Time series of standardised anomalies with respect to the normalisation period 1981-2010 were obtained for each dataset, covering sea ice cover, sea surface temperatures, tropical precipitation, sea level pressure, the stratospheric polar vortex, snow cover, sunspot activity, volcanic activity and carbon dioxide concentrations. In addition, using 500 hPa geopotential height data from the ERA5 reanalysis, time series of jet speed and latitude were derived and the top three principal empirical orthogonal functions of atmospheric circulation variability for the North Atlantic and European sector

The datasets are sets of standardised values and anomalies for different predictors of atmospheric circulation variability, which can be fed into NARMAX machine learning models to generate forecasts of the three leading empirical orthogonal functions of atmospheric circulation variability - roughly corresponding to the North Atlantic Oscillation (NAO), East Atlantic Pattern (EA) and Scandinavian Pattern (SCA). In the SF-NARMAX project these values were used to generate NARMAX forecasts for June, for July/August, and for meteorological winter (December/January/February), which were then compared with actual outcomes, to help assess the reliability of the NARMAX models.