Abstract
Paleoclimate data assimilation has recently emerged as a promising technique to estimate past climate states. Here we test two of the underlying assumptions of paleoclimate data assimilation as applied so far: (1) climate proxies can be modeled as linear, univariate recorders of temperature and (2) structural errors in GCMs can be neglected. To investigate these two points and related uncertainties, we perform a series of synthetic, paleoclimate data assimilation-based reconstructions where “pseudo” proxies are generated with physically based proxy system models (PSMs) for coral δ18O, tree ring width, and ice core δ18O using two isotope-enabled atmospheric general circulation models. For (1), we find that linear-univariate models efficiently capture the GCM's climate in ice cores and corals and do not lead to large losses in reconstruction skill. However, this does not hold for tree ring width, especially in regions where the trees' response is dominated by moisture supply; we quantify how the breakdown of this assumption lowers reconstruction skill for each proxy class. For (2), we find that climate model biases can introduce errors that greatly reduce reconstruction skill, with or without perfect proxy system models. We explore possible strategies for mitigating structural modeling errors in GCMs and discuss implications for paleoclimate reanalyses.
Original language | English |
---|---|
Pages (from-to) | 1164-1179 |
Number of pages | 16 |
Journal | Journal of Advances in Modeling Earth Systems |
Volume | 8 |
Issue number | 3 |
DOIs | |
State | Published - 1 Sep 2016 |
Externally published | Yes |
Bibliographical note
Publisher Copyright:© 2016. The Authors.
Keywords
- climate reconstruction
- data assimilation
- general circulation models
- paleoclimate
- proxy system modeling