Kukunuri et al. (2025) Synthetic data generation using microwave modeling with efficient application of machine learning for bare land soil moisture retrieval: a case study

Identification

• Journal: Elsevier eBooks
• Year: 2025
• Date: 2025-12-05
• Authors: Anjana N.J. Kukunuri, Ajay Kumar Maurya, Dharmendra Singh
• DOI: 10.1016/b978-0-443-34113-7.00019-5

Research Groups

• Department of Electronics and Communication Engineering, Indian Institute of Technology Roorkee, Roorkee, Uttarakhand, India
• Department of Electronics and Communication Engineering, National Institute of Technology Patna, Patna, Bihar, India

Short Summary

This study develops a multilayer microwave model to generate synthetic backscatter data for bare land, accounting for varying soil properties, and demonstrates its use in training machine learning models like Gaussian Process Regression for soil moisture retrieval without requiring extensive real-world ground truth data.

Objective

• To develop a multilayer model to characterize electromagnetic scattering behavior of soils under varying moisture contents, textures, incidence angles, and frequencies for synthetic backscatter data generation, which can be utilized for training machine learning models for soil moisture estimation without a priori information of the study area.

Study Configuration

• Spatial Scale: Synthetic data generation for bare land soil moisture retrieval, applicable for a wide range of soil conditions and Synthetic Aperture Radar (SAR) configurations.
• Temporal Scale: Not explicitly defined for synthetic data generation; applicable for continuous monitoring of soil moisture through trained machine learning models.

Methodology and Data

• Models used: High-Frequency Structure Simulator (HFSS) for microwave modeling, Gaussian Process Regression (GPR) for machine learning.
• Data sources: Synthetic backscatter data generated through microwave modeling using HFSS.

Main Results

• Successfully generated a comprehensive synthetic backscatter dataset using HFSS, encompassing variations in wavelengths, incidence angles, soil textures, and moisture conditions.
• This synthetic dataset was effectively utilized to train a Gaussian Process Regression (GPR) model for bare land soil moisture retrieval, demonstrating a viable alternative to real-world data scarcity for machine learning model training.

Contributions

• Presents a novel methodology for generating large-scale synthetic microwave backscatter data using HFSS, addressing the critical challenge of data scarcity for training machine learning models in soil moisture retrieval.
• Enables the development of robust soil moisture retrieval models (e.g., GPR) that require less a priori information and are independent of extensive real-world SAR datasets.

Funding

• Not specified in the provided text.

Citation

@article{Kukunuri2025Synthetic,
  author = {Kukunuri, Anjana N.J. and Maurya, Ajay Kumar and Singh, Dharmendra},
  title = {Synthetic data generation using microwave modeling with efficient application of machine learning for bare land soil moisture retrieval: a case study},
  journal = {Elsevier eBooks},
  year = {2025},
  doi = {10.1016/b978-0-443-34113-7.00019-5},
  url = {https://doi.org/10.1016/b978-0-443-34113-7.00019-5}
}