Dicroidium is an extinct genus of fork-leaved seed plants. It is the archetypal genus of the corystosperms, an extinct group of seed plants, often called "seed ferns", assigned to the order Corystospermales or Umkomasiales. Species of Dicroidium, which grew as large trees, were widely distributed and dominant over Gondwana during the Triassic (). Their fossils are known from South Africa, the Arabian Peninsula, Australia, New Zealand, South America, Madagascar, the Indian subcontinent and Antarctica.
Dicroidium is an extinct genus of fork-leaved seed plants. It is the archetypal genus of the corystosperms, an extinct group of seed plants, often called "seed ferns", assigned to the order Corystospermales or Umkomasiales. Species of Dicroidium, which grew as large trees, were widely distributed and dominant over Gondwana during the Triassic (). Their fossils are known from South Africa, the Arabian Peninsula, Australia, New Zealand, South America, Madagascar, the Indian subcontinent and Antarctica.
== Description == left|thumb|Dicroidium odontopteroides fossil leaf, Late Triassic Molteno Formation near Birds River South Africa. Within the form genus classification system used in paleobotany, the genus Dicroidium refers specifically to the leaves. Some authors have suggested dividing Dicroidium up into several genera, including Dicroidiopsis, Diplasiophyllum, Zuberia, Xylopteris, Johnstonia and Tetraptilon, but this is rejected by other authors. The leaves of Dicroidium bifurcate (fork) at their base, which is characteristic of all species. The leaves are highly variable in size and morphology, ranging from simple to tripinnate, with the individual leaflets having varying morphologies, including dissected, lobed, needle-like and entire. Some leaf specimens have more than one type of leaflet morphology, which may have been the result of hybridisation between different species. The venation of the leaves is also highly variable, encompassing taeniopteroid, odontopteroid, alethopteroid and simple morphologies.
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).