The EarthCARE satellite mission targets an improved understanding of the influence of clouds and aerosols on the global radiation budget. Toward this goal, a target accuracy of +/- 10 W/m2 has been defined as threshold for closure between observed top-of-atmosphere fluxes and 3D radiative transfer simulations on spatial domains with an area of 10x10 km2. For our understanding of climate processes and other applications, closure of surface radiative fluxes is also of critical importance, but is not currently covered by the EarthCARE mission concept. Radiative closure is however much more difficult to assess experimentally at the surface than at the top-of-atmosphere in particular due to the limited spatial representativeness of ground-based measurements for larger domains, if instantaneous fluxes are considered. A common approach is to average observations for longer time periods or a large number of similar situations to reduce this sampling uncertainty, but this approach is also susceptible to error cancellation. An alternative is the deployment of a dense network of radiation sensors to better sample the average radiation fluxes across a region of interest. A key advantage is the possibility to investigate deviations and assess closure on a case-by-case basis. Using observations from several past field campaigns with a low-cost pyranometer network, the feasibility of such a closure experiment for surface radiative fluxes based on EarthCARE products and processors is assessed. A method based on optimum averaging/ spatio-temporal Kriging is introduced to determine the sampling accuracy of a sensor network for domain-average instantaneous fluxes. For several typical cloud situations, the number of stations required to reach different target accuracies for the average flux across the EarthCARE closure domain size is determined. Based on these findings, potential instrumental configurations for such an experiment are described.