A model-based evaluation of tropical climate in Pangaea during the late Palaeozoic icehouse

A model-based evaluation of tropical climate in Pangaea during 1

the late Palaeozoic Icehouse 2

Nicholas G. Heavensa,*, Natalie M. Mahowaldb, Gerilyn S. Soreghanc, Michael J. 3 Soreghanc, and Christine A. Shieldsd 4 aDepartment of Atmospheric and Planetary Sciences, Hampton University, 23 E. Tyler St., 5 Hampton, VA 23668, USA 6 bDepartment of Earth and Atmospheric Sciences, Cornell University, Ithaca, NY 14853, USA 7 cSchool of Geology and Geophysics, University of Oklahoma, Norman, OK 73019, USA 8 dClimate Change Research Section, Climate and Global Dynamics Division, National Center for 9 Atmospheric Research, Boulder, CO 80305, USA 10 *Corresponding author at: Department of Atmospheric and Planetary Sciences, Hampton 11 University, 23 E. Tyler St., Hampton, VA 23668, USA. Tel: + 1 757 637 3118. E-mail address: 12 [email protected] (N.G. Heavens). 13

Accepted by Palaeogeogr. Palaeoclimatol. Palaeoecol., 17 February 2015 14

15

Highlights: 16

Modeled late Palaeozoic climate with a climate model with a 3-d ocean 17

Examined factors controlling tropical precipitation in Pangaea 18

Orbital forcing’s control of monsoonal variability of primary importance 19

When ice sheets limited in area, sea level and greenhouse gas levels also important 20

Glaciation of the Central Pangaean Mountains results in widespread tropical aridity 21

22

Keywords: 23

Palaeozoic 24

monsoon 25

climate 26

modeling 27

glaciation 28

Carboniferous 29

Permian 30

31

ABSTRACT 32

The Late Palaeozoic Ice Age (LPIA) is the Earth’s penultimate “icehouse climate.” Geological 33

proxies for tropical Pangaean climate during the LPIA are significantly modulated on 100–400 34

kyr and sub-100 kyr scales. In addition, some geological proxies suggest that equatorial 35

continental areas may have been colder and more arid during LPIA glacial intervals than during 36

late Cenozoic glacial intervals. Furthermore, the relationship between polar and tropical climate 37

variability remains controversial. The climate dynamics underlying these phenomena are only 38

partly understood. Nevertheless, past modeling of LPIA climate has suggested that precipitation 39

near the Equator was modulated by monsoonal circulation that itself was modulated by orbital 40

variability. Here, the Earth’s climate during Asselian-Sakmarian time (299–284 Ma) was 41

simulated with the Community Climate System Model version 3 (CCSM3). These simulations 42

address model climate sensitivity to direct and indirect effects of glaciation as well as variability 43

in the Earth’s orbit. The results of these simulations suggest that sea level and orbitally forced 44

monsoon variability were the primary controls on tropical precipitation when ice sheets were 45

confined to polar latitudes. Any impact of orbital forcing on glaciation therefore could have led 46

to aliasing between glacial and monsoonal effects in particular regions, perhaps explaining 47

contrasting geologically-based interpretations of how tropical climate responded to glaciation. 48

Glaciation of the Central Pangaean Mountains would have led to widespread cold and aridity in 49

equatorial Pangaea, but forming these glaciers would have required an unknown additional 50

climate forcing with an unknown impact on tropical precipitation. 51

52

52

1. Introduction 53

54

Evidence for glaciation during the Late Palaezoic Ice Age (LPIA) (~330–270 Ma) is 55

uncontroversial, but the extent of glaciation remains unclear. Sedimentological indicators of ice 56

are widespread within deposits from high southern paleolatitudes, while low-latitude 57

sedimentary deposition appears to have been strongly affected by high-frequency oscillations in 58

sea level, presumably glacioeustatic (e.g., Fielding et al., 2008; Heckel, 2008; Eros et al., 2012). 59

At one extreme, Isbell et al. (2003a,b) proposed the peak level of ice during the LPIA consisted 60

of mountain glaciation in the southern mid-latitudes and small ice sheets at high latitudes, a state 61

corresponding to a climate broadly warmer than the Plio-Pleistocene. At the other extreme, G.S. 62

Soreghan et al. (2014) proposed that at some point during the LPIA, ice sheets reached the mid-63

latitudes in both hemispheres and glaciation occurred on the slopes of equatorial mountain 64

ranges, a state corresponding to a climate broadly colder than the Plio-Pleistocene. The 65

arguments for the extreme hypotheses are based on sedimentological evidence for glaciation, but 66

geochemically-based paleothermometry can be leveraged to corroborate either view (e.g., 67

Grossman et al., 2012; Giles, 2012). The confirmation of either view or some intermediate 68

hypothesis would have implications for the long-term climate evolution of the Earth and 69

biological evolution during the late Palaeozoic (Montañez et al., 2007; Horton and Poulsen, 70

2009). 71

The parallels with and differences between LPIA and Plio–Pleistocene conditions have 72

attracted attention from climate modelers, who have been interested in how hypothetical 73

mechanisms for Plio–Pleistocene climate variability translated to the circumstances of the LPIA 74

(Kutzbach and Gallimore, 1989; Crowley et al., 1991; Gibbs et al., 2002; Otto-Bliesner, 2003; 75

Poulsen et al., 2007; Peyser and Poulsen, 2008; Horton and Poulsen, 2009; Horton et al., 2007, 76

2010, 2012). There is geological evidence for a monsoonal circulation in both hemispheres of 77

Pangaea that began during the LPIA, for which there is no modern analog (Kutzbach and 78

Gallimore, 1989). Therefore, the dynamics of monsoons and tropical climate in general has been 79

of particular interest. 80

Typical climate model inputs, such as greenhouse gas levels, have large uncertainties for 81

LPIA climates (e.g., Montañez et al., 2007). Therefore, modeling investigations primarily 82

encompass sensitivity studies, which focus on the sensitivity of a given climate model to 83

hypothetical forcing: changes in model inputs. This paper presents a model sensitivity study of 84

LPIA glacial–interglacial variability with the Community Climate System Model version 3 85

(CCSM3) (Collins et al., 2006; Yeager et al., 2006): in a configuration that includes a fully 86

coupled ocean with sea ice (Kiehl and Shields, 2005). Many of the forcings (greenhouse gas 87

levels, ice sheet extent, and orbital variability) considered in previous sensitivity studies are 88

considered as well as previously unexplored forcings such as sea level variability and tropical 89

mountain glaciation. The set of forcings explored was chosen to investigate why loess deposits 90

with sub-100 kyr cyclicity might form at low paleolatitudes: a problem directly relevant to 91

whether the LPIA was broadly warmer or colder than the Plio–Pleistocene icehouse (Soreghan 92

and Soreghan, 2002; M.J. Soreghan et al., 2014). Because these simulations are sensitivity 93

studies, they are not always dynamically consistent. This work therefore should be regarded as a 94

preliminary investigation of LPIA climate with a model of the complexity and computational 95

expense of CCSM3. Regardless of their preliminary nature, these experiments still shed light on 96

possible causes of tropical precipitation variability in Pangaea and their relationship to 97

glaciation. 98

In Section 2, the modeling experiments will be motivated by considering relevant geological 99

evidence about LPIA climate, its uncertainties, and past modeling investigations of LPIA 100

climate. In Section 3, the modeling framework will be described. The results will be presented 101

and discussed in Sections 4 and 5 before being summarized in Section 6. 102

103

2. The LPIA 104

2.1. Geological evidence for LPIA glaciation and temperature 105

106

Recognizable glacial facies (diamictite) and other glaciogenic features (striated pavements, 107

striated cobbles, and erratics) dating from the LPIA occur in sedimentary basins of the former 108

Gondwanan supercontinent in South America, Africa, Australia, India, Antarctica, and the 109

Arabian Peninsula (e.g., Du Toit and Reed, 1927; Wanless and Cannon, 1966; Fielding et al., 110

2008). Additionally, indirect evidence for glaciation occurs in the form of abrupt, cyclical 111

changes in sea level (~ 1 cm a-1), evinced by the well-known coal-bearing cyclothems at low 112

paleolatitudes (Wanless and Shepard, 1936; Veevers and Powell, 1987; Heckel, 1986, 2008). 113

Some of this cyclicity matches Milankovitch periodicities (~100 kyr and ~400 kyr) that have 114

dominated Plio–Pleistocene climate variability (e.g., Schwarzacher, 1989; Birgenheier et al., 115

2009; Davydov et al., 2010; Sur et al., 2010; Schmitz and Davydov, 2012). There is also 116

evidence for significant glacial–interglacial variability during the LPIA on timescales of ~ 5 Myr 117

duration (Montañez et al., 2007; Fielding et al., 2008). (For brevity, alternating periods of ice 118

sheet expansion and contraction on 100–400 kyr timescales will be referred to as glacial–119

interglacial intervals and the Myr timescale variability will be referred to as glacial–interglacial 120

eras.) On both timescales of variability, the late Palaeozoic icehouse could have been subject to 121

complex feedbacks involving the cryosphere, biosphere, and hydrosphere similar to those under 122

investigation in the context of late Cenozoic climate. To date, however, correlation between 123

paleoproxy estimates of pCO2, tropical SSTs, and ice volume has been attempted only for 124

timescales of ~ 1 Myr (Montañez et al., 2007; Giles, 2012). 125

While the presence of land ice during the LPIA is unambiguous, its maximum extent remains 126

equivocal. Based on the absence of sedimentary indicators of land ice during the LPIA in 127

Antarctica and re-interpretation of land ice indicators elsewhere in Gondwanaland as indicative 128

of sea or lake ice, Isbell et al. (2003a,b) argued that LPIA ice volumes were similar to the present 129

day (and therefore below Last Glacial Maximum (LGM) values). To the contrary, it has been 130

argued that there was extensive glaciation across both the western and eastern Central Pangaean 131

Mountains (e.g., Becq-Giraudon et al., 1996; Soreghan et al., 2007, 2008a, b, 2009; Sweet and 132

Soreghan, 2008; Hastenrath, 2009; G.S. Soreghan et al., 2014). G.S. Soreghan et al. (2014) 133

inferred that if these glaciers existed, tropical temperatures near sea level could have been 4.3°–134

20.6° C cooler than the present day, 0°–15° C cooler than the LGM. 135

Geochemically-based paleothermometry can be leveraged to support either view, depending 136

on technique, interpretation, and sampled time interval. Studies that assume that brachiopod or 137

conodont-based oxygen isotope paleothermometers were unaffected by changes in ocean 138

composition other than the global effects of glaciation or local variations in the balance between 139

evaporation and precipitation suggest tropical SSTs approximated those of the Plio–Pleistocene 140

throughout the LPIA (e.g., Grossman et al., 2012; Chen et al., 2013). However, if there has been 141

long-term drift in the oxygen isotope composition and pH of seawater, tropical oceans were ~10º 142

C cooler than the present day (Giles, 2012). Clumped isotope paleothermometry is presumably 143

independent of seawater composition effects (Eagle et al., 2013). This technique estimates 144

temperatures during the Middle Pennsylvanian were 1.3°–4.7° C cooler than the present day, 145

mandating a much smaller change in seawater isotopic composition than inferred by Giles (2012) 146

(Came et al., 2007). Powell et al. (2009) used Mg/Ca brachiopod paleothermometry to infer 147

tropical temperatures 9° C cooler than the present day (assuming a present day tropical mean 148

brachiopod habitat temperature of 18.2° C: Giles, 2012); but drift in the Mg/Ca ratio of the ocean 149

as well as other factors could compromise these estimates. Inferred mineral crystallization 150

temperatures in paleosols from western equatorial Pangaea suggest Pennsylvanian temperatures 151

were up to 6° C colder than the present day (Tabor et al., 2013). 152

Note that the warmer paleotemperature values for the LPIA come from techniques (including 153

clumped isotope) that measure high palaotemperatures during Silurian and Ordovician time. 154

These high paleotemperatures likely preclude the glaciations implied by sedimentology during 155

Late Ordovician and Early Silurian time (Giles, 2012). In addition, apatite phosphate 156

paleotemperatures during Early Ordovician time appear too high for higher organisms in the 157

tropics to survive (Pucéat et al., 2010). All of these uncertainties argue for modeling a wider 158

range of glaciation scenarios for the LPIA. 159

160

2.2. Glaciation and tropical precipitation during the LPIA: geological evidence and modeling 161

162

Qualitative reconstructions of climate from characteristics of LPIA tropical and sub-tropical 163

strata show that variations in tropical precipitation strongly correlated with variations in sea level 164

and/or the extent of glaciation (Köppen and Wegener, 1924; Witzke, 1990; Ziegler et al., 1997; 165

Cecil et al., 2003; Tabor and Poulsen, 2008). The sign of the correlation, however, remains 166

disputed. Either the tropics experienced a wetter and less seasonal precipitation regime during 167

glacial (lowstand) intervals than during interglacial (highstand) intervals (glacial humidity; 168

Miller and West, 1993; Miller et al., 1996; Cecil et al., 2003; Eros et al., 2012) or a drier and 169

more seasonal precipitation regime during glacial (lowstand) intervals than during interglacial 170

(highstand) intervals (glacial aridity: Soreghan, 1994, 1997; Heckel, 1995; Rankey, 1997; 171

Olszewski and Patzkowsky, 2003; Soreghan et al., 2008a; Falcon-Lang and DiMichele, 2010; 172

Allen et al., 2011; Jordan and Mountney, 2012). 173

Tropical Pangaean strata from the LPIA also indicate precipitation variability on 174

timescales shorter than the 100–400 kyr periodicities that dominate the cyclothem record (e.g., 175

Tabor and Poulsen, 2008; Olszewski and Patzkowsky, 2003; M.J. Soreghan et al., 2014). These 176

appear to reflect precession or obliquity changes, perhaps analogous to some late Cenozoic 177

records (e.g., Partridge et al., 1997; Garcin et al., 2006). Variability at precessional frequencies 178

also occurs during deep-time intervals characterized by little or no land ice (e.g., Olsen and Kent, 179

1996). 180

One deposit with confirmed sub-100 kyr variability is loess-paleosol couplets in the 181

Lower Permian Maroon Formation of Colorado (M.J. Soreghan et al., 2014). The Maroon 182

Formation here is one among many examples of thick continental successions of loess and 183

paleosols within 5°–10° of the paleo-equator (see compilation by Soreghan et al., 2008a). In 184

contrast, loess deposits of late Cenozoic age are primarily confined to the mid-latitudes (Pye, 185

1987; Muhs, 2013). 186

These loess deposits are suggestive but not unambiguous evidence that areas near the 187

Equator during the LPIA may have been drier and perhaps colder than during late Cenozoic 188

glacials. Loess deposition is not a proxy for aridity at the deposition site, because loess particles 189

are most effectively trapped by moist or vegetated surfaces (Tsoar and Pye, 1987; Pye, 1995). 190

Yet the short atmospheric lifetime of loess particles (less than a day for dust particles > 10 µm, 191

extrapolating from Seinfeld and Pandis, 1998) implies proximity of aeolian activity to the site of 192

loess deposition. Since aeolian activity is maximized under semi-arid and arid conditions (Pye, 193

1987; Rea, 1994), the contrast between the global distribution of loess in the late Palaeozoic and 194

late Cenozoic icehouses may imply drier equatorial climate in the former era. Under the 195

“probably oversimplified view” that loess is primarily a glaciogenic sediment (Muhs and Bettis, 196

2003), high rates of loess deposition would be common in periglacial regions. Therefore, greater 197

loess deposition at the low latitudes could imply cooler as well as drier climate. Alternate loess 198

production mechanisms include salt weathering, eolian abrasion, and frost shattering (e.g., 199

Derbyshire, 1983; Crouvi et al., 2010). The first two mechanisms would associate loess with 200

aridity. Only the dominance of the third would make loess an unambiguous indicator of cold. 201

Analysis of a wider set of proxies confirms that semi-arid and then arid conditions gradually 202

expanded from west to east across equatorial Pangaea during the latter half of the LPIA (305–203

270 Ma). Some have attributed greater aridity to the emergence of climate from icehouse 204

conditions, rather than the influence of cold equatorial temperatures (Tabor and Poulsen, 2008; 205

Boucot et al., 2013). Yet understanding remains incomplete about: (1) why conditions in latest 206

Pennsylvanian and earliest Permian time in western equatorial Pangaea were semi-arid in the 207

first place; (2) what drove both 100–400 kyr and sub-100 kyr precipitation variability; (3) what 208

controlled the phasing of precipitation with glaciation; and (4) how the precipitation dynamics of 209

equatorial Pangaean climate could have co-existed with equatorial cold. 210

Previous quantitative climate modeling of the phasing of precipitation with glaciation (and by 211

implication, 100–400 kyr precipitation variability) mainly supports the hypothesis of glacial 212

humidity (Otto-Bliesner, 2003; Poulsen et al., 2007; Peyser and Poulsen, 2008). Prior modeling 213

has considered indirect effects of glaciation on climate such as greenhouse gas variability and 214

vegetation changes (Peyser and Poulsen, 2008; Horton et al., 2010) but has neglected a 215

mechanism that would result in glacial aridity: sea level change, which would change the 216

availability of evaporable (and thus precipitable) water as well as the albedo of flooded or 217

drained areas (Heckel, 1995; Bishop et al., 2009; Eros et al., 2012). 218

Sub-100 kyr variability also has been addressed. Horton et al. (2012) questioned the simple 219

paradigm of glacial humidity/aridity in which high-latitude ice sheet extent correlates or anti-220

correlates with tropical precipitation, instead demonstrating the importance of interactions 221

between orbital variability and the monsoon in explaining tropical precipitation variability. The 222

simulations of Horton et al. (2012) nevertheless respond to higher pCO2 and lower ice volume 223

conditions with greater aridity at the Equator. 224

Modeling has yet to explain semi-arid equatorial climate under icehouse conditions. One set 225

of simulations implies that key loess-accumulating regions in western Pangaea, such as those 226

highlighted by M.J. Soreghan et al. (2014), would receive 700–1400 mm a-1 of precipitation 227

during ice-free interglacial intervals and then an additional 700–1100 mm a-1 during glacial 228

intervals in which land ice on Gondwanaland advanced equatorward of 30° S (Peyser and 229

Poulsen, 2008). These are extremely wet conditions for aeolian activity (Rea, 1994). If glaciation 230

of the Central Pangaean Mountains occurred, it might affect tropical precipitation by acting as a 231

cold trap for moisture, suppressing orographic convection, or changing the tropical circulation by 232

displacing the maximum heating off the Equator as investigated by Lindzen and Hou (1988). 233

However, models have yet to produce glaciation of the Central Pangaean Mountains (CPM), 234

even under extremely low pCO2 (Soreghan et al., 2008b). 235

Building upon these prior results, the simulations described here address the climate 236

dynamics underlying tropical Pangaean precipitation during the LPIA. These simulations explore 237

the effects of greenhouse gas variability, land ice extent, sea level, and variability in the Earth’s 238

orbital parameters, while mostly neglecting vegetation change and uncertainties in 239

paleogeography, such as the altitude of the CPM. 240

241

3. Simulations 242

243

3.1. General characteristics 244

245

All simulations were performed using CCSM3 in a configuration that couples active 246

components that simulate the atmosphere, ocean, land (prescribed vegetation), and sea ice. The 247

atmosphere model has a horizontal resolution of 3.75° x 3.75° (T31), while the ocean model has 248

25 layers and a nominal horizontal resolution of 3°. This configuration of CCSM3 is nearly 249

identical to that used by Kiehl and Shields (2005). 250

Yeager et al. (2006) characterized the biases of this configuration, showing that, relative 251

to both observations and higher-resolution versions of CCSM3, the T31 version underestimates 252

Atlantic meridional heat transport and overestimates Northern Hemisphere sea ice cover. Like 253

higher-resolution versions of CCSM3 and other global models of its generation (Dai, 2006), the 254

T31 version simulates a double ITCZ precipitation signature in the eastern Pacific. 255

The impact of model biases on this study cannot be quantified. There is little doubt that 256

CCSM3 (at any resolution) is deficient in simulating some aspects of atmospheric and oceanic 257

dynamics. Translating these biases into a different paleogeography, let alone the range of forcing 258

considered in this study, is an underdetermined problem. 259

All simulations use a fixed solar constant of 1333 W·m-2 (97.5% of present) in agreement 260

with the model of Gough (1981). Each simulation uses a prescribed spatially uniform aerosol 261

distribution with a visible optical depth of 0.08, which is lower than that used (0.28) by Kiehl 262

and Shields (2005) and is approximately equal to the pre-industrial non-anthropogenic aerosol 263

optical depth estimated by Mahowald et al. (2011). The optical properties of this background 264

aerosol are broadly similar to sulfate aerosol. Heavens et al. (2012) reviewed the challenges of 265

implementing less-idealized aerosol schemes in deep time. 266

267

3.2. Paleogeography 268

269

The paleogeography used corresponds to the Asselian and Sakmarian Stages of the 270

Permian. (299–284 Ma). This time interval began with a glacial era of widespread glaciation, 271

high-frequency climate variability, and low pCO2 and ended with an interglacial era of high 272

pCO2 (Isbell et al., 2003a; Montañez et al., 2007). All simulations, except for the control 273

simulations of pre-industrial Late Holocene climate, use basic paleogeographic information 274

appropriate for this era (Fig. 1; Ziegler et al., 1997; Blakey, 2008; Section S1), including the 275

estimated extent of 100 m sea level rise during interglacials (Rygel et al., 2008). The ocean grid 276

remains the same for all simulations. 277

278

3.3. Base layer, circular orbit sensitivity, and orbital sensitivity simulations 279

280

Most of the model experiments are the circular orbit sensitivity simulations (Table 1: 281

Cases 3–7, 12), which were constructed as mixtures of different scenarios for greenhouse gas 282

levels, sea level, vegetation and glaciation. The scenarios for glaciation and vegetation were 283

generated on the basis of the results of two “base layer simulations” (greenhouse.base and 284

icehouse.base) corresponding to two distinct greenhouse gas scenarios, one (GG250) 285

approximating hypothetical conditions at the beginning of the Asselian-Sakmarian and the other 286

(GG2500) approximating hypothetical conditions at the end. (Section S2 describes each scenario 287

and how it was generated; the Table 1 key provides a qualitative description of each scenario, 288

and Fig. 2 illustrates the vegetation and glaciation scenarios.) Each simulation, except the 289

simulations testing sensitivity to orbital parameters, was run until the trend in the energy 290

imbalance of the model system for the last 100 years of the run was less than +/- 0.5 W·m-2 and 291

the surface temperature trend was within +/- 0.05 K·(100 a)-1. The orbital sensitivity simulations 292

(Cases 8–11) test climate sensitivity to variability in the Earth’s obliquity and longitude of 293

perihelion at the Earth’s peak eccentricity and were run for 100 years beyond the circular orbit 294

sensitivity simulation used as their initial condition (icehouse.glaciation.polar). 295

296

Fig. 1. Basic paleogeography (Asselian–Sakmarian) used in the modeling experiments. The 297 locations of the highest terrain and highstand flooding (SL100) used in some experiments also 298 are plotted. 299 300

The simulation of Kiehl and Shields (2005) used CCSM3 and a similar paleogeography 301

to the one used here. Analysis of this simulation (Section S3) shows that a global mean surface 302

temperature bias of ~0.5° C is introduced by stopping the simulation at an energy imbalance of 303

+/- 0.5 W·m-2 rather than +/- 0.05 W·m-2 used by Kiehl and Shields (2005). Temperature bias is 304

primarily in polar ocean regions, while precipitation bias is centered on 30º N and 30º S over 305

land areas (Fig. S1). Bias in both fields near the Equator is much less than the global average. 306

This bias analysis, however, is most applicable to simulations that use similar forcing to the 307

simulation of Kiehl and Shields (2005), i.e., ones that use GG2500. 308

0 30 60 90 120 150 180 210 240 270 300 330 360−90

−75

−60

−45

−30

−15

0

15

30

45

60

75

90

Longitude

Latit

ude

Coastline at LowstandHighstand FloodingMountains (topography > 1000 m)

309 310 Fig. 2. Prescribed vegetation, sea level, and glaciation scenarios: a, V2500 with lake biomes from 311 SL100 imposed; b, V250; c, ICEP and ICEE, the point used to extrapolate glaciation in ICEE is 312 marked with a yellow cross. Blank space in biome maps is ocean. Abbreviations for biome types 313 in key below Figs. 1a-b: li: land ice; d: desert; cnef: cool needleleaf evergreen forest; cndf: cool 314 needleleaf deciduous forest; cbdf: cool broadleaf deciduous forest; cmf: cool mixed forest; wnef: 315 warm needleleaf evergreen forest; wbdf: warm broadleaf deciduous forest; wmf: warm mixed 316 forest; tbef: tropical broadleaf evergreen forest; tbdf: tropical broadleaf deciduous forest; s: 317 savanna; eft: evergreen forest tundra; dft: deciduous forest tundra; cfc: cool forest crop; wfc: 318 warm forest crop; cg: cool grassland; wg: warm grassland; t: tundra; esl: evergreen shrub land; 319 dsl: deciduous shrub land; sd: semi-desert; cic: cool irrigated crop; cc: cool crop; wic: warm 320 irrigated crop; wc: warm crop; fw: forest wetland; nfw: non-forest wetland; l: lake. 321 322

Table 1. 323 Simulations of late Paleozoic climate. 324 Identical case numbering is used in Table 5, Fig. 2, and is referred to in the text. The longitude of 325 perihelion is defined following the convention that the longitude of perihelion is 0° at northern 326 vernal equinox. GG250 = A greenhouse gas scenario corresponding to low pCO2 conditions 327 during Asselian time (250 ppmv) and therefore to icehouse conditions; GG2500 = A greenhouse 328 gas scenario corresponding to high pCO2 conditions during Sakmarian time (2500 ppmv) and 329 therefore to greenhouse conditions; SL0 = A sea level scenario approximating lowstand 330 conditions; SL100 = A sea level scenario approximating highstand conditions (100 m sea level 331 rise); V250 = A vegetation scenario based on GG250 conditions; V2500 = A vegetation scenario 332

based on GG2500 conditions; ICEP = a glaciation scenario involving a single ice sheet in the 333 high-latitudes of the Southern Hemisphere; ICEE = a glaciation scenario consistent with 334 glaciation in the Ancestral Rocky Mountains. 335 336 CaseNumber

Name Greenhouse Gas Scenario

Sea Level

Vegetation Scenario

Glacial Configura-tion

Obliquity (°)

Eccentricity

Longitude of Perihelion (°)

1 greenhouse.base GG2500

SL100

Uniform

None 23.5 0 0

2 icehouse.base GG250 SL0 Uniform

None 23.5 0 0

3 greenhouse.noglaciation

GG2500

SL100

V2500 None 23.5 0 0

4 icehouse.highsealevel.highCO2

GG2500

SL100

V250 None 23.5 0 0

5 icehouse.highsealevel

GG250 SL100

V250 None 23.5 0 0

6 icehouse.noglaciation

GG250 SL0 V250 None 23.5 0 0

7 icehouse.glaciation.polar

GG250 SL0 V250 ICEP 23.5 0 0

8 ih.g.orb1 GG250 SL0 V250 ICEP 22.0 0.057 270




12 icehouse.glaciation.equatorial

GG250 SL0 V250 ICEE 23.5 0 0

337

The orbital parameter sensitivity simulations (ih.g.orbx) typically have energy 338

imbalances ~ -0.5 W·m-2 (Table 2). Note that in all simulations, the solstices and equinoxes 339

occur on the same day of year they would in a present day simulation. To reference the 340

simulations to modern climate, two late Holocene pre-industrial control simulations (pic1 and 341

pic2), which met the energy imbalance and temperature trend criteria described above, were 342

performed. These simulations differ only in their vegetation prescriptions (Section S2.3) and 343

meet the same criteria used to evaluate the stability of the base layer simulations. 344

345

Table 2. 346 Mean energy imbalance in years 71–100 of the orbital sensitivity simulations 347 Simulation Energy Imbalance (W m-2)

ih.g.orb1 -0.46

ih.g.orb2 -0.87

ih.g.orb3 -0.09

ih.g.orb4 -0.47 348 349

4. Results 350 351 4.1 Sensitivity of global climate and tropical Pangaean precipitation to forcing 352 353 4.1.1. Greenhouse gases 354 355

356

Global mean characteristics divide the late Palaeozoic climate simulations into three 357

groups (Fig. 3 and Table 3): greenhouse climate simulations (Fig. 3, Cases 1, 3–4), with tropical 358

temperatures 5–6° C warmer than the present day; icehouse climate simulations with no or only 359

high-latitude glaciation (Fig.3, Cases 2, 7–11) and tropical temperatures 0–1° C cooler than the 360

present day; and icehouse climate simulations with low-latitude glaciation and tropical 361

temperatures 8° C cooler than the present day (Fig. 3, Case 12). (Present day tropical 362

temperatures are assumed to be 1° C warmer than pre-industrial to one significant figure (Deser 363

et al., 2010). (See Table 3 for complete case numbering.) 364

The sensitivity to greenhouse gas forcing can be evaluated by differencing 365

icehouse.highsealevel.highCO2 and icehouse.highsealevel, since these simulations only differ in 366

greenhouse gas forcing. The change in global annual mean surface temperature attributable to 367

using the higher-level greenhouse gas scenario (GG2500) rather than the lower-level greenhouse 368

gas scenario (G250) is then 8.7 °C. The sensitivity of surface temperature to a doubling of CO2 is 369

2.2 to 2.3°C, which is approximately the sensitivity of CCSM3’s T31 configuration (2.32 °C) 370

when applied to present-day geography (Kiehl et al., 2006). 371

Higher greenhouse gas levels also result in higher precipitation globally. This effect 372

averages 40-50 mm per doubling of pCO2,eq at the greenhouse gas levels encompassed by G2500 373

and G250. The increase in precipitation is not spatially uniform over tropical Pangaea. 374

Precipitation near the axis of the Central Pangaean Mountains in the model domain (Fig. 1) 375

greatly increases under the higher greenhouse gas levels of G2500 (Fig. 4a). 376

Table 3 377 Annual mean results 378 Identical case numbering is used in Table 1, Figure 1, and is referred to in the text. 379

Case Number1

Name Sfc. T (°C)

Land

Ocean

Tropical

Precip. (mm)

Land

Ocean

Tropical

Sea Ice (% Ocean)

1 greenhouse.base 24.1

21.9

24.8 31.2 1210 920 1310

1490 0.02

2 icehouse.base 16.1

12.3

17.5 24.7 1000 600 1110

1280 2.2

3 greenhouse.noglaciation

23.4

21.4

24.2 30.9 1190 840 1310

1480 0.05

4 icehouse.highsealevel.highCO2

22.1

19.9

22.8 30.2 1160 790 1290

1460 0.14

5 icehouse.highsealevel 13.4

7.7 15.4 23.3 980 590 1110

1300 4.8

6 icehouse.noglaciation

14.5

10.1

16.1 24.2 960 520 1110

1260 4.1

7 icehouse.glaciation.polar

13.0

4.9 15.9 24.2 940 490 1100

1260 4.4

8 ih.g.orb1 12.9

4.9 15.7 24.4 950 490 1100

1260 5.3

9 ih.g.orb2 12.6

4.7 15.4 24.3 940 490 1100

1260 6.4

10 ih.g.orb3 13.1

5.1 15.9 24.0 940 500 1100

1250 4.6

11 ih.g.orb4 13.1

5.1 15.9 24.0 950 500 1100

1260 4.5

12 icehouse.glaciation.equatorial

1.8 -8.4 5.3 18.4 750 250 920 1090 24.4

13 pic1 13.8

8.4 16.1 25.4 990 760 1090

1330 6.8

14 pic2 13.4

7.5 15.8 25.3 980 740 1090

1320 7.1

380 381

381

382

Fig. 3. Mean temperature and precipitation analysis of the late Paleozoic climate simulations, as 383 labeled. Error bars mark the 2 σ confidence interval of the mean (σ=the standard error of the 384 mean). Case numbering matches that of Tables 1 and 3 and the in-text references. 385

386

386

Fig. 4. Precipitation sensitivity of the simulation ensemble over tropical Pangaea to given

387

forcings. The output comes from differencing mean annual precipitation between the simulations

388

indicated. Only sensitivities with amplitudes greater than twice the standard uncertainty

389

estimated from inter-annual variability are plotted. a, Due to greenhouse gas changes

390

(icehouse.highsealevel.highCO2-icehouse.highsealevel); b, Due to high-latitude land ice

391

(icehouse.glaciation.polar-icehouse.noglaciation); c, Due to low-latitude land ice

392

(icehouse.glaciation.equatorial-icehouse.glaciation.polar); d, Due to sea level rise

393

(icehouse.highsealevel-icehouse.noglaciation); e, Due to vegetation change

394

(greenhouse.noglaciation-icehouse.highsealevel.highCO2); f, Due to obliquity increase under

395

eccentric orbital conditions with perihelion at southern summer solstice (ih.g.orb3-ih.g.orb1); g,

396

Due to obliquity increase under eccentric orbital conditions with perihelion at northern summer

397

solstice (ih.g.orb4-ih.g.orb2); h, Due to precession at low obliquity (ih.g.orb2-ih.g.orb1). Land

398

areas are indicated with speckling.

399

400

400

4.1.2. Glaciation 401 402

The presence of land ice generally reduces precipitation. This effect is evident locally 403

(Fig. 5b-d) and in a global average sense (e.g. Fig. 3e, Case 6 vs. Cases 7 and 12). The effect of 404

high-latitude land ice on precipitation over tropical Pangaea is weak (Fig. 4b), but low-latitude 405

land ice changes the distribution as well as the intensity of precipitation in its vicinity (Fig. 4c). 406

The weakness of the effect of high-latitude ice on precipitation suggests that the uncertainties of 407

the high-latitude ice distribution in longitude (Section S2.4) are mostly irrelevant to the results. 408

In most of the simulations, the model resolves a rain shadow between the Ancestral Rocky 409

Mountains and the Alleghanian orogenic belt (~5° N, 160°, approximately modern Iowa, Illinois, 410

and Minnesota) that strongly contrasts with warm and wet conditions throughout the Central 411

Pangaean Mountains (Figs. 5a-c). However, this rain shadow mostly disappears when there is 412

significant ice on the Central Pangaean Mountains (Fig. 5d). The glaciated Central Pangaean 413

Mountains become arid, while areas to the north and south, including the former rain shadow 414

become wetter (Fig. 4c). 415

Land temperatures are lower in simulations with glaciation (e.g., 416

icehouse.glaciation.polar), particularly at the location of the land ice (Fig. 3b, Case 7 vs. Case 6 417

and Case 12; cf. Fig. 6b and Figs. 6c-d). At high latitudes, this effect results from greater 418

reflection of shortwave radiation by permanent ice in the summer months, which is partly 419

compensated by a reduction in outgoing longwave radiation due to cooler surface temperatures 420

(Fig. 7a). Without the ice sheet, shortwave reflection still remains high due to cloud cover (Fig. 421

7b), which decreases when the ice sheet is present (Fig. 7c). Although it is irrelevant to the 422

shortwave radiation budget in polar night, the imposition of high latitude land ice does not 423

change albedo during the winter (Fig. 7d). Polar Gondwanaland is snow-covered during this time 424

of the year. 425

At low latitudes, the cooling primarily results from increased shortwave reflection from 426

the ice due to its albedo throughout the year (Fig. 8d) as well as increased longwave emission 427

around the equinoxes (Fig. 8a). The increased shortwave reflection is mostly a surface rather 428

than a cloud effect (Fig. 8b). The increased longwave emission promotes a significant decrease 429

in high clouds at some seasons (Fig. 8c), which reduces re-emission of longwave radiation from 430

the upper troposphere. 431

Regardless of the effects of equatorial ice on the surface radiation budget, the equatorial 432

glacial configuration (ICEE) is unstable. (Note that the model is constrained to an ice-like albedo 433

even when ice is unstable.) Annual mean temperatures over the Central Pangaean Mountains are 434

only < 0° C over its highest (2000+ m) peaks (0° N, 170°) (Figure 6d). Summer (JJA for most of 435

the Central Pangaean Mountains) temperatures are even more unfavorable for glaciation. Lower 436

greenhouse gas levels, higher aerosol loading, or some other negative radiative forcing would be 437

necessary to keep land ice stable over the Central Pangaean Mountains (or on the islands on the 438

western side of Palaeo-Tethys), as also found by Soreghan et al. (2008b). This simulation 439

therefore is only useful for understanding the climate during the decay phase of equatorial 440

glaciation after the additional forcing is removed, which could significantly differ from the 441

climate under the additional forcing. 442

443 Fig. 5. Annual mean precipitation (mm) for the labeled simulations. Areas with precipitation less 444 than 1 mm are plotted as if precipitation there was 1 mm. 445

446

446 Fig. 6. Annual mean temperature (°C) for the labeled simulations. The red dashed line marks the 447 0° C isotherm. 448 449

450

J F M A M J J A S O N D−200

−150

−100

−50

0

MonthCha

nge

in N

et R

ad. a

t TO

A (W

m−2

)

Incoming ShortwaveOutgoing Longwave


−80

−60

−40

−20

0

20

Month

Clo

ud Im

pact

on

Net

SW R

ad. a

t TO

A (W

m−2

)

No Polar GlaciationPolar Glaciation


−60

−40

−20

0

20

Month

Cha

nge

in C

loud

ines

s (%

)

HighLow

J F M A M J J A S O N D0

0.2

0.4

0.6

0.8

1

Month

Albe

do

No Polar GlaciationPolar Glaciation

a b

c d

Fig. 7. Effects of high latitude glaciation on the radiation budget and cloud cover over the high

451

latitude Gondwanan ice sheet. The averaging is restricted to glaciated grid cells that are 100%

452

land. a, Mean difference in net radiation flux at the top of the atmosphere between simulations

453

with and without polar glaciation (icehouse.glaciation.polar-icehouse.noglaciation) (W m-2); b,

454

Mean difference between the full and clear-sky diagnostic shortwave radiation fluxes at the top

455

of the atmosphere (W m-2). No Polar Glaciation indicates icehouse.noglaciation and Polar

456

Glaciation indicates icehouse.glaciation.polar; c, Mean difference in fractional cloud cover (%)

457

at the top of the atmosphere between simulations with and without polar glaciation

458

(icehouse.glaciation.polar-icehouse.noglaciation). Note that this is an absolute, not relative

459

difference. High cloudiness is defined as cloud cover from the top of the atmosphere to 400 hPa.

460

Low cloudiness is defined as cloud cover from 1000–700 hPa; d, Mean direct shortwave albedo.

461

462

463

Fig. 8. Effects of high latitude glaciation on the radiation budget and cloud cover over the low

464

latitude ice in equatorial Pangaea. The averaging is restricted to glaciated grid cells that are

465

100% land.. a, Mean difference in net radiation flux at the top of the atmosphere (TOA) between

466


−100

−50

0

50

MonthCha

nge

in N

et R

ad. a

t TO

A (W

m−2

)

Incoming ShortwaveOutgoing Longwave


−80

−60

−40

−20

0

20

Month

Clo

ud Im

pact

on

Net

SW R

ad. a

t TO

A (W

m−2

)

Polar Glaciation OnlyEquatorial Glaciation


−40

−20

0

20

40

Month

Cha

nge

in C

loud

ines

s (%

)

HighLow

J F M A M J J A S O N D0

0.2

0.4

0.6

0.8

1

Month

Albe

do

Polar Glaciation OnlyEquatorial Glaciation

a b

c d

simulations with and without polar glaciation (icehouse.glaciation.equatorial-

467

icehouse.glaciation.polar) (W m-2); b, Mean difference between the full and clear-sky diagnostic

468

shortwave radiation fluxes at the top of the atmosphere (W m-2). Polar Glaciation Only indicates

469

icehouse.glaciation.polar and Equatorial Glaciation indicates icehouse.glaciation.equatorial; c,

470

Mean difference in fractional cloud cover (%) at the top of the atmosphere between simulations

471

with and without polar glaciation (icehouse.glaciation.equatorial-icehouse.glaciation.polar).

472

High cloudiness is defined as cloud cover from the top of the atmosphere to 400 hPa. Low

473

cloudiness is defined as cloud cover from 1000–700 hPa. Note that this is an absolute, not

474

relative difference; d, Mean direct shortwave albedo.

475

476 477 4.1.3 Sea level 478 479

Higher sea level lowers surface temperatures, particularly on land (Fig. 3b, Case 5 vs. 480

Case 6). Higher sea level slightly increases precipitation globally, but primarily changes how 481

precipitation is partitioned between land and ocean. The unglaciated icehouse simulation with 482

high sea level (icehouse.highsealevel) has global annual precipitation 20 mm greater than the 483

equivalent simulation with low sea level (icehouse.noglaciation) (Table 3, Cases 5 and 6), with 484

all of the increase due to higher precipitation over land. In tropical Pangaea, this increase is most 485

significant in southeastern equatorial Pangaea, ~ 1000 km inland from the coast (Fig. 4d). As 486

hypothesized, precipitable water does increase over land when sea level rises. In southeastern 487

equatorial Pangaea, precipitable water is up to 10 mm higher in January, which is large in 488

comparison to the average changes elsewhere or at other seasons (Fig. 9a). High cloudiness 489

associated with convection increases in tropical Pangaea in southern hemisphere summer (Fig. 490

10c), preferentially reducing outgoing longwave radiation (Fig. 10b). Flooding in polar 491

Gondwanaland results in increased low cloudiness in summer (Fig. 10d), which results in greater 492

reflection of shortwave radiation (Fig. 10a). This effect is the principal explanation why land 493

temperatures cool when sea level rises but is of limited relevance to deglaciating climates, 494

because the effect of the retreating ice sheet on cloudiness is neglected. 495

496

Fig. 9. Difference in precipitable water (mm) due to a 100 m sea level rise

497

(icehouse.highsealevel-icehouse.noglaciation) in the labeled months.-

498

499

500

Fig. 10. Mean difference in radiation and cloud diagnostics due to a 100 m sea level rise

501

((icehouse.highsealevel-icehouse.noglaciation). Four types of averages are presented: Global

502

(global), Land (over all 100% land grid points); Tropical Pangaea (over all 100% land grid points

503

between 146.25º–191.25º longitude and 30° N and 30° S); and Polar Gondwanaland (over all

504

100% land grid points south of 60° S); a, Net shortwave radiation flux at the top of the

505

atmosphere (W m-2); b, Net longwave radiation flux at the top of the atmosphere (W m-2); c,

506

High cloud cover (%). Note that this is an absolute, not relative change; d, Low cloud cover (%).

507

508

4.1.4 Vegetation 509 510

Model sensitivity to vegetation change can be estimated by differencing the standard 511

greenhouse simulation and the high greenhouse gas level/high sea level icehouse simulation 512

(greenhouse.noglaciation and icehouse.highsealevel.highCO2). A change in vegetation scenario 513

from V250 to V2500 results in a global average warming of ~1.5° C and 50 mm more 514

precipitation over land (Table 3). Over tropical Pangaea, there is little significant precipitation 515

sensitivity to vegetation change (Fig. 4e). Precipitation increases ~100 mm about 15° N, a 516

change which cannot be definitively attributed to local biome change (Figs. 2a–b). Local biome 517

change does have a strong effect on precipitation in parts of the sub-tropics. Annual mean 518

precipitation increases by ~500 mm in southeast Pangaea (30° S, 210°) (Fig. 4e), where 519

xerophytic shrubland in V250 is replaced by tropical broadleaf evergreen forest in V2500 (Figs. 520

2a–b). 521

522

4.1.5 Orbital variability 523 524

Temperature variability among the orbital sensitivity simulations and the simulation from 525

which it is derived (icehouse.glaciation.polar) approximates the estimated systematic bias due to 526

the equilibration criteria (e.g., Fig. 3a, Cases 7-11). The same is true for precipitation (e.g., Fig. 527

3b). Nevertheless, orbital forcing affects low-latitude Pangaean precipitation, particularly the 528

extent and aridity of the sub-tropical deserts (Figs. 4f–h and 11a–d). The longitude of perihelion 529

exerts a stronger effect than obliquity (compare Figs. 4f and 4h), suggesting summer insolation 530

drives tropical precipitation. 531

A closer analysis based on differencing simulations suggests that the effects of obliquity 532

on precipitation in tropical Pangaea are weak. They are most consistent in the northern 533

hemisphere at ~ 15° N (Figs. 4f–g), where higher obliquity results in higher precipitation. Higher 534

obliquity also slightly reduces precipitation off the equatorial east coast. 535

The orbital variability simulations also sample the effects of the precession of the 536

longitude of perihelion, which enhances or mutes the amplitude of seasonal insolation variability, 537

when eccentricity is non-zero. Focusing on low obliquity, which favors the development and 538

maintenance of high-latitude ice, higher insolation at northern summer solstice enhances 539

precipitation north of the Equator and reduces it south of the Equator (Fig. 4h). The effect 540

reverses for the opposite longitude of perihelion. This sensitivity is particularly strong along the 541

northwest coast of tropical Pangaea, which corresponds to the Donets Basin and the Russian 542

Platform (Fig. 4h). 543

Near the equator and close to the sub-tropics, the seasonality of precipitation is consistent 544

between the orbital forcing simulations shown in Fig. 11 and their corresponding circular orbit 545

simulation (icehouse.glaciation.polar). Precipitation peaks in summer and fall and diminishes in 546

winter (e.g., Fig. 12b). Even at 5° N, relatively wet for most of the year, DJF precipitation 547

averages no more than 30 mm (Fig. 12a). Summer precipitation is enhanced relative to a circular 548

orbit simulation in the sub-tropics and at the equator, when perihelion is at summer solstice and 549

diminishes when perihelion falls at winter solstice (e.g., Fig. 12b). The relative precipitation 550

change attributable to this effect is greater in the sub-tropics than at the equator, but the absolute 551

change is greater (cf. Fig. 12a vs. 12b). In a classification scheme like that of Rea (1994), the 552

difference equates to change between sub-humid and humid conditions at 5° N and between 553

hyperarid and arid conditions at 20° N. This seasonality of precipitation and the enhancement of 554

precipitation under higher summer insolation characterizes a monsoonal climate, whose 555

dynamics will be investigated in the next section (Section 4.2). 556

557

557 Fig. 11. Annual mean precipitation (mm) for the labeled simulations. Areas with precipitation 558 less than 1 mm are plotted as if precipitation there was 1 mm. Comparing the left and right set of 559 panels demonstrates longitude of perihelion sensitivity, while comparing the top and bottom sets 560 demonstrates obliquity sensitivity. 561

562

562 Fig. 12. Comparison of the seasonal variability of Pangaean sector (146.25º–191.25º longitude) 563 non-water area (sum of ocean and lake fraction is equal to 0%) precipitation in the icehouse 564 simulation that uses the polar ice configuration (icehouse.glaciation.polar) (labeled “CIRC”) and 565 the orbital parameter sensitivity experiments derived from it (ih.g.orbx) (labeled “ORBX”). The 566 sub-panels correspond to the latitudes labeled. Note that the markers are structured so that higher 567 obliquity is colored red, lower obliquity is colored blue, and the triangle points in the direction of 568 the hemisphere of summer perihelion. 569 570 571 4.2 Monsoonal variability 572 573

A monsoon is a seasonally reversing wind system. In the case of the late Cenozoic South 574

Asian Monsoon, monsoonal flow during summer displaces the ITCZ poleward in a narrow 575

longitudinal band, resulting in higher precipitation in some areas closer to the summer 576

hemisphere tropic. This monsoon is associated with a strong cross-equatorial westerly jet at 577

~1500 m (850 hPa) above the surface (Bunker, 1965; Joseph and Sijikumar, 1998; Webster and 578

Fasullo, 2003). This feature appears, albeit poorly resolved, in climate models and re-analysis 579

data (Meehl and Arblaster, 2003; Meehl et al., 2006), so the difference in zonal wind velocity in 580

DJF MAM JJA SON0

50100150200250300350400450500

Prec

ip. (

mm

)Pangaean Precip. at 5 N

CIRCORB1ORB2ORB3ORB4

DJF MAM JJA SON

10−1

100

101

102

103

Prec

ip. (

mm

)

Season

Pangaean Precip. at 20 N

DJF MAM JJA SON0

100200300400500600700800900

1000

Prec

ip. (

mm

)

Season

Pangaean Precip. at 5 S

DJF MAM JJA SON

10−1

100

101

102

103

Prec

ip. (

mm

)

Season

Pangaean Precip. at 20 S

a b

c d

the lower troposphere (850 hPa) between the equator and the summer hemisphere tropic is a 581

good proxy for monsoon intensity (Wang et al, 2001). 582

Seasonally reversing wind systems that resemble the South Asian Monsoon are present in 583

the late Paleozoic climate simulations (Fig. 13). In the ordinary greenhouse simulation 584

(greenhouse.noglaciation), a monsoonal jet over Pangaea is apparent in both summer 585

hemispheres (Figs. 13a and 13e), which marks this monsoon as the classic “megamonsoon” of 586

Kutzbach and Gallimore (1989) wherein monsoonal circulation occurs in both hemispheres at the 587

same longitude. Westerly flow near the Equator prevails in the South Indian Ocean during 588

southern summer, but is much weaker than in the opposite hemisphere during its summer 589

(Turner and Annamalai, 2012). 590

The monsoonal jet diminishes in the analogous icehouse simulation 591

(icehouse.noglaciation) (Figs. 13b and 13f). A distinct megamonsoon develops over Paleo-592

Tethys in both simulations (cf. Figs. 13a and 13b; Figs. 13e and 13f). This circulation has 593

minimal impact on tropical Pangaean climate, so it will not be discussed further. 594

A simple monsoon circulation index serves to measure variability in the intensity of the 595

Pangaean monsoon among simulations: the difference between the summer (JJA/DJF) mean 850 596

hPa zonal wind at 9° N/S latitude and mean 850 hPa zonal wind at 24° N/S latitude over the 597

longitude range of 146.25º–191.25º. This index is analogous to those defined by Wang et al. 598

(2001), which use low-level zonal wind shear as an integrative proxy for convective heating, 599

moisture convergence, and rainfall associated with a given monsoon. 600

As measured by this index, greenhouse gas levels and the presence of ice sheets set 601

background monsoon intensity in a climate under circular orbital conditions (Fig. 14a, 602

greenhouse.base vs. icehouse.base, icehouse.glaciation.equatorial vs. 603

icehouse.glaciation.polar). Changing orbital forcing can significantly enhance or diminish 604

monsoon intensity (e.g., Figs. 14a-b, icehouse.glaciation.polar vs. the ih.g.orb simulations). 605

High summer insolation strengthens the monsoon, while low summer insolation suppresses it 606

(Fig. 14a, ih.g.orb2 vs. ih.g.orb1, ih.g.orb4 vs. ih.g.orb3). Unlike the monsoons in the 607

simulations under GG2500 greenhouse gas levels (including icehouse.highsealevel.highCO2), 608

the icehouse monsoons excited by orbital forcing are not megamonsoons, since the northern 609

hemisphere monsoon intensifies whenever the southern hemisphere monsoon diminishes and 610

vice versa (Figs. 14a–b, ih.g.orb1 vs. ih.g.orb2). 611

The monsoon circulation indices correlate significantly with annual precipitation at 20° N 612

and 20° S (Figs. 14c–d), especially if the simulation with equatorial glaciation 613

(icehouse.glaciation.equatorial) is excluded as an outlier. Such exclusion is justified, because 614

precipitation in the simulations with equatorial glaciation at 20° N and 20° S exceeds by an order 615

of magnitude that in simulations with comparably weak monsoons (Figs. 14c–d). In addition, the 616

precipitation regime in a rain shadow area in west-central equatorial Pangaea radically changes 617

in these simulations (Section 4.1.3). Therefore, some mechanism in 618

icehouse.glaciation.equatorial produces high amounts of sub-tropical precipitation in the 619

absence of a classical monsoonal circulation (see Section 4.3). 620

Changes in monsoonal intensity exert the leading control on tropical Pangaean 621

precipitation in climates with global mean temperatures similar to the present day (the 622

simulations with GG250 and/or ICEP). A stronger monsoon draws the ITCZ over the northern 623

tropics (rather than displacing it from the Equator), while a weaker monsoon focuses it near the 624

Equator. This effect can be illustrated by meridional streamfunction analysis over Pangaea, in 625

which the monsoon manifests as a seasonally varying cross-equatorial meridional cell (excited 626

monsoon with a cross-equatorial cell and/or weakened summer hemisphere cell: Figs. 15b–c; 627

suppressed monsoon without such a cell: Figs. 15a and 15d). 628

629

630 Fig. 13. Seasonal variability in mean zonal winds at 850 hPa (m/s) in two simulations with 631 strong (greenhouse.noglaciation) and weak (icehouse.noglaciation) Pangaean megamonsoons, as 632 labeled. 633 634

635

635 Fig. 14. a, Inter-simulation comparison of Northern Hemisphere monsoon index: difference 636 between mean JJA Pangaean sector (146.25º–191.25º longitude) 850 hPa zonal winds at 9 º N 637 and 24 º N, dots indicate 2 s.e. range within the simulation; b, as in Fig. 8a. but for the Southern 638 Hemisphere; c, Northern Hemisphere monsoon index vs. zonal average mean annual 639 precipitation over non-water areas of the Pangaean sector at 20º N, linear correlation coefficient 640 (excluding icehouse.glaciation.equatorial) and p-value between monsoon index and log10 641 precipitation are displayed at the top of the panel; d, as in Fig. 14c., but for the Southern 642 Hemisphere. 643 644

645

645 Fig. 15. Seasonal mean zonal average meridional streamfunction in the Pangaean sector 646 (146.25º–191.25º longitude). Contour frequency is 5 × 1010 kg·s-1. The plotting convention 647 follows the left-hand rule; positive streamfunction therefore indicates clockwise meridional flow. 648 649 650 4.3. The tropical circulation and precipitation regime under deep icehouse conditions 651 652

653

The precipitation regime and tropical Pangaean atmospheric circulation in the most 654

glaciated simulation (icehouse.glaciation.equatorial) differ from other simulations (Section 4.1). 655

Annual mean precipitation near the tropics of Cancer and Capricorn in this simulation 656

approximates annual mean precipitation in simulations with strong monsoons (Figs. 5d and Figs. 657

14c–d). The monsoon index, however, is low (Figs. 14a–b). 658

Winds at 850 hPa at ~15º latitude in the summer hemisphere are westerly in strong 659

monsoon simulations (left panels in Fig. 13) and strongly easterly in weak monsoon simulations 660

(right panels in Fig. 13). Yet when the Central Pangaean Mountains are glaciated, strong 661

meridional flow from the mountains toward the summer hemisphere dominates weak easterly 662

zonal wind flow at 15º (Figs. 16d and 17d). In strong monsoon simulations, similar meridional 663

flow is present but not as longitudinally widespread (Fig. 17a). By analogy with the monsoon 664

circulation, these winds may drive the equatorial band of heavy rainfall (the ITCZ) farther into 665

the summer hemisphere (Figs. 13d and 13h). Consistent with this explanation, the summer 666

hemisphere tropical meridional overturning cell over Pangaea occurs farther poleward and/or is 667

stronger than in the simulations with polar glaciation and suppressed monsoons (cf. Figs. 18a and 668

18b; Figs. 18c and 18d). Over the Central Pangaean Mountains themselves, precipitation is very 669

low and convection is inhibited (Figs. 8c, 16d, 17d). The result is a vastly different climate than 670

any other simulated, in which the sub-tropics are anomalously wet, while the Equator is 671

anomalously dry, such that hyperarid to semi-arid conditions are widespread from 0°–20° N (Fig. 672

5d). 673

674

675

Fig. 16. Seasonal mean (DJF) convective precipitation (mm) and winds (m·s-1) in the labeled 676 simulations. 677 678

679

Fig. 17. Seasonal mean (JJA) convective precipitation (mm) and winds (m·s-1) in the labeled 680 simulations. 681 682

683

Fig. 18. Seasonal mean zonal average meridional streamfunction in the Pangaean sector 684 (146.25º–191.25º longitude), as labeled. Plotting conventions are identical to those of Fig. 15. 685 686

687

687 5. Discussion 688 689 5.1. Overview 690 691

692

In this section, the implications and robustness of these simulations for understanding 693

tropical Pangaean climate during the LPIA are addressed (see Section 2): (1) cycles at less than 694

100 kyr timescales, which are attributed to orbital forcing of the Pangaean “megamonsoon” 695

system (Section 5.2); (2) the phasing of precipitation variability between glacials and 696

interglacials (Section 5.3); and (3) the possibility that the Pangaean tropics were exceptionally 697

cool and arid during some intervals of the LPIA, which led to the deposition of dust/loess near 698

the Equator (Section 5.4). 699

700 701 5.2. The changing of the monsoons 702 703 704

Changes in the Pangaean monsoon are the primary control on the spatial pattern of 705

tropical precipitation variability within most of the model simulations, and thus of tropical 706

precipitation away from the Equator. The idea that the Pangaean monsoon explains the strong 707

seasonality in tropical climate evident in the sedimentary record from Carboniferous to Jurassic 708

time is well established (see discussions in Kutzbach and Gallimore, 1989; Parrish, 1993). The 709

results of this investigation add the following two refinements. 710

First, the Pangaean megamonsoon weakens under and in proportion to colder global 711

climate (Figs. 14a–b). Thus, the intensity of the monsoon and its effects on the sedimentary 712

record would have tracked the large pCO2 changes on 5 Myr timescales reported by Montañez et 713

al. (2007). Indeed, if tropical temperatures during the LPIA were ever ~7º C cooler than present 714

day (consistent with some interpretations of paleotemperature proxies: Powell et al., 2009; Giles, 715

2012; as well as the hypothesis of upland glaciation: Soreghan et al., 2008a,b, 2009; G.S. 716

Soreghan et al., 2014), monsoonal circulation would have collapsed. This result comes with a 717

caveat: the simulations presented here cluster into three groups in global temperature, so only 718

three data points in monsoon intensity vs. global temperature are sampled. 719

Second, seasonal insolation variability when the Earth’s orbit was eccentric would have 720

been sufficient to strengthen a cold climate monsoon to warm climate intensity by changing 721

cross-equatorial thermal contrast (Figs. 14a–b). However, this mechanism strengthens the 722

summer monsoon only in the hemisphere where perihelion occurs during summer. The summer 723

monsoon in the other hemisphere is suppressed, just as in the coldest climate simulated (Figs. 724

14a–b). Strata from tropical Pangaea that record monsoonal precipitation or winds should record 725

strong variability at Milankovitch frequencies associated with variations in eccentricity as well as 726

precession. And indeed, this phenomenon is observed in a variety of settings as cyclicity on 727

timescales less than 100 kyr (e.g., Olszewski and Patzkowsky, 2003; Tabor and Poulsen, 2008; 728

M.J. Soreghan et al., 2014). Cycles of this frequency are also observed in non-glacial climates 729

(Olsen and Kent, 1996), but the weakening of the monsoon in cold, but not glacial climates 730

(icehouse.noglaciation) could also explain this. 731

One major concern with the robustness of these results is that all of the modeling 732

experiments (including the orbital sensitivity experiments) likely have inputs that are out of 733

equilibrium with another. For example, the high latitude ice sheets in the ICEP configuration 734

may have expanded or contracted over periods much longer than the integration time of the 735

model. Indeed, ICEP is not definitively in equilibrium with the climate in 736

icehouse.glaciation.polar, nor any other simulation. It is an estimate of where snow would 737

accumulate based on snowfall in another simulation (Section S2.4). At the most, it is cold 738

enough not to melt (Fig. 6c). As another example, the circular orbit and orbital sensitivity 739

modeling experiments only use two fixed vegetation scenarios. Using fixed scenarios simplifies 740

attribution of temperature and precipitation sensitivity to particular mechanisms and conserves 741

model integration time, but it neglects a variety of possible feedbacks that would occur if ice 742

sheets and vegetation were allowed to dynamically adjust to the climate. However, the sensitivity 743

experiments do allow identification and partial quantification of these feedbacks, so that 744

responses to large changes in forcing can shed light on the dynamics that control the effects of 745

uncertainties in those forcings, as when trying to explain why surface temperatures cool in 746

response to ice sheets (Section 4.1.2). The results are therefore robust for the specific boundary 747

condition set used, but that boundary condition set is not self-consistent. 748

All of the monsoonal dynamics identified in this study are partly known from previous 749

modeling studies, which lend confidence to the robustness of the modeling results. Peyser and 750

Poulsen (2008) simulated intensification of the Pangaean monsoon with increased pCO2. Horton 751

et al. (2012) simulated orbital forcing of the Pangaean monsoon in LPIA climate experiments 752

with GENESIS that used dynamically adjusted glaciation and vegetation. Orbital forcing of 753

monsoons has been widely simulated in late Cenozoic climate (see discussion by Kutzbach et al., 754

2008) and recently explored in idealized modeling in a series of papers by Merlis et al. 755

(2013a,b,c). 756

Therefore, dynamically consistent monsoonal circulations during LPIA time can be 757

reproduced in a variety of modeling frameworks. Here at least, modeling has come to a 758

consensus that enables testing against the geological record. 759

760 5.2. Glacial aridity or glacial humidity? 761

762 763

In a climate where land ice was primarily polar (Isbell et al., 2003a,b), the results 764

presented here suggest that glacial–interglacial variability in tropical precipitation would have 765

been driven primarily by the indirect effects of glacial growth and collapse: i.e., (1) greenhouse 766

gas level changes driven by carbon cycle changes, and (2) sea level changes attributable to shifts 767

in glacial volume and loading. The direct effect of such polar ice on tropical precipitation is 768

small and scattered (Fig. 4b) in comparison to predicted increases in tropical precipitation due to 769

greenhouse gas increases (Fig. 4a) or increased precipitation in areas ~1000 km from the coast 770

due to sea level rise (Fig. 4d). 771

Because greenhouse gas levels and sea level decreased during glacials, the simulations 772

suggest that glacials were drier than interglacials in most areas, as both relate to moisture 773

availability. Higher pCO2 produces higher surface temperatures and thus greater evaporation. 774

Higher sea level produces greater evaporation in inundated regions, and thus reduced isolation of 775

continental interiors from moisture (Figs. 9a–d). Under a late Cenozoic-like glacial–interglacial 776

interval, sea level effects would predominate. The amplitude of pCO2 change would be ~100 777

ppmv and thus have proportionally smaller effects than shown in Fig. 4a, while sea level change 778

would be similar to that considered in Fig. 4d. 779

As Horton et al. (2012) noted, however, the orbital forcing that drives ice sheet changes 780

at high latitudes also affects the monsoon at low latitudes. In addition, interactions between the 781

monsoon and higher precipitation due to sea level rise seem possible. Higher precipitation is 782

strongly focused in southeastern tropical Pangaea (Fig. 4d). In this region, the increase in 783

precipitable water is highest in southern summer (Fig. 9a), which is the optimal season for 784

monsoonal precipitation. Thus, aliasing between orbitally forced glacial–interglacial variability 785

and orbitally forced monsoons discussed in Section 5.1 could result in wetter glacial intervals in 786

specific regions during which precipitation was more strongly controlled by surface convergence 787

than moisture availability (as in present day Florida). These simulations alone cannot untangle 788

these competing effects, but additional modeling could. 789

There are practical implications for this aliasing. Eros et al. (2012) inferred glacial-stage 790

humidity from the Upper Palaeozoic record of the Donets Basin (far eastern equatorial Pangaea). 791

In the simulations presented here, precipitation in this region is relatively insensitive to sea level 792

change, since moisture availability is probably not a control on precipitation in an east coastal 793

area along the path of the easterlies. Note that high values of precipitation sensitivity to sea level 794

are displaced far from the eastern Pangaean coastline (Fig. 4d), suggesting that sea level change 795

preferentially affects precipitation in more inland areas rather than the coastal areas actually 796

flooded or drained. 797

Thus, regardless of sea level, the prevailing winds would have been laden with moisture 798

evaporated from Paleo-Tethys. Precipitation instead would be controlled by the location of the 799

convergent flow (and thus the mechanical forcing necessary to drive moist convection). Eastern 800

equatorial Pangaea is particularly sensitive to this effect (Fig. 4h). As validated by Horton et al. 801

(2012) for late Paleozoic conditions, ice sheet ablation and sea level rise would take place under 802

an eccentric orbit with longitude of perihelion around southern summer solstice. These are the 803

very same conditions that favor a strong summer monsoon over the Donets Basin. Thus, strata in 804

the Donets Basin would carry the signal of extremely wet conditions just before transgression, 805

which would appear to be a response to the maximum extent of glaciation but actually would be 806

a response to precession forcing of the monsoon. Validation of this idea is possible: The model 807

ensemble predicts that precession-driven precipitation variability in the Russian Platform (to the 808

north of the Donets Basin) operated in the opposite phase to the Donets Basin (Fig. 4h). 809

810 5.3. Central Pangaean Mountain Glaciation: A Pathway to Equatorial Aridity and Dust 811 Deposition 812 813

814

In the most glaciated simulation (icehouse.glaciation.equatorial), glaciation of the 815

Central Pangaean Mountains suppresses precipitation over the Central Pangaean Mountains but 816

generates strong winds that flow from the Central Pangaean Mountains into the summer 817

hemisphere. These winds displace the ITCZ poleward, enhancing subtropical precipitation, even 818

when the monsoonal circulation is weak. The result is that conditions most favorable to dust 819

lifting (annual precipitation <100-200 mm) are widespread at the Equator (Fig. 5d) in contrast to 820

humid conditions in all other simulations (Figs. 5a–c) or in other modeling studies (e.g., Peyser 821

and Poulsen, 2008; Horton et al., 2012). In this climate, it is easy to explain the occurrence of 822

weakly chemically weathered equatorial dust/loess sources from the Central Pangaean 823

Mountains (Soreghan et al., 2008; M.J. Soreghan et al., 2014). 824

The strong meridional winds that flow from the mountains to displace the ITCZ 825

poleward, like the monsoonal circulation, are most likely driven by large-scale thermal contrast. 826

In this case, however, the contrast is not between the winter and summer sub-tropics but between 827

the cold, glaciated Central Pangaean Mountains and the strongly heated non-glaciated land 828

surface near the summer hemisphere tropic. This circulation is confined to a single hemisphere, 829

since the glaciated Central Pangaean Mountains are located at the Equator. If this explanation is 830

correct, the glaciated Central Pangaean Mountains, if displaced to the south of the Equator (as in 831

Pennsylvanian time), would enhance the thermal forcing driving the northern hemisphere 832

monsoonal circulation and weaken the thermal forcing driving the southern hemisphere 833

monsoonal circulation. The ensemble of simulations does not allow resolution of the contribution 834

of the Angaran and Gondwanan ice sheets or the evolution/exact configuration of the Central 835

Pangaean Mountains to tropical circulation and precipitation changes. 836

Dynamics aside, the icehouse.glaciation.equatorial simulation cannot be the final word 837

on the subject. As noted in Section 4.1.2, it only considers the effects of glaciating the equatorial 838

mountains. The ice there is unstable; suggesting an additional source of negative radiative 839

forcing (such as very low pCO2 or high aerosol optical depth) would be necessary to allow 840

equatorial glaciation. It is not obvious whether this additional forcing would increase or decrease 841

aridity. If tropical upland glaciation cannot be validated or its forcing is inconsistent with 842

equatorial aridity, the alternative hypothesis would be that mobilization of dust/loess sourced 843

from the Central Pangaean Mountains occurred at subtropical latitudes or in the double rain 844

shadow at ~5° N, 160° common to most simulations. 845

846

847

6. Summary 848 849

850

Using a fully coupled atmosphere-ocean-land-sea ice model, climate during the Asselian 851

and Sakmarian Stages of the Early Permian was simulated to test various mechanisms that could 852

have contributed to precipitation variability over tropical Pangaea during the LPIA. 853

The results of these simulations support the idea that equatorial Pangaea and the Earth 854

generally were drier when ice sheets were bigger relative to times when ice sheets were smaller. 855

Glacial aridity in equatorial Pangaea largely relates to sea level fall, which deprived continental 856

interiors of important moisture sources for precipitation. If glaciation of the Central Pangaean 857

Mountains occurred, the resulting suppression of precipitation in their vicinity would have led to 858

more extreme glacial aridity in equatorial Pangaea, which could explain widespread tropical 859

dust/loess deposition during the LPIA. However, these simulations do not explore values of 860

glacial extent intermediate between polar and equatorial glaciation, so it is possible that some ice 861

sheet configurations might result in wetter conditions for equatorial Pangaea. 862

No matter the extent of continental glaciation, the precipitation regime in the Pangaean 863

tropics away from the Equator would have been strongly affected by monsoonal variability 864

forced by the precession cycle, confirming the main results of Horton et al. (2012) in a fully-865

coupled model with different physical parameterizations of cloud processes, but without dynamic 866

adjustment of glaciation and vegetation to climate. Monsoonal variations driven by precessional 867

forcing is therefore the likely explanation for sub-100 kyr cyclicity evident in some sedimentary 868

deposits during this period. Indeed, the dynamics of the Pangaean megamonsoon in simulations 869

of late Paleozoic climate appear robust enough for rigorous testing against long-term 870

sedimentary records hypothesized to have strong monsoonal sensitivity. 871

Plausible rhythms of orbital forcing and high-latitude glaciation also could explain the 872

appearance of more humid conditions during glacials (lowstands) in particular areas of equatorial 873

Pangaea. In some parts of the tropics, glacial-interglacial variability would have been primarily 874

driven by glacioeustatic sea level rise (leading to wetter conditions) and glacioeustatic sea level 875

fall (leading to drier conditions). In other parts of the tropics, precipitation would have been 876

relatively insensitive to changes in sea level, which is plausible if the limiting factor on moist 877

convection in these areas was mechanical forcing (e.g., convergent flow) rather than moisture 878

availability, as in other areas. In those cases, glacial–interglacial precipitation variability would 879

be controlled by the dynamics of the monsoon. Because the same orbital configuration that 880

results in initial glacial retreat also results in a strong southern summer monsoon, areas sensitive 881

to this monsoon would be wettest in the strata associated with the glacial maximum. Caution 882

therefore should be exercised when extrapolating outcrop and regional-scale climate variability 883

to climate variability in the tropics as a whole. 884

885

Acknowledgments 886

N.G. Heavens thanks J.T. Kiehl (NCAR) for assistance in designing and implementing the model 887

simulations. The authors also thank the Editor and two anonymous reviewers for their 888

constructive comments on the initial draft of this manuscript. This work was supported by the 889

National Science Foundation: EAR-0745961, EAR-0932946, and EAR-1003509 to N.M. 890

Mahowald; EAR-0746042 to G.S. Soreghan and M.J. Soreghan; EAR-1338331 to G.S. 891

Soreghan, and EAR-1337363 to N.G. Heavens. High-performance computing support was 892

provided by NCAR's Computational and Information Systems Laboratory, sponsored by the 893

National Science Foundation. 894

895

Appendix A. Supplementary data 896 897

Supplementary Information 897 898 899 S1. Paleogeographic base map 900 901

The template for the input paleogeography was the Early Permian (280 Ma) 902

reconstruction of R. Blakey (2008, n.d.), hereafter B280. B280 was obtained as an image file 903

without digital elevation information. The input paleogeography follows B280 with respect to 904

coastlines and some aspects of bathymetry, such as the placement of trenches along subduction 905

zones (bathymetry of -6000 m). A hypothetical spreading center in the Panthalassic Ocean in 906

B280 was not included, since there is little consensus from other sources about where spreading 907

centers were located in the Panthalassic (see Ziegler et al., 1997). 908

Most of the Panthalassic Ocean is given flat bottom bathymetry of -4000 m. The other 909

major ocean basin, the Paleo-Tethys Sea, is given flat bottom bathymetry of -3250 m. B280 910

suggests Palaeo-Tethys was shallower than the Panthalassic, and Celâl Sengör and Atayman 911

(2009) have proposed that the abyssal regions of Palaeo-Tethys and the Panthalassic were 912

isolated as early as the Carboniferous. The seaway between Angara and northern Pangaea is 913

given flat bottom bathymetry of -2500 m on its northwestern end, which gradually shallows 914

toward the southeast. This bathymetry is broadly consistent with B280 and the arguments of 915

Gradstein et al. (2004) that this seaway was closed by the beginning of Pennsylvanian time. 916

Furthermore, B280 and Scotese et al. (1997) maps the subduction of the Panthalassic plate 917

underneath this seaway, suggesting the oceanic lithosphere of the seaway is warmer, less dense, 918

and thus thicker (shallower bathymetry) than the Panthalassic plate. A similar rationale justifies -919

2500 m flat bottom bathymetry for the seaway to the south of Palaeo-Tethys. 920

The mapping of shallower bathymetry near coastlines generally follows B280. The least 921

deep bathymetry likely was ephemerally flooded during glacioeustatic cycles, so it is given a 922

topography of 50 m. B280 maps a shallow interior sea in Gondwanaland between modern South 923

America and Africa. There is evidence for marine faunal exchange between the Parana Basin of 924

Brazil and the Karoo Basin of South Africa in the Early Permian that ends by Late Permian time 925

(Ziegler et al., 1997), so the base map follows B280. 926

927

Fig. S1. a–c, Digital elevation (m) models for the Earth for the present day (0 Ma) (NOAA 928 NGDC, 2006); the Asselian-Sakmarian (299-284 Ma) (this study); and the Changhsingian (251 929 Ma) (Kiehl and Shields, 2005); d, Distribution of topography with altitude for the digital 930 elevation models in a–c and the Asselian-Sakmarian map with the polar glacial configuration 931 (ICEP). 932 933

The mapping of topography is a compromise between B280 and the ice-free topographic 934

reconstruction during Asselian (292 Ma) time of Rowley (2008), hereafter R292. Areas mapped 935

in R292 as having topography of 0 m, 200 m, 1000 m are given topography of 100 m, 200 m, 936

and 1500 m respectively. An additional category of 600 m terrain is mapped on the basis of a 937

−6000−4000−2000 0 2000 4000 6000109

1010

1011

1012

1013

1014

1015

Altitude (m)

Area

in B

in (m

2 )

Distribution

ModernA−SChA−S+ICEB

Longitude

Latit

ude

Modern

0 60 120 180 240 300 360−90

−60

−30

0

30

60

90

LongitudeLa

titud

e

Asselian−Sakmarian ~294 Ma

0 60 120 180 240 300 360−90

−60

−30

0

30

60

90

Longitude

Latit

ude

Changsinghian ~251 Ma

0 60 120 180 240 300 360−90

−60

−30

0

30

60

90

−10000

−8000

−6000

−4000

−2000

0

2000

4000

6000a b

c d

base map for the Changhsingian created by D. Rowley and used by Kiehl and Shields (2005), 938

hereafter KS251. R292 and B280 disagree somewhat in the placement of high terrain. Mountain 939

ranges in B280 also tend to be wider than R292. The base map generally follows B280 with 940

respect to these matters, though all high terrain can be approximately identified with a 941

topographic feature described by Ziegler et al. (1997). Following Slingerland and Furlong 942

(1989), the core of the Central Pangaean Mountains (approximate 2000 m contour of R292) are 943

mapped at an altitude of 4000 m, which is comparable to the modern central Andes. Finally, a 944

box average filter was used to smooth the base map at 3.5º x 3.5º resolution to obtain reasonable 945

slopes. 946

Figs. S1a–c compares three digital elevation models (DEMs): a DEM for the present-day 947

Earth (NOAA NGDC, 2006) deformed to 0.5º x 0.5º, a resolution identical to the other base 948

maps, and KS251. Fig. S1d compares the distributions of topography and bathymetry for the 949

three DEMs. Note that both the base map and KS251 are highly unrealistic, especially in 950

distributing deep bathymetry and high topography. While the present-day DEM used includes 951

topography due to glaciation on Greenland and Antarctica, serious discrepancies between the 952

Modern DEM and the Early Permian DEM remain even if topography due to land ice is added 953

(Fig. S1d). 954

S2. Simulation scenarios 955 956

S2.1. Greenhouse gas scenarios 957 958

The two greenhouse gas scenarios used (Table S1) span the range of pCO2 inferred by 959

Montañez et al. (2007) for Asselian-Sakmarian time. Glacial-interglacial variability in pCO2 and 960

pCH4 during Plio–Pleistocene time was strongly correlated, so pCH4 levels in the greenhouse gas 961

scenarios were adjusted in proportion to pCO2 levels according to a linear relation derived from 962

the Vostok ice core data of Petit et al. (1990). There is certainly glacial-interglacial variability in 963

pN2O (Petit et al., 1990), but it is less well correlated with CO2, so pN2O was set to its pre-964

industrial value. While the pCH4 in the GG2500 climate scenario may be excessively high, the 965

uncertainties in radiative forcing due to uncertainties in the reconstructed values of pCO2 during 966

the Asselian-Sakmarian (e.g., Montañez et al., 2007) exceeds those that arise from uncertainty in 967

the levels of the other greenhouse gases. 968

Table S1. 969

Greenhouse gas scenarios 970

Scenario pCO2 (ppmv) pCH4 (ppbv) pN2O (ppbv)

GG2500 2500 5570 311

GG250 250 530 311

971

Using the first and third equations for CO2 and the equation for CH4 in Table 6.2 of IPCC 972

(2001), the change in instantaneous radiative forcing between the greenhouse gas scenarios 973

GG2500 and GG250 range from 14.1 to 16.7 W·m-2, which is equivalent to 3.8 to 4.0 times the 974

radiative forcing due to the doubling of CO2 from 385 ppmv to 770 ppmv. 975

976 S2.2. Sea level scenarios 977

978

The advance or retreat of shallow seas by glacioeustatic processes is simulated by 979

representing all areas at altitudes less than 100 m as lakes in the land component of CCSM3 980

(SL100 configuration) or leaving the land component unmodified in this respect (SL0 981

configuration). (The names of simulations and configurations are italicized for the sake of 982

clarity.) This approximation of lakes as epeiric seas has the advantage that it conserves salt and 983

reduces numerical instability (Rosenbloom et al., 2011), approximating the effects of interior 984

seaways and semi-enclosed basins on the hydrological cycle without impairing the global 985

function of the ocean model component. It has the disadvantage of neglecting the exchange of 986

heat and salt between epeiric seas and the ocean that would take place, affecting both 987

evaporation from the epeiric seas and the circulation of the global ocean. All lakes are uniformly 988

50 m deep, the depth at which the effect of lake depth on surface heat fluxes tends to saturate 989

(Bonan, 1995; Oleson et al., 2004). 990

991 S2.3 Vegetation scenarios 992 993

The results of the base layer simulations, uniform present-day global average soil 994

properties, and the pCO2 values for the base layer simulations were used to force the equilibrium 995

vegetation model BIOME4 (Kaplan et al., 2003). The predicted biomes then were converted to 996

equivalent biome types used by the land component of CCSM3 to generate two vegetation 997

scenarios, a greenhouse climate scenario (V2500), which was generated from the output of the 998

base layer greenhouse climate simulation (greenhouse.base), and one for icehouse climates 999

(V250), which was generated from the output of the base layer icehouse climate simulation 1000

(icehouse.base) (Figs. 2a–b). This process could have been iterated in each individual model 1001

experiment to generate vegetation in equilibrium with the model climate. However, vegetation is 1002

a poorly known boundary condition, so changing it between model experiments would have 1003

complicated attribution of sensitivity of model climates to other experimental variables (such as 1004

land ice). 1005

Equilibrium vegetation models such as BIOME4 do not account for processes such as 1006

wildfire that strongly affect the modern biome distribution (Bond et al., 2005). Also note that 1007

BIOME4 assumes the physiological properties of late Cenozoic vegetation, which likely differ 1008

greatly from those of late Palaeozoic vegetation (e.g., Wilson and Knoll, 2010). Grassland 1009

biomes are retained rather than treated as barren ground, because non-grass plants that were 1010

adapted to similar climatic conditions to present-day grasses likely occupied those biomes, 1011

although these plants may have been less productive and/or less tolerant of moisture stress than 1012

grasses (W. DiMichele, personal communication, 2011). Sensitivity simulations performed by 1013

Horton et al. (2010) suggest that the effects of prescribing bare ground as opposed to grasslands 1014

are minimal. 1015

BIOME4 also must be forced by the hydraulic properties of the soil. As a baseline, the 1016

global mean properties in the present-day BIOME4 input file were assumed. The results of the 1017

BIOME4 simulations were insensitive to halving and doubling each soil hydraulic property 1018

separately. 1019

A similar bias is introduced by only using two vegetation scenarios. Vegetation would 1020

change along with climate. Park et al. (2012) showed that the presence of greener vegetation in 1021

CCSM3 results in cooler and wetter conditions. In tropical climates, more strongly 1022

evapotranspiring biomes will experience higher precipitation (Boyce et al., 2009). Thus, the 1023

simulations presented here tend to underestimate the magnitude of tropical precipitation change 1024

due to a particular forcing but simplifies mechanistic attribution of precipitation change. 1025

The vegetation scenarios only specify biomes. When implemented in CCSM3, the 1026

phenology of leaf area index, stem area index, and of other biome-specific parameters are 1027

prescribed by using the biome mean under present-day conditions. Note that present-day biomes 1028

of the same type can have diverse phenological and other properties. 1029

The vegetation scenario for the first pre-industrial control simulation (pic1) is the one 1030

provided with CCSM3. The vegetation scenario for the other pre-industrial control simulation 1031

(pic2) is constructed from the pic1 equilibrium climate analogously to the vegetation scenarios in 1032

the late Palaezoic simulations. 1033

1034

S2.4 Glaciation scenarios 1035 1036

A polar glacial configuration (ICEP) was developed from data in the base layer 1037

simulations. First, the mean daily liquid equivalent snow depth climatology of the icehouse 1038

climate base layer simulation (icehouse.base) identifies where snow lasted at least 160 days per 1039

year in the relatively warm model climate: an estimate of the potential distribution of land ice 1040

under colder climate conditions. Second, a network algorithm was used to assemble potential 1041

land ice areas into discrete ice sheets. The volume/height of the ice sheet were parameterized 1042

using the formulas given in Isbell et al. (2003b), except that the glacial area used in these 1043

formulas was multiplied by a factor of 3.24, in order to increase the volume of smaller ice sheets 1044

that would be poorly resolved by the simulations. Third, the algorithm was tuned to generate a 1045

desired sea level change/land ice distribution (Table S2). 1046

The glacial configurations are used to prescribe the locations of land ice in CCSM3. Land 1047

ice overrides the original vegetation and its altitude is arithmetically added to the topographic 1048

basemaps. Larger ice sheets are affected by the change in the area formula of Isbell et al. 1049

(2003b). Land ice is ~30% higher in altitude than the unmodified formula (3500 m vs. 2700 m 1050

for the polar glacial configuration: ICEP). However, the maximum altitudes of the ice sheets are 1051

similar to the predictions of ice sheet model simulations for late Paleozoic conditions (Soreghan 1052

et al., 2008b) as well as late Cenozoic ice sheets (Lythe and Vaughan, 2001). Realistic prediction 1053

of ice sheet evolution in a GCM remains highly challenging even for the LGM (e.g. Charbit et 1054

al., 2007), so there is no guarantee that a more rigorous protocol for generating the ice sheet 1055

configuration would be more or less biased than the protocol implemented. 1056

The polar glacial configuration (ICEP) consists of Gondwanan ice sheets confined to 1057

high latitudes, as proposed by Isbell et al. (2003a,b) and supported by a recent re-assignment of 1058

glacial deposits in Saudi Arabia to higher paleo-latitude (Melvin et al., 2010). In this 1059

configuration (Fig. S2c), Gondwanan land ice is minimal at ~ 240º longitude, which corresponds 1060

to Antarctica, where the presence of a large ice sheet has been disputed (Isbell et al., 2003a). 1061

This minimum arises directly from a minimum in snowfall, which is robust feature of late 1062

Paleozoic simulations in CCSM3 generally, including those of Kiehl and Shields (2005). This 1063

snowfall minimum is associated with a semi-permanent high-pressure center in spring and fall, a 1064

limiting factor for snowfall. The absence of ice sheets in late Cenozoic eastern Siberia long has 1065

been attributed to similar causes (e.g., Gerasimov and Markov, 1939). 1066

A small patch of land ice occurs in Angara (the northern continent). There is disputed 1067

evidence for glaciation in Angara (present-day Siberia) during the Permian (see discussion by 1068

Shi and Waterhouse, 2010), but Northern Hemisphere glaciation is included for consistency 1069

rather than on the basis of geological evidence. 1070

A second glacial configuration (ICEE) investigates the effects of equatorial glaciation at 1071

altitudes as low as ~500–1000 m (Becq-Giraudon et al., 1996; Sweet and Soreghan, 2008; 1072

Soreghan et al., 2008a,b, 2009) (Fig. S2c). Coupled climate-ice sheet models cannot explain how 1073

glaciers could flourish near the Equator while higher-latitude ice sheets remained relatively small 1074

(Soreghan et al., 2008b). The first approximation to such a land ice configuration is to prescribe 1075

low-latitude ice wherever geological evidence indicates it was present as well as in any land area 1076

that would have been colder. 1077

To estimate where those land areas would be, the results of an icehouse simulation forced 1078

by the polar glacial configuration (icehouse.glaciation.polar) (Table 1) were used to calculate the 1079

maximum seasonal mean temperature, denoted December–January–February (DJF), March–1080

April–May (MAM), June–July–August (JJA), September–October–November (SON); at the 1081

paleo-location of the proglacial/periglacial facies identified by Soreghan et al. (2009) (Fig. 2c). 1082

Land ice was then imposed at all areas whose highest seasonal mean temperature was lower than 1083

this temperature (25.6 °C). Topography was left unmodified, since the albedo effects of the ice 1084

were of primary interest. 1085

Table S2 lists key details of the land ice configurations. Note that these glacial 1086

configurations are not in equilibrium with the model climate and only weakly respond to model 1087

climate (they darken slightly where they would be unstable as they lose fresh snow), as would be 1088

true of any prescribed glacial configuration in CCSM3. 1089

1090

Table S2. 1091

Land ice configurations 1092

The sea level equivalent estimates do not account for isostatic adjustment. The present day ice 1093

volume estimate is from Williams and Ferrigno (2005). Note that while ICEE has no additional 1094

glacial topography relative to ICEP, keep in mind that the ice volume and equivalent sea level 1095

change of land ice covering the same area as ICEE would likely be much greater than shown 1096

here. 1097

1098

Name Area (106 km2) Volume (106

km3)

Sea Level

Equivalent (m)

ICEP 17 44 107

ICEE 79 44 107

Present 16 33 80

1099

1100 S2.5 Orbital Parameter Scenarios 1101 1102

The orbital parameters of the Earth cannot be precisely reconstructed beyond a few tens 1103

of million of years, though the statistical distribution of a particular parameter can be estimated 1104

over the duration of Solar System history (Laskar et al., 2011). Orbital parameters therefore were 1105

chosen that would explore the extreme phase space of the system. The value of eccentricity used 1106

(0.057) comes from the maximum eccentricity value in an eccentricity cycle at ~63.4 Ma in the 1107

La2010 model (Laskar et al., 2011). The minimum eccentricity value in the cycle was nearly 1108

zero, making it a good illustration of the maximum possible difference between a circular and 1109

eccentric orbit over an eccentricity cycle. The obliquity range of 22°–25° was chosen to center 1110

around the present day value of 23.5°. This range is slightly different than the range of 21.8°–1111

24.5° for the last 100 Ma inferred by Laskar et al. (2004), but the absolute difference in obliquity 1112

is similar. 1113

1114

S3. Sensitivity to equilibration criteria 1115 1116

To make a rough estimate of the sensitivity of the results to choosing less rigorous equilibration 1117

criteria than Kiehl and Shields (2005), output for the Latest Permian Simulation (LPS) reported 1118

by Kiehl and Shields (2005) was obtained. The point in the LPS where the trend in the energy 1119

imbalance of the model system for the last 100 years of the run was less than +/- 0.5 W·m-2 and 1120

the surface temperature trend was within +/- 0.05 K·(100 a)-1 was identified. The 30 years of 1121

simulation following this point was then analyzed, as was done for the LPIA simulations. The 1122

same analysis was performed for the last 30 years of the LPS, in which energy imbalance was 1123

within +/-0.05 Wm-2. The bias due to the equilibration criteria was estimated by taking the 1124

difference between the analyzed fields for the normal period with the weaker equilibration 1125

criteria from the last 30 years of the LPS (Fig. S1). 1126

1127 Fig. S2. Differences in analyzed fields between: (1) the earliest normal period of 30 years from 1128 the Latest Permian Simulation of Kiehl and Shields (2005) that would have met the equilibration 1129 criteria of this study, and (2) the normal period of 30 years at the end of the Latest Permian 1130 Simulation: a, absolute difference in mean annual surface temperature (°C); b, relative difference 1131 in annual mean precipitation (%); c, Absolute difference in zonal average annual mean surface 1132 temperature bias; d, Zonal average relative difference in annual mean precipitation (%). 1133 1134

References 1135

1136

Allen, J.P., Fielding, C.R., Gibling, M.R., Rygel, M.C., 2011. Fluvial response to paleo-1137

equatorial climate fluctuations during the late Paleozoic ice age. Geol. Soc. Amer. Bull., 123, 1138

1524–1538. 1139

Becq-Giraudon, J.-F., Montenat, C., Van Den Driessche, J., 1996. Hercynian high-altitude 1140

phenomena in the French Massif Central: tectonic implications. Palaeogeogr. Palaeoclimatol. 1141

Palaeoecol., 122, 227–241, 1996. 1142

Birgenheier, L P, Fielding, C.R., Rygel, M.C., Frank, T.D., Roberts, J., 2009. Evidence for 1143

Dynamic Climate Change on Sub-106-Year Scales From the Late Paleozoic Glacial Record, 1144

Tamworth Belt, New South Wales, Australia. J. Sed. Res., 79, 56–82, 1145

doi:10.2110/jsr.2009.013. 1146

Bishop, J.W., Montañez, I.P., Gulbranson, E.L., Brenckle, P.L., 2009. The onset of mid-1147

Carboniferous glacial eustasy: Sedimentologic and diagenetic constraints, Arrow Canyon, 1148

Nevada. Palaeogeogr. Palaeoclimatol. Palaeoecol., 276, 217–243. 1149

Blakey, R.C., 2008. Gondwana paleogeography from assembly to breakup–A 500 m.y. odyssey. 1150

In: Fielding, C.R., Frank, T.D., Isbell, J.L. (Eds.), Resolving the Late Paleozoic Ice Age in Time 1151

and Space. Geol. Soc. Am. Spec. Pap., 441, 1–28. 1152

Blakey, R.C., n.d. Rectangular Global Maps. Northern Arizona University, 1153

http://www2.nau.edu/rcb7/rect_globe.html, last access: 6 March 2012. 1154

Bonan, G.B., 1995. Land-atmosphere CO2 exchange simulated by a land surface process model 1155

coupled to an atmospheric general circulation model. J. Geophys. Res., 100, 2817–2831. 1156

Bond, W.J., Woodward, F.I., Midgley, G.F., 2005. The global distribution of ecosystems in a 1157

world without fire. New Phytol., 165, 525–538. 1158

Boucot, A.J., Chen, X., Scotese, C.R., 2013. Phanerozoic Paleoclimate: An Atlas of Lithologic 1159

Indicators of Climate. SEPM, Tulsa, OK. 1160

Boyce, C. K., Brodribb, T.J., Feild, T.S., Zwieniecki, M.A., 2009. Angiosperm leaf vein 1161

evolution was physiologically and environmentally transformative. Proc. Roy. Soc. B, 276, 1162

1771–1776. 1163

Bunker, A.F. 1965. Interaction of the summer monsoon air with the Arabian Sea (preliminary 1164

analysis). In: Proceedings of the Symposium on Meteorological Results of the International 1165

Indian Ocean Expedition, Bombay, India, 22–26 July 1965. Indian Meteorological Department, 1166

New Delhi, pp. 3–16. 1167

Came, R.E., Eiler, J.M., Veizer, J., Azmy, K., Brand, U., Weidman, C.R., 2007. Coupling of 1168

surface temperatures and atmospheric CO2 concentrations during the Palaeozoic era. Nature, 1169

449, 198–202. 1170

Cecil, C.B., Dulong, F.T., West, R.R., Stamm, R., Wardlaw, B., Edgar, N.T., 2003. Climate 1171

controls on the stratigraphy of a Middle Pennsylvanian cyclothem in North America. In: Cecil, 1172

C.B. and Edgar, N.T. (Eds.), Climate controls on stratigraphy. SEPM Soc. Sedim. Geol. Spec. 1173

Publ., 77, 151–182. 1174

Celâl Sengör, A.M., Atayman, S., 2009. The Permian Extinction and the Tethys: An Exercise in 1175

Global Geology. Geol. Soc. Am. Spec. Pap., 448, 1–85. 1176

Charbit, S., Ritz, C., Philippon, G., Peyaud, V., Kageyama, M., 2007. Numerical reconstructions 1177

of the Northern Hemisphere ice sheets through the last glacial-interglacial cycle. Clim. Past, 3, 1178

15–37, doi:10.5194/cp-3-15-2007. 1179

Chen, B., Joachimski, M.M., Shen, S-Z., Lambert, L.L., Lai, X.-L., Wang, X.-D., Chen, J., Yuan, 1180

D.-X., 2013. Permian ice volume and palaeoclimate history : Oxygen isotope proxies revisited. 1181

Gondwana Res., 24, 77–89. 1182

Collins, W.D., Bitz, C.M., Blackmon, M.L., Bonan, G.B., Bretherton, C.S., Carton, J.A., Chang, 1183

P., Doney, S.C., Hack, J.J., Henderson, T.B., Kiehl, J.T., Large, W.G., McKenna, D.S., Santer, 1184

B.D., Smith, R.D., 2006. The Community Climate System Model Version 3 (CCSM3). J. 1185

Climate, 19, 2122–2143. 1186

Crouvi., O., Amit, R., Enzel, Y., Gillespie, A.R., 2010. Active sand seas and the formation of 1187

desert loess. Quat. Sci. Rev., 29, 2087–2098. 1188

Crowley, T.J., Baum, S.K., Hyde, W.T., 1991. Climate Model Comparison of Laurentide and 1189

Gondwanan Glaciations. J. Geophys. Res., 96, 9217–9226. 1190

Dai, A., 2006. Precipitation characteristics in eighteen coupled climate models. J. Climate, 19, 1191

4605–4630. 1192

Davydov, V.I., Crowley, J.L., Schmitz, M.D., Poletaev, V.I., 2010. High-precision U-Pb zircon 1193

age calibration of the global Carboniferous time scale and Milankovitch band cyclicity in the 1194

Donets Basin, eastern Ukraine. Geochem. Geophys. Geosyst., 11, Q0AA04, doi: 1195

10.1029/2009GC002736. 1196

Derbyshire, E., 1983. Origin and characteristics of some Chinese loess from two locations in 1197

China. In: Brookfield, M.E., Ahlbrandt, T.S. (Eds.). Eolian Sediments and Processes. Elsevier, 1198

pp. 69–70. 1199

Deser, C., Phillips, A.S., Alexander, M.A., 2010. Twentieth century tropical sea surface 1200

temperature trends revisited. Geophys Res. Lett., 37, doi:10.1029/2010GL043321. 1201

Du Toit, A.L., Reed, F.R.C., 1927. A Geological Comparison of South America with South 1202

Africa. Carnegie Institute of Washington, Washington, DC. 1203

Eagle, R.A., Eiler, J.M., Tripati, A.K., Ries, J.B., Freitas, P.S., Hiebenthal, C., Wanamaker Jr., 1204

A.D., Taviani, M., Elliot, M., Marenssi, S., Nakamura, K., Ramirez, P., Roy, K., 2013. The 1205

influence of temperature and seawater carbonate saturation state on 13C/18O bond ordering in 1206

bivalve mollusks. Biogeosci., 10, 4591–4606. 1207

Eros, J.M., Montañez, I.P., Osleger, D.A., Davydov, V.I., Nemyrovska, T.I., Poletaev, V.I., 1208

Zhykalyak, M.V., 2012. Sequence stratigraphy and onlap history of the Donets Basin, Ukraine: 1209

Insight into Carboniferous icehouse dynamics, Palaeogeogr. Palaeoclimatol. Palaeoecol., 313–1210

314, 1–25. 1211

Falcon-Lang, H.J., DiMichele, W.A., 2010. What happened to the Coal Forests during glacial 1212

phases? Palaios, 25, 611–617. 1213

Fielding, C.R., Frank, T.D., Isbell, J.L., 2008. The late Paleozoic ice age–a review of current 1214

understanding and synthesis of global climate patterns. In: Fielding, C.R., Frank, T.D., and 1215

Isbell, J.L. (Eds.), Resolving the Late Paleozoic Ice Age in Time and Space. Geol. Soc. Am. 1216

Spec. Pap., 441, 343–354. 1217

Garcin, Y., Vincens, A., Williamson, D., Guiot, J., Buchet, G., 2006. Wet phases in tropical 1218

southern Africa during the last glacial period. Geophys. Res. Lett., 33, L07703, 1219

doi:10.1029/2005GL025531. 1220

Gerasimov, I.P., Markov, K.K., 1939. The Glacial Period on the USSR Territory. USSR 1221

Academy of Sciences Publishers, Moscow. 1222

Gibbs, M.T., Rees, P.M., Kutzbach, J.E., Ziegler, A.M., Behling, P.J., Rowley, D.B., 2002. 1223

Simulations of Permian Climate and Comparisons with Climate-Sensitive Sediments. J. 1224

Geology, 110, 33–55. 1225

Giles, P.S., 2012. Low-latitude Ordovician to Triassic brachiopod habitat temperatures (BHTs) 1226

determined from δ18O[brachiopod calcite]: A cold hard look at ice-house tropical oceans. Palaeogeogr. 1227

Palaeoclimat. Palaeoecol., 317–318, 134–152. 1228

Gough, D.O., 1981. Solar interior structure and luminosity variations. Sol. Phys., 74, 21–34. 1229

Gradstein, F.M., Ogg, J.G., Smith, A.G., 2004. A Geologic Time Scale 2004. Cambridge 1230

University Press, Cambridge, UK. 1231

Grossman, E.L., 2012. Oxygen Isotope Stratigraphy. In: Gradstein, F.M., Ogg, J.G., Schmitz, 1232

M., Ogg, G. (Eds.), The Geologic Time Scale 2012. Elsevier, Amsterdam, pp. 195–220. 1233

Hastenrath, W., 2009. Past glaciation in the tropics. Quat. Sci. Rev., 28, 790–798. 1234

Heavens, N.G., Shields. C.A., Mahowald, N.M., 2012. Sensitivity of a Deep Time Climate 1235

Simulation to Aerosol Prescription, J. Adv. Mod. Earth Sys., M11002, 1236

doi:10.1029/2012MS000166. 1237

Heckel, P.H., 1986. Sea-level curve for Pennsylvanian eustatic marine transgressive-regressive 1238

depositional cycles along midcontinent outcrop belt, North America. Geology, 14, 330–334. 1239

Heckel, P.H., 1995. Glacial-eustatic base-level-Climatic model for late middle to late 1240

Pennsylvanian coal-bed formation in the Appalachian basin. J. Sed. Res., 65, 348–356. 1241

Heckel, P., 2008. Pennsylvanian cyclothems in Midcontinent North America as far-field effects 1242

of waxing and waning of Gondwana ice sheets. Geol. Soc. Amer. Spec. Pub., 441, 275–289. 1243

Horton, D.E., Poulsen, C.J., 2009. Paradox of late Paleozoic glacioeustasy. Geology, 37, 715–1244

718. 1245

Horton, D.E. Poulsen, C.J., Pollard, D., 2007. Orbital and CO2 forcing of late Paleozoic 1246

continental ice sheets. Geophys. Res. Lett., 34, L19708. 1247

Horton, D.E., Poulsen, C.J., Pollard, D., 2010. Influence of high-latitude vegetation feedbacks on 1248

late Palaeozoic glacial cycles. Nature Geosci., 3, 572–577. 1249

Horton, D.E., Poulsen, C.J., Montañez, I.P., DiMichele, W.A., 2012. Eccentricity-paced late 1250

Paleozoic climate change. Palaeogeogr., Palaeoclimatol. Palaeoecol., 331–332, 150–161. 1251

IPCC: Climate change 2001: The scientific basis. Contribution of Working Group I to the Third 1252

Assessment Report of the Intergovernmental Panel on Climate, Houghton, J. T., Ding, Y., 1253

Griggs, D. J., Noguer, M., Van der Linden, P. J., Dai, X.,Maskell, K., Johnson, C. A. (Eds.), 1254

2001. Cambridge University Press, Cambridge and New York. 1255

Isbell, J.L., Lenaker, P.A., Askin, R.A., Miller, M.F., Babcock, L.E., 2003a. Reevaluation of the 1256

timing and extent of late Paleozoic glaciation in Gondwana: Role of the Transantarctic 1257

Mountains. Geology, 31, 977–980. 1258

Isbell, J.L., Miller, M.F.,Wolfe, K.L., Lenaker P.A., 2003b. Timing of late Paleozoic glaciation 1259

in Gondwana: Was glaciation responsible for the development of northern hemisphere 1260

cyclothems? In: Chan, M.A., Archer, A.W. (Eds.), Extreme depositional environments: Mega 1261

end members in geologic time. Geol. Soc. Am. Spec. Pap., 370, 5–24. 1262

Jordan, O.D., Mountney, N.P., 2012. Sequence Stratigraphic Evolution and Cyclicity of An 1263

Ancient Coastal Desert System: The Pennsylvanian-Permian Cutler Beds, Paradox Basin, Utah, 1264

U.S.A. J. Sed. Res., 82, 755–780. 1265

Joseph, P.V., Sijikumar, S., 2004. Intraseasonal Variability of the Low-Level Jet Stream of the 1266

Asian Summer Monsoon. J. Atmos. Sci., 17, 1449–1458. 1267

Kaplan, J. O., 17 others, 2003. Climate change and Arctic ecosystems: 2. Modeling, paleodata-1268

model comparisons, and future projections. J. Geophys. Res., 108, 8171, 1269

doi:10.1029/2002JD002559. 1270

Kiehl, J.T., Shields, C.A., 2005. Climate simulation of the latest Permian: Implications for mass 1271

extinction. Geology, 33, 757–760. 1272

Kiehl, J.T., Shields, C.A., Hack, J.J., Collins, W.D., 2006. The Climate Sensitivity of the 1273

Community Climate System Model Version 3 (CCSM3). J. Climate, 19, 2584–2596. 1274

Köppen, W., A. Wegener, 1924. Die Klimate die geologischen Vorzeit. Gebrüder Borntraeger, 1275

Berlin, Germany. 1276

Kutzbach, J.E., Gallimore, R.G., 1989. Pangaean Climates: Megamonsoons of the 1277

Megacontinent, J. Geophys. Res., 94, 3341–3357. 1278

Kutzbach, J.E., Liu, X., Liu, Z., Chen, G., 2008. Simulation of the evolutionary response of 1279

global summer monsoons to orbital forcing over the past 280,000 years. Clim. Dyn., 30, 567–1280

579. 1281

Laskar, J., Robutel, P., Joutel, F., Gastineau, M., Correia, A.C.M., Levrard, B., 2004. A long 1282

term numerical solution for the insolation quantities of the Earth. Astron. Astrophys., 428, 261–1283

285. 1284

Laskar, J., Fienga, A., Gastineau, M., Manche, H., 2011. La2010: A new orbital solution for the 1285

long-term motion of the Earth, Astron. Astrophys., 532, A89. 1286

Lindzen R.S., Hou A.Y., 1988. Hadley circulations for zonally averaged heating centered off the 1287

equator. J. Atmos. Sci, 45, 2416–2427. 1288

Lythe, M.B., Vaughan, D.G., BEDMAP Consortium, 2001. BEDMAP: A new ice thickness and 1289

subglacial topographic model. J. Geophys Res., 106, 11335–11351. 1290

Mahowald, N.M., Ward, D.S., Kloster, S., Flanner, M.G., Heald, C.L., Heavens, N.G., Hess, 1291

P.G., Lamarque, J.-F., 2011. Aerosol impacts on climate and biogeochemistry. Ann. Rev. 1292

Environ. Resour., 36, 45–74. 1293

Meehl, G.A., Arblaster, J.M., 2003. Mechanisms for projected future changes in south Asian 1294

monsoon precipitation. Clim. Dyn., 21, 659–675. 1295

Meehl, G. A., Arblaster, J. M., Lawrence, D. M., Seth, A., Schneider, E. K., Kirtman, B. P., Min, 1296

D., 2006. Monsoon regimes in the CCSM3. J. Climate, 19, 2482–2495. 1297

Melvin, J., Sprague, R.A., Heine, C.J., 2010. From bergs to ergs: The Late Paleozoic Gondwanan 1298

glaciation and its aftermath in Saudi Arabia. In López-Gamundí, O.R., Buatois, L.A., Late 1299

Paleozoic glacial events and postglacial transgressions in Gondwana. Geol. Soc. Am. Spec. Pap., 1300

468, 37–80. 1301

Merlis, T. M., Schneider, T., Bordoni, S., Eisenman, I., 2013a. Hadley circulation response to 1302

orbital precession. Part I: Aquaplanets. J. Climate, 26, 740–753. 1303

Merlis, T. M., Schneider, T., Bordoni, S., Eisenman, I., 2013b. Hadley circulation response to 1304

orbital precession. Part II: Subtropical continent. J. Climate, 26, 754–771. 1305

Merlis, T. M., Schneider, T., Bordoni, S., Eisenman, I., 2013c. The tropical precipitation 1306

response to orbital precession. J. Climate, 26, 2010–2021. 1307

Miller, K.B., West, R.R., 1993. A reevaluation of Wolfcampian cyclothems in northeastern 1308

Kansas: significance of subaerial exposure and flooding surfaces. Kansas Geol. Surv. Bull. 235, 1309

1–26. 1310

Miller, K.B., McCahon, T.J., West, R.R., 1996. Lower Permian (Wolfcampian) paleosol- bearing 1311

cycles of the U.S. Midcontinent: evidence of climatic cyclicity, J. Sed. Res., 66, 71–84. 1312

Montañez, I.P., Tabor, N.J., Niemeier, D., DiMichele, W.A., Frank, T.D., Fielding, C.R., Isbell, 1313

J.L., Birgenheier, L.P., Rygel, M.C., 2007. CO2-Forced Climate and Vegetation Instability 1314

During Late Paleozoic Glaciation. Science, 315, 87–91. 1315

Muhs, D.R., 2013. The geologic records of dust in the Quaternary. Aeol. Res., 9, 3–48. 1316

Muhs, D.R., Bettis, E.A., III, 2003. Quaternary loess-paleosol sequences as examples of climate-1317

driven sedimentary extremes. Geol. Soc. Amer. Spec. Pap., 370, 53–74. 1318

NOAA NGDC: 2-Minute Gridded Global Relief Data (ETOPO2v2) June, 2006, 2006. 1319

http://www.ngdc.noaa.gov/mgg/fliers/01mgg04.html, last access: 6 March 2012. 1320

Oleson, K., Dai, Y., Bonan, G., Bosilovich, M., Dickinson, R., Dirmeyer, P., Hoffman, F., 1321

Houser, P., Levis, S., Niu, G.-Y., Thornton, P., Vertenstein, M., Yang, Z.-L., Zeng, X., 2004. 1322

Technical Description of the Community Land Model (CLM). NCAR Technical Note 461. 1323

1324

Olsen, P. E., Kent, D. V., 1996. Milankovitch climate forcing in the tropics of Pangea during the 1325

Late Triassic. Palaeogeogr. Palaeoclimatol. Palaeoecol.,122, 1–26. 1326

1327

Olszewski, T.D., Patzkowsky, M.E., 2003. From Cyclothems to Sequences: The Record of 1328

Eustasy and Climate on an Icehouse Epeiric Platform (Pennsylvanian-Permian, North American 1329

Midcontinent). J. Sed. Res., 73, 15–30. 1330

Otto-Bliesner, B., 2003. The Role of Mountains, Polar Ice, and Vegetation in Determining the 1331

Tropical Climate During the Middle Pennsylvanian: Climate Model Simulations. In Cecil, C.B. 1332

and Edgar, N.T. (Eds.), Climate controls on stratigraphy. SEPM Soc. Sedim. Geol. Spec. Publ., 1333

77, 227–238. 1334

Park, C. E., Ho, C.H., Jeong, S.-J., Kim, J., Feng, S., 2012. The potential of vegetation feedback 1335

to alleviate climate aridity over the United States associated with a 2xCO2 climate condition. 1336

Clim Dyn., 38, 1489–1500. 1337

Parrish, J. T., 1993. Climate of the Supercontinent Pangea, J. Geology, 10, 215–233. 1338

Partridge, T.C., DeMenocal, P.B., Lorentz, S.A., Paiker, M.J., Vogel, J.C., 1997. Orbital Forcing 1339

of Climate over South Africa: A 200,000-Year Rainfall Record From the Pretoria Saltpan. Quat. 1340

Sci. Rev., 16, 1125–1133. 1341

Petit, J.R., Mourner, L., Jouzel, J., Korotkevich, Y.S., Kotlyakov, V.I., Lorius, C., 1990. 1342

Palaeoclimatological and chronological implications of the Vostok core dust record. Nature, 343, 1343

56–58. 1344

Peyser, C.E., Poulsen, C.J., 2008. Controls on Permo-Carboniferous precipitation over tropical 1345

Pangaea: A GCM sensitivity study. Palaeogeog. Palaeoclimatol. Palaeoecol., 268, 181–192. 1346

Poulsen, C.J., Pollard, D., Montañez, I.P., Rowley D., 2007. Late Paleozoic tropical climate 1347

response to Gondwanan deglaciation. Geology, 35, 771–774. 1348

Powell, M.G., Schöne, B.R., Jacob, D.E., 2009. Tropical marine climate during the late 1349

Paleozoic ice age using trace element analyses of brachiopods. Palaeogeogr. Palaeoclimatol. 1350

Palaeogeogr. Palaeoecol., 280, 143–149. 1351

Puceat et al. (2010) Pucéat, E., Joachimski, M.M., Bouilloux, A., Monna, F., Bonin, A., 1352

Motreuil, S., Morinière, P., Hénard, S., Mourin, J., Dera, G., Quesne, D., 2010. Revised 1353

phosphate–water fractionation equation reassessing paleotemperatures derived from biogenic 1354

apatite. Earth Planet. Sci. Lett., 298, 135–142. 1355

Pye, K., 1987. Aeolian Dust and Dust Deposits. Academic Press, London, UK. 1356

Pye, K., 1995. The nature, origin, and accumulation of loess. Quat. Sci. Rev., 14, 653–668. 1357

Rankey, E.C., 1997. Relations between relative changes in sea level and climate shifts: 1358

Pennsylvanian – Permian mixed carbonate-siliciclastic strata, western U.S.A., Geol. Soc. Amer. 1359

Bull., 109, 1089–1100. 1360

Rea, D.K., 1994. The Paleoclimatic Record Provided by Eolian Deposition in the Deep Sea. Rev. 1361

Geophys., 32, 159–195. 1362

Rosenbloom, N., Shields, C., Brady, E., Levis, S., Yeager, S., 2011. Using CCSM3 for 1363

Paleoclimate Applications. NCAR Technical Note 483. 1364

Rowley, D., 2008. 2003 Revised Permian Paleogeographic Maps. 1365

http://geosci.uchicago.edu/~rowley/Rowley/Revised_Permian_Maps.html, last access: 6 March 1366

2012. 1367

Rygel, M.C., Fielding, C.R., Frank, T.D., Birgenheier, L.P., 2008. The Magnitude of Late 1368

Paleozoic Glacioeustatic Fluctuations: A Synthesis. J. Sed. Res., 78, 500-511. 1369

Schmitz, M.D., Davydov, V.I., 2012. Quantitative Radiometric and Biostratigraphic Calibration 1370

of the Pennsylvanian–Early Permian (Cisuralian) Time Scale and Pan-Euramerican 1371

Chronostratigraphic Correlation. Geol. Soc. Amer. Bull.,124, 549–577. 1372

Schwarzacher, W., 1989. Milankovitch type cycles in the Lower Carboniferous of NW Ireland, 1373

Terra Nova, 1, 468–473. 1374

Scotese, C. R., 1997. Paleogeographic Atlas. PALEOMAP Progress Report, 90-0497, 1375

Department of Geology, University of Texas at Arlington, Arlington, Texas. 1376

Seinfeld J. H., Pandis S. N., 1998. Atmospheric Chemistry and Physics: From Air Pollution to 1377

Climate Change. J. Wiley, New York. 1378

Shi, G.R., Waterhouse, J.B., 2010. Late Paleozoic global changes affecting high-latitude 1379

environments and biotas: An introduction. Palaeogeogr. Palaeoclimatol. Palaeoecol., 298, 1–16. 1380

Slingerland, R., Furlong, K.P., 1989. Geodynamic and geomorphic evolution of the Permo-1381

Triassic Appalachian mountains. Geomorphology, 2, 23–37. 1382

1383

Soreghan, G.S., 1994. The impact of glacioclimatic change on Pennsylvanian cyclostratigraphy. 1384

In: Embry, A.F., Beauchamp, B., Glass, D.J., (Eds.), Pangea—Global environments and 1385

resources. Can. Soc. Petrol. Geol. Mem., 17, 523–543. 1386

Soreghan, G. S., 1997. Walther’s Law, climate change, and upper Paleozoic cyclostratigraphy in 1387

the Ancestral Rocky Mountains. J. Sed. Res., 67, 1001–1004. 1388

Soreghan, G.S., Giles, K.A., 1999. Amplitudes of Late Pennsylvanian glacioeustasy. Geology, 1389

27, 255–258. 1390

Soreghan, G.S., Soreghan, M.J., 2002. Atmospheric dust and algal dominance in the late 1391

Paleozoic: a hypothesis. J. Sed. Res., 72, 457–461. 1392

Soreghan, G., Sweet, D., Marra, K., Eble, C., Soreghan, M., Elmore, R., Kaplan, S., Blum, M., 1393

2007. An Exhumed Late Paleozoic Canyon in the Rocky Mountains. J. Geology, 115, 473–481. 1394

Soreghan, G.S., Soreghan, M.J., Hamilton, M.A., 2008a. Origin and paleoclimatic significance 1395

of tropical silt in the Late Paleozoic icehouse. Palaeogeogr. Palaeoclimatol, Palaeoecol., 268, 1396

234–259. 1397

Soreghan, G.S., Soreghan, M.J., Poulsen, C.J., Young, R.A., Eble, C.F., Sweet, D.E., 1398

Davogustto, O.C., 2008b. Anomalous cold in the Pangaean tropics, Geology, 36, 659–662. 1399

Soreghan, G.S., Soreghan, M.J., Sweet, D.E., Moore, K.D., 2009. Hot Fan or Cold Outwash?: 1400

Hypothesized Proglacial Deposition in the Upper Paleozoic Cutler Formation, Western Tropical 1401

Pangaea. J. Sed. Res., 79, 495–522. 1402

Soreghan, G.S., Sweet, D., Heavens, N.G., 2014. Upland Glaciation in Tropical Pangaea: 1403

Geologic Evidence, and Implications for Late Paleozoic Climate Modeling. J. Geology, 122, 1404

137–163. 1405

Soreghan, M.J., Heavens, N., Soreghan, G.S., Link, P.K., Hamilton, M.A., 2014. High-resolution 1406

record of Permian atmospheric circulation changes inferred from geochemical, magnetic, and 1407

sedimentologic indicators in the Maroon Formation. Geol. Soc. Amer. Bull., B30840.1. 1408

Sur, S., Soreghan, G., Soreghan, M., Yang, W., Saller, A., 2010. A record of glacial aridity and 1409

Milankovitch-scale fluctuations in atmospheric dust from the Pennsylvanian tropics. J. Sed. Res., 1410

80, 1046–1067. 1411

Sweet, D.E., Soreghan, G.S., 2008. Polygonal cracking in coarse clastics records cold 1412

temperatures in the equatorial Fountain Formation (Pennsylvanian-Permian, Colorado), 1413

Palaeogeogr. Palaeoclimatol. Palaeoecol., 268, 193–204. 1414

Tabor, N.J., Poulsen, C.J., 2008. Palaeoclimate across the Late Pennsylvanian-Early Permian 1415

tropical paleolatitudes: A review of climate indicators, their distribution, and relation to 1416

palaeophysiographic climate factors. Palaeogeogr. Palaeoclimatol. Palaeoecol., 268, 293–310. 1417

Tabor, N.J., DiMichele, W.A., Montañez, I.P., Chaney, D.S., 2013. Late Paleozoic continental 1418

warming of a cold tropical basin and floristic change in western Pangea. Int. J. Coal Geol., 119, 1419

177–186. 1420

Tsoar, H., Pye K., 1987. Dust transport and the question of desert loess formation. 1421

Sedimentology, 34, 139–153. doi: 10.1111/j.1365-3091.1987.tb00566.x. 1422

Turner, A. G., Annamalai, H., 2012. Climate change and the summer South Asian monsoon. 1423

Nature Clim. Change, 2, 587–595. 1424

Veevers, J.J., Powell, C.M., 1987. Late Paleozoic glacial episodes in Gondwanaland reflected in 1425

transgressive-regressive depositional sequences in Euramerica. Geol. Soc. Am. Bull., 98, 475–1426

487. 1427

Wang, B., Wu, R., Lau, K.-M., 2001. Interannual Variability of the Asian Summer Monsoon: 1428

Contrasts between the Indian and the Western North Pacific-East Asian Monsoons. J. Climate, 1429

14, 4073–4090. 1430

Wanless, H.R., Cannon, J.R., 1966. Late Paleozoic glaciation. Earth Sci. Rev., 1, 247–286. 1431

Wanless, H.R., Shepard, F.P., 1936. Sea level and climatic change related to Late Paleozoic 1432

cycles. Geol. Soc. Am. Bull., 47, 1177–1206. 1433

Webster, P.J., Fasullo, J., 2003. Monsoon: Dynamical Theory. In: Holton, J.R., Curry, J.A., Pyle, 1434

J.A. (Eds.), Encyclopedia of Atmospheric Sciences. Academic Press, San Diego, CA, pp. 1370–1435

1385. 1436

Williams, R.S., Jr., Ferrigno, J.G., 2005. Satellite image atlas of glaciers of the world. United 1437

State Geological Survey Fact Sheet 2005–3056, http://pubs.usgs.gov/fs/2005/3056. 1438

Wilson, J.P., Knoll, A.H., 2010. A physiologically explicit morphospace for tracheid-based water 1439

transport in modern and extinct seed plants. Paleobiology, 36, 335–355. 1440

Witzke, B.J., 1990. Palaeoclimatic constraints for Palaeozoic Paleolatitudes of Laurentia and 1441

Euramerica. In: McKerrow, W.S. and Scotese, C.R. (Eds.), Palaeozoic Palaeogeography and 1442

Biogeography, Geol. Soc. Mem., 12, 57–73. 1443

Yeager, S.G., Shields, C.A., Large, W.G., Hack, J.J., 2006. The Low-Resolution CCSM3. J. 1444

Climate, 19, 2545–2566. 1445

Ziegler, A.M., Hulver, M.L., Rowley, D.B., 1997. Permian World Topography and Climate. In: 1446

Martini, I.P., Late Glacial and Postglacial Environmental Changes: Quaternary, Carboniferous-1447

Permian, and Proterozoic. Oxford University Press, Oxford, UK, pp. 111–146. 1448

1449 1450

A model-based evaluation of tropical climate in Pangaea during the late Palaeozoic icehouse

Documents

Transcript of A model-based evaluation of tropical climate in Pangaea during the late Palaeozoic icehouse