Analysis of persistence in daily monsoon rainfall over India.

In earlier studies the authors have established that monsoon rainfall up to the time scale of at least 5 days is dependent on its past value. In this study we examine the dependence in daily rainfall for 12 stations spread all over India, and in daily rainfall of 10 meteorological regions during the summer monsoon; 1) by analysing the rainfall as a stochastic point process by fitting Markov chains of orders 1 to 4 to the station data and 2) by fitting autoregressive and autoregressive moving average models to the spatially averaged data. The significant conclusion is that the Markov chain of order 3 fits the runs of wet and dry spells better than chains of lower order, and that the autoregressive model of order 1 fits the spatially averaged rainfall amounts satisfactorily as compared to the higher order autoregressive of ARMA models. Examination of rainfall intensities during spells of various lengths has revealed that the rain intensity is greater during spells of 2 or 3 days than on isolated days. The rain intensity is greater on the days lying in the middle of a wet spell than on isolated days or on the first and the last days of a longer wet spell