Advisers say AI can help spot whales, but satellite access and sampling design remain bottlenecks

Participants at the Marine Mammal Commission's virtual session agreed that automated detection using machine learning offers a promising path to scale satellite monitoring of marine mammals, but cautioned that imagery access and sampling design remain limiting factors.

Peter and others observed that satellites can locate animals effectively when individuals are clumped but struggle with dispersed populations. "AI models are the future," one adviser said, noting that automated detection reduces manual review burdens but still requires extensive training data. Participants highlighted practical bottlenecks: (1) the high cost and limited access to very‑high‑resolution commercial satellites; (2) the need to develop sampling‑design criteria that link required image coverage to population patchiness; and (3) the value of combining satellite work with ground teams to validate detections in turbid or variable lagoons.

Speakers suggested test cases where satellite imagery has higher chances of success—Mexican lagoons with lower cloud cover, and nearshore estuarine dolphin populations—and encouraged pilots that combine satellite imagery, on‑the‑water photography and local observers. The Commission's recent funded projects (UAS, close‑kin and spatial CMR studies) were highlighted as complementary investments that can improve abundance estimates while satellite methods mature.

The group asked staff to track potential pilot sites and funder interest; no formal decisions were taken during the session.