Machine Learning unravels the protracted role of India-Eurasia collision in the uplift of the Tibetan plateau

The Tibetan Plateau, Earth's largest and highest plateau, boasts an extraordinarily thick continental crust (60-80 kilometers) and an average elevation exceeding 4000 meters. Unraveling the plateau's uplift history, vital for comprehending Earth's Cenozoic history and its environmental impacts, has long been a subject of debate. While prior studies predominantly attribute the plateau's formation to the India-Asia collision, 45-59 million years ago, its timing and underlying mechanisms remain contentious. Airy isostasy as a response to crustal thickening during the Indian-Asian collision was considered the main factor for the uplift of the Gangdese terrain, the important portion of the Tibetan. Trace elemental ratios, e.g. Sr/Y and (La/Yb)n ratios, of the bulk magmatic rocks were the main geochemical indexes to recover the thickening history. However, the resultant crustal thickness and the consequent geodynamics recovered by different indexes remain controversial. Here, we compile the geochemical data for the volcanic rocks from global young arcs and continental orogens and built a supervised Machine Learning model to estimate crustal thickness. The reliability of this new model was tested, and the crustal thickening history of Gangdese terrain was recovered with it. The results reveal that the Gangdese terrane maintained a global-average thickness during the early stage of the India-Asia collision, which was not sufficient to support the uplift to >3000 m, as revealed by the recent paleoaltimeter data, through Airy isostasy.  This challenges the conventional belief of rapid uplift due to crustal thickening upon the Indian-Asian collision. Instead, our results suggest a protracted uplift process that parallels crustal thickening, reshaping our understanding of this iconic geological feature.