For a typical space-borne algorithm for the estimation of surface UV-B daily doses, the implications of a limited temporal sampling of cloud properties and of the neglect of three-dimensional radiative effects are addressed. By means of six synthetic diurnal cloud cycles, different temporal samplings of cloud properties have been investigated, while surface and atmospheric properties have been kept constant. The results have been compared with benchmark calculations of the daily doses using a 15 min time step. The neglect of the horizontal photon transport imposes limitations to the estimation of daily doses since it leads to statistical uncertainties up to 25% and biases ranging from 3% to 9%. Additionally, when a reduced temporal sampling is used, maximum uncertainties increase up to 85% when probing the cloud field only every 4 hours. Even in this case, however, 25--40% of the daily doses in the model domain have an accuracy between -20% and +20%. The time around noon turns out to be of crucial importance for a precise estimation of UV daily doses. Therefore usually a limited number of cloud probes around noon is particularly efficient, although in case of high temporal and spatial cloud variability throughout the day three cloud probes can still be too few to prevent the occurrence of large (>80%) deviations. Finally, by spatial averaging, a large improvement of the agreement between reference values and daily doses computed with a given time sampling can be achieved.