Influence of reference period on evaluation of temperature output from the EURO-CORDEX climate ensemble using E-OBS data

This note investigates the fit of daily temperature projections over Fennoscandia for 1950-
2005 for nine combinations of global and regional climate models from EURO-CORDEX.
The evaluation is performed by comparing temperature anomalies from the climate models
to observed temperature anomalies from the E-OBS dataset using integrated quadratic
distance (IQD). Summer and winter are evaluated separately. Different alternatives for a
reference period are explored; 30-year periods, 20-year periods and 10-year periods.
The evaluation results are robust for the 30-year periods, in both seasons. When the
length of the reference period decreases, the evaluation results become less and less robust.
It is hard to say if 20 years are sufficient so it is better to use 30 years. A reference
period of 10 years is too short because the evaluation results highly depend on the chosen
period.