Comparison of models and indicators for categorizing soft wheat fields according to their grain protein contents

The ability to deliver wheat grain with a specific protein content is a major determinant of the profitability of wheat grain production. Various crop models have been developed to predict yield and grain protein content on a field scale. They can be used to predict each year, before harvest, the yields and grain protein contents of the different fields in a collecting area, leading to an optimization of the grading process into low and high protein standards. Indicators have been developed for nitrogen management at the field scale. They can be used to predict grain quality because grain protein content depends strongly on the crop nitrogen uptake during the vegetative growth of the wheat crop. The aim of this study was to evaluate the accuracy of two indicators, nitrogen nutrition index and chlorophyll content of leaves (in SPAD units), and of two models for categorizing fields according to their grain protein contents. A data set including field measurements over 3 years was used to estimate the sensitivity and specificity of the models and indicators using the receiver operating characteristic evaluation procedure. High values of sensitivity and specificity were obtained for the two indicators, and decision thresholds leading to low false negative and false positive proportions were identified. Nitrogen nutrition index showed the best results for the three grain protein content thresholds considered. Conversely, sensitivity and specificity values obtained for the two models were low. Combining model predictions and indicator values did not give better discrimination than the use of a single indicator.