During a sabbatical leave in 1990-91 in order to teach Glaciology in two universities of Southern China, the Kunming Polytechnical School (Yunnan) and the Geological Institute of Cheng du (Sichuan), the author had the opportunity during the Winter vacations and later in May, to study an underground river system flowing in the southern part of Yunnan, in the region of Jianshui, 150km south of Kunming, the capital of the province.In that area, large and thick outcrops of carbonate rocks occur, their ages ranging from Early Paleozoïc to Middle Mesozoïc. The Himalayan orogeny produced (and still produces) a strong uplift of the whole area, accompanied by intense rock deformations. Furthermore, that part of the country lying under the Cancer tropic, it means that large amounts of rain occure mainly in Summer (monsoon). These two factors, a warm and humid climate, plus limestone weakened by fractures, implied an intense karstification that began during the Paleozoïc and which still goes on presently. It leads to a deepening of the valleys, and to rivers frequently « tunneling » through the relief. When the uptift is momentarily slowing down, lateral erosion takes place, leading to widening of the valleys. The vertical and/or lateral erosion gives birth to what is called a « qiufeng » landscape.In 1991, during two short expeditions, we mapped a certain number of tunnel-caves, i.e. underground rivers leaving one cave to enter another one, both being separated by a « tiankeng », a kind of giant window, like a polje with very steep slopes, that phenomenon occuring several times on the same river.On the other hand, rock samples have been brought back to the University of Geneva. Their thin sections show that these caves have developped within Triassic carbonates, and not Permian ones, as shown on local geological maps.