Fig. 3: Hypothesis testing of the forcing mechanisms driving the δ¹⁸O_ice long-term trend during the Holocene.

a Time series of seasonal mean solar insolation at 9°S (W m-2). b Seasonal cycle of solar insolation at 9°S (Huascarán latitude) for 0 ka (dashed black) and 9 ka (dashed red), and at 20°S (Sajama latitude) for 0 ka (solid black) and 9 ka (solid red). c Time series of proxy δ¹⁸O_ice records (‰) from Huascarán (black: left axis), Sajama (red: left axis), and δD_ice records from Illimani⁵⁹ (gray: right axis). d Time series of model annual mean δ¹⁸O_v at the ice core altitude of Huascarán (black) and Sajama (red). Solid lines are for the arithmetic annual mean (\({\delta }^{18}{O}_{v}^{{uw}}\)) and dashed lines are the annual mean weighted by monthly precipitation (\({\delta }^{18}{O}_{v}^{{pw}}\)). e Correlation coefficient between δ¹⁸O_ice and 9°S solar insolation for each month throughout the Holocene at each ice core site. f Correlation coefficient between model arithmetic annual mean δ¹⁸O_v and 9°S solar insolation for each month throughout the Holocene. Note that time series in (a), (c), and (d) are shown as anomalies from the climatological mean. g Seasonality of precipitation at Huascarán derived from PMIP4-CMIP6 simulations (including the iTRACE) for the mid-Holocene (red) and PI (black). Solid lines are the results of multi-model ensembles, and dashed lines are for each model. h Annual precipitation differences (PI minus mid-Holocene) over Huascarán and upstream Amazon region derived from PMIP4-CMIP6 simulations (including the iTRACE). Each dot represents a single model.