Article Preview
Top1. Introduction
Rainfall time series forecasting has, of late, emerged as one of the most vital challenges in water resources planning and management. Forecasting of rainfall variable is employed for flood and drought prevention, reservoir operation, contract negotiation, in addition to irrigation scheduling (Araghinejad et al., 2011). Rainfall is an intricate atmospheric process, that is generally space and time dependent and it becomes very hard to forecast. In view of the apparent random characteristics of rainfall sequences, they are usually expressed by a stochastic process (Chinchorkar et al.). Conventional techniques like time series regression, exponential smoothing, and autoregressive integrated moving average (ARIMA), have been found extensively offered for stochastic time series analysis. Especially, the ARIMA brand is characteristic of time series models and has been able to attain amazing level of popularity. The conventional time series modeling techniques have extended a helping hand to the scientific community from time immemorial; Nevertheless, they have been able to offer only a rational level of accuracy and have been found to face music from the postulations of stationary and linearity (Brockwell and Davis, 1994; Zhang, 2003). A ruthless rainfall prediction technique with the help of severe rainfall artificial intelligence (SRAI), endowed with the quality of forecasting the time sequence deviation and spatial distribution of harsh rainfall in a basin area, is in the offing. A backdrop which triggers the forecasting tasks for relentless rainfall prediction by means of weather forecaster is the underlying goal of the SRAI (Oishi et al.). The forecast of atmospheric parameters is a sine-qua-non for many an application. Prominent among them embrace climate monitoring drought detection, planning in power industry, aviation industry and communication pollutions dispersal (Pal et al., 2003) etc.
Rainfall prediction is a difficult job in the climate dynamics and climate prediction hypothesis (Wu, 2011; Wu and Jin, 2009; Hong, 2008). Generally, rainfall forecasting is a very complicated nonlinear pattern, involving constraints such as pressure, temperature, wind speed and its meteorological traits of the precipitation zone and the like. In the good old days, several techniques have been employed for forecasting the rain fall prediction. Certain machine learning techniques utilized to forecast the rainfall (Cristianini et al., 2000) reveal the rainfall prediction by means of support vector machine methods. Many soft computing approaches were used for rainfall forecasting .The ANN forecasts the overall Indian summer monsoon rainfall with diverse Meteorological parameters as brand inputs (Venkatesan et al., 1997). Several investigations have been done for the purpose of quantitative precipitation forecast (QPF) by means of diverse techniques encompassing numerical weather prediction (NWP) brands and remote sensing observations (Davolio et al., 2008), statistical models (Nayagam et al., 2008), non-parametric nearest-neighbors method (Toth et al., 2000), and soft computing based methods including artificial neural networks (ANN), support vector regression(SVR) and fuzzy logic (FL) (Brath et al., 2002).