Predicting Alfalfa Nutritive Value with Canopy Reflectance and Environmental Factors.

In-field estimations of alfalfa nutritive value can inform management decisions to optimize forage quality and production. However, acquisition of timely information at the field scale is limited using traditional measurements such as destructive sampling and assessment of plant maturity. In other crops, remote sensing technologies (e.g. measurement of canopy reflectance) have enabled rapid measurements of crop health and other agronomic parameters at the field scale, however, efficient and accurate remote sensing tools for alfalfa management have not been developed. A study was initiated in 2014 at the University of Minnesota Rosemount Research and Outreach Center to determine the viability of published spectral vegetative indices, and develop new indices, to remotely predict alfalfa maturity and forage nutritive value. Canopy reflectance (350‐2500 nm) was measured in conjunction with destructive sampling of alfalfa on 3 to 4 day intervals throughout the growth of a stand. The same parameters were measured in 2015 following a series of offset cutting intervals in order to sample a wide range of maturity under consistent conditions. The full range of reflectance data was processed using stepwise regression, as well as variable selection using the Akaike Information Criterion to identify individual wavebands most correlated with alfalfa nutritive value. Eight wavebands were selected to develop linear models for predicting CP (R² = 0.88) and Neutral Detergent Fiber Digestibility (NDFD, 48 hour in‐vitro)) (R² = 0.84). Cumulative Growing Degree Units since last harvest (GDU_base) was used as a covariate for improved model fit, and enabled the inclusion of fewer wavebands while improving predictability. Using three wavebands and GDU_base as the model inputs, strong predictions of CP (R²= 0.91) and NDFD (R² = 0.89) were maintained. Cross‐validation was performed applying the same model (trained on 2015 data) to the 2014 alfalfa CP (R² = 0.87). These results identify new remote sensing tools capable of providing rapid and accurate predictions of forage nutritive value at the field scale for timely and precise cutting management.