Keywords: short–term rainfall, multi–linear regression, auto regressive integrated moving average, ARIMA, artificial neural networks, ANNs, model tree, daily precipitation, precipitation mapping, rain forecasting, data driven techniques, daily rainfall
Daily precipitation mapping and forecasting using data driven techniques
Accurate prediction of precipitation is a challenging task because of the non–linearity and randomness involved in the processes. Most of the conventional models fail to capture this non–linearity and randomness. In the present study, data driven techniques like artificial neural network (ANN) and model tree (MT) are used for predicting the short–term precipitation. Several models are developed using these techniques by changing the parameters and the results are compared with conventional techniques such as multi–linear regression (MLR) and auto regressive integrated moving average (ARIMA) models. The effect of smoothing and pruning is also studied in MT models. The study shows that MT performed better than other techniques with a correlation of 0.70. The advantage of MT over ANN is that, the MT gives several linear equations, which can be easily integrated into other models. Thus, it may be concluded that the highly erratic daily rainfall could be mapped and predicted better using data driven techniques.