observational errors