Is ENSO Changing? Assimilation of Paleoclimate Data Into the Cane-Zebiak Forecast Model

There is a quotation from Niiler on the wall in Room 105 of the Oceanography Building at Lamont-Doherty that reads “Modelers want to do modeling, and observationalists want to make observations”, but Mark always pushed the use of both models and observations to learn about the real world. In this project, we set out to assimilate paleoclimatic observations into the Cane-Zebiak (CZ) intermediate complexity ENSO forecast model, to study if, why and how ENSO changes as greenhouse forcing continues to build. This model might be too simple by modern, GCM-based standards but remains as skillful as state of the art models for ENSO forecasting. We present results and analysis for offline and online assimilation products and error estimates, based on two paleoclimatic datasets: (1) the tree-ring-based reconstruction of NINO SST indices (Cook and Cane, 2024) employed as local observations ; (2) an aggregation of CZ-domain and global coral-based marine paleoclimatic observations modeled as CZ prognostic variables. Initial findings suggest that if there is a greenhouse forced response, it is overshadowed by the substantial unforced variability, a finding consistent with that from Ch 4 of IPCC AR6 (nicely communicated here) and from AOGCM simulations most skillful for multivariate present day ENSO simulation. Two unintended but consequential findings of our work are that age model errors in high-resolution coral paleoclimate observations can be significant, and observational duplicates may occur in mass aggregations of such records. Both issues could produce spurious results with inflated skill scores.