galah is an R interface to biodiversity data hosted by the Global Biodiversity Information Facility (GBIF) and its subsidiary node organisations. GBIF and its partner nodes collate and store observations of individual life forms using the ‘Darwin Core’ data standard.

Installation

To install from CRAN:

install.packages("galah")

Or install the development version from GitHub:

install.packages("remotes")
remotes::install_github("AtlasOfLivingAustralia/galah")

Load the package

library(galah)

Configuration

Begin by choosing which organisation you would like galah to query, and providing your registration information for that organisation.

galah_config(atlas = "GBIF",
             username = "user1",
             email = "email@email.com",
             password = "my_password")

The full list of supported queries by organisation is as follows:

Fig 1: Organisations and APIs supported by galah

Getting data

galah is a dplyr extension package; rather than using pipes to amend a tibble in your workspace, you amend a query, which is then sent to your chosen organisation. These pipes differ from traditional syntax in two ways:

they begin with a function - usually galah_call() - instead of a tibble
they end with one of dplyr’s evaluation functions, usually collect()

So an example query might be to find the number of records per year:

galah_config(atlas = "Australia")

galah_call() |>            # open a pipe
  filter(year >= 2020) |>  # choose rows to keep
  count(year) |>           # count the number of rows
  collect()                # retrieve query from the server

## # A tibble: 7 × 2
##   year     count
##   <chr>    <int>
## 1 2024  11889930
## 2 2023  11007491
## 3 2022   9430065
## 4 2025   9142677
## 5 2021   8695248
## 6 2020   7311836
## 7 2026    309836

Or to find the number of categories present in a dataset, for example how many species are present:

galah_call() |>
  identify("Crinia") |>   # filters by taxonomic names
  distinct(speciesID) |>  # keep only unique values
  count() |>
  collect()

## # A tibble: 1 × 1
##   count
##   <int>
## 1    17

You can ‘glimpse’ a data download before you run it, to check all the data you need is included:

galah_call() |>
  identify("Eolophus roseicapilla") |> 
  filter(year == 2010) |>
  glimpse() |>
  collect()

## Rows: 21,984
## Columns: 8
## $ taxonConceptID   <chr> "https://biodiversity.org.au/afd/taxa/9b4ad548-8bb3-486a-ab0a-905506c463ea", "https://biodiversity.org.au…
## $ eventDate        <dbl> 1.272672e+12, 1.289002e+12, 1.291014e+12
## $ scientificName   <chr> "Eolophus roseicapilla", "Eolophus roseicapilla", "Eolophus roseicapilla"
## $ decimalLatitude  <dbl> -25.98833, -37.83032, -35.41707
## $ decimalLongitude <dbl> 152.0442, 144.9812, 138.6868
## $ basisOfRecord    <chr> "HUMAN_OBSERVATION", "HUMAN_OBSERVATION", "HUMAN_OBSERVATION"
## $ dataResourceName <chr> "BirdLife Australia, Birdata", "eBird Australia", "eBird Australia"
## $ occurrenceStatus <chr> "PRESENT", "ABSENT", "ABSENT"

And, once satisfied that your parameters are correct, download the records themselves:

galah_call() |>
  identify("Eolophus roseicapilla") |> 
  filter(year == 2010) |>
  select(eventDate, decimalLatitude, species) |>
  collect()

## # A tibble: 21,984 × 3
##    eventDate decimalLatitude species              
##    <dttm>              <dbl> <chr>                
##  1 NA                  -36.5 Eolophus roseicapilla
##  2 NA                  -38.2 Eolophus roseicapilla
##  3 NA                  -37.0 Eolophus roseicapilla
##  4 NA                  -37.7 Eolophus roseicapilla
##  5 NA                  -35.6 Eolophus roseicapilla
##  6 NA                  -31.1 Eolophus roseicapilla
##  7 NA                  -38.2 Eolophus roseicapilla
##  8 NA                  -38.2 Eolophus roseicapilla
##  9 NA                  -38.2 Eolophus roseicapilla
## 10 NA                  -38.2 Eolophus roseicapilla
## # ℹ 21,974 more rows

This works because many of the functions in dplyr are “generic”, meaning it is possible to write extensions that apply them to new object classes. In our case, galah_call() creates a new object class called a data_request for which we have written new extensions. This means that galah will not interfere with your use of filter() and friends on your tibbles. Supported dplyr verbs that modify queries are as follows:

arrange.data_request()
count.data_request()
distinct.data_request()
filter.data_request()
glimpse.data_request()
group_by.data_request()
select.data_request()
slice_head.data_request()

Additional verbs are:

apply_profile()
geolocate() or st_crop.data_request()
identify.data_request()
unnest()

It is good practice to download your data in as few steps as possible, to minimize impacts on the server, and to ensure you can get a single DOI for your data. See the download data reproducibly vignette for details.

Finding information

Building queries using filter() requires that you know two things:

what fields (columns) are present in the dataset you are searching
what values exist for those fields

Finding this information requires looking for metadata:

request_metadata(type = "fields") |>
  collect()

## # A tibble: 639 × 3
##    id                  description               type  
##    <chr>               <chr>                     <chr> 
##  1 abcdTypeStatus      <NA>                      fields
##  2 acceptedNameUsage   Accepted name             fields
##  3 acceptedNameUsageID Accepted name             fields
##  4 accessRights        Access rights             fields
##  5 annotationsDoi      <NA>                      fields
##  6 annotationsUid      Referenced by publication fields
##  7 assertionUserId     Assertions by user        fields
##  8 assertions          Record issues             fields
##  9 assertionsCount     <NA>                      fields
## 10 associatedMedia     Associated Media          fields
## # ℹ 629 more rows

You can browser this tibble using View() or search it using filter(). Once you have found a field that you want to include in your query, you can find values for that field using unnest():

request_metadata() |>
  filter(fields == "cl22") |>
  unnest() |>
  collect()

## # A tibble: 11 × 1
##    cl22                        
##    <chr>                       
##  1 New South Wales             
##  2 Victoria                    
##  3 Queensland                  
##  4 South Australia             
##  5 Western Australia           
##  6 Northern Territory          
##  7 Tasmania                    
##  8 Australian Capital Territory
##  9 Macquarie Island            
## 10 Coral Sea Islands           
## 11 Ashmore and Cartier Islands

Different types of metadata are available; see ?request_metadata for a full list.

Wrapper functions

While dplyr syntax is very flexible, there are cases where it is easier to simply say the sort of data you want, rather than create a database query to implement it. For this reason, several common use cases have their own wrapper functions.

The atlas_ family of functions act like collect(), but enforce a particular type of data to be returned, such as record counts:

galah_call() |>
  filter(year == 2025) |>
  atlas_counts()   # note no need for a `count()` function

## # A tibble: 1 × 1
##     count
##     <int>
## 1 9142677

Or occurrences:

galah_call() |>
  identify("Eolophus roseicapilla") |>
  filter(year == 2000,
         cl22 == "Australian Capital Territory") |>
  atlas_occurrences() |>
  print(n = 6)

## # A tibble: 2,032 × 9
##   recordID         scientificName taxonConceptID decimalLatitude decimalLongitude eventDate           basisOfRecord occurrenceStatus
##   <chr>            <chr>          <chr>                    <dbl>            <dbl> <dttm>              <chr>         <chr>           
## 1 0026d29f-b6ab-4… Eolophus rose… https://biodi…           -35.4             149. 2000-08-07 00:00:00 HUMAN_OBSERV… PRESENT         
## 2 0062d446-007b-4… Eolophus rose… https://biodi…           -35.3             149. 2000-03-10 00:00:00 HUMAN_OBSERV… PRESENT         
## 3 00a62ee0-1e08-4… Eolophus rose… https://biodi…           -35.2             149. 2000-01-29 00:00:00 HUMAN_OBSERV… PRESENT         
## 4 00ab2f4d-326f-4… Eolophus rose… https://biodi…           -35.4             149. 2000-09-25 00:00:00 HUMAN_OBSERV… PRESENT         
## 5 00ae4631-ea59-4… Eolophus rose… https://biodi…           -35.3             149. 2000-02-12 00:00:00 HUMAN_OBSERV… PRESENT         
## 6 00b6c8ec-e7b9-4… Eolophus rose… https://biodi…           -35.2             149. 2000-02-05 00:00:00 HUMAN_OBSERV… PRESENT         
## # ℹ 2,026 more rows
## # ℹ 1 more variable: dataResourceName <chr>

atlas_species() replaces the need for distinct() call, while atlas_media() is a shortcut to a complex workflow that incorporates both data and metadata calls. Finally, metadata calls can be made more efficiently using the show_all() and show_values() functions. These take the same arguments as the type argument in request_metadata(), but use non-standard evaluation, so they don’t require quotes. They are also evaluated immediately rather than lazily:

show_all(fields)

## # A tibble: 639 × 3
##    id                  description               type  
##    <chr>               <chr>                     <chr> 
##  1 abcdTypeStatus      <NA>                      fields
##  2 acceptedNameUsage   Accepted name             fields
##  3 acceptedNameUsageID Accepted name             fields
##  4 accessRights        Access rights             fields
##  5 annotationsDoi      <NA>                      fields
##  6 annotationsUid      Referenced by publication fields
##  7 assertionUserId     Assertions by user        fields
##  8 assertions          Record issues             fields
##  9 assertionsCount     <NA>                      fields
## 10 associatedMedia     Associated Media          fields
## # ℹ 629 more rows

You can check the look up information vignette for further details.

Quick start guide

Martin Westgate & Dax Kellie

2026-02-11

Installation

Configuration

Getting data

Finding information

Wrapper functions