Prediction of foliar nutrient concentrations and yield of avocado using on-farm and laboratory based hyperspectral imaging

Abstract

Purpose Current methods to assess tree nutrition and yield are time-consuming. This study explored hyperspectral imaging (HSI) as a rapid method for predicting foliar nutrient concentrations and yield at four developmental stages of avocado cropping. Methods Hyperspectral images were captured of the tree canopy, and single leaves on and off the tree, in the orchard using a handheld camera. The same leaves were re-imaged in a laboratory using both handheld and benchtop cameras. Partial least squares regression (PLSR) models were developed from each image type to predict foliar nutrient concentrations using pooled datasets of all four sampling occasions. PLSR models were also developed to predict yield using the single dataset from each sampling occasion as well as the pooled dataset. Results Concentrations of nitrogen (R 2 Test = 0.68; RMSE = 0.24%), phosphorus (R 2 Test = 0.66; RMSE = 0.01%) and potassium (R 2 Test = 0.65; RMSE = 0.12%) were most accurately predicted using images captured off the tree in the orchard using a handheld camera. Calcium concentration (R 2 Test = 0.78; RMSE = 0.29%) was most accurately predicted using images captured by a benchtop camera. Yield was predicted using images of the canopy as early as the fruitlet stage at 6 months before harvest (R 2 Test = 0.85; RMSE = 24.9 kg tree −1 ) and at later developmental stages (R 2 Test = 0.86 and 0.88; RMSE = 10.7 and 10.6 kg tree −1 ). Conclusion HSI provided an early and rapid method to assess crop nutrition and yield, which will help to make precise fertiliser applications and to arrange harvest and postharvest resources earlier in the cropping cycle.

Bibliographic Information