European multidecadal solar variability badly captured in all centennial reanalyses except CERA20C

Long-term historic climate datasets are valuable tools to investigate climate variability, validate climate models and contextualize anticipated climate change. Surface solar radiation is one particularly relevant variable, with implications on policy decisions (e.g. performance of solar panels) and fundamental questions in climate science (e.g. regarding the energy budget). While all current twentieth century reanalyses provide surface solar radiation, we demonstrate that most of them fail to capture multidecadal surface radiation variability in Europe. To this end, we systematically compare the reanalyses 20CRv2c, 20CRv3, ERA20C and CERA20C and the free model run ERA20CM. We show that only CERA20C captures dimming (1949-1979) and brightening (1979-2009) in line with station observations, satellite-era reanalyses and established theory. The lack of multidecadal surface radiation variability in 20CRv2c/v3 is plausible given the use of constant aerosols. In contrast, ERA20CM, ERA20C and CERA20C are forced with time-varying aerosols. Despite this common forcing, ERA20CM and ERA20C surprisingly show no trends in clear-sky fluxes over the dimming and brightening periods while CERA20C shows significant trends. We discuss different potential explanations for this discrepancy (model versions, ocean coupling and ensemble size) and conclude that none of them provides a convincing explanation. Our results therefore imply that only CERA20C is suitable for assessments of surface solar radiation variability on multi-decadal timescales. This particularly applies to impact studies, for example, on long-term potentials of solar power generation.