Understanding the Efficacy of U-Net & Vision Transformer for Groundwater Numerical Modelling

TL;DR

This study compares U-Net, U-Net with Vision Transformers, and Fourier Neural Operator for groundwater modeling, showing U-Net variants excel in accuracy and efficiency, especially with sparse data.

Contribution

It introduces a comparative analysis of U-Net, U-Net + ViT, and FNO models for groundwater modeling, highlighting the superior performance of U-Net-based models in data-scarce conditions.

Findings

U-Net and U-Net + ViT outperform FNO in accuracy.

U-Net models are more efficient in sparse data scenarios.

U-Net-based models show promise for real-world groundwater applications.

Abstract

This paper presents a comprehensive comparison of various machine learning models, namely U-Net, U-Net integrated with Vision Transformers (ViT), and Fourier Neural Operator (FNO), for time-dependent forward modelling in groundwater systems. Through testing on synthetic datasets, it is demonstrated that U-Net and U-Net + ViT models outperform FNO in accuracy and efficiency, especially in sparse data scenarios. These findings underscore the potential of U-Net-based models for groundwater modelling in real-world applications where data scarcity is prevalent.