Comparing early to middle Miocene terrestrial climate simulations with geological data N. Herold1,*, R.D. Müller1, and M. Seton1 1EarthByte Group, School of Geosciences, University of Sydney, Sydney, NSW 2006, Australia ABSTRACT ent. Of these, the largest source of uncertainty EXPERIMENT DESIGN arguably arises from model boundary condi- The early to middle Miocene was signifi - tions, where values for large areas of the globe Simulations are conducted using the National cantly warmer than present, particularly at are predicted or interpolated from relatively Center for Atmospheric Research (NCAR) high latitudes. Relatively few climatic details few data points (e.g., for vegetation or sea- Community Atmosphere Model 3.1 (CAM; are known about this time period compared surface temperature [SST]). Collins et al., 2006) and Community Land with earlier (e.g., Cretaceous/Eocene) and In atmosphere only General Circulation Model 3.0 (CLM; Vertenstein et al., 2004). Both later (e.g., Quaternary) intervals. In this Models (GCMs), SST is the most infl uential CAM and CLM are confi gured for T31 resolu- study terrestrial proxy data are quantita- boundary condition on the simulated climate, tion, representing 3.75° in longitude and ~3.75° tively compared with three simulations of as oceanic responses to the atmosphere are in latitude. CAM uses a hybrid sigma-pressure early to middle Miocene climate (20–14 Ma) precluded. Over the past decade, increasing vertical coordinate system with 26 levels. CLM carried out using the National Center for evidence of diagenesis in low latitude fossils simulates temperature and water variables for Atmospheric Research (NCAR) Community of planktonic foraminifera have revealed a sig- ten layers of soil and up to fi ve layers of snow. Atmosphere Model 3.1 (CAM) and Com- nifi cant cold bias in δ18O derived temperatures Each land grid cell is prescribed up to four out munity Land Model 3.0 (CLM). Three dif- (Pearson et al., 2001; Williams et al., 2005). of an available 15 vegetation types, constituting ferent meridional sea-surface temperature This has called into question previous SST a biome. Each vegetation type differs in albedo, gradients are prescribed in order to test a reconstructions (e.g., O’Connell et al., 1996; root distribution, aerodynamic, and photosyn- range of plausible climates. Our simulations Sloan and Thomas, 1998). Sensitivity stud- thetic parameters. CAM and CLM reproduce yield generally cooler and more arid condi- ies show that atmospheric and implied surface modern-day climate reasonably well and cou- tions than indicated by the proxy record. ocean characteristics can differ dramatically pled to the NCAR ocean and sea-ice models, Mean model-data discrepancies for precipi- with varying low and/or high latitude SSTs have been used to simulate deep-time paleocli- tation and temperature decrease from −320 (Huber and Sloan, 2000; Rind, 2000). There- mate intervals (e.g., Kiehl and Shields, 2005). to −170 mm/yr and −0.5 to −0.4 °C, respec- fore, as a second-order indication of accuracy, However, despite improvements over earlier tively, when tropical sea-surface tempera- it is useful to constrain reconstructed SSTs by versions, several biases remain in the models, tures are increased by ~4 °C from inferred comparing the terrestrial climate they force in including underestimation of implied ocean Miocene values to near modern values. The a model with the proxy record (e.g., Lunt et heat transport, excessive Northern Hemisphere poor agreement with respect to mean annual al., 2008a; Sloan and Barron, 1992). However, low cloud production, and anomalous continen- precipitation may be attributed to the pre- such a method assumes proxy data are accurate tal precipitation (Collins et al., 2006; Dickinson clusion of an interactive ocean model and/or to within the degree of error of climate models, et al., 2006). model bias. and representative of the grid cells they occupy This study analyzes results from three simu- in model space (discussed later). lations forced with newly constructed bound- INTRODUCTION Here we focus on three simulations of early ary conditions. Inherently, a trade-off exists to middle Miocene climate, a period of distinct between the spatial and temporal resolution of Numerical simulations of Earth’s past cli- global warming terminated by abrupt cooling proxy data and the limited information available mate are often compared with the proxy record in the late-middle Miocene. Each simulation is for the Miocene makes a longer time interval in order to determine the accuracy of the model forced with a different meridional SST gradient favorable. Therefore published information pri- solution. Despite the typically large uncer- based on plausible variations to published δ18O marily for the late-early to early-middle Mio- tainties associated with the reconstruction of and Mg/Ca data. To determine the meridional cene are utilized to reconstruct vegetation and boundary conditions, model formulation, and SST gradient that best matches terrestrial proxy SSTs (ca. 20–14 Ma). derivation of proxy data, this remains the only data we compare model-data discrepancies Each simulation differs only in regard to SST measure of a model’s ability to accurately between simulations. We subsequently discuss distribution. Twelve months of zonally vary- reproduce climates different to that of the pres- uncertainties in our results and draw compari- ing SSTs are prescribed in each simulation to sons with other studies of Miocene climate. replace a mixed-layer ocean model supported *[email protected] Geosphere; December 2010; v. 6; no. 6; p. 952–961; doi: 10.1130/GES00544.1; 7 fi gures; 2 tables. 952 For permission to copy, contact [email protected] © 2010 Geological Society of America Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/6/6/952/3340516/952.pdf by guest on 30 September 2021 Miocene climate simulations compared with geological data by CAM. While prescribing SSTs precludes seasonality during warmer climates—due to and (2) the extent to which proxy SST esti- important feedbacks between the ocean and ice-albedo feedbacks—the magnitude of these mates are preserved in the fi nal reconstruction. atmosphere we consider it more useful than differences is reduced by 30% (after O’Connell Using modern-day SSTs from CAM we cal- applying a mixed layer ocean model to approxi- et al., 1996). Finally, the east-west SST gradient culate mean annual values at locations where mate Miocene conditions. At the peak of Mio- from each modern-day month is applied to each Miocene proxy data exist (Fig. 1 caption; using cene warming, deep water and polar SSTs were respective Miocene month. Although modern- modern-day longitudes and latitudes). Using more than 5 °C warmer than present (Lear et al., day east-west gradients may not be represen- the least-squares method, a Gaussian curve 2000), implying that sea ice could only be sea- tative of Miocene surface conditions, zonally is fi tted to these values to reconstruct zonal sonal. Such conditions have proven diffi cult to constant SSTs may result in larger model-data mean SSTs, as is done for P-MIO (Fig. 2). reproduce using mixed layer or coupled ocean discrepancies (Sloan et al., 2001). Furthermore, A minimum value of −1.8 °C is applied (the models even when applying higher greenhouse the similarities between Miocene and modern- freezing point of seawater). The mean absolute gas concentrations (e.g., Huber and Sloan, 2001; day geography, along with partial reconstruc- difference between the reconstructed meridi- Steppuhn et al., 2007). tions of Miocene surface currents (Kennett et onal SST gradient and a zonal mean of the Our fi rst simulation is forced with SSTs based al., 1985; Wright and Thunell, 1988), imply modern-day CAM SST distribution is 0.9 °C. solely on published proxy data (referred to as large-scale surface conditions, such as boundary Therefore, given a limited number of accurate P-MIO). A Gaussian curve is fi tted to proxy currents, were similar to the modern. Therefore, SSTs we propose it is possible to reconstruct data using the least-squares method (Fig. 1). we believe modern-day east-west gradients pro- an approximate meridional temperature gradi- These values are extended zonally to produce vide a closer approximation to Miocene SSTs ent for a given time period. However, we note a global mean annual SST data set. Monthly than zonally constant values. Sea ice is assumed the current margin of error for proxy data, par- zonal SSTs are calculated based on modern-day to be absent in both hemispheres given mini- ticularly at low latitudes, does not allow a high seasonality. At each latitude, modern-day zonal mum SSTs are above freezing. degree of accuracy for this kind of reconstruc- mean SST for each month is subtracted from the To examine the validity of reconstructing tion (Crowley and Zachos, 2000). modern-day annual zonal mean from the same SSTs based solely on proxy data, we test two To determine the extent to which proxy latitude; these differences are then applied to sources of uncertainty in P-MIO: (1) the lim- SSTs are preserved in P-MIO we perform a site the Miocene data set to construct 12 months of ited number of data points used to constrain the comparison (Fig. 3). The absolute mean differ- zonally uniform SSTs. To account for reduced mean annual meridional SST gradient (Fig. 1), ence between proxy SSTs and the correspond- ing values in P-MIO is 3.5 °C. This is largely attributed to the variation in proxy SSTs around the reconstructed meridional gradient (Fig. 1). Assuming magnitudes of east-west temperature variation were similar in the Miocene as they are today, the range of proxy SSTs around the reconstructed meridional gradient should be up to an order of magnitude smaller (c.f.