Abstract: By means of sliding-window Chebyshev expansion, a series of expansion coefficients and time variables of the polynomial is obtained. Through correlation analysis, a set of forecast factors and elimination factors with clear physics meaning and good correlativity is derived. Based on the multi-form combination of predictors, a forecast model of regional heavy rainfall during the preceding flood season in Fujian Province is developed. The results are as follows: it is rational and feasible to construct a forecasting model of regional heavy rainfall during the preceding rainy season in Fujian Province with Chebyshev expansion coefficients of the height field; the independent field factors can be quantificationally derived from the features contained in the element field through Chebyshev orthogonal polynomials; by using the Chebyshev expansion technique with sliding windows, the multi-scale characteristics may be obtained as a result of making the utmost use of the field information, thus to gain more comprehensive knowledge of predictands.