Abstract
The scarcity of proxies and calibration models for quantitatively reconstructing millennial timescale sea-sonal temperature tremendously constraints our understanding of the Holocene thermal variation and its driven mechanisms.Here,we established two global warm-season temperature models by applying deep learning neural network analysis to the branched tetraether membrane lipids originating from surface soil and lacustrine sediment bacteria.We utilized these optimal models in global well-dated lacustrine,peatland,and loess profiles covering the Holocene.All reconstructions of warm-season temperatures,consistent with climate model simulations,indicate cooling trends since the early Holocene,primarily induced by decreased solar radiation in the Northern Hemisphere due to the precession peak at the early.We further demonstrated that the membrane lipids can effectively enhance the future millennial sea-sonal temperature research,including winter temperatures,without being restricted by geographical location and sedimentary carrier.
基金项目
National Natural Science Foundation of China(42230503)
NETL
NSTL
万方数据