Range shift is the primary short-term response of species to rapid climate change but it is hampered by natural or anthropogenic habitat fragmentation. Fragmented habitats expose different critical areas of a species niche to heterogeneous environmental changes resulting in uncoupled effects. Modelling species distribution under complex real-life scenarios and incorporating such uncoupled effects has not been achieved yet. Here we identify the most vulnerable areas and the potential cold refugia of a top-predator with fragmented niche range in the Southern ocean by integrating genomic, ecological and behavioural data with atmospheric and oceanographic models. Our integrative approach constitutes an indispensable example for predicting the effect of global warming on species relying on spatially and ecologically distinct areas to complete their life-cycle (e.g., migratory animals, marine pelagic organisms, central-place foragers) and, in general, on species constrained in fragmented landscapes due to continuously-growing anthropogenic pressure.