Fig. 4: Summary of causal associations of the climate-proxy records as revealed by convergent cross-mapping (CCM) and time displacements of the ‘CCM skills’.
From: Atmospheric CO2 forcing on Mediterranean biomes during the past 500 kyrs
Precipitation and forest cover at Tenaghi Philippon based on the log(Ca/Fe) and tree-pollen percentage datasets, respectively; orbital parameters from ref. 34 atmospheric CH4 (ref. 31) and CO2 (ref. 13) concentrations from Antarctic ice cores; position of the Intertropical Convergence Zone (ITCZ) and dust influx based on the Ba/Al and Ti/Al ratios from ODP Site 967, respectively (ref. 33); North Atlantic surface-water temperature based on Neogloboquadrina pachyderma ‘sinistral’ percentages from ODP Site 983 (ref. 32). Significance of causal associations was determined when ‘CCM skills’ of the proxy timeseries were outside the 5th and 95th percentiles computed from the randomly generated timeseries and the area under curve (AUC) value was higher than 0.8 (Supplementary Fig. 4; Supplementary Table 2). Bold arrows indicate significant causality (AUC > 0.9), whereas dashed arrows indicate weak causality (0.9>AUC > 0.8); crossed-out variables have no causal associations with the datasets from Tenaghi Philippon.
