Other useful functions

Setting up your folder structure

While most of ARUtools focuses on processing recordings once they are transferred from ARUs, setting up a folder structure can greatly increase efficiency in transferring files.

To set up a folder structure you will need the hierarchical structure you wish to use and the list of sites/arus you are processing.

site_list <-
  example_sites |>
  tidyr::separate(Sites, into = c("plot", "site"), sep = "_", remove = F) |>
  dplyr::select(site_id = Sites, plot, site)


tmp_dir <- tempdir(check = T) |> paste0("/ARUtools/")
dir.create(tmp_dir)

create_directory_structure(
  hexagons = site_list$plot,
  units = site_list$site_id,
  base_dir = tmp_dir
)

This should create a series of folders with plot at the main level and site in the subdirectory below that.

list.dirs(tmp_dir, full.names = F)

Note that this will not work for all project structures and you should think carefully about how you want to set up your file names, folder structure and spatial information before you deploy any ARUs.

Wind processing

One issue that can cause difficulty in interpretation of acoustic recordings is wind. Wind can mask bird songs and even is a potential danger to interpreters’ ears.

The University Of Salford Acoustics Research Centre developed a softare program WindNoiseDetection that detects wind in wave files.

I have developed a fork of the software that has added the ability to run multiple files at once using parallel processing and to provide a list of files to process.

Running the program requires fairly complex setup in Windows as is uses C and C++ and requires Cygwin to run.

However if you do get it running, ARUtools includes a couple helper functions to process your metadata and set it up for running with WindNoiseDetection.

wind_files <-
  wind_detection_pre_processing(
    wav_files = example_clean$path,
    output_directory = "./wind_files/",
    site_pattern = create_pattern_site_id(
      p_digits = c(2, 3), sep = "_",
      s_digits = c(1, 2)
    ),
    write_to_file = F, chunk_size = NULL
  )

The output is a list of vectors that include the path to the wave files (filePaths), the input wave filenames (filenames), and the list of sites to append to the output results (sites).

Once you have run WindNoiseDetection you can read the results in using wind_detection_summarize_json().

example_json <- system.file("extdata", "P71-1__20210606T232500-0400_SS.json", package = "ARUtools")

wind_summary <- wind_detection_summarize_json(example_json)
dplyr::glimpse(wind_summary)
#> Rows: 1
#> Columns: 7
#> $ totalwindless <dbl> 268.58
#> $ pwindless     <dbl> 0.8966715
#> $ n             <int> 5
#> $ length        <dbl> 299.53
#> $ mean_windless <dbl> 53.716
#> $ path          <chr> "/cygdrive/P/Path/To/WaveFile/NL/P71/P71-1/20210606_Napk…
#> $ jsonF         <chr> "P71-1__20210606T232500-0400_SS.json"

Assign tasks

To assign tasks you will need to either download the task template from ‘WildTrax’ or alternatively you can use the new wildRtrax::wt_make_aru_tasks() function.

You will also need a template for observers with the number of hours they will be interpreting. This doesn’t have to match exactly the time in a project as the relative amounts are used.

in_tasks <- fs::file_temp("Input_task_file", ext = ".csv")
task_template <- wildRtrax::wt_make_aru_tasks(
  example_clean |>
    dplyr::mutate(
      recording_date_time = date_time,
      file_path = path, location = site_id,
      length_seconds = 300
    ),
  output = in_tasks,
  task_method = "1SPT", task_length = 300
)

template_observers
#> # A tibble: 4 × 2
#>   transcriber                     hrs
#>   <chr>                         <dbl>
#> 1 Charles Dickens                 5  
#> 2 John Von Jovie                  1  
#> 3 John Yossarian                  6.5
#> 4 Rodion Romanovich Raskolnikov   2.3

Once you have the files you need, you can run wt_assign_tasks() to randomly assign tasks to interpreters based on the amount of effort they can put in.

task_output <- wt_assign_tasks(
  wt_task_template_in = task_template,
  wt_task_output_file = NULL,
  interp_hours = template_observers,
  interp_hours_column = hrs,
  random_seed = 65416
)

task_output$task_summary
#> # A tibble: 4 × 5
#>   transcriber                   hrs_assigned   hrs   phrs updated_hrs_remain
#>   <chr>                                <dbl> <dbl>  <dbl>              <dbl>
#> 1 John Yossarian                       1.08    6.5 0.439               5.42 
#> 2 Rodion Romanovich Raskolnikov        1.08    2.3 0.155               1.22 
#> 3 Charles Dickens                      1.17    5   0.338               3.83 
#> 4 John Von Jovie                       0.167   1   0.0676              0.833

You can alternatively use the Shiny app Shiny_select by running the following:

shiny::runGitHub("dhope/Shiny_select")