Application of random forest to classify weather observation into rainfall using GNSS receiver

In the GNSS meteorology, it is known that the zenith total delay time obtained from the positioning process in GNSS receivers is showing potential rainfall intensity, however, its precision for rainfall nowcasting is not practically high due to high false alarm. In order to enhance the precision of rainfall nowcasting, this paper employs sensor fusion for collecting various kind of information obtained from not only GNSS but also meteorological sensor. A machine learning technique is employed to classify many weather conditions into precipitation or not. In this paper, the classification performance is investigated as the random forest algorithm is applied for binary classification. Better performance can be obtained and the seasonal difference is clearly shown compared to without using a sensor fusion technique.