days defining the temporal length of closure. If annual_closure = TRUE closure periods will be split at year boundaries. If annual_closure = FALSE the closure periods will ignore year boundaries.


character; column name of the variable in x containing the date. This column should either be in Date format or convertible to Date format with as.Date().


character; names of one of more columns in x that define a site, typically the location (e.g. latitude/longitude) and observer ID.


integer; the number of digits to round latitude and longitude to. If latitude and/or longitude are used as site_vars, it's usually best to round them prior to identifying sites, otherwise locations that are only slightly offset (e.g. a few centimeters) will be treated as different. This argument can also be used to group sites together that are close but not identical. Note that 1 degree of latitude is approximately 100 km, so the default value of 6 for ll_digits is equivalent to about 10 cm.


A data.frame filtered to only retain observations from sites with the allowed number of observations within the period of closure. The results will be sorted such that sites are together and in chronological order. The following variables are added to the data frame:

  • site: a unique identifier for each "site" corresponding to all the variables in site_vars and closure_id concatenated together with underscore separators.

  • closure_id: a unique ID for each closure period. If annual_closure = TRUE this ID will include the year. If n_days is used an index given the number of blocks of n_days days since the earliest observation will be included. Note that in this case, there may be gaps in the IDs.

  • n_observations: number of observations at each site after all filtering.


In addition to specifying the minimum and maximum number of observations per site, users must specify the variables in the dataset that define a "site". This is typically a combination of IDs defining the geographic site and the unique observer (repeat visits are meant to be conducted by the same observer). Finally, the closure period must be defined, which is a period within which the population of the focal species can reasonably be assumed to be closed. This can be done using a combination of the n_days and annual_closure arguments.

See also

Other modeling: format_unmarked_occu()


# 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) filter_repeat_visits(ebd_zf, n_days = 30)
#> # A tibble: 181 x 37 #> site closure_id n_observations checklist_id last_edited_date country #> <chr> <chr> <int> <chr> <chr> <chr> #> 1 L105… 2012-9 2 S11680564 2017-10-19 12:4… Singap… #> 2 L105… 2012-9 2 S11718518 2018-02-01 09:2… Singap… #> 3 L127… 2012-8 3 S11664225 2017-05-27 11:0… Singap… #> 4 L127… 2012-8 3 S11675087 2017-05-27 11:0… Singap… #> 5 L127… 2012-8 3 S11668164 2017-05-27 10:2… Singap… #> 6 L136… 2012-0 10 S9492400 2017-05-27 10:3… Singap… #> 7 L136… 2012-0 10 S9503604 2017-05-27 10:3… Singap… #> 8 L136… 2012-0 10 S9513531 2017-05-27 10:3… Singap… #> 9 L136… 2012-0 10 S9629042 2017-05-27 10:4… Singap… #> 10 L136… 2012-0 10 S9652476 2017-05-27 10:3… Singap… #> # … with 171 more rows, and 31 more variables: country_code <chr>, state <chr>, #> # state_code <chr>, county <chr>, county_code <chr>, iba_code <chr>, #> # bcr_code <int>, usfws_code <chr>, atlas_block <chr>, locality <chr>, #> # locality_id <chr>, locality_type <chr>, latitude <dbl>, longitude <dbl>, #> # observation_date <date>, time_observations_started <chr>, #> # observer_id <chr>, sampling_event_identifier <chr>, protocol_type <chr>, #> # protocol_code <chr>, project_code <chr>, duration_minutes <int>, #> # effort_distance_km <dbl>, effort_area_ha <dbl>, number_observers <int>, #> # all_species_reported <lgl>, group_identifier <chr>, trip_comments <chr>, #> # scientific_name <chr>, observation_count <chr>, species_observed <lgl>