In this research, the multiple linear regression models for the prediction of the Changma onset in Korea have been developed. The predictors are based on correlation analysis between Changma onset dates over 30 years (1973–2002), and the winds and geopotential heights at high, middle, and low levels of the troposphere during the preceding 6 months (December to May). The clearest correlation was observed at 850 hPa geopotential height field in the preceding April among 6 months. The fundamental meaning of this model indicates that Changma in Korea comes earlier when the Mascarene high becomes languished and the tropical storms around India and western North Pacific subtropical high (WNPSH) become strengthened. The reliability of this model has been tested and proved by statistical cross-validation analysis, and additionally, has been analysed for differences in the 850 hPa stream flows for those years that Changma comes earlier and later. From the preceding April onwards, the Mascarene high tends to become weakened so that the cross-equatorial flow is strengthened along the African east coast. The cross-equatorial flow fortifies the tropical storms around Indian as it moves to the east. In the long run, the flow is influenced so that the WNPSH is more developed to the north, influencing Changma rain band to move north faster.
To determine why Changma comes later in Korea, the water equivalent of accumulated snow depth (WEASD) difference between the late and early onset phases in the preceding April has been analysed. The result indicated that WEASD had a positive value for most East Asian regions. Since the snow reflects solar radiation by the Albedo effect, it decreases the difference in temperature between continents and oceans by decreasing the surface air temperature. The result has been confirmed by analysing the sensible heat net flux (SHTFL) difference between the two phases. In this way, small temperature gaps between continents and oceans weaken the WNPSH allowing Changma rain band to more slowly move north.