Bayesian Integration of Temperature Trends and Amplitudes Across China During the Past 10,000 Years: Implications for Regional Sustainability

Long-term temperature reconstructions provide essential context for evaluating regional climate change, but synthesis across China is complicated by uneven spatial coverage, seasonal sensitivity, and proxy-specific biases. Here, we compiled 93 published quantitative temperature records spanning the past 10,000 years and applied a Bayesian multilevel regression framework to reconstruct national and subregional temperature trajectories. Annual mean, summer, and winter temperatures were compared among four climatic regions, proxy groups, and sedimentary archives. Within-stage temperature ranges were assessed using both primary and robust-extrema estimators. National annual mean and summer integrations indicate generally warmer early–middle Holocene conditions followed by late-Holocene cooling, whereas winter results remain exploratory because only five records are available. Regional trajectories were heterogeneous, with stronger cooling in northern monsoon China, larger within-stage proxy-derived ranges in the northern non-monsoon region, weaker changes in southern monsoon China, and pronounced seasonal differences on the Tibetan Plateau. Large branched glycerol dialkyl glycerol tetraethers (brGDGTs) and loess amplitudes were substantially reduced by robust estimation, suggesting sensitivity to isolated extremes and archive-specific processes rather than direct atmospheric temperature changes. These results demonstrate that regional, seasonal, proxy, and archive-specific baselines are essential for interpreting paleoclimate evidence and its implications for ecosystem and resource management.