library(knitr)
# Knit hook to truncate output.
hook_output <- knit_hooks$get("output")
knit_hooks$set(output = function(x, options) {
    lines <- options$output.lines
    if (is.null(lines)) {
        return(hook_output(x, options)) # pass to default hook
    }
    x <- unlist(strsplit(x, "\n"))
    more <- "..."
    if (length(lines) == 1) { # first n lines
        if (length(x) > lines) {
            # truncate the output, but add ....
            x <- c(head(x, lines), more)
        }
    } else {
        x <- c(more, x[lines], more)
    }
    # paste these lines together
    x <- paste(c(x, ""), collapse = "\n")
    hook_output(x, options)
})
# Comments as #> not ##
knitr::opts_chunk$set(
    comment = "#>"
)
# All the API and download calls for remote data used in building this
# vignette is mocked and served up locally from the inst/api_data directory.
safedata:::mock_api(on = TRUE)

The safedata package makes it easy to search for and use datasets collected at the SAFE Project. It provides an interface to download data files and packaged record metadata and then functions to load data worksheets and add taxonomic and spatial data where available.

For further information on the publication and structure of data through the SAFE Project and within the safedata package, see the Overview vignette: vignette("overview", package = "safedata").

Installing safedata

The safedata package is available from CRAN:

install.packages("safedata")

The development version can also be installed from GitHub:

devtools::install_github("ImperialCollegeLondon/safedata@develop")

Package dependencies

The safedata package requires the following packages:

The SAFE data directory

The safedata package makes use of a local directory to store downloaded data, index and metadata files (vignette("overview", package = "safedata") for details) . These files are needed for the safedata functions to work correctly, so the first step in using safedata is to set the location of the directory and the package will remind you to do this when it is loaded.

library(safedata)

Initialising a SAFE data directory

If this is the first time you are loading safedata -- or if you simply want to have two separate SAFE data directories -- then you need to create a new, empty safedata directory. You will need the URL of the server that holds the dataset metadata for your data community.

# Create a temporary location for use in the vignette
safe_dir_path <- file.path(tempdir(), "my_safe_directory")
# Create a new safedata directory
create_safedata_dir(
    safe_dir_path,
    url = "http://example.safedata.server"
)

This will create the directory and download the current index files. You cannot use an existing directory: the package wants to start with a fresh, empty directory. Note that the directory path is stored in options():

getOption("safedata.dir")

Using an existing SAFE data directory

Once you have a SAFE data directory, the set_safedata_dir function is used to tell the safedata package where to look for index and data files:

set_safedata_dir(safe_dir_path)

You will see that this function checks with the metadata server for updates to the key index files. This can be turned off for offline use (set_safedata_dir('~/my_safe_directory', update=FALSE)). The function also validates local data files: it checks the MD5 hash of local data file copies against the MD5 of the published file.

Finding data

You can browse datasets published using the safedata system either at the project Zenodo site or from its metadata server. For example, for the SAFE Project these links are:

If you have done this, or have a dataset DOI from another source, then you can look up the dataset directly.

However, if you want to search the dataset metadata or the taxa and locations covered by datasets then there are a set of search functions built into the safedata package.

Search functions

The safedata package contains a set of search functions to explore datasets. These functions make use of a metadata index stored on the SAFE Project website and so need an internet connection to work. These search functions provide structured access to the same metadata shown in project description text but also provide extended taxonomic and spatial searches.

The functions are:

All of these functions return a safe_record_set objects, which is just a data frame containing validated record ids and access information and so you can use the normal data frame indices (e.g. recs[1,]) to select particular records.

soil_datasets <- search_text("soil")
print(soil_datasets)

Taxon search details

Published datasets contain a taxonomic index of any organisms referred to within the data - see here for details of the Taxa worksheet containing the index.

The taxa in this index, along with all of the parent taxa in the taxonomic hierarchy leading up to that those taxa, are added to a taxonomic database on the SAFE Project website. The search_taxa() function searches that index to identify all the datasets that contain a particular taxon.

ants <- search_taxa("Formicidae")
print(ants)

The taxonomic index is built around the GBIF and NCBI taxonomic databases. The GBIF database only uses the following core taxonomic levels: kingdom, phylum, class, order, family, genus, species and subspecies. Although the NCBI uses a much wider range of taxonomic levels, we only include these core backbone ranks for NCBI taxa, although we add superkingdom, which is used for bacteria and viruses.

It is also possible to search the taxon index using the taxon index number from one of GBIF or NCBI. For example, to search for the GBIF ID for Formicidae (4234) .

ants <- search_taxa(taxon_id = 4342, taxon_auth = "GBIF")

Spatial search details

Datasets also have to provide a full index of sampling locations used in the data. Sampling locations are either linked to existing sampling locations included in the SAFE gazetteer or users can identify new sampling locations and provide location data if possible.

The search_spatial() function allows users to search for datasets by sampling locations. Accepted location names from the gazetteer can be used to search for datasets but users can also provide their own search geometries using the Well Known Text format. The search includes simple GIS capabilities to look for sampling within a given distance of the query location.

# Datasets that include sampling within the location of experimental block A
within_a <- search_spatial(location = "BL_A")
# Datasets that sampled within 2 km of the Maliau Basin Field Study Centre
near_maliau <- search_spatial(wkt = "POINT(116.97394 4.73481)", distance = 2000)

Note that WKT coordinates should be supplied as WGS84 longitude and latitude - typically the output of GPS receivers - but the database uses the local projected coordinate system for all distance calculations and GIS operations.

Combining searches

It is possible to combine searches using logical operators (& and |), which simply find the intersection and unions of sets of SAFE records.

# Three searches
fish <- search_taxa("Actinopterygii")
odonates <- search_taxa("Odonata")
ewers <- search_authors("Ewers")
# Logical combinations
aquatic <- fish | odonates
aquatic_ewers <- aquatic & ewers
all_in_one <- (fish | odonates) & ewers

Another approach is to restrict the records that will be searched in the online database. You can pass a search result into a second search result to only search within those records. This cuts down on the amount of information that has to be retrieved from the server, so might be faster on poor networks. However, this obviously only works for repeated narrowing of a search, so has less functionality.

fish <- search_taxa("Actinopterygii")
ewers <- search_authors("Ewers", ids = fish)

Look up a specific dataset

Datasets are identified by their record number, which is the number included in both the dataset DOI and Zenodo URL. All of the following point to the same dataset:

Note that all metadata is available for all records, regardless of whether they are open, embargoed or restricted. This includes field descriptions and taxon and location sampling so that users can assess whether a dataset is going to be useful even if it is not yet openly available.

Once you have the details of a dataset you are interestested in then you can validate a dataset reference to access metadata and available data, using the validate_record_ids() function. This function does the following:

  1. checks that the record is valid,
  2. checks whether the record number is a record id, referring to a specific version of a dataset, or a concept id, which identifies all the versions of a dataset. In the example code below, two of the values are record ids, so the appropriate concept id is located and printed, and one is a concept id, so no specific version number is given.
  3. checks whether the data are currently available, and
  4. provides an interface to download and import the related data files.

Like the search functions, the output is safe_record_set object. Note that you can validate multiple references at once.

recs <- validate_record_ids(c(
    "https://doi.org/10.5281/zenodo.3247631",
    "10.5281/zenodo.3266827",
    "https://zenodo.org/record/3266821"
))
print(recs)

In addition, all of the main functions in safedata that expect to be passed a dataset id will run validate_record_ids() on their inputs, so you can simply use those URLs directly with those functions without needing to specifically create a safe_record_set yourself.

Displaying dataset metadata

Printing a safe_record_set object displays a deliberately compact summary of a set of record ids. There are three function that show the detailed metadata for records at three levels:

show_concepts()

The show_concepts() function displays concept level metadata about a set of record ids. This includes the (most recent) dataset title and a short summary of the versions available under the dataset concept. Note that the output is not restricted just to the set of record ids given to the function: it shows metadata for all versions for each of the concept ids included.

show_concepts(recs)

show_record()

This function shows metadata for a specific version of a dataset: if you give it a concept ID then it will display the available versions for that concept. Otherwise, the function prints out information about the dataset with that record id: it includes the dataset title, status and other dataset level metadata and then a summary of the data worksheets contained in the dataset.

Note that - because a safe_record_set is just a data frame with some extra information attached - you can use the usual data frame indexing to select a row to pass to other functions. Running show_record() also requires an internet connection when a record is first examined: the package downloads a JSON file of the record metadata and stores it in the SAFE data directory.

show_record(1400562)

show_worksheet()

This function shows metadata for a named worksheet within a specific record. The default is to show a compact table of field names, field types and truncated descriptions:

show_worksheet(1400562, "EnvironVariables")

There is also an extended display (extended_fields=TRUE) that will print out a list of all the available metadata for each field.

show_worksheet(1400562, "EnvironVariables", extended_fields = TRUE)

Downloading data

Once you have found records for which you want to explore the actual data, then you need to download the data files for the dataset from Zenodo to work with it. There are two ways to do this

Bulk loading data

The download_safe_files() function can be used to download the data files for the data records in a safe_record_set, such as those created using a search. Alternatively, you can use the URL of a safedata record or simply use the record id for the dataset.

The function will check which datasets are currently available and download them to the safedata directory. The default behaviour is to present a brief report on the number and size of available files to be downloaded and ask for confirmation before actually doing anything. The confirm=FALSE argument can be used to download without confirmation, which can be useful in non-interactive use. The function will also download the JSON metadata for the specified datasets.

download_safe_files(within_a)
## 26 files requested from 26 records
##  - 0 local (0 bytes)
##  - 4 embargoed or restricted (2.2 Mb)
##  - 22 to download (43.6 Mb)
##
## 1: Yes
## 2: No
##
## Selection:

By default, the download_safe_files() function downloads all of the files associated with the record. In addition to the core Excel file, this will also download any additional data files which may contain primary data that is either not suited to the Excel format or which provides additional information, such as PDFs. Although some additional file formats are likely to be readable in R, thesafedata package does not currently provide a mechanism to load them automatically.

download_safe_files(3697804, xlsx_only = FALSE)
# 2 files requested from 1 records
#  - 1 local (11.3 Kb)
#  - 0 embargoed or restricted (0 bytes)
#  - 1 to download (74.1 Kb)
#
# 1: Yes
# 2: No
#
# Selection: 1
# 2 files for record 3697804: 1 to download
#  - Downloaded: Sampling_area_borders.xlsx

The function will warn you if the local copies of data files have been altered and the refresh=TRUE argument can be used to restore data files to the version of record. Note that this will delete local changes.

Loading individual datasets

The load_safe_data function is used to load a named data worksheet into R. If the datafile containing the worksheet is not currently downloaded, then load_safe_data will automatically download the file if available. In this example, confirm=FALSE is passed onto download_safe_files to remove the interactive confirmation before downloading.

beetle_abund <- load_safe_data(1400562, "Ant-Psel", confirm = FALSE)

The safedata object

Once a worksheet is loaded, it is stored as an object of class safedata. This is just a data frame with some additional attribute data and it will in general behave just like any other data frame - the additional attributes are used for further data processing and adding brief metadata to the str and print methods.

Some data formatting takes place based on field types: categorical variables are converted to factors; dates and datetimes are converted to POSIXct and times are converted to chron::time objects.

str(beetle_abund)
print(beetle_abund)

The display of safedata objects is kept deliberately simple to avoid cluttering the screen with metadata. You can always view additional metadata for a loaded worksheet by using the show functions directly on the loaded safedata object:

show_concepts(beetle_abund)
show_record(beetle_abund)
show_worksheet(beetle_abund)

Dataset taxa

There are a number of functions that can be used to work with the taxa in a dataset. Some generate simple tables of taxa:

  1. get_taxa(): This function loads a dataframe containing all of the taxa used within a dataset, with fields including the core GBIF taxonomic levels, the taxonomic label used within the dataset and the taxonomic status of the each taxon. You can load a taxonomic dataframe from a safe_record_set row or using an existing loaded safedata object.

    r beetle_taxa <- get_taxa(beetle_abund) str(beetle_taxa)

  2. add_taxa(): This function adds taxonomic details to an already loaded data worksheet.

    r beetle_morph <- load_safe_data(1400562, "MorphFunctTraits") beetle_morph <- add_taxa(beetle_morph) str(beetle_morph)

  3. The get_taxon_coverage function can be used to get a taxon table of all taxa currently referenced in all datasets.

    r all_taxa <- get_taxon_coverage()

The other functions convert taxon tables into graphs (vertices and edges) and phylogenetic trees using the GBIF taxonomic backbone to represent phylogeny. The data validation should ensure that the taxa in a dataset can be connected as a phylogenetic tree, but this is not always the case. For this reason, these functions use a more general graph conversion and then checks whether the result can be converted to a phlyogenetic tree. Technically, that is checking that the graph is a connected, simple, directed acyclic graph.

  1. The get_taxon_graph function converts a taxon table to a graph. The function also does some extra validation, and will give warnings when the taxon table has to be adjusted to represent the taxonomic hierarchy and worksheet taxa properly. The result is an object of class igraph and the vertices have attributes containing the original table data.

  2. The igraph_to_phylo function tests whether a taxon graph can be converted to a phylogeny and returns a phylo object (package ape). The tips and internal nodes are labelled with taxon names but the phylo structure does not store other node information.

  3. get_phylogeny(): This is simply a wrapper that runs both functions above in turn to go straight to a phylogeny.

    r library(ape) beetle_phylo <- get_phylogeny(1400562) plot(beetle_phylo, show.node.label = TRUE, font = 1, no.margin = TRUE)

Dataset locations

Nearly all SAFE datasets will include observations at spatial locations, and these datasets must include a Locations worksheet used as an spatial index for research activities. There are three functions that can be used to work with locations in a dataset. All of these functions use the sf package to represent the GIS geometry of locations and which provides an extensive toolset for further spatial analysis.

  1. load_gazetteer(): The gazetteer is one of the three key index files saved in the SAFE data directory and updated when set_safedata_dir() is run. It includes sampling locations drawn from across a wide range of projects running at SAFE and is intended to hold all locations that are likely to see repeated sampling. Locations included the gazetteer can be used directly as known locations in datasets, although data providers can also include new locations.

    r gazetteer <- load_gazetteer() print(gazetteer)

  2. get_locations(): This function returns an sf object containing the locations used within a dataset. For known locations, the GIS data for the location are taken directly from the gazetteer. If the locations are new sampling sites, then GIS data provided in the dataset is used. Note that it is possible for dataset providers to create a new locations without including GIS data - these will be represented using empty GIS geometries.

    By default, the returned sf object will only include the location name used in the dataset, the gazetteer name for known sampling sites and an indication of whether the location is new or known, but gazetteer_info=TRUE can be used to include the gazetteer attributes for known locations.

    r library(sf) beetle_locs <- get_locations(1400562) print(beetle_locs) fragments <- subset(gazetteer, type == "SAFE forest fragment") par(mar = c(3, 3, 1, 1)) plot(st_geometry(fragments), col = "khaki", graticule = TRUE) plot(st_geometry(beetle_locs), add = TRUE, col = "red", pch = 4)

  3. add_locations(): This functions adds location data to an already loaded worksheet. The result is a safedata object that is also an sf object.

    r beetle_env <- load_safe_data(1400562, "EnvironVariables") beetle_env <- add_locations(beetle_env) print(beetle_env) plot(beetle_env["Cover"], key.pos = 4, breaks = seq(0, 100, by = 5))

Inserting datasets

If you want access to a dataset that is currently embargoed or restricted, then you can approach the authors listed on the Zenodo record to ask for permission to use the data. If you are then provided with the files for the dataset, then they need to be inserted into your local SAFE data directory so that they can be accessed by the safedata functions.

The insert_dataset function does this: it will check to see if a set of files are part of specified Zenodo record and then copy them into the correct places in the current SAFE data directory.

files <- system.file("api_data", "template_ClareWfunctiondata.xlsx",
    package = "safedata"
)
insert_dataset(1237719, files)
dat <- load_safe_data(1237719, "Data")
str(dat)

File details and accessing files

At the moment, the safedata package only handles loading data from the core Excel files. We do intend to add functions and recipes to access other files stored within a dataset in the future. At the moment, if you have downloaded all the files associated with a dataset then the get_file_details function allows you to quickly see a list of the files in a dataset, whether you have downloaded a local copy and the absolute file path of those files.

get_file_details(1237719)
# Turn off URL mocking and tidy up
safedata:::mock_api(on = TRUE)
unlink(safe_dir_path, recursive = TRUE)


ImperialCollegeLondon/safe_data documentation built on Jan. 27, 2024, 9:51 a.m.