The current paper presents the first detailed investigation of open-ocean dispersal of hatchlings and juveniles of the critically endangered western Pacific leatherback turtle Dermochelys coriacea populations nesting in New Guinea. Dispersal patterns were simulated by releasing particles drifting passively, or almost passively, into a state-of-the-art World Ocean circulation model. Analysis of the simulation results combined with sighting, genetic, bycatch, and adult satellite tracking information reveals that: (1) Hatchlings emerging from the main New Guinea nesting beaches are likely to be entrained by highly variable oceanic currents into the North Pacific, South Pacific, or Indian Oceans. Those drifting into the Indian Ocean likely suffer very high mortality. This suggests that, as ocean current variability determines the partition of hatchlings into different dispersal areas, it also largely influences juvenile survival rate at the population level. (2) Within 1 to 2 yr, most passively drifting juveniles reach temperate oceanic regions where the water temperature in winter drops well below the minimum temperature likely tolerated by such small individuals. This leads us to hypothesize that, after an initial period of mostly passive drift, juveniles initiate active swimming towards lower (warmer) latitudes before winter and back again towards higher latitudes, where food abounds, during spring. Such seasonal migrations would significantly slow the eastward progression of individuals circulating in the North Pacific current. This slower drift scenario better explains the size distribution of leatherbacks observed, or incidentally caught by pelagic fisheries, in the North Pacific. This dispersal mechanism combining passive drift with active habitat-driven seasonal migrations might well apply to many other sea turtle populations and deserves further study.