2. Material & Methods

Conservation biology is a broad discipline encompassing different thematics. For the selection of articles, we used conservation terms that could easily be placed under the umbrella of two classification systems as keywords: the broader CBD threat classification system, and the more specific International Union for the Conservation of Nature threat (Table 1) and conservation measures (Table 2) classification systems (Annex A). Additionally, we were also looking for papers that focused on monitoring, as we felt that, although not specifically included in the original categories, automation of species identification, remote sensing through camera trapping, etc. are methods that show major growth potential in machine learning usage. Extensive consideration was put into what terms to use in the search string to strike a balance between sensitivity and specificity. As per the definition of ML stated above (see section 1.), our focus was on methods popularised recently (Table 3), with those more traditionally associated with statistics (e.g. linear regression in its most basic forms) not being considered.

2.1 Systematic search

A standardised literature search was performed in Clarivate Analytics’Web of Science on 8 June 2021 in Helsinki, Finland, using the Mozilla Firefox web browser running on Windows 10. The final search string used was composed of 33 conservation terms and 23 ML terms. For convenience, multiple searches were made, each containing only one conservation term combined with every machine learning term. Effectively, our search string was: