Quantifying structural and proxy composition uncertainty in Common Era climate field reconstructions using online paleo data assimilation

Abstract. Paleodata assimilation combines proxy observations with climate model information to produce spatially complete reconstructions of past climate, but uncertainty is often represented only by posterior ensemble spread within a single reconstruction setup. Here, we generate annually resolved global reconstructions of near surface air temperature, sea surface temperature, precipitation, evaporation and sea level pressure over the 1–2012AD Common Era using an online Sequential Importance Resampling particle filter coupled with Linear Inverse Models trained from CMIP5 and CMIP6 simulations. We quantify uncertainty with an 80 member factorial ensemble varying proxy system model calibration, model prior and bias correction, and a 50 member ensemble testing sensitivity to tree ring and coral proxy inclusion. We distinguish posterior internal uncertainty, structural uncertainty and proxy composition uncertainty. The reconstruction agrees best with instrumental products for large scale temperature and sea surface temperature fields, while hydroclimate and circulation variables show more heterogeneous skill. Structural uncertainty dominates the amplitude of secular global mean temperature and North Atlantic sea surface temperature variability, mainly through proxy system model calibration, whereas prior choice and bias correction become more important for tropical and circulation indices. The uncertainty framework identifies several coherent reconstructed features: robust industrial era global mean warming, a pre industrial warm optimum around 600–700 CE following colder conditions around 400–600 CE, a low frequency North Atlantic cooling toward the late Little Ice Age, and selected tropical Pacific anomalies around 950, 1200 and 1400 CE. Recent tropical Pacific changes include a positive Niño3.4 tendency and a strong negative Pacific Walker circulation excursion near 2000 CE, the latter suggesting a potentially unusual circulation state relative to the reconstructed Common Era background. Tree rings and corals provide complementary constraints, showing that proxy composition affects not only uncertainty width but also the interpretation of reconstructed climate features.