Abstract: Based on analyzing the shortages of current methods considering runoff distribution through a year when partitioning runoff into abundant and low state, a new method, weighting method based on factor analysis, for obtaining weight values of each month in a yearI was presented. This method was applied to partition yearly average flow of Wuzhou hydrological station on the main branch of Xijiang River into abundant and low state, and then Markov process method was applied to study the probability characteristic of state series. The results showed that self-conservative characteristics weakened orderly as low, normal and abundant state; no matter what initial state runoff was, the transition probabilities decreased orderly as low, normal and abundant state; during long evolvement of runoff, the emergence probabilities reduced orderly as low, normal and abundant state.

Key words: partition of abundant and low state, distribution through a year, weighting method based on factor analysis, Markov process