Remote Sensing | March 31, 2023
Researchers from the Henan Engineering Research Center of Spatial Information Processing in China conducted a study focusing on soil moisture prediction, an essential factor connecting agriculture, ecology, and hydrology. Surface soil moisture (SSM) prediction plays a crucial role in irrigation planning, water quality monitoring, water resource management, and estimating agricultural production. To assess SSM in agricultural areas, the researchers utilized multi-source remote sensing.
However, limited field-measured SSM samples can severely impact the accuracy of SSM inversion using remote sensing data. To address this challenge, the researchers proposed an SSM inversion method suitable for small sample sizes. They employed the alpha approximation method to expand the measured SSM samples, providing more training data for SSM inversion models.
Feature parameters were extracted from Sentinel-1 microwave and Sentinel-2 optical remote sensing data, and three methods—Pearson correlation analysis, random forest (RF), and principal component analysis—were used for feature optimization. Three machine learning models suitable for small sample training—RF, support vector regression, and genetic algorithm-back propagation neural network—were built for SSM retrieval.
Experimental results demonstrated that after sample augmentation, SSM inversion accuracy improved. The combination of RF for feature screening and RF for SSM inversion yielded the highest accuracy, with a coefficient of determination of 0.7256, a root mean square error of 0.0539 cm3/cm3, and a mean absolute error of 0.0422 cm3/cm3. The proposed method was successfully applied to invert regional SSM, showcasing its effectiveness with small sample sizes.