Effect of Trend or Technology Variables and Record Period on Prediction of Corn Yields with Weather Variables

Abstract

Reliable crop-weather-technology models are needed for yield prediction and for the determination of climatic risks in crop production. Accurate and timely information on crop yields would help both industry and government anticipate and more efficiently allocate food supply and create a more stable grain price for individual farm planning. Critical analyses of several multiple regression models for predicting county average corn (Zea mays L.) yields with weather information in Indiana showed that yield predictions are very sensitive to the trend variable(s) used to account for technology and also to the specific period of record used to estimate the regression parameters. Models using a function of year as the variable to represent technology trend showed great fluctuation in their parameter estimates and in their yield predictions from year to year. Predicted yields with average weather fluctuated up to 22 bushels per acre from one year to the next in the 1970's as each additional year was added to the fitted model. The use of nitrogen applied to corn land as a technology marker in the regression model reduced the variability in the estimates of the technology and weather coefficients and improved the yield predictions. The identification of the principle environmental factors with a single derived Energy-Crop Growth (ECG) index facilitates the search for other technological and management variables to interpret better the weather-technology interaction in corn production.