Sensor fusion and downscaled climate projections reveal climate refugia in the California Channel Islands.

Conservation actions such as habitat restoration and translocation require spatially informed, quantitative decision-making. We modeled habitat suitability for nine priority landbird species of the Channel Islands to understand habitat preferences, inform conservation planning and assess future climate impacts. In the absence of recent airborne lidar, we derived vegetation structure from spaceborne lidar, radar and optical remote sensing data combined with downscaled climate observations to train machine-learning models. Models performed well (average AUC = 0.84) under spatial cross-validation. Incorporating citizen science data improved Boyce Index performance for eight of the nine species, though AUC declined for 6 out of 9 species, likely reflecting spatial bias in opportunistic records. A radar-derived vegetation index was consistently influential, while maximum temperature strongly affected species restricted to Santa Cruz Island. Most species are projected to lose suitable habitat under two near-future (2040-2069) climate scenarios, except Grasshopper sparrow, Island scrub jay and Rufous-crowned sparrow. We identified climate change refugia to inform spatial conservation priorities.