Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 Proxy records from the Miocene epoch (∼23-5 Ma) indicate a warmer climate than today in spite of lower atmospheric carbon dioxide (CO2) concentrations in the range of preindustrial levels. As yet the simulation of a warm Miocene climate with these low CO2 values has proven to be a challenge. In this study we present climate simulations of the Late Miocene (11-7 Ma) with a preindustrial CO2 level, using a coupled atmosphere-ocean general circulation model (AOGCM). The simulated global mean surface temperature of ∼17.8 °C represents a significantly warmer climate than today. We have analyzed the relative importance of tectonic and vegetation changes as forcing factors. We find that the strongest temperature increase is due to the Late Miocene vegetation distribution, which is more than three times stronger than the impact induced by tectonic alterations. Furthermore, a combination of both forcing factors results in a global temperature increase which is lower than the sum of the individual forcing effects. Energy balance estimates suggest that a reduction in the planetary albedo and a positive water vapor feedback in a warmer atmosphere are the dominating mechanisms to explain the temperature increase. Each of these factors contributes about one half to the global temperature rise of ∼3 K. Our results suggest that a much warmer climate during the Late Miocene can be reconciled with CO2 concentrations similar to preindustrial values.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Geological archives from different localities around the globe demonstrate that the climate during the Miocene was more humid and warmer than today [e.g., Steppuhn et al., 2007; Zachos et al., 2001]. Although absolute values of global temperature changes are difficult to assess, proxy based sea surface temperature reconstructions and paleovegetation data combined with Late Miocene climate-vegetation model simulations suggest that the global mean annual surface temperature might have been as warm as ∼17–19°C [Lunt et al., 2008a; Pound et al., 2011]. This temperature range is significantly warmer than estimates for the modern climate ∼14.5°C [e.g., Hansen et al., 2010]. Furthermore, the Late Miocene is characterized by a substantial mid and high latitude warming of ∼6 K and ∼8 K, respectively, while the temperature changes in the tropics are less pronounced [Flower and Kennett, 1994; Steppuhn et al., 2006] (the use of ‘warming’ is not meant to indicate a time-directional sense). As a consequence the meridional temperature gradient was weaker than at present during the Miocene. These global-scale temperature changes are assumed to have been mainly related to increased CO2 concentrations and alterations in the tectonic setting, including altered ocean gateways such as the Panama seaway and the eastern Tethys region [Mikolajewicz and Crowley 1997; von der Heydt and Dijkstra, 2006; Micheels et al., 2009]. Furthermore, changes in the vegetation distribution [Dutton and Barron, 1997] and vegetation feedbacks [Henrot et al., 2010] plus the sea ice-albedo feedback mechanism have been suggested to contribute to warmer temperatures and more equable climates around the globe with a lower meridional temperature gradient. However, the importance of these factors in shaping the global Miocene temperature signature is poorly understood. So far these contributions have been mainly tested and evaluated individually in sensitivity studies employing ocean and atmosphere/slab ocean model set-ups [Mikolajewicz and Crowley, 1997; Micheels et al., 2009; You et al., 2009] neglecting important feedbacks in the coupled atmosphere/ocean system. Only a limited number of studies have investigated a Miocene climate with a coupled AOGCM configuration [Micheels et al., 2011; von der Heydt and Dijkstra, 2006]. However, these studies rely on relatively high CO2 levels of 360 ppmv and 710 ppmv, respectively, to simulate a warm Miocene climate. This is in contrast to CO2 reconstructions based on marine geochemical proxies such as δ13C of marine phytoplankton [Pagani et al., 2005] and δ11B of marine carbonate [Demicco et al., 2003; Pearson and Palmer, 2000] which indicate relatively low and stable values of ∼280 ppmv. During the Late Miocene CO2 values might have been even lower than preindustrial levels as suggested by boron/calcium ratios in planktic foraminifera [Tripati et al., 2009]. Based on the apparent paradox of warm climates with low and relatively stable CO2 levels in combination with pronounced global temperature variations [Zachos et al., 2001], the role of CO2 as a forcing factor in Miocene climate change has been discussed controversially [Ruddiman, 2010]. It has been hypothesized that alternative forcing factors at relatively low CO2 values were potentially more effective in the geological past [Royer, 2008].
 In this work we examine the climate impact of tectonic and vegetation conditions representative of the Late Miocene. Furthermore, we investigate the most important mechanisms and feedbacks that dominate temperature differences relative to the pre-industrial climate. In contrast to previous coupled modeling simulations with elevated CO2 levels we apply an atmospheric CO2 concentration at a preindustrial level of 278 ppmv.
 We utilize the comprehensive AOGCM ECHAM5-MPIOM without any flux corrections [e.g., Jungclaus et al., 2006]. The atmosphere model ECHAM5 was used at T31 resolution (∼3.75°) with 19 vertical levels. The ocean model MPIOM was run at an average resolution of ∼3° with 40 vertical layers. Vegetation is a fixed factor represented by specifying different land surface parameters like albedo, roughness length, vegetation ratio, leaf area index and maximum soil water capacity. This model approach has been applied and validated with proxy data in an investigation of heat transport mechanisms for the Late Miocene [Micheels et al., 2011]. We use the same boundary conditions (Figure 1) with respect to the tectonic setting and the vegetation distribution, but a lower CO2 concentration of 278 ppmv and minor land-sea modifications, e.g., a closed Hudson Bay [Smith et al., 1994]. For the ocean we adopted the present day bathymetry to the Late Miocene land sea mask including an open Central American seaway [Collins et al., 1996] with a depth of 500 m. For preindustrial and Miocene boundary conditions we perform the simulations CTRL and MIO. To evaluate the relative importance of tectonic and vegetation conditions as forcing factors we have run two additional simulations TECT and VEG, respectively. Furthermore, the isolated application enables us to separate the monocausal impact of the respective boundary conditions from synergistic effects of both forcing factors if applied in combination as done in simulation MIO. This experimental strategy is based on a factorization of boundary conditions [Stein and Alpert, 1993] for the climate simulation in MIO, accounting for a broader representation of the complex interactions and synergy within the coupled climate system [Berger, 2001]. Further details regarding the factor separation analysis are explained in the auxiliary material.
 Simulation TECT adopts the tectonic conditions of MIO, while it is run with present-day vegetation. In experiment VEG we apply the Late Miocene vegetation reconstruction to modern tectonic conditions. The factor separation analysis is completed by an investigation of the global surface temperature change, based on a zero-dimensional energy model [e.g., Budyko, 1969], to disentangle and quantify the dominant mechanisms and feedbacks that govern the Late Miocene climate.
 All simulations are integrated for 1500 years and orbital parameters are kept constant at present-day values. All model output shown in this study represents climatological mean values of 100 years at the end of each equilibrium simulation run for 1500 years. Further details of the model configuration, the boundary conditions (including the vegetation reconstruction) and the energy balance model are given by Micheels et al. [2007, 2011] and the auxiliary material.
 For full Miocene conditions in experiment MIO, we simulate a global surface air temperature of ∼17.8 °C, which is warmer than the modern climate and the preindustrial control run (CTRL, ∼14.8 °C). Besides a global temperature rise of ∼3 K the spatial temperature differences between MIO and CTRL are heterogeneous (Figure 2a). The most pronounced increase of the mean annual temperature occurs over Greenland exceeding 17 K. Other regions showing a strong warming are the Tibetan Plateau, north Africa and northern Australia. Furthermore, larger coastal and offshore regions of the Weddell Sea experience a strong warming of up to ∼10 K. In particular the adopted Miocene paleo-vegetation with taiga and cold deciduous forest in Greenland and the occurrence of grassland to savanna vegetation instead of the modern Sahara desert cause lower land surface albedo values (Figure 2b), contributing to the warming in these regions. In combination with warming-induced ice-albedo changes (Figure 2c), the resulting temperature response is particularly pronounced in the high-latitudes of both hemispheres. As a result the zonally averaged meridional surface air temperature gradient in the Northern Hemisphere decreases by more than 5 K, although the maximum northward ocean heat transport in the ocean declines from 2.28 PW to 1.88 PW (1 PW = 1 Petawatt = 1015 Watt). The decrease in the global oceanic heat transport is mainly related to a weakening of the Atlantic meridional overturning circulation (AMOC) (Figure S1) with a decrease of the maximum overturning from ∼18 Sv to ∼12 Sv (1 Sv = 106 m3 s−1) and a reduction of the associated northward heat transport in the Atlantic from 0.96 PW to 0.75 PW. The meridional heat transport decrease in the ocean is partly compensated by a strengthening of the maximum northward heat transport in the atmosphere that increases from 5.08 PW to 5.21 PW, which points to the relevance of additional mechanisms to explain the Miocene temperature signature. A quantitative comparison of the modeled mean annual temperature in simulation MIO to terrestrial proxy data estimates (Figure S2) and previous studies indicates a relatively good agreement with the Late Miocene climate as shown in the auxiliary material.
 To examine the relative importance of tectonic and vegetation conditions as forcing factors f we can compare our simulations TECT and VEG with the preindustrial climate simulation CTRL. The two simulations TECT (Figure 3a) and VEG (Figure 3b) clearly show that changes in the tectonic setting and vegetation distribution strongly modify the global temperature pattern. The global-mean surface temperature increase of ∼3 K in MIO can be attributed to the impact of the Late Miocene vegetation (fVEG = +2.5 K) and the effect of tectonic changes (fTECT = +0.7 K), while the synergy between them is negative (fSYN = −0.2 K). An explicit calculation of the respective values is provided in the auxiliary material.
 At a regional scale the factor separation reveals that the warming over Greenland can be explained by a similar contribution of both forcing factors. The warming in the modern north Australia region (Figure 3a) is largely controlled by a southward displacement of the continental landmass during the Miocene [Heine et al., 2010] and the influence of the relatively warm tropical ocean during the Miocene. Furthermore, the occurrence of tropical forest and the southward expansion of savanna vegetation during the Miocene correlate with a warmer climate over large parts of Australia (Figure 3b) due to a lower land surface albedo. The warming at the Tibetan Plateau (Figure 3a) is mainly related to surface elevation changes, while the warming in the Sahara region (Figure 3b) is due to vegetation changes. In these two regions the synergy between the forcing factors (Figure 3c) is slightly negative. The negative synergy is more pronounced over Greenland. In contrast to these regions the high-latitude North Atlantic and the Labrador Sea are also strongly influenced by a regional warming as a result of the positive synergy between the tectonic and the vegetation forcing. At high latitudes the relative importance of the negative synergy effect is particularly pronounced in the Weddell Sea, the Barents Sea and the Sea of Okhotsk (Figure 3c). The strong synergy impact in these high-latitudes regions of both hemispheres can be explained by a pronounced temperature sensitivity of the sea ice cover and associated feedbacks that are already triggered by global warming in TECT and VEG. Therefore, the combined impact of both forcing factors can result in a much weaker warming at a regional scale, also representing an important effect at the global scale as indicated by the negative sign of fSYN.
 Besides the general warming of large parts of the globe (Figure 2a) there are also areas in MIO that experience a temperature decrease. Especially the North Atlantic realm south of Iceland cools up to ∼1.7 K. This temperature decrease and the cooler temperatures in the eastern sector of the subtropical gyre can be largely attributed to the impact of tectonic changes causing a weak bi-polar north-south temperature pattern in the Atlantic (Figure 2a). This pattern can be attributed to basin-scale temperature response to a weakened AMOC and an associated northward heat transport reduction in the Atlantic, which characterizes the difference between MIO and CTRL (Figure S1). Similar temperature patterns associated with AMOC changes have been identified in a wide range of paleoclimate investigations ranging from glacial/interglacial [Knorr and Lohmann, 2007] to Pliocene [Lunt et al., 2008b] conditions. Besides the inter-hemispheric temperature changes, pronounced high-latitude temperature gradients with strong warming over Greenland and cooler temperatures in the Labrador Sea and the Barents Sea develop in experiment TECT (Figure 3a). These characteristics point to a strong influence of large-scale atmosphere circulation changes as a consequence of a lowered topography of the Rocky Mountains and the Greenland ice sheet in these regions. At a global scale the surface temperature increase of ∼0.7 K in TECT is close to the climate sensitivity to changes in Miocene paleotopography of Herold et al. , who estimated an increase of 0.5 K.
 So far our investigations suggest that large changes in surface albedo (Figure 2b) are important to explain the Miocene temperature increase. To better understand the global radiation balance and to quantify the impact on this temperature rise, we compare the planetary albedo (α) and the effective long wave emissivity (ɛ) of MIO and CTRL. These quantities are derived from the globally averaged radiative fluxes [Smith et al., 2006] in our simulations. In CTRL α and ɛ amount to 0.317 and 0.584, respectively. The planetary albedo in MIO is reduced by ∼0.014, which causes less shortwave reflection by the atmosphere and a warming. The effective longwave emissivity in MIO also decreases by ∼0.012 reducing the ratio between the longwave radiation emitted at the earth surface and leaving the top of the atmosphere. The reduction of ɛ is mainly governed by a ∼22% increase of the vertically integrated water vapour content in the atmosphere, which strengthens the greenhouse effect. Based on a zero-dimensional energy balance model [e.g., Budyko, 1969] the impact of α and ɛ can be quantified, each causing about one half of the global warming of ∼3 K.
 Our simulated Late Miocene temperature increase at relatively low CO2 values is stronger than previous estimates assuming modern CO2 levels that show only a limited temperature rise of ∼≤1.5 K for the Late Miocene [e.g., Micheels et al., 2011]. This suggests that the thermal response is stronger if Late Miocene boundary conditions are applied to a preindustrial climate state. This points to an underlying nonlinearity as indicated by the negative synergy term, which is primarily due to a larger potential for sea ice changes and associated feedbacks for a cooler climate. Furthermore, significant changes in the high-latitude ocean mixed-layer depths between CTRL and MIO during the winter months of both hemispheres indicate the importance of three-dimensional ocean changes on the thermal response of the coupled system (not shown). Based on the combination of the factor separation analysis and the energy balance estimates we can infer a cause and effect relationship that highlights the role of vegetation changes. According to the factor separation the vegetation impact is more than three times stronger than the effect of tectonic changes. The vegetation distribution can directly influence the planetary albedo. In contrast the planetary albedo is only indirectly affected by the lower land surface elevation, e.g., via warming-induced ice-albedo variations.
 In summary our coupled AOGCM investigations suggest that a Miocene climate warmer than today can be explained without elevated CO2 levels relative to preindustrial values. Although our results should not be interpreted as direct evidence for low CO2 during the Miocene and essentially all existing CO2 reconstructions may be biased, our study supports Miocene CO2 estimates in the preindustrial range as indicated by marine geochemical proxies [Demicco et al., 2003; Pagani et al., 2005; Pearson and Palmer, 2000]. This implies that comprehensive coupled atmosphere/ocean models are able to reproduce ‘enigmatic’ [Ruddiman, 2010] climate states as recorded in geological archives. We suggest that future modeling studies using alternative AOGCMs should help to evaluate the robustness of the presented Late Miocene temperature estimates at preindustrial CO2 levels. Furthermore, so far state-of-the-art model approaches accounting for dynamical vegetation changes are limited in their ability to reproduce a warm Late Miocene climate at low CO2 (Figure S3 and Table S1). Therefore, future modeling efforts need to incorporate advanced numerical vegetation schemes and ice sheet models in order to quantify the associated feedbacks and synergies in the fully coupled earth system.
 We thank P. Pearson and S. Barker for helpful discussions and two anonymous reviewers for their constructive comments that helped to improve the study significantly. This work was supported by the German Research Foundation (DFG) grant (FOR 1070) within the research unit ‘Understanding Cenozoic Climate Cooling: The Role of the Hydrological Cycle, and Vegetation Changes’ and the Federal state of Hessen within the LOEWE initiative.
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.