Counting Species Faster Than They Vanish

Earth's biodiversity is declining even under the most optimistic models of global change, with current rates of species loss demanding urgent action. But conservation science and conservation practice are disconnected — basic research rarely translates into on-the-ground conservation action, and conservation practitioners rarely have access to cutting-edge technologies. Existing technologies for species identification, population monitoring, tracking, behavior analysis, and habitat assessment are not cost-effective or scalable to the levels needed. Critical knowledge gaps persist in understanding how functional diversity of organisms interacts with and responds to environmental change. The result: conservation decisions are made with incomplete data, and the efficacy of conservation actions is rarely evaluated systematically.

The IPBES Global Assessment estimates that 1 million species face extinction. The Kunming-Montreal Global Biodiversity Framework committed 196 nations to protecting 30% of land and ocean by 2030, but monitoring progress toward these targets requires technologies that do not yet exist at scale. Global spending on conservation is estimated at $124–143 billion/year, but most of this is invested without rigorous evidence of effectiveness. Pollinators alone provide ecosystem services worth $235–577 billion/year — their decline threatens global food security. Migratory species connect ecosystems across continents, and their conservation requires real-time tracking and habitat assessment across vast areas that exceed the capacity of current monitoring systems.

Camera traps generate millions of images but manual identification is a bottleneck — AI identification exists but accuracy drops sharply for rare species, in poor lighting, and for partial images, exactly the conditions that matter most for conservation. Satellite remote sensing provides landscape-scale habitat data but cannot detect most species or assess ecological condition. GPS tracking provides detailed movement data for individual animals but tags are expensive ($200–$5,000/unit), limited to larger animals, and have finite battery life. Acoustic monitoring is promising but requires species-specific classifiers that must be trained from scratch for each region. eDNA sampling can detect species presence from water or soil samples but cannot estimate population size or health. The fundamental problem is that each technology provides one narrow data stream, and integrating multiple streams into actionable conservation intelligence has not been achieved at scale.

AI for cost-effective wildlife identification, tracking, and behavior analysis that works reliably for rare species under field conditions. Technology solutions that integrate basic research with conservation practice in a closed feedback loop — where conservation actions are monitored, evaluated, and refined based on evidence. Big data approaches co-developing species-specific conservation strategies integrating community objectives. Autonomous marine and terrestrial monitoring platforms that combine multiple sensor modalities (visual, acoustic, chemical, genetic). Low-cost, long-duration tracking devices suitable for small-bodied organisms (insects, small birds, bats).

A student team could develop an AI classifier for a local wildlife monitoring camera trap dataset, focusing specifically on improving accuracy for rare or difficult-to-identify species (nighttime images, partial views, juveniles) and benchmarking against existing tools like MegaDetector. Alternatively, a team could design and prototype a low-cost acoustic monitoring station ($50–$100 target) capable of continuous outdoor recording and on-device species classification, testing it in a local ecosystem. Relevant skills: computer vision, machine learning, ecology, electronics, acoustic signal processing, conservation biology.