Can Deep Learning Extract Useful Information about Energy Dissipation and Effective Hydraulic Conductivity from Gridded Conductivity Fields?

We confirm that energy dissipation weighting provides the most accurate approach to determining effective hydraulic conductivity (Keff) of a binary K grid. A deep learning algorithm (UNET) can infer Keff with extremely high accuracy (R2 > 0.99). The UNET architecture could be trained pattern from an image distribution, although it was less for cases highly localized structures controlled flow. Furthermore, learned even if not directly on this information. However, weights were represented within in way immediately interpretable by human user. This reiterates idea ML/DL algorithms are make some hydrologic predictions accurately, they must designed and provide each user-required output their results used improve our understanding systems.