Synthetic quaternary aquifers dataset based on Bumberg Quarry (Facies and porosity)

The dataset contains a series of facies realizations generated with EROSim, being a synthetic representation of upper Aare Valley aquifers in Switzerland. We consider the available geological studies on the Bümberg quarry, located in the canton of Bern, Switzerland, and considered a suitable analog for the type of geological heterogeneity expected in such depositional sequences. Facies To represent geological facies types and structures in the training data we adopt EROSim (Schorpp et al., 2025), a surface-based facies modeling tool, to generate 20 different 3D realizations, considering the same surfaces statistics proposed by Schorpp et al. (2025) and derived from the site geological survey of Menga (2021; unpublished Master's thesis). The dataset considers four different gravel dominated facies types: Class 0 (Yellow): Bedload sheet aggradation (C,P,G (h,l,n,i,m)) - cobbles, pebbles, gravel; horizontally to low angle stratification, normally graded or massive. Class 1 (Blue): Transverse bar migration (C,P(p,t,x)) - cobbles and pebbles cross-stratified Class 2 (Brown): Scour fill (C,P,G,Ss) - Chaotic setting of cobbles, pebbles, gravel and sand Class 3 (Green): Gravelly dunes (PS, GS(p,t,x)) - planar-cross stratification mix of pebbles and sand C (cobbles), P (pebbles), G (gravels), S (sands)o (open framework), i (imbricated), h (horizontal stratification), n (normally graded), r (reverse graded), l (low angle stratification), p (planar‐cross stratification), t (trough‐cross stratification), s (draping troughs), x (cross‐stratification), m (massive) Porsity The porosity distribution for each facies class was estimated from the closes analogous in the work of Heinz et al. (2002). Class 0: the closes analog is the family of the poorly sorted and alternating gravel lithofacies (Gcx; Gcm; Gcm,i; Gcg,a in the origianl paper). Following Heinz et al. (2002), we consider eventual presence of matrix between clasts, with variable grainsize from silt to fine-medium sands. Thus, we indicate a relatively large range of porosities from 0.08 to 0.25. We represent this distribution with a Gaussian with mean at 0.15 and std of 0.03, truncated for values lower than 0.08. Class 1: We consider this class very close to Class 0, but with a relatively larger grain-size and better layer stratification. Thus, we consider this class belonging to the higher-porosity group of the gravel lithofacies identified by Heinz et al. (2002), and define its porosity distribution with a gaussian centered at 0.2 with a standard deviation of 0.2. Class 2: the scour pool fills identified by Menga (2021) are closest to the sedimentary lithofacies of Heinz et al. (2002) having average porosity = 0.26 and standard deviation of 0.02. We used this distribution for our TI. Class 3: these gravelly dunes show significant content of sand and cross-stratification, indicating higher porosity values. We consider this facies to belong to classes with the highest porosity ranges in Heinz et al. (2002), 0.36 on average, with a standard deviation of 0.03. Continuity patterns of porosity values have been modelled through spherical variograms, considering the stratigraphic description of the facies to define their range and nugget and angle (arbitrarily assumed). Follwing the facies modeling from Schorpp et al. (2025), we consider one horizontal direction (assumed orthogonal to flow direction) to have a shorter correlation range. The resulting dataset is a set of 35 3D volumes of facies distribution generated with EROSim, and 35 corresponding porosity distributions volumes modelled using sequential Gaussian simulation (independently for each facies). Shape of volumes:Cell count = 200 × 200 × 100Cell resolution (in real field applications) = 0.85 × 0.85 × 0.14 \ m^3 2D imagesPixels = 152 × 48Extension = 129.2 × 6.72 \ m^2