Habitat temperature is often assumed to serve as an effective proxy for organism body temperature when making predictions of species distributions under future climate change. However, the determinants of body temperature are complex, and organisms in identical microhabitats can occupy radically different thermal niches. This can have major implications of our understanding of how thermal stress modulates predator-prey relationships under field conditions. Using body temperature data from four different sites on Santa Cruz Island, California, we show that at two sites the body temperatures of a keystone predator (the seastar Pisaster ochraceus) and its prey (the mussel Mytilus californianus) followed very different trajectories, even though both animals occupied identical microhabitats. At the other two sites, body temperatures of predator and prey were closely coupled across a range of scales. The dynamical differences between predator and prey body temperatures depended on the location of pairs of sites, at the extremes of a persistent landscape-scale weather pattern observed across the island. Thus, the well understood predator-prey interaction between Pisaster and Mytilus cannot be predicted based on habitat-level information alone, as is now commonly attempted with landscape-level ('climate envelope') models.