Six‐state discrete simple Markov chain models are applied to the 5‐day spatial rainfall features during the summer monsoon for 30 years (1964–1993) over 33 meteorological subdivisions of India to understand the persistence behaviour of the spatial coverage of rainfall and the underlying time‐evolutionary processes on the synoptic scale. The stochastic models are cross validated on 5 years of independent data (1994–1998) by evaluating various measures of forecast skill. It is revealed that the spatial coverage of rainfall does exhibit first order Markov persistence for all features but diminishes for longer intervals for some features.
The stochastic matrix, together with the climatic information about the spatial coverage of 5‐day rainfall features, could be an aid to the operational forecaster to judge the evolution of specific areal rainfall events for qualitative prediction on a medium to extended range scale over a subdivision. Copyright © 2000 Royal Meteorological Society