Using Red and Red-Edge Reflectance-Based Vegetation Indices for Stalk Yield Prediction of Energy Cane at Different Harvest Dates.

High biomass production and narrow window of time for sensing are among the challenging aspects of adopting optical remote sensing technology in energy cane production. The saturation problem associated with canopy closure limits its application and use at the later growth stage of energy cane. A field experiment was initiated in 2012 at the LSU AgCenter Sugar Research Station in St. Gabriel, Louisiana to evaluate the feasibility of using spectral reflectance-based vegetation indices to estimate energy cane yield. The experiment with a 2x4 factorial treatment structure was arranged in split plot in a randomized block design with four replications. The treatments included two energy cane varieties (Ho 02-113 and US 72-114) and four N rates (0, 56, 112, and 224 kg N ha^-1). Canopy reflectance readings were collected using Jaz® hyperspectral spectrometer (300 to 1100 nm) from a 1.0 m² area. Nine spots per plot were flagged; reflectance readings were taken from each spot from 3 weeks after N (WAN) application until 16 WAN. Fifteen stalks were collected once a month beginning two months prior to the harvest date to determine the plot yield. Vegetation indices (VIs) computed from reflectance readings at wavebands 670 nm (red) and 705 nm (red-edge) consistently produced linear relationships with stalks across harvest dates. Simple ratio (SR) and normalized difference vegetation index (NDVI) for both red and red-edge obtained better correlation with stalk yield at 6 WAN application (plant cane and second ratoon) and 14 WAN (first ratoon). When pooled across harvest dates, SR and NDVI obtained a coefficient of determination (r²) with millable stalk ranging from 0.26 to 0.36. Future work should focus in building and improving the database using data from a wide array of growing conditions.