Time series prediction of rain rate during rain events at a tropical location

Abstract

A new model is proposed to predict rain rate series during a rain event at a tropical location. The model is based on Gaussian distribution of conditional occurrence of rain rate with a particular value of the rain occurring earlier. The mean and standard deviation of the distribution are modelled with the measured data. The rain rate at a particular time instant is predicted from the knowledge of previous samples. The predictor has tested well with a maximum mean error within 20%. This model produces predicted rain rate series whose first and second-order statistics fit very accurately with those of the experimental data. The method can also be applied with increased error, even if rain rate measurements are missing for certain periods of time. Comparisons with other models are also shown.