Diagnosing nutrient deficiencies in Kinnow using a yield-quality optimization approach and machine learning

While Diagnosis and Recommendation Integrated System (DRIS) and Compositional Nutrient Diagnosis (CND), have been widely used for diagnosing nutrient imbalances, they often overlook fruit quality, which is equally vital for market success. To address this limitation, we proposed a novel, more precise methodology that integrates both yield and quality into diagnostic frameworks. Leaf and soil samples from 351 Kinnow orchards in Abohar, Punjab (India), were analyzed during 2017–2019 and 2022. Populations were classified based on aggregate score of yield and fruit quality attributes, derived through ‘Analytical Hierarchy Process’. Nutrient norms were refined using Mahalanobis distance and cate-nelson analysis. Results revealed severe deficiencies of Mg (98.3%), Mn (98.3%), P (82.8%), S (79.3%), and Zn (75.9%), along with moderate deficiencies in N (51.7%) and Ca (56.9%). Crucially, the modified CND-clr (centred log ratio) approach outperformed traditional DRIS/CND (RMSEP: 21.45 vs. 13.44), demonstrating enhanced diagnostic precision. The production constraints related to soil were further identified using a random forest algorithm, demonstrated high soil pH, EC, and CaCO₃ as key limiting factors. This framework offers a more precise approach to nutrient management in global fruit cultivation. A novel multicriteria approach integrating yield and fruit quality improved nutrient diagnostic efficacy in Kinnow.Modified CND-clr outperformed traditional DRIS and CND methods, offering more precise nutrient thresholds for dual yield-quality optimization.Machine learning revealed high soil pH, EC, CaCO₃, and widespread S, P, Mg, and Zn deficiencies as key constraints limiting Kinnow productivity in South-Western Punjab. A novel multicriteria approach integrating yield and fruit quality improved nutrient diagnostic efficacy in Kinnow. Modified CND-clr outperformed traditional DRIS and CND methods, offering more precise nutrient thresholds for dual yield-quality optimization. Machine learning revealed high soil pH, EC, CaCO₃, and widespread S, P, Mg, and Zn deficiencies as key constraints limiting Kinnow productivity in South-Western Punjab.