Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Effect of Different Optimization Strategies to Physics-Constrained Deep Learning for Soil Moisture Estimation

Jianxin Xie, Bing Yao, Zheyu Jiang

TL;DR

A physics-constrained deep learning (P-DL) framework to integrate physics-based principles on water transport and water sensing signals for effective reconstruction of the soil moisture dynamics is proposed.

Abstract

Soil moisture is a key hydrological parameter that has significant importance to human society and the environment. Accurate modeling and monitoring of soil moisture in crop fields, especially in the root zone (top 100 cm of soil), is essential for improving agricultural production and crop yield with the help of precision irrigation and farming tools. Realizing the full sensor data potential depends greatly on advanced analytical and predictive domain-aware models. In this work, we propose a physics-constrained deep learning (P-DL) framework to integrate physics-based principles on water transport and water sensing signals for effective reconstruction of the soil moisture dynamics. We adopt three different optimizers, namely Adam, RMSprop, and GD, to minimize the loss function of P-DL during the training process. In the illustrative case study, we demonstrate the empirical convergence of Adam optimizers outperforms the other optimization methods in both mini-batch and full-batch training.

The Effect of Different Optimization Strategies to Physics-Constrained Deep Learning for Soil Moisture Estimation

TL;DR

A physics-constrained deep learning (P-DL) framework to integrate physics-based principles on water transport and water sensing signals for effective reconstruction of the soil moisture dynamics is proposed.

Abstract

Soil moisture is a key hydrological parameter that has significant importance to human society and the environment. Accurate modeling and monitoring of soil moisture in crop fields, especially in the root zone (top 100 cm of soil), is essential for improving agricultural production and crop yield with the help of precision irrigation and farming tools. Realizing the full sensor data potential depends greatly on advanced analytical and predictive domain-aware models. In this work, we propose a physics-constrained deep learning (P-DL) framework to integrate physics-based principles on water transport and water sensing signals for effective reconstruction of the soil moisture dynamics. We adopt three different optimizers, namely Adam, RMSprop, and GD, to minimize the loss function of P-DL during the training process. In the illustrative case study, we demonstrate the empirical convergence of Adam optimizers outperforms the other optimization methods in both mini-batch and full-batch training.
Paper Structure (12 sections, 13 equations, 3 figures, 1 table)

This paper contains 12 sections, 13 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Research Methodology
  3. Richardson-Richards Equation
  4. Physics-constrained deep learning (P-DL)
  5. Optimization Techniques
  6. Mini-batch/batch Gradient Descent (GD)
  7. Root Mean Square Propagation (RMSProp)
  8. Adaptive Moment Estimation (Adam)
  9. Experimental Results
  10. Convergence and performance observations
  11. P-DL performance using full batch training
  12. Conclusions

Figures (3)

  • Figure 1: Illustration of the proposed P-DL framework for soil moisture prediction.
  • Figure 2: The comparison of training loss $\mathcal{L}$ of P-DL model using mini-batch (a) and full batch (b) with different optimization methods. (c) Hydraulic conductivity function (HCF), (d) Water retention curves (WRC) generated from the P-DL model and van Genuchten model.
  • Figure 3: (a)The benchmark water content distribution $\theta$, (b) the predicted $\hat{\theta}$ obtained from P-DL trained by full-batch Adam optimizer, (c) the discrepancy mapping between the ground truth and the prediction.