Compare Herbarium Sources

Overview

When working with plant specimen data from multiple biodiversity repositories (e.g., GBIF, JABOT, speciesLink), users often encounter overlapping records across sources. barRoso offers the barroso_cat() function to merge, harmonize, and optionally deduplicate these datasets based on collection codes.

This article demonstrates how to:

• Merge multiple data sources into a unified data frame
• Combine raw speciesLink downloads with GBIF downloads
• Prioritize a preferred source when herbaria overlap

Function: barroso_cat()

combined_df <- barroso_cat(
  list_sources = list(
    GBIF = gbif_data,
    speciesLink = splink_data,
    JABOT = jabot_data
  ),
  keep_source = "GBIF"
)

Arguments

• list_sources: A named list of data frames, one per biodiversity source. The names (e.g. GBIF, speciesLink) must be unique, as they are used to identify the source of each record.
• keep_source: Optionally, the name of the preferred source (e.g. "GBIF"). The records of any herbarium also present in the preferred source are removed from the other sources.

If no source is specified, the function merges all sources, retaining potential duplicates for further reconciliation with barroso_flag_duplicates().

Example

library(barRoso)

# Load datasets downloaded from GBIF and speciesLink
gbif <- read.csv("gbif.csv")
splink <- read.csv("speciesLink.csv", fileEncoding = "latin1")

# Merge, giving preference to GBIF for overlapping herbaria
combined_df <- barroso_cat(
  list_sources = list(
    GBIF = gbif,
    speciesLink = splink
  ),
  keep_source = "GBIF"
)

table(combined_df$datasource)

For example, when combining 9,806 GBIF records and 4,432 speciesLink records of Ormosia, 1,259 speciesLink records of 74 herbaria also available in GBIF are removed, resulting in 12,979 records (9,806 from GBIF and 3,173 from speciesLink).

How It Works

1. Raw speciesLink columns are renamed into Darwin Core

Each source is first formatted with barroso_dwc(), which can also be used alone.

Raw speciesLink downloads use lowercase column names that differ from the Darwin Core terms used by GBIF. barroso_cat() detects them and renames them automatically, so records of both sources end up in the same columns:

speciesLink	Darwin Core
collector	recordedBy
collectornumber	recordNumber
collectioncode	collectionCode
institutioncode	institutionCode
catalognumber	catalogNumber
identifiedby	identifiedBy
typestatus	typeStatus
yearcollected, monthcollected, daycollected	year, month, day
stateprovince	stateProvince
latitude, longitude	decimalLatitude, decimalLongitude
species	specificEpithet

In speciesLink, the column species is just the specific epithet (e.g. arborea), while in GBIF it is the binomial (e.g. Ormosia arborea). So the speciesLink species becomes specificEpithet, and a binomial species column is created, as in GBIF.

2. REFLORA and JABOT values are standardized

Records downloaded with refloraR and jabotR already have Darwin Core column names, but some of their values are written differently from GBIF. barroso_cat() recognizes these sources by their content, whatever their names in list_sources, and standardizes them:

Source	Column	Before	After
REFLORA	identifiedBy	Administrador (the system user, in most records)	NA
REFLORA	eventDate, dateIdentified	2/12/1940, --/5/1880, 9/10/1990 até 12/10/1990	1940-12-02, 1880-05, 1990-10-09/1990-10-12
REFLORA	typeStatus	HOLOTYPUS, ISOTYPUS, TYPUS	holotype, isotype, type
REFLORA	institutionCode of RB	RB	JBRJ, as in GBIF and JABOT
JABOT	eventDate, dateIdentified	2008-9-26, 22/4/2020, 4/2020	2008-09-26, 2020-04-22, 2020-04

library(refloraR)
library(jabotR)

reflora <- reflora_records(herbarium = c("RB", "K"), taxon = "Ormosia", save = FALSE)
jabot <- jabot_records(herbarium = c("MBM", "UPCB"), taxon = "Ormosia", save = FALSE)

combined_df <- barroso_cat(
  list_sources = list(GBIF = gbif, speciesLink = splink, REFLORA = reflora, JABOT = jabot),
  keep_source = "GBIF"
)

The REFLORA collection of RB is fed by JABOT, so the same RB specimens are often found in both sources, and also in GBIF. With keep_source = "GBIF", they are kept just once.

3. Overlapping herbaria are removed when keep_source is given

Overlapping herbaria are detected with collectionCode. When keep_source = "GBIF", the records of any herbarium present in GBIF are removed from the other sources. Herbaria shared only among the other sources (e.g. present in both speciesLink and JABOT, but not in GBIF) are kept in all of them, and their duplicates can then be flagged with barroso_flag_duplicates().

4. Columns are aligned

• Columns missing in some sources are filled with NA.
• Columns with different types across sources (e.g. year as numbers in GBIF and as text in speciesLink) are combined as text, instead of failing.
• A new column datasource stores the source of each record (e.g. "GBIF", "speciesLink").

Tips

• Name each data source in list_sources, e.g. list(GBIF = gbif, speciesLink = splink)
• Ensure each dataset includes a collectionCode column, otherwise overlapping herbaria cannot be detected for that source
• Use keep_source = NULL if you want to preserve all records
• Use barroso_std() after combining to clean the remaining fields and flag duplicates

cleaned_df <- barroso_std(combined_df, flag_duplicates = TRUE)

See Also

• barroso_cat() reference
• Detect and Flag Duplicates