Can sensors and crop models predict the need for late-season nitrogen for protein enhancement in spring wheat?

Background: Grain protein content impacts the market value of hard red spring wheat in North America. A post-anthesis application of aqueous UAN can increase protein levels. Predicting grain protein content prior to flowering would help growers determine the need for this extra nitrogen application. Our objective was to determine how useful pre-flowering NDVI values and crop models might be in predicting grain protein.

Methods: Field experiments with several N rates were conducted in multiple environments from 2013 to 2016. NDVI was measured at several crop growth stages. Regressions with NDVI and protein were calculated. Since low grain protein commonly occurs when yields are high, the DSSAT crop growth model was used as a means of predicting yield based on weather inputs.

Results: NDVI was found to be predictive of grain protein in some seasons (r2>50) but not others. Normalized NDVI was more likely to be predictive than NDVI alone. Predicting low protein levels (less than 12%) was more likely than moderate to high levels. The OSSAT model was more effective in predicting yield when predicted weather data was used at later growth stages.

Discussions: NDVI values tend to saturate before protein reached 13%, reducing its value to predicting only very low protein levels. Since weather plays such an important role in yield development, the DSSAT model is best at predicting a range of yield outcomes rather than a specific outcome.

Conclusion: Normalized NDVI values can be used in predicting very low protein levels. This finding could be adapted to the field level by using a drone that collects NDVI in fields with an N rich strip. The DSSAT model can predict yield prior to flowering but the outcome is subject to variable post-anthesis weather that in many environments may be highly variable and difficult to predict, limiting its usefulness.