Abstract
Seawater temperature is an essential quantity for paleoclimatological and paleoecological studies. A potential archive that can provide century-long, temporally well-constrained and high-resolution temperature proxy data is available in the form of bivalve shells. However, the number of well-accepted and robust temperature proxies contained in shells is limited to stable oxygen isotopes and carbonate clumped isotopes. Many studies have therefore investigated the possibility to reconstruct temperature from element/Ca properties, specifically Sr/Ca ratios in case of aragonitic shells. As demonstrated here, in agreement with thermodynamic expectations and the lattice strain model, shell Sr/Ca of laboratory-grown Arctica islandica specimens is strongly positively coupled to water temperature. If ultrastructure-related bias is mathematically eliminated, up to 75% of the variability in shell Sr/Ca data can be explained by water temperature. However, in field-grown specimens, this relationship is superimposed by other environmental variables that can hardly be quantified and mathematically eliminated. The explained variability of Sr/Ca is reduced to merely 26% and the prediction uncertainty too large for reliable temperature estimates. Most likely, the equable, less biased conditions in the laboratory resulted in the production of a more uniform shell ultrastructure (with larger and more elongated biomineral units) which in turn was associated with less variable Sr/Ca values and a stronger link to water temperature. Without a detailed understanding and quantification of the factors controlling ultrastructural variations in field-grown bivalves, it remains impossible to employ shell Sr/Ca of wild A. islandica specimens for precise temperature estimates, merely a qualitative temperature reconstruction seems feasible.