Spatio-Temporal Modeling and Forecasting of Corn Yield.

A timely and accurate crop yield forecast enables better decision on crop selection, soil and crop management, marketing, storage, transport, and assessing risk associated with these activities. Crop yield is closely related to the crops genetics, seasonal weather, inputs, and soil and crop management. Therefore, modeling and forecasting yield is possible with knowledge of these variables. Among these variables that are essential to model yield, information on crop genetics, inputs, and soil and crop management can be accurately collected as early as time of sowing. However, all weather information pertinent to forecasting yield is only available when it is close to harvest. Therefore, accuracy in early prediction is always questionable. Several time series models and regression models with autocorrelated errors are proposed to fit corn yield data. The objectives of our research is to select optimal model for corn yield forecasting using historic yield and climatic data based on multiple statistical techniques and suggest appropriate model to forecast yield for different regions. The merits and disadvantage of each technique, as well as the possible spatial-temporal correlation adjustment in the modeling will be discussed.