Pre-harvest mango yield prediction using artificial neural networks based on leaf nutrient variability

Accurate pre-harvest mango yield prediction provides valuable insights for improving productivity, reducing food waste, enhancing food security, and supporting the farmer livelihoods. In this respect, nine mango orchards, which had different agricultural practices, were selected to attain the essential data to integrate yield and leaf nutrient variability using an artificial neural network model. In the mango leaves, noticeable variations were detected in concentrations of nitrogen (N), phosphorus (P), potassium (K), magnesium (Mg), calcium (Ca), chlorophyll a (Chl a), chlorophyll b (Chl b), and total carbohydrates (Carbs) fraction. The yield variation between seasons is high; the ON season gave a high yield, and the OFF season gave a low yield. The results revealed that at the pre-harvesting time, the relationship between the carbohydrate: nitrogen (C/N) ratio and the yield against the growing season had a reverse trend. The artificial neural network (ANN) mango yield model was created using eight inputs representative of the nutrient status of leaves. The ANN model achieved an accurate match in predicting mango yield from investigated parameters, with an R² value of 0.975 using a testing dataset, and the mean absolute percentage error (MAPE) was 3.02%. The concentration of Chl a, Chl b, and the Carbs fraction had the greatest contribution in predicting mango productivity. It was concluded that the ANN model performed adequately and captured the non-linear effects of the interaction between the nutrition status of the mango leaves and mango productivity.