A global analysis of pollen-based reconstructions of land climate changes during Dansgaard–Oeschger events

Abstract. Dansgaard–Oeschger (D–O) warming events are comparable in magnitude and rate to the anticipated 21st century warming. As such, they provide a good target for evaluation of the ability of state-of-the-art climate models to simulate rapid climate changes. Despite the wealth of qualitative information about climate changes during the D–O events, there has been no attempt to date to make quantitative reconstructions globally. Here we use frequency-corrected Tolerance-weighted Weighted Averaging Partial Least Squares regression (fxTWA-PLS) to reconstruct mean temperature of the coldest month, mean temperature of the warmest month, and a plant-available moisture index across multiple D–O events between 50 and 30 ka based on available pollen records across the globe. The reconstruction of plant-available moisture is corrected for the impact of changing atmospheric CO2 concentrations on plant water use efficiency. These reconstructions show that the largest warming occurred in northern extratropics, especially Eurasia, while western North America and the southern extratropics were characterised by cooling. The change in winter temperature was significantly larger than the change in summer temperature in the northern extratropics and the tropics, indicating that the D–O warming events were characterised by reduced seasonality, but there was no significant difference between the summer and winter temperature changes in the southern extratropics. The antiphasing between northern and southern extratropical changes, and the west-east pattern of cooling and warming in North America were generally consistent across the eight D–O events examined, although coherency is greatest during the strongest events. There was no globally consistent pattern between changes in moisture and changes in temperature. These reconstructions can be used to evaluate the spatial patterns of changes in temperature and moisture in the transient simulations of the D–O events planned as part of the Palaeoclimate Modelling Intercomparison Project.