The main goal on this step collect information of available resources about species interaction data, study cases and standard development, creating a list and providing a detailed description of funcionalities, advantages and disadvantages of each resource.
We analyzed 392 datasets from GloBI sources and those found with Google Dataset Search.
The Google Dataset Search was used to search for interaction datasets not indexed by GloBI, using the follow search string:
"species interaction" OR "biotic interaction" OR "interaction network" OR "trophic interaction" OR "food web" OR "plant-pollinator" OR "predator-prey"
The search returned 81 datasets which are not indexed by GloBI.
The analysis of the datasets were conducted in order to understand how people organize and share interactions data. For each dataset the follow metadata are extracted (when available):
- Metadata about the related scientific publication (i.e. DOI, title, year, abstract, publication type, keywords, authors, journal).
- Repository where the dataset is stored
- General data structure/schema used, classified in following categories:
- TABULAR/INTERACTION_RECORDS: interactions are represented by individual records (one interaction per line/row). Each record contains data about both interaction partners (source and target species).
- TABULAR/CHECK_LIST: interactions are given as a list of interaction partners for a specific taxon (not mentioned in the dataset)
- DWCA: Darwin Core Archives which may use associatedTaxa, associatedOccurrences, ResourceRelationship or another DwC term to stored interaction data.
- ADJ MATRIX: adjacency matrix are used to share interactions network where the first row/column are species name or identifiers. The adjancency matrix can be binary or non-binary matrix, and usually a 0 represents no interaction and a value greater than 0 indicates an interaction between species at given row/column.
- G(V,E): a network interactions where a list of nodes (species) and a list of links (interactions) are given separately
- NATURAL_LANGUAGE: interactions are given in natural language (textual) within publications.
- CUSTOM_API: custom API's comprising different schemas are grouped together.
- Dataset file format (e.g. CSV, ZIP, HTML, RData)
- Dataset Metadata (ie. title, description, keywords, year, metadata URL)
- Interaction types classified according to Relation Ontology.
- Interaction level:
- species-species: species interaction. No spatial or temporal information about the interacting species are given, just taxa.
- occurrence-species: Spatial or temporal information are given to just one of the interacting partners.
- occurrence-occurrence: Spatial or temporal information are given for both interacting partners.
- Relational model:
- single: data is store in just one file (flat-table).
- multiple: data is store in multiple files (relational).
Proportion of DwC-A that uses associatedTaxa, associatedOccurrences and ResourceRelationship. Data were not investigated in order to verify if interaction information are given using this terms/class. It just summarizes the proportion of DwC-A which have non-empty values of any of those terms/class.
