Instrumental observations have highlighted strong interconnections within the Antarctic climate system between atmosphere, ocean and sea ice over the past few decades. Our goal in this study is to leverage these links by combining marine and terrestrial records with climate modeling results, using data assimilation, to reconstruct past climate changes in West Antarctica and the surrounding seas. Through comparing two data assimilation experiments that use the same terrestrial proxy records with annual and decadal temporal resolutions, we confirm that the covariance relationship between several key climate variables (surface air temperature (SAT), sea ice, and atmospheric circulation), used in the data assimilation is time scale dependent. Our experiments show that a data assimilation step of 10 years appears to be a good compromise. Additionally, we clarify the contributions of continental and marine records to the reconstructions through two separate data assimilation experiments: one using the continental proxies only and another using both continental and marine records. Our results suggest that the latter is more useful, providing a stronger and more direct constraint on the evolution of the ocean-atmosphere coupled system, and is more reliable, as suggested by the validation against independent sea ice-related marine records. Based on our reconstruction for multiple variables covering the past two millennia, we confirmed a widespread cooling trend occurred over West Antarctica and surrounding seas over the period 900–1800 CE. This reconciles the inconsistency between some existing reconstructions and climate model results, in particular over the Antarctic Peninsula. This cooling could be explained by a direct thermodynamic response to external forcing, with a dominant contribution of the volcanic forcing over the period. Our results thus highlight that West Antarctica and surrounding seas are sensitive to external forcings at this time scale and that the temperature evolution there is coherent with global changes at the centennial scale over the past millennium.< Réduire