Using hand-held infrared spectroscopy to guide harvest decisions

The use of hand-held NIR spectroscopy for in-field DM measurements to guide harvest decisions requires knowledge of calibration methodology, model accuracies, the relevance of various maturity indices and sampling strategies. Typical calibration performance for NIR DM estimations included R2CV>0.85 and RMSECV<0.96 on individual cultivar models, and R2=0.86 RMSEP = 0.92 on a global (all cultivar) model predicting an external validation set. Fruit DM increased relatively linearly (at approx. 0.07-0.15% DM each day) depending on growing conditions, such that DM at harvest maturity, as judged by fruit shape or flesh colour, varied by growing condition, but nonetheless offered a tool for forward estimation of harvest date and relative maturity of blocks on a farm. For NIR-DM surveys, a protocol is recommended involving sampling fruit from various positions on the tree canopy, and across the orchard block, to a sample number determined by the SD of DM and a desired estimate error. Geospatial data referenced to NIR-DM readings has potential for determining “hot/cold spots”. A recommendation is made for an initial stratified sampling across each block for estimation of block average NIR-DM, followed by measures of 15 tagged fruit per block on at least three occasions separated by intervals of at least a week to determine a daily DM increase. This rate of increase can be applied to the initial survey measurement of block average DM to determine harvest maturation date. © 2019 International Society for Horticultural Science. All rights reserved.