The Pliocene epoch is considered the most recent analog of modern warming because CO2levels were similar to the present. To explore the carbonate minerals formed in the warmer Pliocene epoch, we studied two continuous sections of the Red Clay Formation on the Chinese Loess Plateau (CLP) by X-ray Diffraction (XRD), Fourier Transform Infrared Spectroscopy (FTIR), Scanning Electron Microscopy (SEM) and Stable Isotope Mass Spectrometry. The Red Clay Formation on the CLP exhibits diagnostic FTIR absorption features of calcite and protodolomite. This allowed quantification of the two carbonate minerals by the FTIR method. Using the FTIR method we estimate the average concentration of protodolomite in Bajiazui is 3.6% whereas the Duanjiapo section is 6.0%. Protodolomite occurrence is more consistent and the concentration is higher from ∼6.5–4.2 Ma B.P. and decreases markedly from 4.2–2.6 Ma B.P. Red Clay protodolomite is depleted in bothδ13CPDB and δ18OPDB, ranging from −4.1‰ to −10.4 and from −6.7‰ to −11.6, respectively, and has a slightly higher δ18O value than the calcites. SEM observations show that Red Clay protodolomite is composed of euhedral rhombic crystals that range from 1–20 μm in diameter, grow into the soil voids and coexist with authigenic calcite and palygorskite. These observations imply that the protodolomite grew in situ and is authigenic from pedogenesis. Dolomitization in the Red Clay sequence appears to be the result of overcoming kinetic barriers. We propose that in the Red Clay a warm climate with seasonal dry conditions leads to the formation of calcrete from soil pore waters thereby enriching the pore solutions with respect to Mg2+ and significantly increasing the Mg/Ca ratio bringing about the formation of protodolomite.
 The thick and nearly continuous eolian deposits of the Asian dust on Chinese Loess Plateau (CLP) are constituted of the alternating Quaternary loess-paleosol horizons [Liu, 1985] and the underlain Red Clay Formation that started since about 7 Ma [Ding et al., 1999; Sun et al., 1997]. As one of the most abundant and also the most variable minerals in the eolian dust deposits [Liu, 1985], carbonate minerals have been systematically investigated in the Pleistocene loess-paleosol sequence [Chen et al., 1996; Liu, 1985; Sheng et al., 2008]. Work on the identification of Pleistocene loess carbonate minerals species indicates that calcite is the primary mineral. Loess carbonate has two origins: one is detrital and the other is authigenic [Liu, 1985]. Authigenic carbonate formed through pedogenesis, after the deposition of the eolian dust that forms the loess [Han et al., 1997; Sheng et al., 2008]. However, in contrast to the detailed work on carbonate minerals from Pleistocene loess [Chen et al., 2007; Han et al., 1997; Li et al., 2007; Rowe and Maher, 2000], little work has been done on the carbonates from the underlying Pliocene Red Clay. The Red Clay soil complex contains couplets of pedogenic B horizons and horizontal carbonate nodule horizons [Ding et al., 1999]. Few attempts have been made to determine Red Clay carbonate genesis or distinguish differences between carbonate species. As a result, the lack of the relative abundance of different carbonate mineral species in the Red Clay is a potential hindrance to subsequent research.
 The two common carbonate species found in the eolian deposits on CLP are calcite and dolomite [Liu, 1985]. But, in the red clay protodolomite is also present. The objective of this study is to provide details of the temporal distribution of proto-dolomite on CLP from late Miocene to Pliocene including its relative abundance and origin as indicated by micro-morphology and stable isotope geochemistry from two typical Red Clay sequences.
2. Geological Settings and Sampling
 Two Red Clay sections, one at Bajiazui and one at Duanjiapo have been investigated in a north to south transect on the CLP (Figure 1). These sections are about 2° latitude from each other. Both the mean annual temperature (MAT) and mean annual precipitation (MAP) of southern section is higher than the northern section (Table 1) according to modern meteorological observations [Lu et al., 1998; Wang et al., 2006].
Table 1. Climatic and Geographic Conditions of Duanjiapo and Bajiazui Red Clay Sections
 The southern section is located at Duanjiapo in Lantian County (Table 1). This Red Clay sequence has an age range from 2.58–7.21 Ma B.P. [Chen et al., 2001] and the section has been subjected to strong pedogenesis. The soil ranges in color from brown to light red. Dozens of carbonate nodule horizons were identified in the field. Calcareous pseudomycelium, a pedo-feature that describes a branching, thread-like morphology [Khokhlova et al., 2008], have been reported from a depth of 1.6 m in the upper portion [Sun et al., 1997] of the section. This is a feature that is believed to result from downward leaching from overlying sediments.
 The northern section, the Bajiazui section, is located 16 km north of Xifeng County (Table 1). This Red Clay sequence has a temporal range from 2.58–6.57 Ma B.P. [Sun et al., 1997]. Compared to the Duanjiapo section the degree of pedogenesis is less. The soil color ranges from light red to reddish yellow. In the middle to lower portions of this section field work identified 4 carbonate horizons about 0.5–1 m in thickness. A minor amount of carbonate pseudomycelium have been reported at the depth of 10.9 m in the upper portion [Sun et al., 1997].
3. Experiments and Analyses
3.1. X-Ray Diffraction (XRD)
 The mineralogy of Red Clay bulk samples was determined by XRD from randomly oriented samples. A Rigaku D/max was used for XRD analysis with Cu-Ka radiation, a voltage of 40 kV, and a current of 25 mA. Prior to analysis each sample was oven-dried at 40°C for 2 days and ground to <63μm using an agate mortar to thoroughly disperse the minerals. No chemical pre-treatment was employed. Samples were scanned from 3° to 70° with a step size of 0.02°.
 XRD is commonly used to identify minerals in geological samples. But, XRD has many disadvantages including long analysis time, a requirement that the tested material be well crystallized, and generally a detectability limit of over 1% by weight. As a result, in this study XRD was applied primarily to assist in the detection of protodolomite in our soil samples and was not used as the primary detection method. As noted below, XRD analyses indicate that all the dolomite found in these samples is a crystalline protodolomite, that is, a calcium rich dolomite with a disordered lattice in which magnesium and calcium ions occur in the same layer instead of in alternative layers as in ordered dolomite. Hence, the XRD method was used qualitatively, not quantitatively, as a tool to check our quantification system based on FTIR (see below).
3.2. Scanning Electron Microscopy (SEM)
 In order to examine Red Clay soil micro-morphology, SEM analysis was carried out using a JSM-6700F field emission SEM at a magnification of 10,000x. Elemental analysis was performed with an attached Oxford Instruments EDX system to distinguish protodolomite, calcite and palygorskite in the same soil sample. The tested samples were Pt coated.
 FTIR was applied to identify and quantify calcite and dolomite. Compared to XRD methods, FTIR has shorter analysis time per sample, less than 2 min, and, most importantly, it has a much lower detection limit [Ji et al., 2009; Vagenas et al., 2003]. Furthermore, because FTIR is based on molecular vibration, it is not affected by the degree of crystallinity and can be applied to amorphous as well as semi-crystalline and crystalline material. FTIR spectra display IR absorption features resulting from molecular vibration of the chemical components in a sample thereby allowing quantitative determination of specific minerals [Smith, 1995, 1998].
 Sample preparation and analysis followed procedures described in Ji et al. . Each sample analyzed with the FTIR was ground in agate mortar for more than 15 min and dried in an oven at 105°C for 24 h to remove water. After grinding and drying, 200 mg of the powdered sample was packed into a cylindrical cup. Prepared samples were then analyzed with a Thermo Nicolet FTIR 6700 with a diffuse reflectance attachment at the mid-infrared range from 400 to 4000 cm−1. Each sample was scanned 128 times to optimize spectral quality. The spectral background correction material used in our analysis was analytical grade KBr. The reflectance intensity of pure KBr was set to 100% and the reflectance intensity of tested samples was corrected to a percentage relative to the KBr.
3.4. Vacuum Gasometric Measurements
 We choose a series of samples from the two sections which varied significantly in their calcite FTIR spectral signal for vacuum gasometric analysis. Prior to analysis we determined that these samples were dolomite-free based on a zero value of the FTIR spectral absorption band indicative of dolomite, 728 cm−1 (see below). The 19 samples we chose for calcite calibration had an FTIR band indicative of calcite, 713 cm−1, that ranged from 0–24.6 (Table A1). We then measured the samples' weight percent calcite using the gasometric technique described by Jones and Kaiteris . The samples were ground for 15 min to <63 μm and dried in oven for more than 24 h at 50°C. Then 200 mg of the ground sample was reacted with 85% phosphoric acid for 90 min under vacuum. The pressure generated by the reaction was measured on a pressure manifold. Percent calcium carbonate was calculated by comparing the pressure generated from samples to the pressure generated by pure CaCO3, after correcting temperature to standard conditions of 25°C.
 To calibrate the FTIR for protodolomite we chose 15 samples that exhibited significant variation in the protodolomite absorption band at 728 cm−1 (Table A2). The samples were ground to <63 μm and reacted with 0.2 M acetic acid for 4 h to remove any calcite; during this reaction dolomite showed little or minor dissolution. To verify that the calcite had been totally removed from the samples, we confirmed that the FTIR calcite band at 713 cm−1 was zero. After removal of calcite, the powder was dried in the oven for more than 24 h at 50°C. Two hundred mg of the dried powder was then reacted under vacuum in a 90°C water bath with 85% phosphoric acid for 15 min to make sure that protodolomite was completely dissolved [Hodell and Curtis, 2008]. The pressure generated in the reaction was measured by a pressure manifold after natural cooling of the reaction vessels to ambient laboratory temperature. Percent protodolomite was calculated by comparing the pressure generated by our samples with the pressure generated by pure dolomite after adjusting to standard conditions of 25°C.
3.5. Stable Isotope Measurements
 We determined the stable isotopic composition of calcite and protodolomite for BJZ 3185, BJZ 3520, DJP 2609, DJP 2615 and DJP 2621. The samples were ground to <63 μm. For calcite, 200 mg of powdered sample was reacted directly with 100% phosphoric acid for 20 min at 50°C. However, for protodolomite the ground sample was pretreated with 0.2 M acetic acid for 4 h to remove calcite. During this pretreatment the protodolomite exhibited a minor degree of dissolution. We used the FTIR spectral absorption band at 713 cm−1, which is caused by the calcite CO32−vibration, to make sure that all the calcite contained in the powder was removed. After removing calcite, 200 mg of the calcite-free powder was reacted with 100% phosphoric acid for 24 h at 50°C to dissolve the dolomite. The carbon dioxide from both the calcite and protodolomite was gathered and analyzed with a Thermo Finnigan DeltaPlus at the State Key Laboratory of Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences. The δ18O and δ13C data reported here are relative to the PDB standard. The precision of the analysis is less than ±0.1‰ (1σ) and ±0.04‰ (1σ) for δ18O and δ13C, respectively.
4.1. XRD Results
 XRD results indicate that the typical mineral composition of Red Clay protodolomite-rich layers includes quartz, K-feldspar, plagioclase, calcite and protodolomite (Figure 2). Red Clay proto-dolomite has a d104 that is 0.2902 nm, not 0.2886 nm typical of stoichiometric dolomite. It does not have the ideal Mg:Ca = 1:1 or the ordered alternating Mg and Ca layer structure. The strong protodolomite intensities, which are shown in Figure 2, are consistent with the FTIR 728 cm−1 protodolomite strong absorption feature (Figure 4b).
4.2. SEM Micromorphology of Red Clay
 We analyzed a nodule, sample DJP 2633, collected from lower Duanjiapo section (53.6 m) that is high in protodolomite with the SEM. Elemental distribution results from this nodule are listed in Table 2. The nodule reveals a primary pedogenic texture. Quartz and feldspar in the nodule are coarse-grained and interlayered with the clay matrix. The soil matrix is poorly consolidated and has high porosity; void/porosity space appears to have grown during protodolomite crystal formation (Figure 3e). Layered silicate-clay minerals are also common (Figure 3d).
Table 2. Elements of Coexisting Protodolomite and Palygorskite by EDX
 Groups of rhombic euhedral protodolomite, designated as Dol in Figure 3, are common as are fine, curly, trichoid rod coatings (Figures 3a and 3c). Previous work [Chen et al., 2005] confirmed that the rod coating was palygorskite. The rhombic euhedral protodolomite, ranges from 1 to 20 μm in diameter (Figures 3e–3g) and depends on the void space size to grow (Figure 3a). Protodolomite diameters are variable, but commonly they are all rhombic euhedral in shape. The layered clays and the fine, curly, trichoid rod clays coexisted (Figure 3h).
4.3. FTIR Results
4.3.1. Identification of Calcite and Dolomite by FTIR
 In the mid-infrared range (400–4000 cm−1), our results show 4 basic types of molecular vibration motions for pure dolomite and calcite (Figure 4a). According to the corresponding vibrational modes, they are: symmetric stretching (v1), out of plane bending (v2), asymmetric stretching (v3) and in plane bending (v4) [White, 1974]. Their combined vibrational modes for CO32−, (v1 + v3) and (v1 + v4), can also be detected. Among all the modes detected, the v4 motion mode, 712–732 cm−1, is the most sensitive band for quantifying calcite (713 cm−1) and protodolomite (728 cm−1).
 We used the 728 cm−1 absorption band feature to distinguish protodolomite from calcite. For example, protodolomite absorption features were detected from BJZ410, BJZ1095, DJP1619, DJP2249 and DJP2657, while a minor, but still observable protodolomite signal was detected in BJZ3010 and BJZ2556 (Figure 4b). The same strategy was employed to distinguish calcite (Figure 4b). The calcite absorption feature v4 (713 cm−1) shifts toward a lower wave number because of its much lower Mg content than protodolomite [Bottcher et al., 1997].
4.3.2. Quantification of Calcite and Dolomite by FTIR
 In order to determine the concentration of coexisting calcite and dolomite we employed a linear regression equation with the vacuum gasometric measurements as the independent variable and band areas of the corresponding band in the FTIR spectrum as the dependent variable. In the calcite regression the weight percent calcite as determined gasometrically was estimated using the FTIR 713 cm−1 absorption feature producing an R2 of 0.92 (Figure 5a). Using a similar approach, a protodolomite regression produced an R2 of 0.96 (Figure 5b). These regression results indicate that 713 cm−1 and 728 cm−1 are appropriate FTIR absorption bands for quantifying calcite and protodolomite. The best measuring ranges for these regressions are 0–40% for calcite and 0–25% for protodolomite by weight (Figures 5a and 5b). Beyond these ranges the equation extrapolates and the results are less reliable. Comparison of observed concentrations from the gasometric measurements versus the estimated values from the FTIR equations for both calcite (Table A3) and protodolomite (Table A4) exhibits good consistency. However, the FTIR method of determining protodolomite estimates a consistently lower percentage, possibly the result of minor dissolution of protodolomite during removal of coexisting calcite.
 To check the precision of calcite and protodolomite concentrations estimated using the FTIR absorption band area, XRD intensity ratios, Cal/Q (calcite/quartz) and Dol/Q (dolomite/quartz), were regressed against the estimated dolomite and calcite content using the regression equations above (Figures 5c and 5d). D values of XRD peaks for quartz, calcite and protodolomite are d = 0.334 nm, 0.303 nm and 0.290 nm, respectively. The regression of Cal/Q versus estimated calcite content and Dol/Q versus estimated protodolomite content produces R2s of 0.90 and 0.95, respectively. Therefore, XRD ratios are in a good linear relationship with estimated calcite and protodolomite content. These results verify the regression equations based on FTIR.
4.3.3. Distribution of Protodolomite in Bajiazui and Duanjiapo Red Clay Sections
 Protodolomite and calcite concentrations of the Duanjiapo samples are shown in Figure 6. Among the total of 471 samples, 136 contain protodolomite (Table A5). The protodolomite concentrations is primarily between 0.7–25%, with an average of 6.0%. However, there are eight samples with a protodolomite concentration higher than 25%, which exceeds our regression maximum (Figure 6). The Duanjiapo section was divided into two parts according to the protodolomite concentration: the upper part (0–32 m, 2.58–4.2 Ma B.P.) and the lower (32–61.3 m, 4.2–7.21 Ma B.P., Figure 6). Protodolomite is concentrated mainly in the lower part where the concentration is generally higher than 3% (6.8% average); in the upper part of the section protodolomite concentration averages only about 1.5%.
 Protodolomite and calcite concentrations of the samples from Bajiazui are also shown in Figure 6. Among the 314 samples analyzed, 109 samples contain protodolomite (about 1/3, Table A5). The protodolomite concentrations range between 0.5–24.9%, average 3.6% and increase down the section. Based on variations in protodolomite content, the Bajiazui Red Clay section was also divided into two parts: the upper part (2.9–39.2 m, 2.58–4.2 Ma B.P.) and the lower (39.2–63.12 m, 4.2–6.57 Ma B.P., Figure 6). In the upper part protodolomite averages 2.3%, whereas in the lower part it averages 4.6%.
4.4. Stable Isotope Records of Red Clay
 The δ18O values of our protodolomite samples are ranged from −7.8 to −11.6‰ and averaged at −9.9‰ for DJP (n = 26, σ = ±1‰); for the BJZ section δ18O values ranged from −6.7 to −9.6‰ and averaged at −8.5‰ (n = 17, σ = ±0.65‰). Compared with the protodolomite, the calcite δ18O value is depleted, −10.7‰ (n = 26, σ = ±0.78‰) for DJP and −9.5‰ for BJZ (n = 17, σ = ±0.47‰) (Table 3).
Table 3. δ18O and δ13C Isotope Ratios of Red Clay Protodolomite and Calcite
δ13C values of our protodolomite are −8.8‰ (n = 26, σ = ±1‰) for DJP and −5‰ (n = 17, σ = ±0.4‰) for BJZ. For calcite, δ13C values are −9.3‰ (n = 26, σ = ±0.7‰) for DJP and −5.4‰ (n = 17, σ = ±0.3‰) for BJZ (Table 3).
5.1. Origin of Red Clay Protodolomite: Indications From SEM and C-O Isotopes
 Micromorphological characteristics of Red Clay protodolomite reveal that it is drastically different from the Pleistocene loess dolomite. Pleistocene loess dolomites are detrital, originating from rock fragments [Li et al., 2007]. The loess dolomites are >45 μm, well ordered crystals that lack sharp corners or edges; most of them are well rounded suggesting they have endured abrasion due to long distance wind transportation [Li et al., 2007]. Thus, detrital dolomite grains in loess contain information about material in the loess source region. Red Clay protodolomites, in contrast, are poorly ordered and exhibit a perfect rhombic crystal shape (Figures 3f and 3g) that differs from the rounded, transport-influenced shape of dolomite in Pleistocene loess indicating that Red Clay protodolomite is authigenic.
 Besides the implications as to the origin of the protodolomite, SEM analysis also provides direct information about paleoclimate. The existence of micritic rhombic euhedral carbonate crystals (Figures 3f and 3g) indicates pedogenesis [Deutz et al., 2002] and is characterized by the dissolution and recrystallization of carbonate minerals in Red Clay. Additionally, SEM analysis indicates that detrital eolian carbonates are absent in the Red Clay where they were likely dissolved and reprecipitated. In contrast, Pleistocene carbonates are composed of both detrital grains >45 μm and well crystallized, authigenic, fine carbonate grains <2 μm [Sheng et al., 2008]. Hence, the preservation and types of carbonates suggests a stronger primary pedogenic effect in Red Clay than in Pleistocene loess.
 Detrital and authigenic dolomites have contrasting C–O isotopic values [Rabenhorst et al., 1984]. According to previous research on Pleistocene loess from the CLP, loess dolomite was inherited from the mountain belts of northwestern China, whose δ13CPDB isotope ratios range from 0 to 4‰ [Li et al., 2007]. However, our study reveals that Red Clay protodolomite δ13CPDB isotopic ratios range from −4.1 to −10.4‰ (Figure 7). The strong negative δ13CPDB of the Red Clay protodolomite differs significantly from inherited, detrital dolomite in Pleistocene loess and modern desert sand.
 Red Clay protodolomite δ18OPDB isotopic ratios average −9.9‰ in DJP and −8.5‰ in BJZ. The depletion of Red Clay δ18OPDB and their value range show a striking similarity to Pleistocene loess primary pedogenic carbonate (Figure 7), which is mostly calcite. Typical loess authigenic carbonate δ18OPDB ranges from −4 to −10‰ [Chen et al., 1996; Han et al., 1997; Sheng et al., 2008]. The very close δ18OPDB isotopic composition suggests that Red Clay protodolomite and loess primary pedogenic carbonate have a similar origin. Red Clay protodolomite is a primary pedogenic carbonate.
 Red Clay primary pedogenic protodolomite's C–O isotope ratio ranges plot in the meteoric water zone according to the classification of Warren  because both δ13CPDB and δ18OPDB ratios are highly negative. Protodolomite δ18OPDB ratios for DJP averaged −9.9‰ (n = 26, σ = ±1‰), and −8.5‰ for BJZ (n = 17, σ = ±0.65‰). However, under moderately evaporative conditions, typical pedogenic dolomite (that is <2 μm) in the saline soils of Alberta Canada [Kohut et al., 1995], the δ18OPDB ratios were enriched and increased to −3.3‰ (n = 3). In the highly evaporative environments as Qinghai salt lakes [Yu and Kelts, 2002] or South African hypersaline pans [Mauger and Compton, 2011], authigenic dolomite exhibits extremely positive δ18OPDB ratios, averaging about +5.3‰ (n = 2) and +4.1‰ (n = 8), respectively (Figure 7). Thus, as the negative δ18OPDB ratios indicate, Red Clay protodolomite is isotopically different from those formed under highly evaporative conditions and is more closely related to those formed from meteoric water and seasonally dry conditions.
5.2. Formation of Red Clay Protodolomite
 Dolomite is a common occurrence in ancient strata, but is only rarely present in modern carbonate sediments [Holland and Zimmerman, 2000]. Moreover, attempts to synthesize dolomite under the earth's surface temperature and pressure have been unsuccessful [Land, 1998]. These observations produced the long standing geological mystery - the “dolomite problem” - that has attracted geologists' attention for over 100 years. As currently viewed, the principal methods of low temperature dolomitization are the mixing zone model [Badiozamani et al., 1977; Cander, 1994; Humphrey and Quinn, 1989], the sabkha model [Gunatilaka, 1991], and bacterial mediation models [Sánchez-Román et al., 2008].
 However, in spite of the many models of dolomitization, there is a general consensus that the control on dolomitization is a kinetic barrier and not thermodynamic state [Arvidson and Mackenzie, 1999]. From a thermodynamics point of view, the formation of secondary dolomite could be expressed by the reaction equation:
The equation above refers to the direct precipitation of dolomite from solution. Thermodynamically, the basic chemical reaction requirements should be that the IAP of dolomitization fluid is >KSPDolomite. For the equation above, IAPDolomite = [Ca2+][Mg2+][HCO31−]4/[pCO2]2[pH2O]2. Factors such as higher temperature and seasonally dry conditions will decrease soil pCO2 and pH2O and promote the reaction toward right side. Other factors including fluid chemical composition of Mg2+, Ca2+, HCO3−/CO32− and ionic strength will also affect the rate of dolomitization. According to Hardie , the disordered Ca-rich dolomite has a KSPdolomite = 10−16.5 and [Mg2+]/[Ca2+] = 3.31 at 25°C and 1 atm pressure. Fluid supersaturated with Mg suitable for dolomitization is widely distributed on CLP. The largest river on CLP, the Yellow River, has an IAP = 10−10.7 using [Mg2+], [Ca2+] and [HCO3−] data from Gaillardet et al.  and Roy et al.  and an IAP = 10−10.5 [Zhang et al., 1995]. These values are orders of magnitude larger than KSPdolomite = 10−16.5 and therefore thermodynamically favorable. But, dolomitization on CLP mainly depends on kinetic factors, not aqueous thermodynamic equilibrium.
 Among the kinetic factors it has been suggested that elevated Mg/Ca ratio could disrupt the hydration spheres around Mg2+ when the solution is hypersaturated and highly saline [Burton and Machel., 1992]. In addition, temperature is a key factor; dolomite nucleation and crystal growth relies on solution temperature [Arvidson and Mackenzie, 1999]. Other factors, such as microbial mediation are possible. Typical spherical, highly crystalline phase dolomite was synthesized through microbial mediation from 25–45°C [Sánchez-Román et al., 2008].
 One of the striking features of Red Clay protodolomite are the concentrated layers of dolomite (dolocrete) separated by thick horizons of highly concentrated calcite (calcrete) (Figure 6). At low temperatures, the precipitation of calcite and dolomite is competitive [Arvidson and Mackenzie, 1999], and the Mg/Ca ratio determines which species will precipitate [Folk and Land, 1975]. Calcrete or dolocrete commonly forms in arid soils by soil solution/precipitation. In arid areas, Mg2+ and Ca2+ are leached downward from the upper soil layers during rainfall, whereas during the dry season, intense evaporation causes calcite to precipitate and calcrete to form. The calcrete horizon is crucial for dolocrete formation for two reasons. First, the compact calcite occurs as cementation thereby lowering the permeability of the soil [Mowers and Budd, 1996], preventing solutions from percolating downward and allowing stagnant soil solutions to form above the calcrete. Second, during seasonal dry and Ca precipitation in calcrete, the Mg2+concentration increases markedly in soil solution and the solutions become Mg-rich increasing the Mg/Ca ratio.
 The favorable Mg/Ca ratio for dolomite formation is discussed extensively in the literature. For disordered, Ca-rich dolomite to form,Folk and Land  proposed that an Mg/Ca ratio >1 is required, but others have suggested that values as high as 3.31 are required [Hardie, 1987]. In Hawaiian soil, dolomite develops on basalt and precipitates when the Mg/Ca ratio of its soil solution is ∼1 [Capo et al., 2000; Whipkey and Hayob, 2008; Whipkey et al., 2002]. Stream flux on central CLP has a Mg/Ca ratio of 2.34 and the largest river on CLP, the Yellow River has an Mg/Ca ratio of ∼0.85 [Gaillardet et al., 1999; Zhang et al., 1995]. Compared to stream flux, local rainwater has only a minor amount of Mg2+ and Ca2+, with Mg/Ca ratio <0.1. Local precipitation is the only major supply of water to soil. The contrast in Mg/Ca ratio values between rainwater and soil water provides evidence for downward leaching and weathering of Mg-containing minerals, like detrital dolomite, Mg-rich calcite and chlorite which is present in the eolian dust contribution to the soil.
 Compared to the overlying Pleistocene loess-paleosol, the Red Clay displays a higher silicate MgO % (wt) [Xiong et al., 2010] and higher chlorite mineral concentration [Gylesjö and Arnold, 2006], both of which are unstable during weathering. Moreover, the Pleistocene loess deposits on Loess Plateau are characterized by a high content of detrital dolomite [Li et al., 2007], but detrital dolomite is absent in the Red Clay. Because it is unlikely that the composition of eolian dust changed significantly from the Pliocene to Pleistocene [Wang et al., 2007], the lack of dolomite is likely the result of dissolution of carbonate minerals during weathering of the Red Clay. This weathering would provide a Mg2+ source for the formation of protodolomite in Red Clay. Therefore, during weathering more Mg2+ is expected to be released from detrital carbonates and chlorite, resulting in a higher a concentration of Mg2+ and Ca2+and a more highly evolved, Mg-rich soil solution than expected in the overlying loess.
 Red Clay protodolomite is present mainly in voids of about 100 μm (Figure 3e). The rhombic euhedral dolomite crystal groups suggest direct precipitation from soil solution [Whipkey et al., 2002], especially during dry and warm seasons. Due to seasonally aridity, soil water could evolve to have a high Mg/Ca ratio and high salinity resulting in a crystal size on the order of micrometers because of rapid crystallization (Figure 3f). The existence of voids in the soil provided spaces for precipitation and channels for the dolomitizing fluid and the delivery Mg2+, Ca2+ and HCO3− involved in the dolomitization reaction. From a dynamic point of view, besides void volumes the amount of dolomite depends on Mg flux, which is indispensible in any dolomitization model [Compton and Siever, 1986]. In the Red Clay Mg flux is proportional to the weathering contribution from detrital Mg-containing minerals, like detrital dolomite, Mg-rich calcite and chlorite, relying significantly on sufficient rainfall leaching by episodic monsoon events. The underlying calcrete horizons cemented Red Clay soil grains, lowered the permeability of soil layers, prevented further water percolation and guaranteed enough residence time for dolomitization. This resulted in alternating soil layers of dolocrete and calcrete.
 In addition, palygorskite accompanies dolomite in the Red Clay. The occurrence of palygorskite is commonly reported in soils from warm and arid regions, where soil alkalinity is high and Mg2+ and SiO2 (aq) in pore water are active [Singer, 1984]. The Mg2+ and SiO2 (aq) are often the product of local in situ weathering and pedogenesis [Hong et al., 2007]. Hence, the observation that the woven, silky palygorskite aggregates grew on the surface of Red Clay protodolomite in void spaces (Figure 3g) provides an important precipitation condition indicator. Palygorskite formation needs strict chemical conditions and its precipitation strongly depends on [Al3+], [H+] and [SiO2 (aq)] of soil solutions. According to Hong et al. , increasing the alkalinity of soil solutions is crucial for palygorskite to precipitate and higher alkalinity solutions only need lower [Al3+] and [SiO2 (aq)] to precipitate palygorskite. Thus, the existence of palygorskite as a woven coating on the surface of Red Clay protodolomite also indicates arid and alkaline conditions [Singer, 1984], which likely are seasonal when the two minerals precipitated from soil solution in Red Clay.
 Dolocretes as a groundwater cement or a groundwater replacement of primary carbonates are commonly reported in some soil profiles [Nash and McLaren, 2003; Watts, 1980]. But, in the Red Clay on the CLP, dolocretes are primary pedogenic in origin, not groundwater. Primary pedogenic and groundwater carbonates are different in many aspects [Nash and McLaren, 2003; Pimentel et al., 1996]. First, the groundwater replacement would produce carbonate that has meso-crystals or sparry crystals, whereas pedogenesis produces micrites (<5μm in diameter). Second, horizontal elongate carbonate bodies are produced by deep groundwater or non-pedogenic calcretes, but primary pedogeneic carbonates produce pendent, laminar and pisolitic fabrics for carbonate calcretes or dolocretes [Colson and Cojan, 1996]. Third, the horizontal groundwater carbonates are larger and thicker, up to 10 m or more in thickness; primary pedogenic calcretes are typically 0.5–2 m in thickness [Wright and Tucker, 1991]. Fourth, groundwater calcretes and dolocretes are associated with more permeable lithologies [Fu et al., 2004]. Cracks, cavities and channels are necessary in host sediments for groundwater carbonates to precipitate [Khalaf and Gabler, 2008]. Fifth, the groundwater environment lacks biogenic relics such as rhizogenic calcretes and root traces [Khalaf and Gabler, 2008; Pimentel et al., 1996].
 The following observations suggest that Red Clay calcretes and dolocretes are primary pedogenic carbonates: (1) In Red Clay soils porosities were low, ∼3–5% by volume [Guo et al., 2001]. Carbonates of groundwater origin need higher porosities, locally >25%. (2) Carbonate nodules in the Red Clay soil matrix are a few centimeters in diameter, appearing as irregular intervals and dispersed individual nodules in the groundmass. But, typical groundwater carbonates display uniformly massive horizons with a gradational top and base. (3) The carbonate nodules in the Red Clay are composed of micritic crystals (<5 μm), contrasting to phreatic sparry crystals typical of groundwater deposits. (4) The macrostructure of Red Clay carbonates displays laminar, nodular, pisolitic and pendent fabrics. But, groundwater carbonates never show pisolitic or pendent fabrics. (5) Moreover, dark Fe-Mn films (∼10% in volume) were also abundant in lower portions of Red Clay chronosequences. Their reddish oxidation color is similar to Fe-Mn films in the upper portion of Red Clay. This brick red color is distinct from the strongly reduced gray-green color of iron minerals affected by groundwater [Pipujol and Buurman, 1994]. In summary, we propose that the Red Clay dolocretes we studied are primary pedogenic and not phreatic/groundwater carbonates. The development of Red Clay dolocretes is closely related to climatic factors through their influence on pedogenesis.
 Our protodolomite records show that the occurrence of protodolomite gradually decreases upward in both the BJZ and DJP sections (Figure 6) and this decrease is especially prominent in the southern DJP section. This trend is broadly consistent with the global cooling as recorded in oxygen isotopes of marine sediments since the late Miocene [Zachos et al., 2001].
 (1) Red Clay on CLP shows diagnostic FTIR absorption features of calcite and protodolomite, which allowed the two carbonate minerals to be identified and quantified using the FTIR method. Linear regression of the 713 cm−1 and 728 cm−1 absorption band area versus calcite and dolomite concentration, respectively, produces regression equations with R2′s of 0.92 and 0.96.
 (2) In the DJP section 136 samples of the 471samples analyzed contain protodolomite with an average content of 6.0%. In the BJZ 109 of the 314 samples analyzed contain protodolomite with an average content of 3.6%. In both sections protodolomite concentrations increase downward.
 (3) SEM observations indicate that Red Clay protodolomite is composed of euhedral rhombic crystals coexisting with calcite and palygorskite. The crystals vary from 1–20 μm and grow into the soil voids. These morphologic features imply that the protodolomite is authigenic, that is, grew in situ. The depletion of δ18O further indicates that this protodolomite formed in a fresh water environment. Taken together, the above results demonstrate that the protodolomite found in Red Clay originated from pedogenesis.
 (4) Dolomitization in the Red Clay sequence appears to be the result of overcoming kinetic barriers. A high pore water Mg/Ca ratio is the critical kinetic factor controlling Red Clay dolomitization. During Red Clay times warm and seasonally dry conditions leads to the formation of calcrete from soil pore waters thereby enriching the remaining pore solutions with Mg2+ and significantly increasing the Mg/Ca ratio. This elevated Mg/Ca ratio then brought about the formation of protodolomite.
 Regression data for quantifying concentrations of calcite and protodolomite shown in Figures 5a and 5b are listed in Tables A1 and A2. Comparisons of observed concentrations from the gasometric measurements versus the estimated values from the FTIR equations for both calcite and protodolomite are shown in Tables A3 and A4, respectively. Protodolomite concentration ranges within both Duanjiapo and Bajiazui sections are shown in Table A5 with brief statistical descriptions.
Table A1. Calcite Band Areas of Red Clay Samples and the Gasometric Values for Calcite
 This study was funded by the National Basic Research Program of China (grant 2010CB833400) and the National Natural Science Foundation of China (through grants 40973062 and 41021002). We thank the Editor, John S. Compton and other two anonymous reviewers for their thoughtful comments that significantly improved the original manuscript.