Estimating potential evapotranspiration: the effect of random and systematic errors