Answers are grounded in this paper and cite their sources.
Apple flowering strongly influences fruit set and yield in temperate orchards. However, measuring spatial variability in bloom intensity across heterogeneous orchards remains difficult. In the northwestern Himalayan region of India, changing winter chill and spring temperatures are altering flowering patterns. This makes reliable monitoring of flowering increasingly important for orchard management. In this study, we integrate UAV imagery, deep learning, and climate-based phenological metrics to quantify flowering variability across three apple orchards located at different elevations in Himachal Pradesh, India. UAV images collected during the flowering period were analysed using a YOLOv8 instance-segmentation model to detect blossoms. On an independent test dataset, the model achieved an F1-score of 73.6% and mAP@0.5 of 76.3%. Tree-level flower counts derived from the model showed strong agreement with manual counts from 134 trees (R² = 0.88). This confirms that the workflow reliably estimates flowering intensity. To further evaluate segmentation robustness, a Mask R-CNN model was implemented as a comparative approach. The model achieved a precision of 0.6288, recall of 0.8902, F1-score of 0.7370, and mean IoU of 0.7596, indicating higher detection sensitivity but increased false positives under dense canopy conditions. Climatic analysis using chill portions (CP) and growing degree days (GDD) revealed clear phenological differences among the orchards. The higher-elevation site was captured at full bloom during UAV acquisition. In contrast, lower-elevation orchards had already entered the petal-fall stage. As a result, fewer flowers were visible at those sites. These findings show that combining UAV-based blossom detection with chill and heat metrics helps interpret spatial flowering patterns. The comparative evaluation further highlights how model-specific detection behaviour can influence flowering estimates derived from UAV imagery. The approach can support climate-adaptive orchard management and improve phenology monitoring in mountain fruit systems.
No takes yet. Share an insight, caveat, or question.
Shukla et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: