Supplementary Materials for "Ensemble design for seasonal climate predictions: Studying extreme Arctic sea ice lows with a rare event algorithm"

Here we provide the data required to reproduce the results of the paper "Ensemble design for seasonal climate predictions: Studying extreme Arctic sea ice lows with a rare event algorithm" submitted to Earth System Dynamics (preprint available at Sauer al. 2024(a); https://doi.org/10.5194/egusphere-2024-3082). The paper has been produced in the context of a cross-journal special issue. The special issue is led by the European Geosciences Union's (EGU's) Earth System Dynamics and also includes the journals Geoscientific Model Development (GMD) and Nonlinear Processes in Geophysics (NPG). The special issue has the topic "Theoretical and computational aspects of ensemble design, implementation and interpretation in climate science". All the data are produced with the Planet Simulator (PlaSim) (Fraedrich et al. 2005) coupled to a dynamic Large-Scale Geostrophic (LSG) ocean model (Maier-Reimer et al. 1993) and are structured as follows: SEAICE_CTRLRUN_0501-3500.pkl contains daily data from the 3000-year control run (Sauer et al. 2024(b)). The included variables are the pan-Arctic sea ice area (SIA_ARCTIC), pan-Arctic sea ice volume (SIV_ARCTIC), the pan-Arctic mean sea ice thickness (SIT_ARCTIC) and the cumulative area with a sea ice thickness equal or larger than a given threshold ({X}_UPPER). The model years range from year 501 to year 3500. SIA_CTRLENS_YEAR{j}.tar contains for each of the four initial conditions (control run model years j=(1930, 1037, 1083, 1075)) one .npz file with daily pan-Arctic sea ice area data (SIA_CTRL) available from the four 6000-member control ensembles. PICONTROL-AV0_LSM.nc contains the land-sea mask of PlaSim. MEAN_STD_CTRLENS_YEAR{j}.tar contains the control intra-ensemble means and intra-ensemble standard deviations of different quantities for the two initial conditions taken from model year 1930 and 1037 of the control run. The included variables in the *INTEGRATED*.npz files are the pan-Arctic sea ice area (SIA) and sea ice volume (SIV), the surface sensible and latent heat fluxes (HFSS, HFLS), the net longwave and net shortwave surface radiative fluxes (RLS, RSS), the net atmosphere-surface energy flux (RES), the vertical oceanic heat flux at the surface (OFLUXA), the surface downward longwave, upward longwave, downward shortwave, upward shortwave energy fluxes (STRD, STRU, SSRD, SSRU), the vertically integrated water vapour (PRW) and the total cloud cover (CLT). Upward (negative) surface energy fluxes are positive (negative). The first column in the arrays is the February-September time-average, while the subsequent columns are the February-March, March-April, April-May, May-June, June-July, July-August, August-September bimonthly means. The *GRIDDED*.nc files include the February-March, March-April, April-May, May-June, June-July, July-August, August-September bimonthly means of the sea ice area fraction (SIC), the sea ice thickness (SIT), two metre temperature (TAS) and 500 hPa geopotential height (Z500). The prefixes "MEAN" and "STD" correspond to the intra-ensemble means and standard deviations respectively. MIN_ALGENS_YEAR{j}.tar contains data regarding the trajectories leading to the most extremely negative August-September sea ice anomalies available from realization #1 of the year 1930 experiment and available from realization #9 of the year 1037 experiment. The included variables in the *.npz files are the pan-Arctic sea ice area (SIA) and sea ice volume (SIV), the surface sensible and latent heat fluxes (HFSS, HFLS), the net longwave and net shortwave surface radiative fluxes (RLS, RSS), the net atmosphere-surface energy flux (RES), the vertical oceanic heat flux at the surface (OFLUXA), the surface downward longwave, upward longwave, downward shortwave, upward shortwave energy fluxes (STRD, STRU, SSRD, SSRU), the vertically integrated water vapour (PRW) and the total cloud cover (CLT). The data include both the actual values (no prefix/suffix) and anomalies relative to the corresponding control ensemble (prefix/suffix "ANO"). Actual and anomaly values of HFSS, HFLS, RLS, RSS, RES, OFLUXA are positive (negative) if they are directed upwards (downwards). Actual values of STRD, STRU, SSRD, SSRU are positive (negative) if they are directed upwards (downwards). Direction-independent values are used for the anomaly values of STRD, STRU, SSRD, SSRU, i.e., positive (negative) anomalies indicate that the fluxes are larger (smaller) in magnitude compared to the control ensemble mean. The first column in the arrays is the February-September time-average, while the subsequent columns are the February-March, March-April, April-May, May-June, June-July, July-August, August-September bimonthly means. The included variables in the .nc files are the sea ice area fraction (SIC), the sea ice thickness (SIT), two metre temperature (TAS) and 500 hPa geopotential height (Z500). All values are February-March, March-April, April-May, May-June, June-July, July-August, August-September bimonthly time-averages. Anomalies (prefix/suffix "ANO") are defined as the deviation of the actual value (no prefix/suffix) from the control ensemble mean. SIA_kick_ntraj600_k0075_res5d_FEBSEP_p16_YEAR{j}_EXP{i}_resampling.0001_0048.tar includes for each of the four initial conditions (control run model years j=(1930, 1037, 1083, 1075)) respectively one .tar file with one .npz file for each of the ten realizations of the rare event simulation (i.e. a total of 4 x 10 .npz files). The .npz files include daily pan-Arctic sea ice area data (obs_*) and all the information on the selection of the trajectories performed by the algorithm. Quantities indicated by *_ens correspond to the straight output of the rare event algorithm, while *_eff correspond to the same quantity associated to the trajectories in the effective ensemble (see Sauer et al. 2024(a,b) for more details about the rare event algorithm).