Prediction of Sugarcane Sucrose Content with High Resolution Hyperspectral Leaf Reflectance Measurements.

Experiments were conducted to determine if leaf reflectance measurements could be used to predict theoretically recoverable sugar (TRS) levels in sugarcane prior to harvest. Leaf and stalk samples were collected from multi-variety first-ratoon (FR) sugarcane maturity studies in 2005 at three sample dates and from the plant-cane (PC) and FR sugarcane throughout the 2006 harvest season. Reflectance measurements were taken from leaves of each sample, using a dual input, fiber optic spectrometer. Sugarcane juice was extracted from stalk samples and analyzed for its Brix, sucrose, and fiber contents to predict TRS.  Discriminant analysis of the 2005 FR maturity studies showed that leaf reflectance was effective at predicting TRS in 79% and 54% of the cases if varieties were combined using resubstitution and cross validation techniques, respectively.  If the varieties were considered separately, then 100% of the cases could be classified correctly using both resubstitution and cross validation. In the 2006 FR and PC maturity studies when varieties were combined, TRS levels were predicted correctly in 59 and 76% of the cases using resubstitution and in 37 and 36% using cross validation. For individual varieties, TRS was predicted in 100% of the cases in both PC and FR tests using re-substitution techniques and from 65-92% of the cases for FR and in 60-93% of the cases in PC using cross-validation. Results from stepwise regression analyses between leaf reflectance values and TRS indicated that simple models could be developed that described much of the variability present in stalk sucrose levels with the best results obtained for individual varieties. Several spectral regions were identified as important in describing stalk sucrose levels, including regions of the ultraviolet (250-330 nm), blue, green and yellow (450-590 nm), orange and red (590-650 nm), and the near-infrared (740-850 nm).  These combined results indicate that it may be possible to utilize remote sensing techniques to estimate sugarcane maturity (TRS) prior to harvest. This would allow producers to more effectively manage harvest schedules to insure maximum sucrose yields.