Should we account for detectability in population trends?