Abstract
We confirm that energy dissipation weighting provides the most accurate approach to determining the 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 to infer the energy dissipation weighting pattern from an image of the K distribution, although it was less accurate for cases with highly localized structures that controlled flow. Furthermore, the UNET architecture learned to infer the energy dissipation weighting even if it was not trained directly on this information. However, the weights were represented within the UNET in a way that was not immediately interpretable by a human user. This reiterates the idea that even if ML/DL algorithms are trained to make some hydrologic predictions accurately, they must be designed and trained to provide each user-required output if their results are to be used to improve our understanding of hydrologic systems.
Highlights
IntroductionWater movement through the subsurface is controlled largely by the hydraulic conductivity distribution, which can vary over orders of magnitude across multiple scales [5]
We examined the effect of the structure of a binary medium on the effective hydraulic conductivity, Keff, using MODFLOW, a well-known finite difference numerical groundwater flow model
By comparing deep learning (DL) algorithms trained with and without access to energy dissipation weightings’ information, we sought to understand the mechanism by which Keff is inferred by the DL
Summary
Water movement through the subsurface is controlled largely by the hydraulic conductivity distribution, which can vary over orders of magnitude across multiple scales [5]. Recent advances in hydrogeophysics increasingly suggest that the spatial pattern of hydraulic conductivity can be mapped effectively [6,7,8,9]. Coupled with carefully selected point measurements of hydraulic conductivity, these methods offer the promise of real improvements in our ability to accurately model water flow and associated solute transport in the subsurface. One remaining fundamental challenge is how to translate an image of a spatially heterogeneous K field to an upscaled, effective K for use in a flow model
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
