Large ensemble particle filter for proxy-based spatial reconstructions of the last 2000-years climate variability

Proxy records (corals, marine sediments, etc.) documenting the last 2000 years (2K) provide evidences for the wide range of natural variability not captured by recent direct observations. Assessing climate models ability to reproduce such natural variations is crucial to understand climate sensitivity and impacts of future climate change. The length of record is relatively short for investigating slow climate features, especially when considering coupled ocean-atmosphere variability. In order to extend the information contained in proxies from the locations and times to which they pertain, additional information is needed to create a climate field reconstruction. Paleoclimate data assimilation offers a powerful way to extend the instrumental period and better characterize the decadal to secular natural ocean variability by optimally combining the physics described by climate models with information from available observations while taking into account their uncertainties. Here we present a new Proxy Data Assimilation product based on a sequential importance resampling particle filter (PF-SIR) that uses Linear Inverse Modeling as an emulator of GCMs, providing dynamical ocean memory and improving the reconstruction of low-frequency climate variability. The climate reconstructions include robust uncertainty quantification and a set of physically consistent spatial fields useful for dynamical inquiry beyond what is feasible from proxies or climate models alone. We use these new results to explore low-frequency aspects of main coupled variability modes and provide some constrains on climate model simulations for the last millennium.