The original forecasts have been bias-adjusted by applying a quantile mapping method (Gudmundsson et al., 2011) on a monthly basis. ERA5 reanalysis at 0.25° upscaled at the same resolution as the seasonal forecast data, i.e. 1°, was employed as a reference dataset.
The temperature downscaling exploits information from a reference fine-scale temperature climatology (ERA5 at 0.25° resolution). The coarse-scale data are remapped to the target grid and adjusted pixel-by-pixel by an additive constant so that the resulting fine-scale climatology is identical to the reference climatology.
The data processing has been performed using the CSTools R package (Perez-Zanon et al., 2020)