On the Need to Use Multi-scale Paleoclimate Proxy Targets in Paleoclimate Assimilation in Order to Understand and Constrain Projections of Future Climate and Biogeochemistry

Abstract:

Earth system models (ESMs) can be used for forecasting of global climate and biogeochemical cycling. Still, ESMs are usually developed using present and recent historical climates as references, while climate projections indicate that future climates will lie outside these conditions. The longer the climate forecast is, the lower the influence of (reasonable) initial conditions and higher the influence of parameters is, as the model converges to its attractor. Past biogeochemical observations (paleoclimate proxies) can help to constrain parameters in ESMs and so to narrow the uncertainty in climate projections. A way to optimally merge the information in paleoclimate proxies with ESMs is data assimilation (DA), where uncertainty in both ESM parameters and observations is taken into account. Although current ESMs are deterministic, the assimilation is a statistical process, where one needs to consider how to best use the observations.

Let us consider an ESM with given physics and a prior parameter set calibrated with present climate targets. On one hand, local climate and biogeochemical anomalies in a given past climate time window (~30 years) may well be teleconnected with anomalies thousands of kilometres away. This teleconnection may be detected by both, observations and model. On the other hand, both the local proxy and the model may be biased. Also, for ensemble DA, a reduced ensemble size (due to High Performance Computing limits) may result in spurious correlations in the ESM. This advocates for the need to a) use global, and as comprehensive as possible, paleoclimate proxy dataset for crossed Quality Control (QC), and b) include both point observations as well as regional and large scale targets in the assimilation process. By using the Community Earth System Model (CESM), we show an example of a) how assimilation of tropical coral records can lead to a successful recovery of the Atlantic Meridional Overturning Circulation (AMOC) in a biased model scenario, b) how joint evaluation of parameter sensitivities to proxy observations and model minus observation values (innovations) is needed as part of QC, and c) with a comprehensive database, representing the Last Glacial Maximum, how nonlinearity between point observations and model parameters informs observation screening and the need to apply regional targets.