Humans have affected the environment on Madagascar since their earliest arrival, directly or indirectly leading to the extinction of dozens of species. Today, while species may face different threats, many remain unassessed, and species may be disappearing before they are even described.
Using GBIF-mediated occurrence records for all vascular plants in Madagascar from iNaturalist and the Missouri Botanical Garden, the authors of this substantial review compiled a dataset of nearly 10,0000 species, representing 83.5 per cent of the known Malagasy flora, adding data on 57 features , including biome, climate, elevation, forest cover, geography, human footprint and use.
Based on this dataset, they trained a Bayesian Neural Network (BNN) using ~4,000 species for which Red List assessments already existed, then leveraged the model to predict the conservation status and threats for the remaining 5,887 unassessed species.
Across most taxonomic groups, the model predicted a similar distribution of threat categories as in the already assessed species, indicating that current percentages of threatened plants are likely representative of the entire group. Ferns and lycophytes were the notable exception, as the model predicted large deviations, suggesting an underestimate of threatened species within these groups.
To further consolidate existing knowledge on threats to Malagasy flora and fauna, the authors revealed a disproportionately high number of genetically distinct at-risk species. They showed that overexploitation and unsustainable agricultural practices threaten two thirds of all vertebrates and 90 per cent of all plants.
On a positive note, the 10 per cent of Madagascar's land area currently protected covers at least part of the range of 97 per cent of vertebrates and two-thirds of all plants.