TY - JOUR
T1 - A Bayesian, spatially-varying calibration model for the TEX86 proxy
AU - Tierney, Jessica E.
AU - Tingley, Martin P.
PY - 2014/2/15
Y1 - 2014/2/15
N2 - TEX86 is an important proxy for constraining ocean temperatures in the Earth's past. Current calibrations, however, feature structured residuals indicative of a spatially-varying relationship between TEX86 and sea-surface temperatures (SSTs). Here we develop and apply a Bayesian regression approach to the TEX86-SST calibration that explicitly allows for model parameters to smoothly vary as a function of space, and considers uncertainties in the modern SSTs as well as in the TEX86-SST relationship. The spatially-varying model leads to larger uncertainties at locations that are data-poor, while Bayesian inference naturally propagates calibration uncertainty into the uncertainty in the predictions. Applications to both Quaternary and Eocene TEX86 data demonstrate that our approach produces reasonable results, and improves upon previous methods by allowing for probabilistic assessments of past temperatures. The scientific understanding of TEX86 remains imperfect, and the model presented here allows for predictions that implicitly account for the effects of environmental factors other than SSTs that lead to a spatially non-stationary TEX86-SST relationship.
AB - TEX86 is an important proxy for constraining ocean temperatures in the Earth's past. Current calibrations, however, feature structured residuals indicative of a spatially-varying relationship between TEX86 and sea-surface temperatures (SSTs). Here we develop and apply a Bayesian regression approach to the TEX86-SST calibration that explicitly allows for model parameters to smoothly vary as a function of space, and considers uncertainties in the modern SSTs as well as in the TEX86-SST relationship. The spatially-varying model leads to larger uncertainties at locations that are data-poor, while Bayesian inference naturally propagates calibration uncertainty into the uncertainty in the predictions. Applications to both Quaternary and Eocene TEX86 data demonstrate that our approach produces reasonable results, and improves upon previous methods by allowing for probabilistic assessments of past temperatures. The scientific understanding of TEX86 remains imperfect, and the model presented here allows for predictions that implicitly account for the effects of environmental factors other than SSTs that lead to a spatially non-stationary TEX86-SST relationship.
UR - http://www.scopus.com/inward/record.url?scp=84891358221&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84891358221&partnerID=8YFLogxK
U2 - 10.1016/j.gca.2013.11.026
DO - 10.1016/j.gca.2013.11.026
M3 - Article
AN - SCOPUS:84891358221
SN - 0016-7037
VL - 127
SP - 83
EP - 106
JO - Geochimica et Cosmochimica Acta
JF - Geochimica et Cosmochimica Acta
ER -