本文以广西1962—1981年早稻平均产量为样本,采用贝叶斯判别法和回归分析相结合的方法,建立从定性到定量的大尺度产量系列预报模式。产量预报模式分两步建模:第一步选取与气象产量相关密切的大气环流因子,建立贝叶斯定性判别方程,判别未来气象产量属于正值(增)或负值(减),即Y_w≥0或Y_w<0。第二步是在一级定性预报模式的基础上,按历史年两大类气象产量分别选取预报因子,运用统计回归方法,建立二级定量预报模式A和模式B。两级早稻产量预报模式,预报时效长,外延性能好,经历史年检验,拟合率达100％,试报和预报效果良好。 In this paper, the average yield of early rice in Guangxi from 1962 to 1981 was taken as a sample, and the Bayesian discriminant method was combined with regression analysis to establish a large-scale yield forecasting model from qualitative to quantitative. The production forecasting model is modeled in two steps: Step 1: Select the atmospheric circulation factor closely related to the weather output and establish the Bayesian qualitative discriminant equation to determine whether the future meteorological production is positive (increase) or negative (decrease), that is, Y_w ≥0 or Y_w <0. The second step is to select forecast factors based on the first-level qualitative forecasting model and two types of meteorological production in historical year respectively, and use statistical regression method to establish two-level quantitative forecasting model A and model B. The two-stage early rice yield forecasting model has the advantages of long-term forecasting, good epitaxial performance and historical history test. The fitting rate is 100%, and the test and forecast results are good.