Watermelon (Citrullus lanatus (Thunb.) Mansf.) breeding lines were subjected to 26 standardised distinctness, uniformity and stability (DUS) morphological descriptors to estimate the genetic diversity among 165 advanced breeding lines (F4 and BC1F3). Descriptive statistics quantified trait variability across vegetative, floral, fruit and seed characters. The breeding lines were classified through hierarchical agglomerative clustering using Ward's method and squared Euclidean grouped into 13 clusters, revealing broad phenotypic divergence of trait combinations. Clusters of particular breeding interest included dwarf or compact genotypes with white or orange flesh, groups combining high seediness with orange flesh and compact early-growing lines or traits with direct value for parent selection. Both qualitative descriptors (flesh colour, rind pattern, sex expression) and quantitative measurements (fruit size, leaf dimensions, seed number) contributed to cluster discrimination. Statistical analyses and cluster validation were performed in SPSS and cross-validated using R (principal component analysis (PCA) and dendrogram visualisation). Cluster means, trait-wise patterns and identified outlier genotypes inform strategic parent choice for introgression, candidate lines for DUS-based varietal registration and the conversion of rare phenotypes. These results provide an actionable framework to broaden the genetic base of watermelon breeding programmes and to accelerate development of DUS-compliant cultivars with improved fruit quality and resilience.
