Using canopy spectral reflectance to estimate nitrogen use traits in hard winter wheat

Wheat nitrogen use efficiency must be improved to reduce the need for nitrogen (N) fertilizers. This study was conducted to determine if measurement of canopy spectral reflectance (CSR) could be used to non-destructively and indirectly select wheat genotypes with improved nitrogen use traits. Canopy spectral reflectance measurements were collected during grain fill in a 299 genotype trial planted near Ithaca, NE in 2012 and 2013. The objectives of this study were i) evaluate vegetation indices (VI) to determine the best index for indirectly evaluating nitrogen use (NU) traits in the context of a hard winter wheat breeding program ii) determine the ability of genomic prediction models to accurately predict VI phenotypes. Twenty-eight VI were calculated, and the relationship between VI and NU traits was investigated. Vegetation indices were highly heritable in both years and showed significant relationships with the NU traits anthesis biomass, anthesis N yield, mature biomass, grain N yield, grain yield, N harvest index, N utilization efficiency, N uptake efficiency and post anthesis N uptake. Two VIs, Maccioni and Boochs2, performed most consistently and were significantly related to several nitrogen use traits. The results of this study indicate that VIs in particular, Maccioni and Boochs2, could be used in wheat breeding program to non-destructively phenotype for nitrogen use traits. In order to utilize VI phenotypes across an entire breeding program, we proposed that genomic prediction models could be used to predict VI phenotypes based on genomic marker-phenotype relationships. Prediction of VI phenotypes would allow for indirect improvement of NU traits throughout a breeding program rather than in a single experimental trial with one generation of genotypes. Results of this study showed that VI traits can be used successfully in genomic selection models. Prediction accuracy for VIs was higher than the prediction accuracy of NU traits and grain yield. The Boochs2 index was predicted more accurately than the Maccioni index using both within year and across year cross validation. As a result, we recommend that the Boochs2 index be used in GP models to increase genetic gain for NU traits while reducing the time and labor costs of phenotyping