KelpWatch maps California kelp forests in 3m

This is the kind of AI work that matters outside the usual demo cycle: a segmentation model turns raw satellite noise into something managers can actually act on. The value isn’t just prettier maps, it’s moving from broad trends to place-by-place conservation decisions.

• –3-meter mapping surfaces small, sparse fringing kelp beds that 30-meter Landsat records can miss entirely
• –The dataset makes post-heatwave recovery patterns easier to see, including severe losses in Sonoma and Mendocino counties
• –Local drivers like temperature, depth, and habitat fragmentation become measurable at a scale that supports targeted restoration
• –KelpWatch’s open access delivery means the research is usable by scientists, agencies, and community groups instead of staying trapped in a paper