Improving rainfall estimation from ground based radar measurements using neural networks

Neural network is a nonparametric method to represent the relationship between radar measurements and rainfall rate. The performance of neural network based rainfall estimation is subject to many factors such as the representativeness and sufficiency of the training dataset, the generalization capability of the network to new data, seasonal changes, and regional changes. Improving the performance of the neural network in real time context is of great interest. In this paper, the goal is to improve rainfall estimation based on Radial Basis Function (RBF) neural networks. The principal components analysis (PCA) technique is used to reduce the dimensionality of the training dataset. Reducing the dimensionality of the input training data will reduce the training time as well as reduce the network complexity. More importantly, the small scale uncertainty will be removed during PCA such that the network is less likely overfitted. In addition, ¿Rain/No Rain¿ detection is performed using an adaptive neural network running simultaneously with the rainfall estimation neural network. The ¿Rain/No Rain¿ detection can eliminate those ¿No Rain¿ data inputs from the training set.