Generalized Algorithm for Variable-Rate Nitrogen Application in Cereal Grains

Many different mathematical algorithms have been developed and used in conjunction with commercial sensors for sensor-based nutrient management. Several of the N algorithms have led to the precise mid-season prediction of yields and calculation of sidedress N rates. The original Oklahoma State University (OSU) algorithm identified several limitations that were addressed in this study. Based on data analyses from more than 390 winter wheat (Triticum aestivum L.) and 200 corn (Zea mays L.) experiments and analyses of more than 100 N-rich strips, a generalized algorithm (for both corn and wheat) was developed to estimate the optimum N application rate based on spectral measurements. The generalized model adjusts the yield calibration curve for growth stages and better predicts corn and wheat yields. The coefficients of determination of the generalized model explained 5 to 6% less of the model error than the individual regressed data for both crops. Mean absolute error (MAE) was approximately 0.9 Mg/ha greater with the generalized model than with the individually regressed model. The larger MAE with the OSU generalized model was due to sensitivity to location of the inflection point; however, this sensitivity did not impact the calculated fertilizer rates. The generalized model reported here using normalized difference vegetation index sensor measurements collected midseason can be used to apply fertilizer N with changing growth stage for both corn and wheat.