We use models incorporating the normalized difference vegetation index (NDVI) derived from remote sensing satellites to improve soybean yield predictions in 10 major producing states in the United States. Unlike traditional methods that assume an ordinary least squares model applies to all observations, we allow for global flexibility in the relationship between NDVI and soybean yield by using the flexible Fourier transform (FFT) model. FFT results confirm that there is a nonlinear response of soybean yield to NDVI over the growing season. Out-of-sample predictions indicate that allowing for global flexibility with the FFT improves the predictions in time-series prediction and forecasting.
