Model seasonal and proxy spatial biases revealed by assimilated mid-Holocene seasonal temperatures.

Proxy-based reconstructions and climate model simulations of Holocene global annual mean temperatures exhibit divergent trends, leading to the well-known "Holocene temperature conundrum (HTC)". This discrepancy is most pronounced in the mid-to-high latitudes of the Northern Hemisphere (NH) and has been attributed to either proxy seasonal bias or deficiencies in climate models. Paleoclimate data assimilation (PDA), which integrates proxy records with climate model simulations, provides an advanced method for generating global seasonal temperature reanalysis datasets for the mid-Holocene (MH). Assimilated results indicate that MH Eurasian temperatures are largely independent of the choice of model priors and exhibit significant spatial heterogeneity. Compared to the pre-industrial (PI) period, the MH is characterized by winter and annual mean warming in Europe and high-latitude Eurasia, while the rest of Eurasia experiences cooling. However, this spatial heterogeneity is not well represented in model simulations due to a pronounced winter cooling bias at high latitudes, likely resulting from inadequate representations of vegetation and sea ice feedback mechanisms. As Eurasian proxy records are predominantly concentrated in Europe, this regional imbalance introduces a warm bias in reconstructed winter and annual temperatures intended to represent broader Eurasian temperature changes. These results suggest that the HTC may stem from both the uneven spatial distribution of proxy records and the incomplete representation of internal climate feedbacks in current models.