Constraining Atmosphere-Ocean Variability over the Last Millennium

Coupled interactions between oceans and the atmosphere are fundamental to low-frequency variability of the Earth System. While the instrumental record provides an account of this coupled variability over time, the length of record often hinders investigation of the mechanisms of variability on decadal and longer timescales. Recent reconstruction studies using data assimilation provide an objective framework for estimating coupled fields over longer periods, but they often omit dynamical constraints from forecasts between analysis times, affecting their representation of ocean memory and dynamics. Linear inverse models (LIMs) offer an efficient method to incorporate temporal dynamics with forecasts based on coupled global climate models (GCMs). In this work, we present annually-resolved reconstructions during 1000 - 2000 C.E. using the Last Millennium Reanalysis framework, which successfully incorporates online data assimilation (DA) via multivariate LIM forecasts. The introduction of dynamical memory into the climate reconstruction significantly improves the consistency of upper-ocean heat content estimates and its relationship with sea-surface temperature (SST). Additionally, the inclusion of LIM forecasts provides a base dynamical consistency with GCMs when proxy information is limited. We validate reconstruction results against instrumental products, finding broadly positive spatial correlations for SST and high correlations (~0.6 - 0.9) for ocean indices. Compared to previous reconstructions, online reconstructions show cooler large-scale average temperatures, especially over 1000 - 1200 C.E. (~0.25 - 0.5 K cooler), and no clear millennium-scale cooling trend. However, regional estimates of temperature change between early (1000-1250 C.E.) and later (1400 - 1700 C.E.) periods suggest warmer temperatures in high-latitude Europe and colder tropical temperatures during the early period.