International Journal of Environment and Climate Change

Understanding how weather conditions influence mango productivity at different developmental stages is essential for improving yield estimation across contrasting growing environments. This study investigated phenophase-specific weather–yield relationships and evaluated the integration of meteorological and satellite-derived variables for predicting Alphonso mango (Mangifera indica L.) yield across three major production regions of India. Field observations were conducted during the 2025-2026 growing season in 54 orchards distributed across Muthalamada, Kerala (15 orchards), Dharwad, Karnataka (17 orchards), and Konkan, Maharashtra (22 orchards), comprising 109 flowering cohort observations. Meteorological information from ERA5 Land was combined with vegetation, moisture and thermal indicators derived from Sentinel-2 and MODIS satellite observations. Pearson correlation analysis, with Benjamini–Hochberg false discovery rate adjustment, was used to examine relationships between six meteorological variables during six reproductive and fruit development stages and orchard yield. Six regression algorithms were evaluated across seven phenological feature sets using nested orchard-grouped cross-validation and independent testing. Weather–yield relationships varied considerably among regions and developmental stages. Inflorescence emergence showed prominent associations in Muthalamada, while flowering and early fruit development were particularly relevant in Dharwad. In Konkan, solar radiation during the marble-size fruit stage exhibited the strongest positive association with yield (r = 0.770), whereas relative humidity during inflorescence emergence showed the strongest negative association (r = −0.686). Among the 42 feature set model combinations, lasso regression using flowering stage predictors (Feature Set 3) achieved the highest predictive accuracy, with a cross-validation R² of 0.921 and RMSE of 0.334 t ha⁻¹. Independent testing yielded an R² of 0.890 and RMSE of 0.323 t ha⁻¹, with positive predictive performance across all three regions. The findings highlight the importance of developmental timing in weather-yield relationships and demonstrate the potential of integrating phenological, meteorological and remote-sensing information for Alphonso mango yield estimation.