Abstract:Soil moisture is vital for the crop growth and directly affects the crop yield. The conventional synthetic aperture radar (SAR) based soil moisture monitoring is often influenced by vegetation cover and surface roughness. The machine-learning methods are not constrained by physical parameters and have high nonlinear fitting capabilities. In this study, machine-learning methods were applied to estimate soil moisture over winter wheat fields during its growing season. RADARSAT-2 data with quad polarizations and 240 sample plots in the study area were acquired and collected, respectively. In addition to the four linear polarization channels, polarimetric decomposition parameters were extracted to expand the SAR feature space. Three advanced machine-learning models were selected and compared, which were support vector regression, random forests (RF), and gradient boosting regression tree. To improve the performances of the models, three feature-selection methods were compared, which were based on Pearson correlation, support vector machine recursive feature elimination, and RF, respectively. The coefficient of determination (R2) and root-mean-square error (RMSE) were used to compare and assess the performances of those models. The results revealed that polarimetric decomposition parameters were effective in estimating soil moisture, and RF model obtained the highest prediction accuracy (training set: RMSE = 2.44 vol.% and R2 = 0.94; and validation set: RMSE = 4.03 vol.%, and R2 = 0.79). This study finally concluded that using polarimetric decomposition parameters combined with machine-learning and feature-selection methods could effectively estimate soil moisture at a high accuracy, which helps monitor soil moisture across the agricultural field during its growing season.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to estimate soil moisture content (Soil Moisture Content, SMC) during the winter wheat growing season by using machine - learning methods. Specifically, the research aims to: 1. **Explore the potential of combining polarization parameters and backscattering coefficients in soil moisture retrieval in agricultural areas**: By using multiple polarization parameters extracted from the polarization decomposition model and combining them with the backscattering coefficients, the accuracy of soil moisture estimation is improved. 2. **Evaluate the performance of three machine - learning models in soil moisture estimation**: These three models are Support Vector Regression (SVR), Random Forest (RF) and Gradient Boosting Regression Tree (GBRT), and they are evaluated in combination with three feature selection methods (based on Pearson correlation coefficient, Support Vector Machine Recursive Feature Elimination and Random Forest). ### Research Background Soil moisture is an important factor for crop growth and directly affects crop yields. Traditional Synthetic Aperture Radar (SAR) soil moisture monitoring methods are often affected by vegetation cover and surface roughness. However, machine - learning methods are not limited by physical parameters and have a high non - linear fitting ability, so they are applied to the estimation of soil moisture. ### Research Methods 1. **Data Collection**: - **Ground - measured data**: Soil moisture measurements were carried out in a rain - fed agricultural area in southwestern Ontario, Canada, and data from 240 sample points were collected. - **RADARSAT - 2 data**: 8 fully - polarized RADARSAT - 2 images were obtained with a spatial resolution of 8 meters. - **Sentinel - 2 data**: Used to obtain vegetation description parameters of sample points. 2. **Data Pre - processing**: - **RADARSAT - 2 data**: Calibrate the original image, generate the T3 matrix, and perform polarization speckle filtering. - **Sentinel - 2 data**: Perform radiometric correction and atmospheric correction, and calculate NDVI as a vegetation description parameter. 3. **Feature Extraction**: - 30 feature parameter variables were extracted from RADARSAT - 2 data, including four linear polarization channels and polarization decomposition parameters. 4. **Model Construction and Evaluation**: - Use three machine - learning models, SVR, RF and GBRT, for soil moisture estimation. - Combine three feature selection methods (based on Pearson correlation coefficient, SVM - RFE and RF) to gradually increase the number of features and evaluate model performance. ### Main Results - **Feature Selection**: Different feature selection methods rank the importance of features differently. The feature selection method based on RF can achieve a higher prediction accuracy when using fewer features. - **Model Performance**: The Random Forest (RF) model shows the highest prediction accuracy on the validation set (training set: RMSE = 2.44 vol.%, R² = 0.94; validation set: RMSE = 4.03 vol.%, R² = 0.79). ### Conclusion The study finally concludes that using polarization parameters combined with machine - learning and feature - selection methods can effectively estimate soil moisture with high accuracy, which is helpful for monitoring farmland soil moisture during the crop - growing season.

Estimating Soil Moisture Over Winter Wheat Fields During Growing Season Using Machine-Learning Methods

Estimating Soil Moisture over Winter Wheat Fields during Growing Season Using RADARSAT-2 Data

Soil Moisture Retrievals By Combining Passive Microwave And Optical Data

Retrieval of Farmland Surface Soil Moisture Based on Feature Optimization and Machine Learning

Estimation of Soil Moisture Using Multi-Source Remote Sensing and Machine Learning Algorithms in Farming Land of Northern China

Soil Moisture Retrieval Using SAR Backscattering Ratio Method during the Crop Growing Season

Retrieving Soil Moisture Over Soybean Fields During Growing Season Through Polarimetric Decomposition

Improving Water Status Prediction of Winter Wheat Using Multi-Source Data with Machine Learning

Surface Soil Moisture Retrieval on Qinghai-Tibetan Plateau Using Sentinel-1 Synthetic Aperture Radar Data and Machine Learning Algorithms

Sentinel-1 Imagery Incorporating Machine Learning for Dryland Salinity Monitoring: A Case Study in Esperance, Western Australia

Comparison of Three Machine Learning Algorithms for Retrieving Soil Moisture Information from Sentinel-1A SAR Data in Northwest Shandong Plain, China

Wheat yield estimation using remote sensing data based on machine learning approaches

Evaluating soil moisture content under maize coverage using UAV multimodal data by machine learning algorithms

Soil Moisture Retrieval in Agricultural Fields Using Adaptive Model-Based Polarimetric Decomposition of SAR Data

Comparative Analysis of Machine-Learning Models for Soil Moisture Estimation Using High-Resolution Remote-Sensing Data

Prediction of wheat SPAD using integrated multispectral and support vector machines

Retrieving Surface Soil Moisture over Wheat and Soybean Fields During Growing Season Using Modified Water Cloud Model from Radarsat-2 SAR Data

Detecting Water Stress in Winter Wheat Based on Multifeature Fusion from UAV Remote Sensing and Stacking Ensemble Learning Method

Monitoring Soil Moisture over Wheat and Soybean Fields During Growing Season Using Synthetic Aperture Radar

Multilayer Soil Moisture Mapping at a Regional Scale from Multisource Data via a Machine Learning Method

UAV- and Machine Learning-Based Retrieval of Wheat SPAD Values at the Overwintering Stage for Variety Screening