Most of the northeastern region of Brazil (NEB) has a maize production system based on family farming, with no technological advances and totally dependent on the natural rainfall regime, which is concentrated in 4 to 5 months in most parts of the region. This means that the productivity of this crop is low in the NEB. In the northern mesoregions of the NEB, rainfall is concentrated between January and June, in the east of the NEB from April to September, and in the west of the NEB from October to March. The growing season takes place during these semesters. With this in mind, our objective was to develop a model based on canonical correlation analysis (CCA) to predict corn production in the mesoregions of the NEB between 1981 and 2010, using accumulated precipitation per semester as the predictor variable and predicting the observed production in kg/ha. Our results showed that the CCA model presented higher correlations between observed and simulated production than that obtained simply from the direct relationship between accumulated rainfall and production. The other two metrics used, RMSE and NRMSE, showed that, on average, in most mesoregions, the simulation error was around 200 kg/ha, but the accuracy was predominantly moderate, around 29% in most mesoregions, with values below 20% in six mesoregions, indicative of better model accuracy, and above 50% in two mesoregions, indicative of low accuracy. In addition, we investigated how the different combinations between two modes of climate variability with a direct influence on precipitation in the NEB impacted production in these 30 years, with the combination of El Niño and a positive Atlantic dipole being the most damaging to harvests, while years when La Niña and a negative Atlantic dipole acted together were the most favorable. Despite the satisfactory results and the practical applicability of the model developed, it should be noted that the use of only one predictor, rainfall, is a limiting factor for better model simulations since other meteorological variables and non-climatic factors have a significant impact on crops. However, the simplicity of the model and the promising results could help agricultural managers make decisions in all the states that make up the NEB.