Cascade direction and structure of multiscale volatilities in SSECI was investigated empirically. The results indicate that there exist the mutual causal relationships between the coarse and fine volatilities within critical timescale of 4 days. When exceeding the criticality, the causal relationships only display the single direction. The additional two measures of crossed correlation coefficient and power spectrum were simultaneously applied to the cascade characteristics of volatility. The further researches suggest that the lagged coarse volatilities can predict well the fine volatilities over three magnitudes of 101～103min timescale, and that the cascade structure exhibits the single direction from the course to fine volatilities. The framework of four types of cascade model and the basic idea of risk management are also proposed.