Accurate bycatch species identification in commercial fisheries is crucial as fishing pressure susceptibility is different for individual species. Species misidentification within datasets arise from varied experience levels of fishery observers. At Heard Island (HIMI), the Patagonian toothfish longline fishery bycatch comprises four grenadier species. Currently, bycatch limit categories are in separate species pairs: 1) ‘Macrourus caml and M. whitsoni’ and 2) ‘M. holotrachys and M. carinatus’. However, similar characteristics between species within each pair has contributed to observer misidentification of Macrourus species within bycatch records. Therefore, improving observer identification of Macrourus species is vital for accurate bycatch information. We analysed 800 otoliths among the four Macrourus species collected from HIMI's fishery (2015-2021). Macrourus otolith shape analysis and elliptical Fourier construction was performed through shapeR, while random forest models (RFM) coupled initial observer identifications with otolith morphometrics to predict species identifications within each bycatch pair. RFMs predicted M. whitsoni (60% accurate) were highly misidentified as M. caml, with otolith width an important morphometric. Additionally, otolith ellipses of M. whitsoni and M. carinatus displayed prominent dorsal margins. This research highlights RFM effectiveness within bycatch species identification and management, presenting baseline otolith morphometry of four Macrourus species that aids future observer training.