Reflections about regional climate modelling and synoptic scales based on progress over the past 25 years

New generations of climate scientists get involved in downscaling global climate models, and the scientific community is increasingly being populated by scholars with a diversity of backgrounds and come from various scientific cultures. In particular, there is a renewed drive buoyed by an optimism in artificial intelligence and machine learning (AI/ML), which involves cohorts from the field of computation with different experiences to those with a background of meteorology or climatology. Many of the AI/ML papers on downscaling have ignored lessons learned from 30 years of progress within empirical-statistical downscaling, despite there being attempts with AI/ML starting already in the 1990s. There are also different opinions about downscaling within the CORDEX community which may enrich the scientific discourse within synoptic climatology. It is therefore useful to review and discuss what is established, what are the knowledge gaps, and what are the challenges. In this context, a number of questions are revisited that are relevant for the synoptic and mesoscales. They include: What are the models’ minimum skilful scale? What is the difference between downscaling, bias-correction, and interpolation?  Should we regard the (1) downscaling of each data point (“downscaling weather”) or (2) downscaling of information about the curve of statistical distributions, the parameters that determine their shape (“downscaling climate”) as different branches? And what about AI/ML? Caveats concerning non-stationarity still pose a serious problem, and the question of how representative the training data is for a changed future climate is particularly relevant for AI/ML. This is especially important for multi-variable predictors. Some downscaling approaches emphasise the data while others may lean on mathematical theory or physics. A question is how to design tests and evaluations that are fit for purpose when it comes to downscaling global climate models to provide future projections.