I recall reading some years ago about a study of tornado warnings that showed that every additional minute of warning time had significant value in terms of preventable damage and casualties. Presumably it would be similar for earthquakes, but early detection is a much harder problem.