nature.com
Environment , taxonomy, and socioeconomics predict non-imperilment in freshwater fishes
The dataset developed for this study involved the compilation and processing of 12 global data sources (Table 1). Predictive data sources were selected based on their offering of information relevant to one or more of three broad categories: Environmental, Socioeconomic, and Intrinsic. Each of the initial 122 candidate predictive variables in our dataset was also classified into one of 13 sub-categories (i.e., Habitat, Climate, Hydrology, Economy, Development, Footprint, Threats, Impoundments, Taxonomy, Physiology, Life-history, Knowledge, Conservation), based on their domain. A predictor representing knowledge gaps was calculated using unknown attributes, the number of predictive variables with NA values, for each species. For the present study, we only used one response variable (i.e., the latest IUCN Red List conservation status), provided as binary (imperiled and non-imperiled) and ordinal (five classes) responses for random forest and ordinal forest models, respectively. For a more detailed description of each variable in the conflated dataset, see Table S1. Intrinsic and species response data We accessed species conservation data from the IUCN Red List of Threatened Species26 (2024-v2) using the IUCN Red List API: http://apiv3.iucnredlist.org. Data included the latest conservation assessments, identified threats, and necessary management actions for species recovery. Original threat classes (15) were binned into...
