Turkish Journal of Agriculture and Forestry
Robust soil moisture content detection based on data augmentation under different testing conditions
Abstract
Accurate and cost-effective monitoring of soil moisture content is essential for the development of precision agriculture and the optimization of water resource management. While low-frequency acoustic signals are used to detect soil moisture content, models trained with controlled laboratory acoustic data often exhibit significant performance degradation with field data due to their sensitivity to environmental noise, soil heterogeneity, and structural variations. This study proposes a novel framework to enhance model robustness and generalization by integrating a WGAN-GP network data augmentation strategy into the soil moisture content workflow. This framework is designed to generate high-quality and diverse data on different soil types under different moisture conditions, consisting of an augmented dataset with original data. The datasets are handled by different data processing methods to train the Swin-Transformer regression model. We then compare the model performance for different datasets. The baseline method achieves root mean square error (RMSE), mean absolute error, and R2 values of 6.715%, 1.829%, and 0.711 for the field test set. In contrast, the proposed generative adversarial network-augmented method reduces the field test RMSE value to 4.852% and improves the R2 value to 0.837, representing a 28% reduction in RMSE and confirming that the method proposed in this study is more robust at handling data on different soil types.
Author ORCID Identifier
YANG LI: 0000-0002-4268-4004
CHENGHAO LIU: 0009-0009-6740-9349
JING NIE: 0000-0002-3763-9559
JINGBIN LI: 0000-0003-4264-7024
DOI
10.55730/1300-011X.3478
Keywords
Soil moisture content detection, Swin-Transformer image regression, generative adversarial networks, data augmentation
First Page
476
Last Page
489
Publisher
The Scientific and Technological Research Council of Türkiye (TÜBİTAK)
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
LI, Y, LIU, C, NIE, J, & LI, J (2026). Robust soil moisture content detection based on data augmentation under different testing conditions. Turkish Journal of Agriculture and Forestry 50 (4): 476-489. https://doi.org/10.55730/1300-011X.3478
