Development of a Method for Classifying Convective and Stratiform Rains from Micro Rain Radar (MRR) Observation Data Using Artificial Neural Network

Abstract

This study examined the performance of Artificial Neural Network (ANN)-backpropagation to classify rain types from observations of Micro Rain Radar (MRR) in Serpong (6.359oSL; 106.673oEL). The inputs of ANN are radar reflectivity, Doppler velocity, and Liquid Water Content (LWC). Rain events on January 5, 2017; at 16.28 – 21.21 local time were used as training data. The ANN results were validated with rain classified by the Bright Band (BB) and Countour Frequency by Altitude Diagram (CFAD) methods. The most appropriate ANN-backpropagation architecture is the 3-6-1 architecture (input layer-hidden layer-output layer), with an activation-transfer function being competitive and a learning rate of 0.9. The Mean Square Error (MSE) of the training step was 0.0098735, and the average percentage of accuracy for the test step was 94%. A rain event with a single type of rain can be classified accurately by ANN and gives the same results as the CFAD method. Thus, the ANN can be a solution to the shortcomings of the BB method, which sometimes classification results of a single type of rain events is interspersed with another type, which is physically impossible.

Artificial intelligence applications in solar energy

“…The artificial neural network, also known as an ANN, is a simplified form of biological-based neural architecture that is capable of effectively correlating a greater number of uncertain input points to a variety of characteristics [115]- [117]. There is no need for mathematical equations or a sophisticated mathematical base when it comes to the process of building relationships between various factors [118], [119]. ANN models use such an approach.…”