Your link fails to establish that. The main problem with animal behavior surveys is that they tend to report good results when you start with a known earthquake, and then look backwards to find unusual animal behavior before the earthquake. If you instead try to define objective criteria on animal behavior and then use that as a signal to predict earthquakes, the quality of animal behavior as a predictor turns to crap. In a nutshell, unusual animal behavior precedes most earthquakes... and many more events that turn out not to be earthquakes!
That's a fair point; a detector that reliably detects impeding once-in-a-century strong earthquakes as well as every wolf or bear wandering too close to some herd of goats is probably not that useful.
