3TIER internal analysis has shown that downscaling the re-analysis data adds significant value to the hourly and daily correlations when compared to observed data. Improvements in the correlations on the monthly and inter-annual time scale due to downscaling are much less. If the final data are being used in applications where the fidelity of the hourly and daily values are critically important then downscaling can add value. However if these data are being used to perform a long-term correction – then the ability to distinguish a windy year from a non-windy year is more important. Downscaling adds little to no skill in the ability to sort years, especially for the ERA-Interim data which is at significantly higher resolution (80km) and includes many more vertical levels than the NNRP data (280km).

While this result may come as a surprise – it can be better understood in the context of weather versus climate. Models such as WRF are weather simulation models. If actual climate variability is not captured in the re-analysis data, then a weather model can not recreate it. What the weather model can do is simulate weather patterns from the re-analysis data – hence the improvement on the hourly and especially daily time scales.