Cloud shadows drive strong variability in surface solar irradiance. With solar power as the fastest growing energy source worldwide, this variability poses an increasing threat to the stability of the electrical grid, because it drives strong local fluctuations in voltage. Due to the complex nature of cloud formation, a thorough understanding of the spatial extent, duration, and intensity of solar irradiance variability is lacking, and consequently weather forecasts of this variability are poor. This project will address this knowledge gap, by performing a novel field experiment, and by improving the next-generation of weather models guided by the outcome of the experiment. We will set up a spatial grid (1 km2, 50 m spacing) of hundreds of newly-designed sensors and measure surface solar irradiance at 12 wavelengths during one year. This experiment will be embedded into the new Ruisdael Observatory, which will deliver complementary cloud and aerosol observations. We will elucidate the poorly understood links between surface solar irradiance variability and the optical and geometrical properties of clouds. We will further corroborate the findings of our experiment with 3D large-eddy simulation (LES) as a virtual laboratory. This simulation technique permits models to resolve individual clouds and is therefore able to explicitly simulate the 3D interactions between clouds and radiation. While LES enables understanding of cloud-radiation interactions beyond what is possible from field experiments, using it for forecasting remains problematic due to the large computational costs associated with 3D radiation computations. Building upon my experience in LES and GPU computing and guided by the field experiment, I aim to remove this barrier by developing the first LES setup with 3D radiation fast enough for forecasting. This setup will be put to the test and validated against weather station and solar panel production data.