Abstract: This study examined the predictors for the dominant EOF modes of autumn precipitation in Southwest China (SWC) based on the causality of information flow. A statistical model of autumn precipitation in SWC was then established. Finally, the prediction skills of the empirical model were evaluated. The first two dominant modes of autumn precipitation during 1979–2020 in SWC are the basin mode and saddle mode, which were obtained by empirical orthogonal function (EOF). These modes are closely related to the development of eastern and central El Niño. The predictors of the PCs of the first two dominant modes were chosen based on the causality of information flow. A multiple linear stepwise regression with a leave-one-out method was employed to further select the predictors and establish a statistical model. During the training period between 1980 and 2015, the correlation coefficients between the predicted PC1 and PC2 and the actual PCs were found to be 0.89 and 0.83, respectively. Additionally, the sign coincidence rates were determined to be 90% and 83%, respectively. In the forecast years from 2016 to 2020, the predicted PC1 and PC2 are in phase with the actual PCs in four years, with a sign coincidence rate of 80%. During the 36-year training period from 1980 to 2015, the averaged anomalous pattern correlation coefficient (ACC) between the reconstructed precipitation with the predicted PCs and the observed precipitation anomalies is 0.48. The ACC exceeds 0.5 for more than half the years. The regional average temporal correlation coefficient (TCC) is 0.48. Additionally, a similar-year forecast was conducted with predicted PC1 and PC2, which was employed to address the limitations in the reconstructed field regarding precipitation data.