We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Predictive Deep Learning for High‐Dimensional Inverse Modeling of Hydraulic Tomography in Gaussian and Non‐Gaussian Fields.
- Authors
Guo, Quan; Liu, Ming; Luo, Jian
- Abstract
Inverse modeling of hydraulic tomography (HT) is computationally expensive for estimating high‐dimensional hydrogeologic parameter fields. In this work, we develop a novel method called HT‐INV‐NN, which combines dimensionality reduction techniques with a predictive deep learning (DL) model to estimate high‐dimensional Gaussian and non‐Gaussian channel fields. The HT‐INV‐NN model consists of a predictor that directly learns the inverse process from hydraulic head measurements to latent variables of random fields, and a decoder that generates high‐dimensional parameter fields from predicted latent variables. For Gaussian spatially correlated fields, the decoder utilizes principal components derived from spatial covariance, and for non‐Gaussian channel fields, a generative adversarial network (GAN) is trained using generated realizations based on a training image (TI). The predictor is a deep neural network calibrated using the reference data obtained from HT forward simulations, which can be implemented in parallel. HT‐INV‐NN is successfully tested in multiple numerical experiments including steady‐state and transient HT for estimating Gaussian fields in 2D and 3D, as well as binary discontinuous or continuous non‐Gaussian channel fields. The training process is efficient, and the model structure demonstrates robustness for input data with perturbations. The model performance on multiple validation data sets are satisfactory when compared with other numerical and deep learning methods. Plain Language Summary: This study introduces an innovative approach for solving complex groundwater flow problems. In hydraulic tomography, we aim to estimate hydraulic properties of underground aquifers by analyzing water pressure measurements from wells. Traditionally, this process involves computationally intensive simulations and optimization techniques. We propose a predictive deep learning model called HT‐INV‐NN, which directly predicts the unknown hydraulic properties using machine learning algorithms. This eliminates the need for time‐consuming simulations and optimization, making the inverse modeling process much more efficient. We conducted a series of numerical experiments using both Gaussian and non‐Gaussian fields to evaluate the performance of our model. The results demonstrate that HT‐INV‐NN achieves high accuracy in estimating the hydraulic properties, even in challenging scenarios with complex channel patterns and varying smoothness levels. Our findings highlight the effectiveness and versatility of HT‐INV‐NN in handling different geostatistical variations and field characteristics. By leveraging the power of predictive deep learning, our approach offers a faster and more accurate solution to inverse modeling in hydraulic tomography. This has significant implications for groundwater management, environmental monitoring, and decision‐making processes. Overall, our study presents a promising advancement in hydraulic tomography and contributes to the growing body of research on predictive modeling using deep learning techniques. Key Points: HT‐INV‐NN integrates dimensionality reduction and predictive deep learning for high‐dimensional hydraulic tomography inverse modelingThe predictor is a deep neural network that learns the inverse process directly from hydraulic head measurements to latent variables of random fieldsThe decoder utilizes Principal Component Decomposition for Gaussian fields and trains a generative adversarial network for binary continuous and discontinuous non‐Gaussian channel fields
- Subjects
DEEP learning; MACHINE learning; HYDRAULIC models; GENERATIVE adversarial networks; GAUSSIAN processes; HYDRAULIC measurements; TOMOGRAPHY; RANDOM fields
- Publication
Water Resources Research, 2023, Vol 59, Issue 10, p1
- ISSN
0043-1397
- Publication type
Article
- DOI
10.1029/2023WR035408