Artificial intelligence-based classification of flower colour polymorphism for conservation and genetic resource management of <i>Tecomella undulata</i> (Sm) Seem

Abstract Flower colour is a key trait shaping pollination, reproduction and plant–environment interactions. In arid ecosystems, it may also signal adaptations to heat and (Ultraviolet) UV stress. Tecomella undulata , a threatened keystone tree of the Indian Desert, exhibits striking flower colour polymorphism with yellow, orange and red morphs. This study tested whether artificial intelligence (AI) can reliably classify these morphs, thereby supporting conservation efforts. Field surveys were conducted across natural populations in the Thar Desert. An accessible no-code AI platform (Google Teachable Machine) was used for supervised classification of flower and tree images, with unsupervised clustering applied for validation. The AI classifier achieved high accuracy in distinguishing morphs at both flower and tree scales. Morphs showed consistent separation, with orange functioning as an intermediate form. Despite red morphs being more frequent, the presence of yellow and orange morphs contributes essential functional diversity important for pollinator interactions and reproductive resilience. This study demonstrates that no-code AI provides an effective, scalable approach to documenting intraspecific variation in threatened species. By enabling rapid and reliable identification of flower colour morphs, the approach offers practical applications for ex situ conservation, restoration and morph-aware biodiversity management in T. undulata and other arid-zone trees.