[Objective] To discuss how to forecast annual precipitation of the Taihu Lake basin and find regularity of annual precipitation.[Methods] Annual precipitation data of Taihu Lake basin from 1960 to 2010 were analyzed. The annual rainfall forecast model was established by the weighted Markov chain.[Results] The precipitation of the Taihu Lake basin was below average precipitation for years in the 1960s

1970s and 1980s. Based on the verification of the Markov chain characteristics of precipitation

the weighted Markov chain model was used to predict the annual precipitation state and annual rainfall of 2011 and 2012

the relative errors were 6.59% and -10.74%. Analysis on Markov chain of various step size showed that the appearance probability of normal flow year

partial low flow year

partial flow year

low flow year and wet year was 0.268 5

0.206 7

0.184 5

0.166 1 and 0.174 2

respectively.[Conclusion] The relative errors of weighted Markov chain precipitation simulation from 2002 to 2011 shows that the relative errors of wet year and partial low flow year are bigger and the normal flow year is smaller. The appearance probability of normal flow year

partial low flow year and partial flow year are larger.