Characterizing leafless inflorescence structure for thinning strategies in ‘Yu Her Pau’ litchi using factor analysis and K-means clustering

• Statistical analysis is applied to explore litchi inflorescence thinning. • Two functional zones identified using factor analysis and K-means clustering. • Basal 40–50 % of inflorescences contains > 70 % of female flowers and fruit. • Basal 70–75 % of inflorescences present > 90 % of female flowers and fruit. • Results provide a basis for adjusting thinning to meet production goals. Inflorescence thinning is a common fruiting management practice in litchi ( Litchi chinensis Sonn.), particularly for cultivars with large, complex inflorescences such as ‘Yu Her Pau.’ This practice reduces competition for carbon assimilates and enhances the proportion of female flowers and fruit retention; however, its effectiveness varies among years and cultivation practices and previous studies have largely relied on single-variable assessments. This study aims to develop a novel, data-driven and quantitative approach to inflorescence thinning in litchi by analyzing paracladium node-level floral distribution patterns and fruit retention dynamics along the inflorescence axis. In 2013, leafless inflorescences were labeled in a central Taiwan litchi orchard, comprising a total of 107 paracladium nodes. Data were collected on axis and paracladium length, flower count, gender distribution, and fruit retention for each node. Factor analysis (FA) and K-means clustering were sequentially applied to integrate multivariate node-level data. FA reduced the observed paracladium architectural attributes into fewer key factors, which were then used as inputs for K-means clustering to classify inflorescence segments. The labeled inflorescence measured 34.4–41.1 cm in length and contained 34–37 nodes with 4675–7498 flowers. Two rounds of flowering were observed, resulting in 6.0 %–8.9 % female flowers and 2.8 %–6.7 % fruit retention. Two factors explained 78.435 % and 79.953 % of total variance in single- and two-flowering rounds, respectively. Based on FA and K-means clustering, thresholds were identified for the core and flexible reserve zones along the inflorescence axis at 41.4 %–50.2 % and 71.9 %–76.6 % of the axis length from the base. The basal 40 %−50 % of the inflorescence contained at least 70 % of all female flowers and 80 % of retained fruit, whereas the basal 70 %−75 % accounted for >90 % of both. These findings suggest functionally distinct zones for targeted thinning based on structural and reproductive metrics and provide an evidence-based foundation for refining thinning practices, offering potential guidance for both conservative and intensive management strategies depending on production goals and resource allocation.