It might take three: proxy system models as the missing link between proxies and climate models, and their potential for paleoclimate data assimilation

Abstract:

In recent years, data assimilation (DA) has emerged as a competitive method for climate field reconstruction (CFR). DA blends information from climate models with proxy observations to estimate the most plausible climate state that could have given rise to a set of observations, together with uncertainties about this state. The DA framework allows one to explicitly represent proxy complexity via proxy system models (PSMs), which are physically-based, potentially nonlinear models of proxy systems. In contrast, most (regression-based) CFR approaches make a number of strong assumptions that may not be met in practice, including the use of linear temperature-proxy relationships, or ignoring the confounding effects of dating uncertainties in proxy data.

In this study, we combine DA-based CFR techniques with proxy system modeling (PSM) to investigate uncertainties that arise as a result of these assumptions. We use PSM-generated pseudoproxy experiments to address three questions: (1) how much information is lost assuming proxies are linear, univariate responders to temperature? (2) does a misspecified proxy system contribute to poor climatic interpretations? and (3) given perfectly constrained proxy system parameters, what is the role of age uncertainties in blurring the retrieved climate signal?

We find that employing linearized models for complex proxy systems contributes to substantial information loss in climate field reconstructions, while other complications (including age uncertainties) may not prove as problematic in a multi-proxy framework. We investigate the utility of embedding PSMs in DA techniques for paleoclimate state estimation, and use them to explore new uncertainties in paleodata-model comparison associated with linearity and nonstationarity. Implications for the potential of paleoclimate reanalyses are discussed.