Estimation of soybean [Glycine max (L.) Merr.] yield early in the growing season is an appealing idea for both, farmers and soybean-related industries. Prior attempts to predict soybean yield have had limited success, especially when using information early in the growing season. The objective of this study was to evaluate the release date and maturity group (MG) of the cultivar, digital imaging, reflectance, and weather data during successive stages of crop development as explanatory variables in a soybean yield prediction model. The data were collected in the North Central (NC) United States at Arlington, WI (2010-2011), and Lafayette, IN (2011), using 59 MG II cultivars (released 1928-2008) at Wisconsin, and 57 MG III cultivars (released 1923-2007) at Indiana that were planted in performance trials on two planting dates (May and June). A second order polynomial regression analysis followed by ridge regression was used to develop the soybean yield prediction equation. The model accounted for 80% of the yield variability in the NC U.S. data set. An additional dataset not used in the calibration was used to conduct a validation test of the predictive performance of the model. The average difference between the fitted and actual yields in the validation test was 67 kg ha-1. Results from this study suggest that the use of cultivar release year, planting date, MG, near-infrared (NIR), visible red (RED), and Red-edge wavelength bands recorded at 77 d after planting, and weather data 30 d before and after the planting date can closely estimate soybean yields in the Midwest.
ASJC Scopus subject areas
- Agronomy and Crop Science