Removal of fish from gear by marine predators, known as depredation, is a fishery dependent mortality that needs to be included in stock assessments for affected stocks to avoid misestimation of the assessed resource. Toothed whales engage regularly in depredation from longlines, and while in some regions they leave clear marks of depredation activity, in the longline fisheries in the Southern Ocean they often leave no trace of removal, making it necessary to estimate depredation through modelling approaches. Several modelling approaches have been developed over the past decade in affected Southern Ocean fisheries, and in this paper, we examine five applications of common CPUE model structures to the same dataset from the longline fishery around South Georgia. We then compare the estimates of depredation with those based on observed bycatch ratios. The different model structures estimated very similar annual depredation removals, with all approaches averaging around % of the catch removed throughout the entire fishery. While depredation varies spatially, the different modelling approaches consistently highlighted areas where the impact of depredation was highest.