ABSTRACT

Prediction of rainfall on a given time period (daily, monthly, seasonal and annual) is of utmost importance for planning of irrigation and drainage system as also for command area development. With the development of Artificial Intelligence (AI), number of AI methods such as Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System, Fuzzy Logic, Support Vector Machine and Evolutionary Optimization Algorithm are generally applied for rainfall prediction. Out of which, ANN has an ability to obtain complicated non-linear relationship between the variables, which is suitable to predict the rainfall. This paper presented a study on prediction of annual maximum rainfall (AMR) of Gaganbawada, Lanja and Radhanagari using Multi-Layer Perceptron (MLP) and Radial Basis Function (RBF) networks of ANN. For this purpose, the annual maximum series of meteorological data viz., rainfall, minimum and maximum temperature and average wind speed was generated from the daily data observed at Gaganbawada (1950 to 2020), Lanja (1950 to 2021) and Radhanagari (1950 to 2021), and used as an input for prediction of AMR through MLP and RBF. The performance of the MLP and RBF networks applied in rainfall prediction was evaluated by model performance indicators such as correlation coefficient, Nash-Sutcliffe model efficiency and root mean squared error. The study showed that MLP is better suited amongst two networks of ANN applied for prediction of AMR of Gaganbawada, Lanja and Radhanagari.

Keywords: - Correlation Coefficient, Multi-Layer Perceptron, Radial Basis Function, Model Efficiency, Rainfall, Mean Squared Error