I will also note that models are known to often have very different cloud fractions than those observed (e.g. https://agupubs-onlinelibrary-wiley-com.cuucar.idm.oclc.org/doi/full/10.1002/2015JD024699). Directly using the surface longwave and shortwave forcing, as suggested, will be more physically correct as ERA (or any other model you use) should take into account not only cloud fraction, but cloud height, phase, and thickness when calculating the radiative transfer code.