Apple Flowering Response to Climate Variability along the Himalayan Elevation Gradient

Apple flowering in the Himalayan region depends on winter chill and spring heat, which are changing under a warming climate. These changes have increased uncertainty in flowering intensity and timing across different elevations. In this study, high-resolution UAV imagery and a YOLOv8-based segmentation model were utilized to map tree-level flowering intensity across three apple orchards situated along an elevation gradient in the northwestern Himalayas. The YOLO model was found to reliably detect flower clusters and showed strong agreement with manual counts, with an R² value of 0.85. This allowed consistent comparison of flowering intensity across sites. The winter chill was estimated using the Dynamic Model, expressed as chill portions derived from ERA5 Land hourly temperature data. Spring heat accumulation was quantified using growing degree days. Flowering varied clearly with elevation. Mid-hill orchards bloomed earlier and showed lower visible flowering during UAV surveys. Higher-elevation orchards bloomed later and exhibited higher flowering intensity. The winter chill was sufficient at all sites. Flowering responses were mainly controlled by the combined effects of chill and spring heat. The results demonstrate that integrating UAV-based deep learning with climate indices provides a practical framework to assess climate-driven changes in apple phenology in mountain environments. This approach can support climate risk assessment and adaptive orchard management in the face of continued warming.