PALEOGENE DESERTIFICATION OF CENTRAL ASIA SEEN THROUGH THE PALEOSOL RECORD: A TALE OF CLUMPED ISOTOPES AND SOIL PALEOBAROMETRY

The onset of central Asian aridification and subsequent desertification has been variously dated from the late Cretaceous to the early Miocene depending on the paleoclimatic proxy (first eolian sediments, paleobotany, or soil carbonates). The aridification and desertification have been alternatively linked to paleogeographic factors (uplift of the Tibetan Plateau, retreat of the Paratethys) or to global climate change (pCO2 decrease and global cooling through the late Eocene and early Oligocene). Here, we use different geochemical proxies applied to several paleosol sequences covering 20 million years in the Xining Basin, NE Tibet, precisely dated from 43 to 23 million years ago by paleomagnetism. The paleosols were sampled at an ~100kyr resolution for proxy analysis including carbon, oxygen, and clumped isotopes of pedogenic carbonates and carbon isotopic compositions of soil organic matter, complementing previously published pollen and sedimentological data. Our multi-proxy dataset allows us to quantitatively reconstruct soil temperature and soil productivity and determine when NE Tibet shifted to a similar-to-present desert ecosystem.

While pCO2 decreases about twofold through the late Eocene and early Oligocene, our results show that soil productivity in NE Tibet decreases by a factor of ten. Soil productivity shows a fast decline starting around 37 Ma and reaching the (low) productivity of modern central Asian deserts at the 34 Ma Eocene-Oligocene transition. Soil temperatures decrease abruptly at the Eocene-Oligocene Transition, but recover after the Oi-1 glaciation, returning to their greenhouse Eocene values. However, soil productivity never recovers, indicating the permanent set-up of desert conditions. Our results thus indicate that if the uplift of the Tibetan Plateau and the retreat of the Paratethys might have paved the way to aridification in NE Tibet, final desertification was triggered by global cooling. Our findings also have important implications for the use of soil paleobarometers, which commonly consider soil productivity as a constant through past intervals of pCO2 change. We show here that soil productivity can be extremely sensitive directly or indirectly to pCO2 changes and suggest that paleo-pCO2 reconstructions based on soil properties are significantly biased by these effects.