Abstract: Based on L-dimensional original images X₁, X₂,…, X_N defined with L forecasting methods (L≥3), 2-dimensional images Y₁, Y₂, …, Y_N are set in accordance with the restraints that the sum of squares for the deviation of the ordinate from the historic value of the forecast objects should be no greater than the one by the original methods, so that the error E is kept the minimal in the determination of synthesized field that is capable of both objectively integrating forecast information by the original L methods and keeping the fitting rate above the one by the old methods. With transformation by the Lagrangian multiplier and construction of a new function Z, it is now possible to directly have solution on the computer by means of the simplex method.