Long Range Forecasting Indian Summer Monsoon Rainfallby a Hybrid Principal Component Neural Network Model

Abstract

The existing methods based on statistical techniques for long range forecasts of Indian monsoon rainfall have shown reasonably accurate performance, for last 11 years. Because of the limitation of such statistical techniques, new techniques may have to be tried to obtain better results. In this paper, we discuss the results of an artificial neural network model by combining two different neural networks, one explaining assumed deterministic dynamics within the time series of Indian monsoon rainfall (Model I) and other using eight regional and global predictors (Model II). The model I has been developed by using the data of past 50 years (1901–50) and the data for recent period (1951–97) has been used for verification. The model II has been developed by using the 30 year (1958–87) data and the verification of this model has been carried out using the independent data of 10 year period (1988–97). In model II, instead of using eight parameters directly as inputs, we have carried out Principal Component Analysis (PCA) of the eight parameters with 30 years of data, 1958–87, and the first five principal components are included as input parameters. By combining model I and model II, a hybrid principal component neural network model (Model III) has been developed by using 30 year (1958–87) data as training period and recent 10 year period (1988–97) as verification period. Performance of the hybrid model (Model III) has been found the best among all three models developed. Rootmean square error (RMSE) of this hybrid model during the independent period (1988–97) is 4.93% as against 6.83%of the operational forecasts of the India Meteorological Department (IMD) using the 16 parameter Power Regression model. As this hybrid model is showing good results, it is now used by the IMD for experimental long-range forecasts of summer monsoon rainfall over India as a whole.