Aptian-Albian clumped isotopes from northwest China: Cool temperatures, variable atmospheric pCO2 and regional shifts in hydrologic cycle

The Early Cretaceous is characterized by warm background temperatures (i.e., greenhouse climate) and carbon cycle perturbations that are often marked by Ocean Anoxic Events (OAEs) and associated shifts in the hydrologic cycle. Higher-resolution records of terrestrial and marine δ13C and δ18O (both carbonates and organics) suggest climate shifts during the Aptian-Albian, including a warm period associated with OAE 1a in the early Aptian and subsequent “cold snap” 20 near the Aptian-Albian boundary prior to the Kilian and OAE 1b. Understanding the continental system is an important factor in determining the triggers and feedbacks to these events. Here, we present new paleosol carbonate stable isotopic (δ13C, δ18O and Δ47) and CALMAG weathering parameter results from the Xiagou and Zhonggou Formations (part of the Xinminpu Group in the Yujingzi Basin of NW China) spanning the Aptian-Albian. Published mean annual air temperature (MAAT) records of the Barremian-Albian from Asia are relatively cool with respect to the Early Cretaceous. However, these 25 records are largely based on coupled δ18O measurements of dinosaur apatite phosphate (δ18Op) and carbonate (δ18Ocarb), and therefore rely on estimates of meteoric water δ18O (δ18Omw) from δ18Op. Significant shifts in the hydrologic cycle likely influenced δ18Omw in the region, complicating these MAAT estimates. Thus, temperature records independent of δ18Omw (e.g., clumped isotopes or Δ47) are desirable, and required to confirm temperatures estimated with δ18Op and δ18Oc, and to reliably determine regional shifts in δ18Omw. Primary carbonate material was identified using traditional petrography, 30 cathodoluminescence inspection, and δ13C and δ18O subsampling. Our preliminary Δ47-based temperature reconstructions (record mean of 14.9 oC), which we interpret as likely being representative of MAAT, match prior estimates from similar paleolatitudes of Asian MAAT (average ~15 oC) across the Aptian-Albian. This, supported by our estimated mean atmospheric paleo-pCO2 concentration of 396 ppmv, indicates relatively cooler mid-latitude terrestrial climate. Additionally,

Hydrologic cycle models and observations of past warm intervals (e.g., early Cenozoic and greenhouse Cretaceous) indicate an "intensification" of the hydrologic cycle due to enhanced poleward moisture transport associated with global warming (e.g., Carmichael et al., 2015;Hasegawa et al., 2012;Suarez et al., 2011a;White et al., 2001;Poulsen et al., 2007). Likewise, as temperatures cool during Cretaceous climate recovery or during long-term transitions driven by changes in global 70 tectonics and paleogeography, the hydrologic cycle tends to respond with regionally dependent shifts in mean annual precipitation (MAP). For example, Hasegawa et al. (2012) observed hydrologic cycle responses track greenhouse gas (GHG) forcing in Asia during Aptian-Albian. For the Aptian-Albian, models and observations suggest changes in continental interior precipitation during the global "cold snap" (e.g., Mutterlose et al., 2009) and the potential for variable Asian aridity associated with warm/cool cycles (Hasegawa et al., 2010;Hasegawa et al., 2012;Föllmi, 2012;Poulsen et al., 2007;Tabor et 75 al., 2016;Zhou et al., 2008), which may hamper δ 18 O p-based temperature reconstructions for the Aptian-Albian that fail to quantify δ 18 Omw independently of δ 18 Op. To address this deficiency, here we provide new multi-proxy records from the Yujingzi Basin of NW China spanning the Aptian-Albian using δ 13 C, δ 18 O and Δ47 (i.e., clumped isotopes) of terrestrial paleosol carbonates. Additionally, MAP is quantified using chemical weathering ratios, specifically CALMAG (Nordt and Driese, 2010). We combine our new records with organic stable carbon isotope chemostratigraphic records for the site 80 (Suarez et al., 2018) to provide age control to quantitatively interpret shifts in regional temperature, δ 18 Omw, MAP, and global atmospheric paleo-pCO2 associated with the Aptian-Albian. These proxy interpretations are compared to models and records of Cretaceous Asian climate and the global exogenic carbon cycle (i.e., atmospheric paleo-pCO2) to provide new constraints on Aptian-Albian climate, carbon and hydrologic cycles.

Sampling and analyses
The Xiagou and Zhonggou Formations, part of the Xinminpu Group in the Yujingzi Basin of Northwest China (Gansu Province), were sampled in 2011 with the goal of placing the Early Cretaceous paleobiology and geology of this region in a global climate and carbon isotope chemostratigraphic framework (e.g., Suarez et al., 2018). The Xinminpu Group, approximately Early Cretaceous in northwest China, is composed of four formations (ordered stratigraphically bottom to 90 top): Chijinqiao Formation, Chijinpu Formation, Xiagou Formation, and Zhonggou Formation. Outside of the Yujingzi Basin, Xinminpu group strata produce Aptian radiometric dates of 123.0 ± 2.6 to 133.7 ± 1.8 Ma (Li et al., 2013;Kuang et al., 2013). Outcrop sections (Fig. 1) are regionally exposed in the Yujingzi basin at a fossil-rich site informally known as the White Pagoda Site, produced by accommodation of strike-slip motion from Lhasa Block convergence with Asia (Chen and Yang, 1996;Vincent and Allen, 1999). 95 Outcrop sections for White Pagoda were numbered and split into three facies by Suarez et al. (2018): 1) the lowermost facies consisting of sections 1, 2, 2A, 2) an overlying facies of alternating gray and variegated mudstones and muddy sandstones https://doi.org/10.5194/cp-2020-152 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. consisting of sections 3, 3A-H, and 6, and 3) coarse-grained poorly sorted arkosic sandstones and sandy mudstones (Section 4). Here, we utilize sample material from sections (in stratigraphic order) 3, 3A-H, 6, and 4. Suarez et al. (2018) observed 100 carbonate nodules, root-traces, charophytes, turtle remains, ostracodes and gastropods within the middle facies (Sections 3, 3A-H, and 6), with root-traces and nodules extending into the uppermost facies (Section 4). Section 4 had a higher degree of color-mottling, blocky ped-structures and burrows compared to the underlying facies. Facies interpretations for sections sampled for this study indicate fluvio-lacustrine and palustrine environments (i.e., Suarez et al., 2018). For example, sections 3, 3A-H, 6, and 4 exhibit evidence of subaerial exposure (soils indicated by horizonation, slickensides, root traces and 105 carbonate nodule formation), fluvial deposition (lenticular sands fining up with erosive bases), and lacustrine environments (turtle remains, charophytes, ostracodes, thin limestones, and organic-rich mudstones) (Suarez et al., 2018). For the sections of interest, the presence of cracking, slickensides (mukkara structures) and expansive clays suggests wet/dry periods typical of vertisols.

110
The organic stable carbon isotope record at White Pagoda was used as a guide to sample carbonate nodules for analysis, at 8 to 60 meter intervals, consistently with global Aptian-Albian carbon isotope chemostratigraphy (i.e., samples from intervals in which regional trends in organic δ 13 C were consistent with higher-resolution records of Bralower et al. (1999) and Menegatti et al. (1998)). Nodules for analysis were sampled from well below channel sands and surface paleosol horizons in outcrop (i.e., sampled from paleosol B horizons; Tabor and Myers, 2015) to help avoid potential surficial biases on carbonate 115 (e.g., radiative heating in soil depths <50 cm; Burgener et al., 2019). Paleosols from which nodules are sampled are fine grained throughout the section which suggests suitability for clumped isotope-based MAAT interpretation (e.g., Kelson et al., 2020).
Thin sections were cut from hand samples for petrographic analysis and cathodoluminescence (CL) imaging to aid in the 120 identification and isolation of primary carbonate nodule material. Briefly, thin sections were inspected for environmental indicators and microfabrics, and photographed under plane-polarized and cross-polarized light (PPL and XPL, respectively) using an Olympus BX43P petrographic microscope with a SC50 Olympus camera. Thin sections were then CL imaged using a Relion Industries Reliotron III cold-cathode chamber, with operating conditions consisting of a rarified helium atmosphere at 50 milliTorr, accelerating voltage of 10 kV, and beam current of 0.5 mA. Macroscale imaging through the 50 mm top 125 window of the chamber was carried out using a 16 Mpx Canon EOS SL1 DSLR camera with a macro lens suspended over the CL chamber. CL imaging was used to detect any heterogeneities in cation substitution which may indicate alteration, as Mn 2+ tends to substitute for Ca 2+ in reducing conditions generating bright orange luminescence (Habermann et al., 2000;Cazenave et al., 2003).

130
Once located in thin section using petrography and CL, primary nodule carbonate was mapped onto the corresponding thinsection billet, and sampled using a dental drill. In some samples, suspect non-primary carbonate material (e.g., spar) was also   (Brand et al., 2010) were used in the initial steps of data reduction, following recommendations of (Daëron et al., 2016) and 150 (Schauer et al., 2016). We then applied an acid correction factor (0.088‰) appropriate for use with values calculated using IUPAC parameters (Petersen et al., 2019). International standards (i.e., ETH1, ETH2, ETH3, ETH4, IAEA-C1, IAEA-C2, Merck and NBS19) were utilized to further correct Δ47 values. Potentially contaminated data was culled (e.g., sample analyses which exhibit Δ48 excess that tracks variability in Δ47; see Supplementary Material). 155 X-Ray Fluorescence (XRF) measurements were carried out on samples from horizons that appear to be well-developed paleosols, specifically horizons interpreted as B-horizons. Analysis was completed with a Rigaku Primus II WD-XRF spectrometer at the University of Texas at San Antonio. Raw X-ray intensities were calibrated by the analysis of eight USGS certified elemental standards (BIR-1a, COQ-1, DNC-1a, GSP-2, RGM-2, SBC-1, STM-2, W-2a), with an RSD value of 0.036%. Weight percentages were converted into molar weights before application of a chemical index, following Sheldon 160 and Tabor (2009). Al2O3, CaO, and MgO are the oxides used for calculation of the CALMAG (Nordt and Driese, 2010) chemical weathering index (see following section for parameter calculation and proxy details).

Quantitative proxies
Clumped isotopes (i.e., Δ47) have been successfully utilized to estimate temperature in carbonates, leveraging the thermodynamically controlled abundance of isotopically heavy 13 C and 18 O bonded isotopes Schauble et 165 al., 2006). This approach has an advantage over δ 18 O-based temperature estimates, as other controlling variables (e.g., δ 18 Omw) need not be estimated. Δ47 values are translated into calcification temperature following the calibration of Petersen et al. (2019) and we define our temperature uncertainty as 1σ of replicate analyses. Additional temperature calibration approaches (i.e., Ghosh et al., 2006;Bonifacie et al., 2017) and calculation details (i.e., R code for data analysis) are available in the Supplementary Material. However, for this study, in subsequent calculations and figures, we opt for Petersen 170 et al. (2019) Δ47 values and calibration temperatures calculated using the following relationship: Δ47 = (0.0383 ± 1.7 −6 ) × (10 6 / T 2 ) + (0.258 ± 1.7 −5 ) (1) Groundwater δ 18 O is derived from the oxygen isotopic composition of precipitation which is ultimately controlled by factors such as temperature, amount, continentality and seasonality. It can be further modified by processes such as evaporation in 175 paleoenvironments which experience wet/dry cycles. δ 18 O of groundwater (δ 18 Ow) can be determined for pedogenic carbonate calcification once temperature is known and δ 18 Ocarb is measured following Friedman and O'Neil (1977): To estimate mean regional precipitation for the study interval and determine shorter-term precipitation variability in our 180 record, we use the bulk geochemical compositional proxy CALMAG (Nordt and Driese, 2010), which utilizes the gains and losses of elemental oxide abundances as a result of weathering in vertisols. The concentration of aluminum oxide, calcium oxide and magnesium oxide are estimated using XRF and the CALMAG parameter is determined: Mean annual precipitation (MAP) is then determined from the CALMAG parameter based on the Nordt and Driese (2010)  185 calibration: Paleosols have been widely utilized as archives to determine the past concentration of atmospheric pCO2 (Ekart et al., 1999;Cerling, 1991). While requiring a number of assumptions, soil carbonate nodule δ 13 C, when used in tandem with estimates 190 from other proxies (e.g., MAP from CALMAG and respired soil δ 13 CCO2 from δ 13 Corg), provide many of the most robust estimates of Cretaceous atmospheric pCO2 outside of a stomatal approach (Franks et al., 2014), especially because paleosol carbonate nodules are abundant in the rock record. The soil carbonate paleobarometer uses a diffusion model in which atmospheric pCO2 (δ 13 Ca) and respired CO2 from soils (δ 13 Cr) are the dominant controls on soil CO2 (δ 13 Cs) following the mixing model of Cerling (1991) in terms of δ 13 C (Ekart et al., 1999). The relative isotopic influence of atmospheric versus 195 respired soil CO2 on soil CO2 (i.e., the source CO2 for calcite) will therefore be controlled by the concentration of CO2 in the atmosphere, if the concentration of the soil-derived component of total gas at depth, S(z), is accounted for following Ekart et al. (1999): pCO2 = S(z) × ((δ 13 Cs − 1.0044 × δ 13 Cr − 4.4 / (δ 13 Ca − δ 13 Cs)) (5) δ 13 Cs can be determined from δ 13 Ccarb, assuming temperature-dependent fractionation (here we use Δ47-based temperature) 200 between gaseous soil CO2 and carbonate (Romanek et al., 1992). Suarez et al. (2018) correlated sections in this study to bulk carbonate surface marine sections using δ 13 C of organic carbon. We estimate atmospheric δ 13 C (i.e., δ 13 Ca) from a marine section correlated chemostratigraphically with the White Pagoda Site (i.e., Peregrina Canyon, Mexico of Bralower et al. (1999) correlated to White Pagoda by Suarez et al. (2018)), applying a δ 13 CDIC (i.e., δ 13 C of marine dissolved inorganic carbon, DIC) to δ 13 Ca fractionation of -8.23‰ consistent with "greenhouse climate" carbon cycle simulations (i.e., Zeebe, 205 2012), and assuming bulk carbonate δ 13 C for the Peregrina Canyon section is representative of global surface DIC δ 13 C. For δ 13 Cr, we apply the bulk sedimentary organic carbon δ 13 C values of Suarez et al. (2018).
In addition to estimates of δ 13 C for the three carbon reservoirs outlined above, the term S(z), or the depth-dependent contribution of soil-respired CO2, must be determined to compute atmospheric paleo-pCO2. While this term is a significant 210 source of uncertainty due in part to a large range of potential past environmental conditions, Cotton and Sheldon (2012) hypothesized a relationship between summer minimum S(z) and MAP using observations of modern soils: Here, we apply their relationship to compute S(z) from our CALMAG-based MAP estimates. It is important to note that the relationship defined by Cotton and Sheldon (2012) uses a dataset which does not include humid climate soils or vertisols, 215 and it is therefore cautiously applied and discussed in terms of paleoenvironmental influence on our paleo-pCO2 estimates (i.e., we evaluate our atmospheric pCO2 record against a large range in S(z)).

Petrography
Based on carbonate petrography we recognize two distinct microfacies in our samples and split samples into two groups 220 (microfacies (i) and (ii)) to evaluate the origin of stable isotope values (primary vs. secondary; depositional environment) ( Fig. 2; Table 1). Microfacies (i) is characterized by distinct nodules which originated from primarily clayey horizons, consisting of dense micrite with abundant root traces and fractures filled with sparry calcite and microspar calcite (Fig. 2).

Traditional stable and clumped isotopes
Stable isotopes of drill spot samples show a high degree of intrasample homogeneity (Fig. 4). Measurements between 240 University of Kansas and University of Colorado, Boulder are largely consistent with comparable precision (Tables 1 and 2; Fig. 4). δ 13 C values range from −8‰ to −3‰ and δ 18 O ranges from −12‰ to −6‰ for carbonates measured in this study.
Sample 3B-021 displays the most heavy-isotope enriched δ 13 C and δ 18 O values, with δ 13 C more than 2‰ and δ 18 O more than 1‰ greater than all other samples (Fig. 4), despite the relative isotopic low in the δ 13 Corg curve which results in a large Δ 13 C for that sample (Table 1). Carbonate samples tend to be isotopically homogenous (2σ £ 0.6‰ for all sample δ 13 C and δ 18 O, 245 with only 2 samples with 2σ > 0.3‰; Table 1) following Cotton and Sheldon (2012), who proposed a requirement of 2σ < 0.5‰ for δ 13 C and δ 18 O for all samples applied to paleo-pCO2 reconstructions. We discern no relationship between δ 13 C and δ 18 O of carbonates, nor grouping of microfacies by stable isotopic composition (e.g., Fig. 4; Table 1).
Clumped isotope (Δ47) mean sample values range from 0.707 to 0.732 (Table 2) which, following the Petersen et al. (2019)  250 calibration, translates to temperatures ranging from ~10 to 20 ºC, with an average temperature of 14.9 ºC for the entire record. Transient cooling of ~2 to 4 ºC (i.e., down to 11.1 ºC) is observed in the C10 carbon isotope interval, with the warmest temperature occurring immediately following the C10 interval (i.e., warms to 18.8 ºC; Fig. 5).

CALMAG
CALMAG values for all measured samples range from a low of 2% to a high of 70%. Lowest values are either samples that 255 were not identified as B-horizons or likely immature soils which yield values inapplicable to range in calibration (CALMAG less than ~35%; Table S4). If only B-horizon samples applicable to the range in the Nordt and Driese (2010) calibration are considered, maximum variability in CALMAG is ± 12% (Table 3; Table S4). This translates to MAP variability of ± 270 mm/yr over the interval, with mean MAP of 641 mm/yr (i.e., mean CALMAG of 47.5%) for paleosols in which clumped isotopes were also measured (  1 and 2; Fig. 4) indicating primary carbonate was successfully sampled from nodules for clumped isotope analyses (i.e., primary carbonate isotopic composition characterized by drill spot measurements at KU match values from CUB clumped measurements). δ 13 C 265 in carbonate nodules is controlled by soil water DIC which, through time, is ultimately controlled by variation in the other exogenic carbon reservoirs. Carbonate δ 18 O is reflective of regional meteoric water and temperature. Though much coarser resolution, our carbonate δ 13 C largely follows δ 13 Corg which has been tied to global variations in the carbon cycle (Suarez et al., 2018;Ludvigson et al., 2010;Ludvigson et al., 2015;Heimhoffer et al., 2003;Ando et al., 2002), suggesting both carbonate and organic records at the site track global variability in the carbon cycle originally described in Menegatti et al. 270 (1998) and Bralower et al. (1999) (Fig. 5) (e.g., δ 13 Ccarb is highest in the C10 interval). We observe no clear grouping of carbonate stable isotopes by microfacies and all samples contain pedogenic features. This suggests δ 13 Ccarb tracks global variations in the carbon cycle, and δ 18 Ocarb values are reflecting δ 18 O of regional precipitation once temperature is considered.

Interpreting paleoenvironmental biases in Δ47-based temperatures 275
Macroscopic features described in Suarez et al. (2018) along with traditional carbonate petrography suggest a paleoenvironment which experienced wet/dry cycles. These features include redoximorphic color mottling, gilgai structures, fracturing pervasive to varying degrees in carbonate nodules, microspar and spar recrystallization present in voids/fractures, Mn staining and root traces ( Fig. 2 and 3). Microscopic features are consistent with facies interpretations of Suarez et al. (2018) which suggest fluvio-palustrine paleoenvironment. Rhizoliths (i.e., calcified root structures) in nearly all nodule 280 samples (e.g., Fig. 2) indicate that vegetation was present and the carbonate nodules are indeed soil-formed in subhumid to semiarid conditions (Zhou and Chafetz, 2009). Indeed, mean MAP derived from our CALMAG proxy record suggests 712 mm/yr (respective minimum and maximum MAP of 476 and 984 mm/yr for the interval; Fig. 5) and Δ47-based temperatures range from 11.4 ± 4.8 ºC to 18.8 ± 2.2 ºC, consistent with the subhumid to semiarid environments in which soil carbonates commonly form (Zhou and Chafetz, 2009;Birkeland et al., 1999;Breecker et al., 2009). 285 Understanding the timing of carbonate formation in soils is important for interpretation of δ 13 C, δ 18 O and Δ47. The solubility of calcite is the primary controlling factor on carbonate formation, and it is significantly affected by soil CO2 concentration.
Because CO2 concentration is lower in warmer conditions, and drier conditions result in greater concentration of ions, calcite precipitation tends to occur during warm, dry conditions. Numerous early studies have suggested warm season bias in soil 290 carbonate formation and thus the Δ47-derived temperatures (Breecker et al., 2009;Passey et al., 2010). Recent work of Kelson et al. (2020) suggests this may not always be the case for a number of reasons. The presence of vegetation (suggested https://doi.org/10.5194/cp-2020-152 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. by abundant root traces) may shade the soil surface from solar radiation. However, Burgener et al. (2019) and Kelson et al. (2020) found that this effect is rare, and samples for this study were collected from paleosol horizons deep enough (i.e., > 50 cm) to be buffered against the effects of radiative heating (i.e., Burgener et al., 2019). Seasonality of precipitation, 295 evaporation, and evapotranspiration likely affects the degree to which a warm season temperature bias may occur. In a study of modern soils in North America, Gallagher and Sheldon (2016) suggested that only continental climate with rainy seasons in the spring had summer temperature biases. Suarez et al. (2011b) suggested that lower than expected temperatures of Mio-Pliocene soil carbonates from the Chinese Loess Plateau may be the result of a monsoon climate in which the rainy seasons occur during the warmest part of the season and conditions for calcite precipitation occurs prior to or after the warm season. 300 These studies suggest that carbonate nodule clumped isotope-based temperatures revealed from the Xinminpu Group likely represent lower temperatures than mean warm season. In addition, mean clumped isotope-based temperature over the study interval (14.9 ºC) matches Aptian-Albian MAATs derived from phosphate δ 18 O in dinosaur teeth from similar paleolatitudes in Asia (i.e., 15 ºC for Xinminpu group; Amiot et al., 2011). However, our mid-latitude continental interior temperatures 305 reflect the temperature of calcite precipitation and may be biased towards the time of year during which a region experiences its first month without water storage, which varies by regional climate (Gallagher and Sheldon, 2016). Given our paleoenvironmental interpretation of wet-dry seasonality which resulted in vertisol formation at our study location, and proxy-based estimates of MAAT and MAP, the paleoenvironment is likely best-represented by either the "continental" or "semi-arid monsoonal" climates of Gallagher and Sheldon (2016). We note that the modern soil type for the settings of 310 Gallagher and Sheldon (2016) consists of mollisols and thus may not be representative of the vertisols in which nodules used in this study formed. Their "continental" model indicates a decline in water storage in July/August which tends to bias carbonate formation to warmer values. In contrast, the "semi-arid monsoonal" model shows a decrease in water storage in April resulting in a slight cool season bias. However, cool season biases tend to be much smaller in magnitude (less than 4 ºC) than warm season biases (as much as 24 ºC) (Kelson et al., 2020). Therefore, regardless of the interpretation of seasonal 315 biases, our mean temperature based on clumped isotopes (14.9 ºC) suggests very cool conditions in the mid-latitude Asian continental interior during the Aptian-Albian. Any potential warm season bias on our temperature results is unlikely as it would suggest even cooler conditions inconsistent with combined proxy observations. Indeed, the temperatures calculated here are consistent with other regional paleotemperature proxy observations (e.g., Amiot et al., 2011), and counter to the predominantly warm greenhouse climate of the Cretaceous (Föllmi, 2012). 320

Latitudinal gradients of temperature and δ 18 Omw for the Aptian-Albian
Clumped isotope-based temperatures for the White Pagoda site indicate a mean record value of 14.9 ºC which is equivalent to δ 18 Op-based temperature estimates (15 ºC) carried out on dinosaur teeth from formations within the same group (Xinminpu) in NW Asia (Amiot et al., 2011). The groundwater δ 18 O based on our combined clumped isotope and carbonate isotope analyses range from −11.54 to −6.69‰ (VSMOW) and average −9.47, which is somewhat lower than the values of 325 Amiot et al. (2011) Modern climate observations of the study site indicate cool, dry conditions with mean δ 18 Omw of −9.31‰ and −7.66‰ in nearby Zhangye and Lanzhou, respectively (IAEA/WMO, 2020). Largely due to the influence of regional topography (study location elevation: ~1500 m), present day precipitation averages < 100 mm/yr and MATs indicate locally cooler 330 temperatures (9.0 ºC and 10.5 ºC in Zhangye and Lanzhou, respectively; IAEA/WMO, 2020) relative to global zonal averages (15.0 ºC; Rozanski et al., 1993). Aptian-Albian temperatures may similarly be influenced by regional paleotopography, though topographic reconstructions for Asia during the Aptian-Albian are lacking, limiting speculation.
Generally, proxy-based temperatures and δ 18 Omw for the Xinminpu Group tend to fall within zonally averaged general 335 circulation GENESIS-MOM model results (Zhou et al., 2008) given the large range in possible site paleolatitude during the Aptian-Albian. For example, paleogeographic reconstructions indicate paleolatitudes ranging from ~35ºN to ~48ºN for the White Pagoda Site during the Aptian-Albian (Lin et al., 2003;Matthews et al., 2016;Torsvik et al., 2012), which corresponds to simulated temperatures ranging from 9 to 19 ºC and simulated δ 18 Omw ranging from −11.8 to −6.7‰ (Zhou et al., 2008). 340 Combining our new temperature and δ 18 Omw data with that compiled in Amiot et al. (2011), we re-cast latitudinal temperature and δ 18 Omw gradients according to the paleogeography applied in Amiot et al. (2011) (Lin et al., 2003) and using an updated paleogeography based on Matthews et al. (2016) (i.e., Gplates). The updated paleogeography results in higher Early Cretaceous paleolatitudes for all Asian sites included in this compilation (Supplementary Material; Table S1), 345 including a more than +13ºN shift for the Xinminpu group sites (Fig. 6). When placed on the paleogeography of Lin et al. (2003), proxy-based temperature reconstructions for Asia indicate a cool climate relative to latitudinal models of temperature and hydrology (i.e., land surface gradients compiled in Suarez et al. (2011a) including: leaf physiognomy-based gradients of Spicer and Corfield (1992), cool and warm Cretaceous gradients of Barron (1983), and GENESIS-MOM general circulation model gradients of Zhou et al. (2008)). For example, temperature data falls below even the coolest Cretaceous modeled 350 gradient (i.e., Barron, 1983) despite agreement between proxy δ 18 Omw in mid-latitude continental Asia and the modeled cool Cretaceous (Fig. 7 panels a and b). However, if Matthews et al. (2016) paleolatitudes are applied, proxy-based temperatures become better aligned with Cretaceous modeled temperature gradients (Fig. 7 panel c). Additionally, the updated paleolatitudes tend to offset δ 18 Op-based δ 18 Omw estimates in a positive direction relative to the modeled cool Cretaceous δ 18 Omw gradient, aligning these data with a flatter, more modern appearing δ 18 Omw gradient (Fig. 7 panel d) evaporatively-enriched leaf water in herbivores provides one possible mechanism for 18 Op-enrichment (Levin et al., 2006).
Alternatively, the range in paleolatitudes presented here demonstrate the large degree of uncertainty with regards to Early 360 Cretaceous paleogeographic reconstructions of Asia (Supplementary Material; Table S1), which may be driving Aptian-Albian proxy-model disparities.

Atmospheric paleo-pCO2
Cotton and Sheldon (2012) refine procedural guidelines previously established by Cerling and Quade (1993) and Ekart et al. (1999) for use of pedogenic carbonates in reconstructions of atmospheric pCO2 which include maximum limits for Δ 13 C (i.e., 365 δ 13 Ccarb − δ 13 Corg), isotopic heterogeneity and δ 13 C versus δ 18 O covariation. They suggest limiting proxy application to samples with 14‰ < Δ 13 C < 17‰ as modern soils with large Δ 13 C tend to have S(z) values which fall off of the MAP versus S(z) relationship defined by Cotton and Sheldon (2012) and are likely to have been disproportionately influenced by atmospheric δ 13 C. For our atmospheric pCO2 reconstruction, we occluded samples with large Δ 13 C (i.e., samples with Δ 13 C > 18‰; sample 3B-021). We include two samples in our reconstruction (samples 6-003 and 4-038) which have 17‰ < Δ 13 C < 370 18‰ (Table 1). Though this Δ 13 C signature may indicate low-productivity (Cotton and Sheldon, 2012) which can influence the MAP versus S(z) relationship, the presence of abundant root traces in sections 4 and 6 (i.e., Suarez et al., 2018) suggests otherwise. In addition to meeting Δ 13 C criteria, no clear correlation between carbonate δ 13 C and δ 18 O is observed (Fig. 4) and carbonates tend to be isotopically homogeneous ( Fig. 4; Table 4; maximum 1σ of 0.3‰ in all samples for both δ 13 C and δ 18 O). We include two estimates of uncertainty in our atmospheric pCO2 reconstructions to illustrate the influence of S(z) 375 estimates on pCO2: 1) error bars which represent 1σ uncertainty in δ 13 Ccarb and Δ47-based temperatures, and 2) an error envelop which encompasses the prior uncertainty listed in (1) in addition to a range in S(z) for all samples (Table 4). The maximum range in S(z) is set using the relationship of Cotton and Sheldon (2012), applying the maximum MAP value observed in the sections containing samples for pCO2 reconstruction (i.e., 984 mm/yr which translates to S(z) of 5309 ppmv). This maximum value is representative of some maximum modern S(z) values observed in Holocene calcic soils by 380 Breecker et al. (2010) and consistent with summer minimum S(z) values observed in vertisol grasslands by Mielnick and Dugas (2000). Minimum S(z) is set at 2500 ppmv, following the recommended S(z) of Breecker et al. (2010), as this value is consistent with minimum MAP for our record following the relationship of Cotton and Sheldon (2012). As observed previously for the Cretaceous (e.g., Franks et al., 2014), atmospheric paleo-pCO2 derived from pedogenic carbonate stable isotopes tends to lose sensitivity at low atmospheric CO2 concentrations resulting in calculated error which spans negative 385 concentrations. Here, we exclude negative, unrealistic pCO2 values from our record and report these minimums as 0 ppmv (Fig. 5) and note that calculated minimum pCO2 is > −165 ppmv for all samples (Table 4).
Our atmospheric pCO2 reconstruction suggests relatively low (for greenhouse climates) and variable pCO2 over the study interval. This observation is consistent with cool Aptian-Albian temperatures (i.e., MAAT ~15 ºC in midlatitudes as 390 indicated by this study and others). Mean atmospheric pCO2 for the entire record is 396 ppmv and pCO2 generally tracks https://doi.org/10.5194/cp-2020-152 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. temperature variability with low (i.e., < 300 ppmv) pCO2 in the cool C10 interval ramping up section to ~1100 ppmv. Our record is largely in agreement with paleobotanical proxy-based pCO2 reconstructions for the Aptian-Albian, which range from ~500 to 1300 ppmv (Du et al., 2016;Haworth et al., 2010;Passalia, 2009;Aucour et al., 2008). While this study indicates slightly lower pCO2 than other carbon-isotope based records for the Aptian-Albian (e.g., Ekart et al. (1999Wallmann (2001) suggest 700 to 1500 ppmv; Fletcher et al. (2005) suggest 1100 to 1200 ppmv), these records do not all satisfy requirements of Cotton and Sheldon (2012) (e.g., Δ 13 C < 17‰ in the record of Ekart et al. (1999) likely biases to higher atmospheric pCO2), and may lack the sampling resolution to pick up on shorter-term variations.
Additionally, though comparatively offset to lower values, variability in our atmospheric pCO2 reconstruction follows the pattern of Aptian-Albian pCO2 variability observed in other pedogenic and pelagic marine carbonate-based estimates (i.e., 400 gradual decrease in late Aptian with a low at the Aptian-Albian boundary before increasing into the early Albian; Li et al., 2013;Bottini et al., 2015).

Aptian-Albian variations in atmospheric pCO2, climate and the hydrologic cycle
Cooler midlatitude terrestrial temperatures (MAATs of ~15 ºC) are consistent with the post-OAE 1a "cold snap" hypothesis (e.g., Mutterlose et al., 2009) observed in terrestrial (e.g., Amiot et al., 2011) and sea surface temperature records (e.g., both 405 nannofossil indicators and organic GDGT temperature proxy TEX86 show cooling at globally distributed sites; Bottini et al., 2015). Following Friedman and O'Neil (1977), coupled carbonate δ 18 O and Δ47 suggest variations in δ 18 Omw of ± 2.2‰ over the study interval consistent with shifts in the distribution of atmospheric moisture associated with climate change. Hence, our Δ47-based temperature estimates prove useful in interpreting regional climate variability over the interval independent of δ 18 Omw. Climate change-induced variations in Asian continental interior δ 18 Omw during the early Cretaceous would be 410 expected given model results which show shifts in δ 18 Omw of +2 to +4‰ associated with two doublings of atmospheric pCO2 (global average surface warming of 6 ºC) in continental interiors (Poulsen et al., 2007). Our records similarly indicate warming-induced 18 O-enrichment in δ 18 Omw for continental Asia, as atmospheric pCO2, temperature and δ 18 Omw increase following the C10 interval (Fig. 5). While age controls are limited, chemostratigraphic correlations (i.e., Suarez et al., 2018) suggest our record spans several Myr (i.e., C7 to C12 carbon isotope segments after Menegatti et al. (1998) and Bralower et 415 al. (1999); roughly 6 million years). Given the temporal coarseness of our record which likely does not pick up on peak temperature variability, we observe subtle temperature shifts over the interval (i.e., cooling of −2 to −4 °C preceding +4 to +6 °C of warming across the inferred C10 interval), which likely corresponds to the cool interval between OAE 1a and OAE 1b and may include Kilian and/or OAE 1b warmth (e.g., Bottini et al., 2015) around 140 m (Fig. 5).

420
Shifts in the hydrologic cycle reflected in δ 18 Omw and MAP (Fig. 5) track Δ47-based temperatures, suggesting climatecontrolled regional shifts in interior Asian hydrologic cycle. Interestingly, the driest conditions tend to correlate to relative lows in temperature and δ 18 Omw perhaps pointing to variations in the seasonality and/or sourcing of meteoric waters in Asia associated with long-term climate evolution. Compared with background environmental conditions, cooler temperatures (down to 11.1 °C), drier conditions (MAP < 600 mm/yr), and 18 O-depleted meteoric waters are observed in the C10 interval, 425 consistent with models of warming-induced precipitation change for the mid-Cretaceous (e.g., Poulsen et al., 2007) and observations of more widespread dry conditions in Asia during cool Cretaceous intervals (e.g., Hasegawa et al., 2012). Hasegawa et al. (2012) used sedimentological records from Asia to help constrain potential shifts in the descending limb of the Hadley cell related to Cretaceous climate change and compared these observations to of Hadley cell circulation 430 simulations, concluding that as paleo-pCO2 concentration increases, so does the width of the Hadley Cell, but that at ~1000 ppmv paleo-pCO2 and greater, the descending limb of the Hadley cell contracts from 30ºN and ºS to 15ºN and ºS reorganizing the distribution of atmospheric water vapor. This hypothesis is further supported by paleoenvironmental observations (i.e., shifts in lithology and climate-sensitive fossils associated with changes in aridity; Chumakov, 2004;Chumakov et al., 1995), and general circulation models which indicate a 30ºN position of the descending limb of the Hadley 435 cell during Cretaceous greenhouse warmth (Floegel, 2001). The time bins that Hasegawa et al. (2012) investigates are larger than the interval investigated here, but the shifts in atmospheric pCO2 encompass the range of hypothesized thresholds for shifts in Hadley Cell circulation. The relatively cool and dry conditions and 18 O-depleted meteoric waters during the C10 interval (potentially associated with the "cold snap"; Mutterlose et al., 2009) may reflect on a shift in climate and Hadley cell circulation, driven by a decrease in atmospheric pCO2 (e.g., Hasegawa et al., 2012). Indeed, other sedimentological evidence 440 (e.g., glendonites) provide further support for relatively cold conditions at high northern latitudes associated with this interval (Vickers et al., 2019). For our records,, atmospheric pCO2 drops during the C10 interval from pre-and post-segment values within error of 1000 ppmv, to < 350 ppmv in the C10 interval ( Fig. 5; Table 4), well below the critical threshold proposed by Hasegawa et al. (2012). As atmospheric pCO2 increases from the low in the C10 interval to a peak just after, precipitation and temperature increases similar to the proposed climate shifts that Hasegawa et al. (2012) suggest, in a 445 "supergreenhouse" mode. Variability in δ 18 Omw and MAP is observed throughout our study interval, however, which may indicate either multiple fluctuations in Hadley cell circulation across the interval or background regional variability in the early Cretaceous hydrologic cycle in continental interior Asia.

Conclusions
In summary, new continental Asia midlatitude multi-proxy records of Aptian-Albian carbon cycle, climate and hydrologic 450 cycle suggest cool conditions in early Cretaceous midlatitudes (mean of 14.9 ºC; 35ºN to 48ºN paleolatitude depending on applied paleogeographic reconstruction) relative to background Cretaceous greenhouse warmth, consistent with our estimated atmospheric pCO2 (mean value of 396 ppmv) calculated using carbon isotopes in pedogenic carbonates and previous regional MAAT observations (Amiot et al., 2011). Variations in the hydrologic cycle (i.e., decreases in MAP and δ 18 Omw) are associated with transient cooling (−2 to −4 ºC) during the C10 carbon isotope high, consistent with general 455 circulation models which suggest differences in temperature, MAP and δ 18 Omw similar in magnitude to our observations https://doi.org/10.5194/cp-2020-152 Preprint. Discussion started: 21 December 2020 c Author(s) 2020. CC BY 4.0 License. associated with one to two doubling(s) (or in terms of cooling, halving(s)) of atmospheric pCO2. These new paleoclimate parameters may be useful to future climate modeling efforts and for understanding potential variability (cooling and warming; shifts in precipitation) in an otherwise greenhouse climate.