Generally,dynamical analyses of the original systems are limited to some extent by the observational methods.In this paper,by using the method of multivariate singular systems analysis (M SSA),which combines simultaneously information sampling the complete range of the spatial cross correlation function in signals with complex spatio temporal structure,we study the changes of the correlation dimensions and the largest Lyapunov exponents of the multivariate time series in an- observation window.And the characterization of prediction is discussed also with the multi-layer- perceptron.It is shown that when the number of the variates is increased,the complexity is also-- increased and the stability and the characterization of prediction are different.