Experimental Study of Detecting Rainfall Using Microwave Links: Classification of Wet and Dry Periods

• Published in: IEEE journal of selected topics in applied earth observations and remote sensing 2020, Vol.13, p.5264-5271
• Main Authors: Song, Kun, Liu, Xichuan, Zou, Mingzhong, Zhou, Ding, Wu, Haonan, Ji, Feng
• Format: Article
• Language: English
• Subjects: Accuracy; Attenuation; Attenuation measurement; Classification; Classification of wet and dry period; Forecasting; Links; Microwave attenuation; microwave link; Microwave measurement; Microwave theory and techniques; Parameters; Probability theory; Rain; rain detection; Rainfall; Statistical analysis; Statistics; support vector machine (SVM); Support vector machines; Training
• ISSN: 1939-1404; 2151-1535
• DOI: 10.1109/JSTARS.2020.3021555

To improve the accuracy of rainfall estimation by microwave links, this article presents a method for classifying wet and dry periods based on the support vector machine (SVM). The average, minimum, and maximum attenuation measurements in 5 min are applied as the feature vector of the SVM after the analysis of the relation between the statistical parameters of the attenuation measurements from seven microwave links and the wet/dry periods. When the baseline attenuation is needed for retrieving the path-averaged rain rate, the method can classify the wet/dry periods and estimate a dynamic baseline with an optimal combination of the statistical parameters of the attenuation measurements based on the prior training. Experiments are conducted to test the classification method. The results show that the classification accuracy is higher than 0.8, which is a satisfactory result. Most values of the true positive rate are higher than 0.9, which indicates that the method can correctly classify most of the wet periods. Additionally, the values of the false positive rate are less than 0.3, and most of the values are less than 0.2, suggesting that the method incorrectly classifies the dry period as the wet period with a low probability. The results demonstrate that the classification method is capable of classifying the wet and dry periods with a high accuracy, which can help improve the precision of the baseline of microwave links and rainfall estimation.