The data of the beginning of the rainy season of12 weather stations in Dali from 1962 to 2008 and height,wind fields from
NCEP/NCAR were used to analyze the characteristics of spatial and temporaldistribution and anomalous circulation on the beginning of
the rainy season in Dal.i The results ofEOF analysis indicates that there are severalkinds of spatialpatterns about the beginning of the
rainy season in Dal,i and themajor patterns are the earlier(or later) in thewhole area and the southeast-northwestpattern. The results from Morletwavelet transform shows that the beginning of the rainy season there was earlier from the1960s to the middle of the 1970s and later from themiddle of the 1970s to the 1990s then earlierafter the 1990s. Italso shows that the beginning of the rainy season presented periodic variation of 2 and 10 years. It is a“-+-”distribution alongmeridionalover theNorthernHemisphere in the height departure field inMay of the earlieryears of rainy season, buta“+-+”distribution inMay of the lateryears of rainy season.It is obviously differentofwind departure field atupperand lower level in anomalous years of the beginning of rainy season. There is an anticyclonic circulation to strengthen the Iran high at the upper level and a strongerSouthAsiamonsoon circulation in the earlier years of rainy season, but it is the opposite circulation in the later years of rainy season.

Based on traditional statistical method, the multi - level recursive regression method was used to make up the inadequate p roblem that regression coefficient is a constant in conventional statisticalmethod. Using the multi - level recursive regression model,the forecasting p recip itation amounts inMay for 160 weather stations over China are investigated. The results show that the model can reasonably simulate the precip itation amounts over Central China, East China, most of Guangxi and Guangdong p rovinces, a majority of Yunnan and Guizhou p rovinces in the early stage of summer.