Most atmospheric models consider radiative transfer only in the vertical direction (1D), as 3D radiative transfer calculations are too costly. Thereby, horizontal transfer of radiation is omitted, resulting in incorrect surface radiation fields. The horizontal spreading of diffuse radiation results in darker cloud shadows, whereas it increases the surface radiation in clear sky patches (cloud enhancement). In this study, we developed a simple method to account for the horizontal transfer of diffuse radiation. We spatially filter the surface diffuse radiation field with a Gaussian filter, which is conceptually simple and computationally efficient. We applied the filtering to the results of Large-Eddy Simulations for two summer days in Cabauw, the Netherlands, on which shallow cumulus clouds formed during the day. We obtained the optimal filter size by matching the simulation results with detailed high-quality observations (1Hz). Without the filtering, cloud enhancements are not captured, and the probability distribution of global radiation is unimodal, whereas the observed distribution is bimodal. After filtering, the probability distribution of global radiation is bimodal and cloud enhancements are simulated, in line with the observations. We found that small changes in the filter width do not strongly influence the results. Furthermore, we showed that the width of the filter can be parameterized as a linear function of e.g. the cloud cover. Hence, this work presents a proof-of-concept for our method to come to more realistic surface irradiances by filtering diffuse radiation at the surface.