Dynamic controls on zonal variability of sea surface temperatures and implications for paleoclimate reconstruction

Estimates of past sea surface temperature (SST) are requisite to constrain paleoclimatic conditions in deep time. Many Mesozoic and Cenozoic (and virtually all Paleozoic) SST data come from coastal environments (i.e., <500 km from land). Geochemical analyses using marine macrofossils require outcrop exposures and hence are largely restricted to continental margins, and the majority of micro-fossil analyses from ocean drill cores are similarly sited along these margins because both sedimentation rate and preservation potential are higher. However, coast-proximal data are often interpreted assuming they represent larger-scale, globally-relevant conditions (e.g., 'mid-latitude,' 'equatorial,' 'North Atlantic'), and are used at face value to quantify metrics of paleoclimatic interest, such as the meridional (latitudinal) temperature gradient. It is therefore important to understand how representative these locations are of zonal mean SST. This is particularly true during greenhouse climate regimes with suppressed latitudinal gradients (e.g., the Eocene), where longitudinal variability in proxy data can at times exceed the inferred meridional temperature gradient.

Here, we explore the spatial distribution of and dynamical controls on coastal SSTs using modern monthly-resolved climatologically-averaged SST data, focusing on patterns related to the direction and magnitude of a given location's deviation from the zonal mean. We then turn to proxy-derived SST compilations from the Eocene and Pliocene to assess the potential for bias when using sparse coastal data to make inferences about global climate. In modern oceans, significant and systematic deviations from the zonal mean are observed along coastlines, with distinct and contrasting patterns evident along the eastern and western boundaries of ocean basins. We quantitatively relate these patterns to wind-driven ocean circulation processes (i.e., gyre circulation) and create a paradigm with which to more robustly contextualize and interpret annually- and seasonally-resolved paleoclimate proxy data. Applying this paradigm, we reevaluate Eocene and Pliocene proxy-based latitudinal temperature gradients and discuss the extent to which both historical and deep time climate models are able to reproduce them.