*2.2. GCM-RCM Data*

Five high-resolution GCM-RCM simulations (Table 1) produced by the European branch of the Coordinated Regional Climate Downscaling Experiment (EURO-CORDEX) initiative were downloaded from the Earth System Grid Federation (ESGF) database [29]. The cross-validation within a pseudo-reality framework (see Section 2.5) requires all projections to be forced by the same representative concentration pathways (RCP). Here, we chose RCP4.5 as a mid-range emission scenario, which corresponds to end-of-21st century radiative forcing of 4.5 Wm−<sup>2</sup> [30]. Each of the five model simulations has a different GCM and RCM component. One should note that the models were not

selected based on their performance in the present-day conditions but to ensure that the ensemble members are independent from each other to the extent possible.

The model simulations were re-gridded to the regular 0.125◦ × 0.125◦ EURO-CORDEX (EUR-11i) grid using nearest neighbor interpolation before bias adjusting them. The grid box closest to each sub-basin was then used to drive the hydrological model. The difference in the spatial resolution to WFDEI was not taken into account, as most of the results are based on comparisons between the GCM-RCM simulations (see Section 3). For both calibration and validation of the bias adjustment methods, we selected three 32-year periods including a 2-year spin-up period for the hydrological model runs. After excluding the spin-up period, years 1981–2010 were used as the baseline period, while years 2011–2040 and 2061–2090 were used to validate the methods in early and late 21st century conditions.

**Table 1.** List of GCM-RCM simulations selected for this study. The first column shows the abbreviation used in the text, the second and third columns show the GCM and RCM part of each model, respectively and the last column shows the name of the providing institution.

