Forecasts of global crop yields prior to planting have generally been single values, based entirely on past trends. Regression analysis testing a combination of data from ENSO (El Niño/Southern Oscillation) and ARMA models suggests that yield forecasting errors can be reduced, generating more normal distributions of these errors. Keywords: El Niño, ENSO, forecasting crop yields, long range weather forecasting, agricultural modeling, food security, risk management