In biodiversity conservation, species distribution modelling can provide initial guidance on the potential distribution of a taxon of interest and help identify areas worth prioritising. A low number of species occurrences, however, may result in inconsistent predictions for some modelling algorithms.
In this study, researchers set out to inform conservation planning to protect rare insect species in Fiji. Using GBIF-mediated occurences and five environmental variables for the region, the researchers employed an ensemble-based approach to predict the distribution of long-horned beetles (the Cerambycidae family) in the Fiji islands.
Despite an inherent scarcity of Cerambycidae occurrence records, the researchers were able to produce a predictive map based on an ensemble model of the best-performing modelling algorithms, including machine learning-based SVM, randomForest and MaxEnt.
The resulting distribution map pointed to highest taxon suitability on the middle slopes of the central mountains of the islands Viti Levu, Vanua Levu and Taveuni.