Predicting cetacean habitats beyond surveyed regions in Indonesian waters: species distribution model transferability may not always be preferable

Abstract

Abstract Understanding the distribution of cetaceans in Indonesian waters is imperative for their conservation management, however such information is lacking for the country. Our study predicted the species distributions of two cetacean species (common bottlenose dolphin and sperm whale) beyond the surveyed regions in Indonesian Exclusive Economic Zone (EEZ). This was done by using a combination of presence-only data, randomly generated pseudo-absences and environmental predictors variables within the Biomod2 framework in R. Ten potential predictor variables were identified, of which five were selected after correlation tests. Local Random Forest models were built to the extent of four small study regions, and later projected to the whole Indonesian EEZ. The common bottlenose dolphin local models showed preference for areas close to the coast and shallower waters. Sperm whale local predictions were located further into the open waters and at deeper waters. The extrapolated predictions into the Indonesian EEZ, however, showed some unexpected results. The high occurrences for common bottlenose dolphins were not only located close to the islands, but also more into open waters. In contrast, sperm whale distributions have high occurrences near coastal areas and in the vicinity of islands than in the open oceans. This information suggested that the transferability of species distribution models may not always be preferable, because provide low accuracy. Sighting data, choices of variables and model settings influenced the outcome of the extrapolated models. Despite the unpreferable of the extrapolations, the results are still beneficial for cetacean conservation purposes, since the study was able to identify potential habitats in unsurveyed regions.