Our study has focused on improving the prediction skill hrough the construction of multiple regression models using different approaches for generating superensemble forecasts. We have shown that the results of the proposed techniques are clearly better than those of the conventional superensemble method, and the superensemble forecast that is based on the SVD method demonstrates the best result from computation of the covariance matrix. The SVD method removes the singularity problem that the conventional technique appears to retain. In constructing the covariance matrix of the SVD method, we have used only anomalies that are bias-corrected and that have the seasonal cycle-removed. Obviously, the SVD technique explains the variance better than the other techniques. The SVD technique decomposes the covariance matrix in the orthogonal matrix. This orthogonality relation explains the maximum variance and it minimized the covariance matrix error. A postprocessing algorithm based on multiple regression of multi-model solutions toward observed fields during a training period is one of the best solutions for long-term prediction. Due to the cancellation of biases among different models, the forecast superensemble errors are quite small. Our study shows that the proposed techniques reduce the forecast errors below those of the bias-removed ensemble mean and the conventional superensemble technique.