Addendum to figs. 10 and 11 in “Clear-sky shortwave downward flux at the earth's surface: Ground-based data vs. satellite-based data” [J. Quant. Spec. and Rad. Tran. 224 (2019) 247-260]

•Investigated the cause for the discrepancies between the CERES SYN1deg(Ed4A) output monthly mean aerosol optical depths and precipitable water and their BSRN counterparts.•The BSRN original data are available intermittently for technical reasons.•It is found that when the CERES hourly output are similarly sampled, their monthly means are much closer to their BSRN counterparts. While the CERES input aerosol optical depth and precipitable water data are continuously available, the same parameters derived from BSRN site measurements are available only intermittently. Cloudiness disrupts the aerosol optical depth (AOD) retrieval and precipitable water (w) observation sampling. Irregularly sampled records, such as these, can cause systematic biases if the averages of the continuous data are not sampled at the same times as the observed data, as shown by high biases in Fig. 10 in our original paper. This, however, does not suggest that the CERES input aerosol optical depth and precipitable water are systematically significantly higher than ground-based observations. In this addendum, we show that when the monthly means are computed from only those CERES hourly means that have a BSRN match, then the resulting monthly means differ from BSRN-derived parameters to a much lesser extent. To be precise, the biases in the original Fig. 10 decrease 82%, 69% and 52%, respectively; the magnitudes of slopes in the original Fig. 11 decrease by 69% and 65%, respectively. Imposing a cloud fraction less than or equal to 5% further reduces AOD and w mean values but not biases. Nevertheless, the flux bias reduces from -16.9 W m−2 to -4.2 W m−2 after imposing the cloud fraction restraint relative to the RADFlux clear-sky fluxes which appear to represent the driest and clearest conditions.