The 1st validation is based on a run over 166 time steps and studies the spatial average and variance before and after downscaling. The 2nd one checks whether value of a coarse pixel agrees with the average of its corresponding fine pixels.

Results

The validation revealed a problem with OASIS3 in February 2018: The full downscaling procedure introduced a kind of bias into the fine scale field. This prompted an overhaul of the relevant new functionality in OASIS3, which, as of March 2018, is not yet completed.

February 2018

Mean and variance measure the effect of the downscaling on the field. It makes at most 10^{–4 of a relative difference in mean, and 10^10 less so if no Schomburg rules affected the concerned variable. Ground temperature is the exception with a 10}–7 change in mean.
The variance increases across the board. The variables subjected to Schomburg rules experienced a relative increase of around 10, the others 1.