Weak signals from an ecological and statistical perspective

by Frédéric Barraquand

In this talk, I will try to delineate how our ability to detect a difference of ecological relevance depends not only on a signal nature and strength, but also on ecological and observational noise as well as study design. Then I will explore several ways to obtain weak ecological signals despite reasonably large sample sizes: coarse measurement tools and multiple confounding factors. These notions will be illustrated with varied ecological case studies, from bird monitoring to threats to pollinators. I will then discuss recent attempts to infer signs of imminent ecological collapse from so-called early warning signals, that are often a rather weak manifestation of impending large-scale ecological changes.