Dragon fruit (Hylocereus spp.) cultivation is significantly influenced by training systems, which modulate canopy architecture, light interception, and resource allocation. Understanding the physiological and agronomic impacts of different training methods is essential for optimizing yield and fruit quality. This study evaluated three training systems, single-pole (T1), pyramid stand (T2), and T-trellis (T3), on white (Hylocereus undatus Britton & Rose) and red (H. costaricensis Britton & Rose) dragon fruit cultivars, assessing their effects on vegetative growth, reproductive performance, and fruit characteristics. The single-pole system (T1) demonstrated superior performance, promoting balanced growth and higher yields (61 t/ha in T1-white) due to improved light distribution and floral initiation (26 days to flowering). In contrast, the T-trellis system (T3) enhanced vegetative expansion (194 cm plant spread in T3-white) but incurred physiological stress, including elevated canopy temperatures (43 ◦C) and photo damage (yellowing intensity up to 3.88), reducing bud retention and yield (28 t/ha in T3-white). White cultivars exhibited greater adaptability across systems, while red cultivars performed optimally only in T1, achieving superior pulp yield (348 g) and sweetness (16 ◦Brix). The pyramid system (T2) showed intermediate results, balancing moderate yield and manageable canopy structure. The single-pole system (T1) is the most effective for commercial dragon fruit production, maximizing yield and fruit quality, while minimizing stressrelated losses. Cultivar-specific responses highlight the need for tailored training approaches, with white types being more versatile and red types requiring precise management. Future research should explore long-term system viability, climate resilience, and economic feasibility to refine cultivation practices for sustainable production. These findings provide actionable insights for growers to enhance productivity and fruit marketability in diverse agro ecological conditions.
