Kelp forests are the foundation of shallow marine ecosystems in many temperate regions but face threats from climate change and other stressors. Understanding these impacts and identifying potential refuges can help guide conservation efforts. This study uses a long-term dataset (1998–2019) of kelp cover for two canopy-forming species, Ecklonia radiata and Phyllospora comosa, along the coast of Victoria, Australia. Data from three scuba sampling programs were linked to environmental variables such as depth, seafloor structure, wave climate, currents, temperature, and population connectivity using generalized additive mixed-effects models. These models generated predictive maps of kelp cover within Victoria’s marine protected areas (MPAs) and projected future changes under climate scenarios (2090, RCPs 4.5 and 8.5). We assessed kelp cover change (1998–2019), stability, and predicted future dynamics (2019–2090) to classify areas as very unlikely refugia, unlikely refugia, neutral, potential refugia, or likely refugia. A management framework was then developed using these rankings, applied across three case studies at different scales. Our findings demonstrate how species distribution models—both current and future—can identify potential refugia, helping prioritize management actions and guide conservation strategies to protect kelp forests amid climate change.