1. Introduction
The late Miocene period is notable for the occurrence of global cooling and the establishment of modern ecosystems (Herbert et al. Reference Herbert, Lawrence, Tzanova, Peterson, Caballero-Gill and Kelly2016). Globally, there was a remarkable expansion of C4 plants after 8 Ma (Cerling et al. Reference Cerling, Ehleinger and Harris1998), and the drivers of this expansion remain uncertain (Edwards et al. Reference Edwards, Obsorne, Stromberg and Smith2010). C4 plant species compose only 3 % of vascular plant species (Sage, Reference Sage2004), but account for ∼25 % of terrestrial photosynthesis (Still et al. Reference Still, Berry, Collatz and DeFries2003). C4 plants dominate tropical and subtropical grasslands and savannas. They thrive in areas with high temperature, high aridity and low partial pressure of atmospheric carbon dioxide (pCO2). Because C3 and C4 plants have different δ13C values as a result of having different photosynthetic pathways, their relative abundances can be estimated geochemically by the analysis of leaf wax δ13C in sediments (Chikaraishi et al. Reference Chikaraishi, Naraoka and Poulson2004). The late Miocene expansion of C4 plants is thought to have taken place in low-pCO2 environments (Cerling et al. Reference Cerling, Ehleinger and Harris1998), but its triggers are disputed. However, both alkenone δ13C and foraminifera δ11B-based pCO2 reconstructions showed no evidence for a significant drop in pCO2 during the late Miocene period (Pagani et al. Reference Pagani, Freeman and Arthur1999; Sosdian et al. Reference Sosdian, Greenop, Hain, Foster, Pearson and Lear2018). Huang et al. (Reference Huang, Clemens, Liu, Wang and Prell2007) reported positive shifts in leaf wax δ13C and δD in sediments at ODP Site 722 in the Arabian Sea from 11 to 6.3 Ma, and suggested that aridification drove the expansion of C4 plants in the Himalayan foreland and Arabian Peninsula. Tipple & Pagani (Reference Tipple and Pagani2007) concluded that the timing of geographical expansion of C4 plants was not globally synchronous, and thus pointed towards more regional controls such as aridity, rainfall seasonality, growth season temperature, fire disturbance, etc., on the development of C4-dominated ecosystems. On the other hand, Herbert et al. (Reference Herbert, Lawrence, Tzanova, Peterson, Caballero-Gill and Kelly2016) suggested that the pCO2 decrease at 8 Ma, which was not shown by any pCO2 proxy records but was assumed from the global cooling trend, triggered the expansion of C4 plants. Polissar et al. (Reference Polissar, Rose, Uno, Phelps and Demenocal2019) recently reported evidence of synchronous expansion of C4-dominated ecosystems across northwestern and East Africa after 10 Ma, which was not accompanied by aridification, and suggested that the decline of pCO2 was a direct cause of C4 grassland expansion.
In the present study, we analysed the δ13C values of long- and mid-chain n-fatty acids derived from vascular plant leaf wax in sediments from the International Ocean Discovery Program (IODP) Site U1457 in the Indus Fan of the Arabian Sea to characterize the development of C3 and C4 vegetation since 10.6 Ma. Because the sediments at the study site were delivered mainly from the Indus River basin and also possibly from western India (Pandey et al. Reference Pandey, Clift and Kulhanek2016), the fatty acid results reflect vegetation changes in these areas.
2. Materials and methods
2.a. Samples
IODP Site U1457 was drilled in the Laxmi Basin (17° 09.95′ N, 67° 55.80′ E) at a water depth of 3534 m in the Arabian Sea (Fig. 1; Pandey et al. Reference Pandey, Clift and Kulhanek2016). Site U1457 lies offshore of the western margin of India in the Arabian Sea, ∼491 km from the Indian coast and ∼750 km from the modern mouth of the Indus River, which is presumed to be the primary source of sediment to the area, at least during the Neogene Period (Pandey et al. Reference Pandey, Clift and Kulhanek2016). Site U1457 is situated on the western edge of the Laxmi Basin, at the toe of the slope leading up to the structural and topographic high of the Laxmi Ridge.
Map showing the location of IODP Site U1457 and the average δ13C of organic matter calculated by the CARAIB dynamic vegetation model for the present (Galy et al. Reference Galy, Francois, France-Lanord, Faure, Kudrass, Palhol and Singh2008). Locations of ODP Site 722 and Siwalik palaeosol sequences are also indicated.
Five lithological units were defined at Site U1457 (Fig. 2; Pandey et al. Reference Pandey, Clift and Kulhanek2016). Unit I consists of a ∼74 m thick sequence of Pleistocene nannofossil ooze and nannofossil-rich clay. Unit II is ∼194 m thick and is dated to the early Pleistocene. It consists mainly of silty clay and sandy silt interbedded with very thin sandy silt turbidites. Unit III is ∼450 m thick and consists of upper Miocene to lower Pleistocene silty claystone, silty sandstone, nannofossil chalk and nannofossil-rich claystone. Unit IV is ∼227 m thick and consists of a mixture of interbedded lithologies dominated by claystone at the top of the unit and calcarenite, calcilutite, breccia and limestone towards the base of the unit. This unit is dated to the late Miocene. Lower Paleocene Unit V is ∼30 m thick and mostly consists of claystone and volcaniclastic sediment. These sedimentary rocks directly overlie the basaltic basement.
Lithologic column of Site U1457 and the age–depth model based on biostratigraphic constraints.
The Indus Fan acquires most of its sediment load from the high-relief topography of the western Tibetan Plateau, Karakoram and Himalaya (Clift et al. Reference Clift, Campbell, Pringle, Carter, Zhang, Hodges, Khan and Allen2004; Garzanti et al. Reference Garzanti, Vezzoli, Ando, Paparella and Clift2005). In the sediments deposited at Site U457 during the last 600 ka, the 87Sr/86Sr and clay mineral ratios suggest the mixing of sediments derived from the Indus River and Deccan Plateau (Yu et al. Reference Yu, Colin, Wan, Saraswat, Song, Xu, Clift, Lu, Lyle, Kulhanek, Hahn, Tiwari, Mishra, Miska and Kumar2019). In comparison, the low 87Sr/86Sr and high ϵNd values in Site U1457 sediments older than 600 ka changed gradually into the high 87Sr/86Sr and low ϵNd values typical of the Himalayas, suggesting that the sediments were consistently derived from the Indus River before 600 ka, but the changes in the 87Sr/86Sr and low ϵNd values reflect the exposure of rock caused by the uplift of the Himalayas (Clift et al. Reference Clift, Zhou, Stockli and Blusztajn2019).
In the modern condition, the proportions of C4 plants in the watersheds of the Indus and western Indian rivers, including the Narmada and Tapti rivers, are similarly high, whereas C3 plants are more abundant in the upstream areas of the Indus River and the coastal areas of southwestern Peninsular India south of 2° N (Fig. 1; Galy et al. Reference Galy, Francois, France-Lanord, Faure, Kudrass, Palhol and Singh2008). This distribution of C4 plants suggests that the δ13C of long-chain n-fatty acids was potentially affected by environmental changes in sediment source areas as well as changes in the main source areas of sediments.
A total of 75 samples were collected mainly from hemipelagic layers from a composite section between 1 and 990 m CSF-A (Units I to IV) at Site U1457. Cores and samples were stored at ∼4°C until analysis. Samples were freeze-dried and pulverized.
2.b. Age–depth model
The succession of calcareous nannofossil and planktonic foraminifer events indicates that Site U1457 spans the early Paleocene through recent, albeit with a very long hiatus (∼50 Myr) between lower Paleocene and upper Miocene sediments (Fig. 2). There are three other hiatuses around 8 Ma (∼0.3 Myr), 6–4 Ma (∼1.5 Myr) and 2 Ma (∼0.2 Myr). The age–depth model was made based on nannofossil datums listed in Table 1.
Biostratigraphic datums used for the age–depth model of Site U1457 in this study
1 – Pandey et al. (Reference Pandey, Clift and Kulhanek2016); 2 – C. M. Routledge (unpub. M.Sc. thesis, Florida State Univ., 2015).
2.c. Fatty acid δ13C analysis
Lipids were extracted (× 3) from c. 3 g of dried sediment using a DIONEX Accelerated Solvent Extractor ASE-200 at 100°C and 1000 psi for 10 minutes with 11 ml of dichloromethane–methanol (6:4) and then concentrated. The extract was separated into two fractions with column chromatography (aminopropyl silica gel, i.d., 5.5 mm; length, 45 mm): 3 ml dichloromethane–2-propanol (2:1) (neutral fraction) and subsequent 3 ml diethyl ether–acetic acid (96:4) (acid fraction) following Gao et al. (Reference Gao, Burnier and Huang2012). The acid fraction dissolved in 0.3 ml toluene was methylated with 1 ml methanol–acetyl chloride (95:5) under nitrogen gas at 60°C for 12 hours. The methylated acid fraction was supplemented with 1 ml 5 % sodium chloride in water and extracted (× 3) with hexane. The fraction was further purified with SiO2 column chromatography: 3 ml hexane and subsequent 3 ml dichloromethane (methylated acid fraction for analysis).
Gas chromatography (GC) was conducted using an Agilent 6890 series gas chromatograph with on-column injection and electronic pressure control systems, and a flame ionization detector. Samples were dissolved in hexane. Helium was the carrier gas and the flow velocity was maintained at 30 cm s−1. A Chrompack CP-Sil5CB column was used (length, 50 m; i.d., 0.32 mm; thickness, 0.25 μm). The oven temperature was programmed from 50°C to 120°C at 30°C min−1, and from 120°C to 310°C at 5°C/min−1, and then maintained at 310°C for 30 minutes.
The carbon preference index (CPI) and averaged chain length (ACL) of n-fatty acids are defined in this study as:
$$\begin{align}
& \text{CPI=0}.\text{5}\{\left( {{\text{C}}_{\text{26}}}\text{+}{{\text{C}}_{\text{28}}}\text{+}{{\text{C}}_{\text{30}}}\text{+}{{\text{C}}_{\text{32}}} \right)\text{/}\left( {{\text{C}}_{\text{25}}}\text{+}{{\text{C}}_{\text{27}}}\text{+}{{\text{C}}_{\text{29}}}\text{+}{{\text{C}}_{\text{31}}} \right) \\
& \text{ +}\left( {{\text{C}}_{\text{26}}}\text{+}{{\text{C}}_{\text{28}}}\text{+}{{\text{C}}_{\text{30}}}\text{+}{{\text{C}}_{\text{32}}} \right)\text{/}\left( {{\text{C}}_{\text{27}}}\text{+}{{\text{C}}_{\text{29}}}\text{+}{{\text{C}}_{\text{31}}}\text{+}{{\text{C}}_{\text{33}}} \right)\} \\
& \text{ ACL=}\left( \text{24}{{\text{C}}_{\text{24}}}\text{+}\text{26}{{\text{C}}_{\text{26}}}\text{+}\text{28}{{\text{C}}_{\text{28}}}\text{+}\text{30}{{\text{C}}_{\text{30}}}\text{+}\text{32}{{\text{C}}_{\text{32}}} \right) \\
& \text{ /}({{\text{C}}_{\text{24}}}\text{+}{{\text{C}}_{\text{26}}}\text{+}{{\text{C}}_{\text{28}}}\text{+}{{\text{C}}_{\text{30}}}\text{+}{{\text{C}}_{\text{32}}}) \\
\end{align}$$Combined GC-isotope ratio-mass spectrometry (GC/IRMS) for n-fatty acids was carried out using an Agilent 6890 series gas chromatograph with a capillary column coated with DB-5MS (30 m length; i.d. 0.32 mm; 0.25 μm film thickness) combined with a Finnigan MAT delta Plus mass spectrometer through a combustion furnace at 850°C. GC conditions were the same as above. As an internal isotopic standard, n-C36H74 was used to check the condition of measurements. Data were converted to values relative to the Vienna Pee Dee Belemnite (VPDB) using standard delta notation by comparison with CO2 standard gas.
The δ13C value of methanol (source of methylated carbon) used in this study was –34.1 ± 0.2 ‰. Hence, the δ13C value of fatty acids was calculated as:
$${{\rm{\delta }}^{{\rm{13}}}}{{\rm{C}}_{{\rm{free}}}}{\rm{ = }}\left\{ {\left( {{{\rm{C}}_{\rm{n}}}{\rm{ \, +\, 1}}} \right){\rm{ }}{{\rm{\delta }}^{{\rm{13}}}}{{\rm{C}}_{{\rm{ester}}}}{\rm{ \, +\, 34}}{\rm{.1}}} \right\}{\rm{/}}{{\rm{C}}_{\rm{n}}}$$where δ13Cfree = δ13C of free acid, δ13Cester = δ13C of methyl ester and Cn = carbon number of the free acid. Reproducibility of the measurements based on repeated analyses is better than ±0.5 ‰.
3. Results
3.a. The carbon number distribution of n-fatty acids
Normal fatty acids have a bimodal pattern of carbon number distribution showing maxima at C16 and C26/C28 and a strong even carbon number preference (Fig. 3; online Supplementary Material Table S1). The CPI ranged from 2.8 to 5.4, which is common in higher plant leaf waxes (e.g. Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007). The CPI tended to increase with significant fluctuation from 10.6 to 6 Ma, and to decrease slightly after 3.6 Ma (Fig. 4). The carbon number distribution was characterized by relatively abundant mid-chain C24 homologue. The ACL fluctuated on a million-year timescale with maxima around 8, 6, 3–2 and 1.3 Ma (Fig. 4).
Averaged carbon number distribution of n-fatty acids at Site U1457.
Carbon number preference index (CPI), averaged chain length (ACL) of n-fatty acids and the δ13C of n-C24 to n-C32 fatty acids in sediments from Site U1457 during the last 10.6 million years. Arrows indicate the average values of terrestrial C3, aquatic C3 and C4 plants (tC3, aqC3 and C4, respectively; Chikaraishi et al. Reference Chikaraishi, Naraoka and Poulson2004).
3.b. δ13C values of n-fatty acids
The δ13C values of long- and mid-chain n-fatty acids followed different trends from 10.8 to 6.3 Ma (Fig. 4; online Supplementary Material Table S1). The δ13C of long-chain n-C32 fatty acid shifted from −34 to −22 ‰ from 10.4 to 6.3 Ma, while the δ13C of mid-chain n-C24 fatty acid was consistently around −23 to −22 ‰, with the exception of a significant negative excursion around 8 Ma. The δ13C values of n-C24 fatty acid were always higher than those of n-C32 fatty acid before 6.3 Ma. The δ13C values of n-C26 to n-C30 fatty acids were intermediate between those of C24 and n-C32 fatty acids. After 6.3 Ma, the δ13C values of long- and mid-chain n-fatty acids did not change significantly.
4. Discussion
4.a. Major sources of long- and mid-chain n-fatty acids
The δ13C of n-C32 fatty acid shifted from −34 to −22 ‰ from 10.4 to 6.3 Ma (Fig. 4). This positive shift in the δ13C of n-alkanes was also reported in a previous study of the Arabian Sea and was attributed to an increased abundance of C4 plants (Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). However, alternate interpretations are possible based on the δ13C values and concentrations of n-C24 to n-C32 fatty acids in the study samples.
The δ13C values of long- and mid-chain n-fatty acids followed different patterns from 10.8 to 6.3 Ma (Fig. 4). This observation cannot be explained if we assume that C3 and C4 terrestrial plants were the sole sources of these n-fatty acids, as C3 and C4 terrestrial plants have nearly identical patterns of n-fatty acid distribution, and little difference in the δ13C exists among homologues in single species (Chikaraishi et al. Reference Chikaraishi, Naraoka and Poulson2004; Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007). Ficken et al. (Reference Ficken, Li, Swan and Eglington2000) reported that the n-fatty acids of aquatic vascular C3 plants are characterized by a homologous distribution with a maximum around C24 (ACL = 26.6), whereas terrestrial vascular plants have a maxima around C30 (ACL =28.6). Chikaraishi et al. (Reference Chikaraishi, Naraoka and Poulson2004) reported that the free n-fatty acids (C30 and C32) of aquatic vascular C3 plants have δ13C values of −24.8 ± 1.5 ‰, which is close to the value for C4 plants (−21.1 ± 1.1 ‰) and higher than that for terrestrial vascular C3 plants (−38.5 ± 3.4 ‰). Thus, the large difference in the δ13C values of mid-chain (C24 and C26) and long-chain (C30 and C32) fatty acids from 10.8 to 6.3 Ma may be attributable to the contribution of aquatic C3 plants such as freshwater submerged and floating plants and sea grasses.
The plot of the ACL and the δ13C values of long-chain fatty acids demonstrates that most U1457 samples group within a triangle of C3 angiosperm trees, C4 plants and aquatic C3 plants (Fig. 5). Before 6.3 Ma, the ACL and the δ13C values of long-chain fatty acids were negatively correlated (Fig. 5), suggesting that the δ13C values reflected the relative contribution of terrestrial C3 versus aquatic C3 plants in the Indus River basin and western India. The samples of 10.8–8.3 Ma are distributed along the line between terrestrial and aquatic C3 plant end-members, implying a negligible contribution of C4 plants. The samples of 8.1–7.5 Ma are distributed near the line between terrestrial C3 and C4 plant end-members, indicating a significant contribution of terrestrial plants including C4 plants. After 5.8 Ma, the δ13C values remained constant and were independent of the ACL (Fig. 5), suggesting that the δ13C values reflected heavier δ13C values of both aquatic C3 and C4 plants. The relative contribution of terrestrial C3 (ftC3), aquatic C3 plants (faC3) and C4 plants (fC4) was estimated from the end-members of their ACL and C30 and C32 δ13C values following the equations:
$$\eqalign{{\rm{f}}_{{\rm{tC3}}}}{\rm{ \,+\, }}{{\rm{f}}_{{\rm{aC3}}}}{\rm{ \,+\, }}{{\rm{f}}_{{\rm{C4}}}}{\rm{ = 1}} \\ {{\rm{\delta }}^{{\rm{13}}}}{{\rm{C}}_{{\rm{tC3}}}}{{\rm{f}}_{{\rm{tC3}}}}{\rm{ \,+\, }}{{\rm{\delta }}^{{\rm{13}}}}{{\rm{C}}_{{\rm{aC3}}}}{{\rm{f}}_{{\rm{aC3}}}}{\rm{ \,+\, }}{{\rm{\delta }}^{{\rm{13}}}}{{\rm{C}}_{{\rm{C4}}}}{{\rm{f}}_{{\rm{C4}}}}{\rm{ = }}{{\rm{\delta }}^{{\rm{13}}}}{\rm{C}} \\ {\rm{AC}}{{\rm{L}}_{{\rm{tC3}}}}{{\rm{f}}_{{\rm{tC3}}}}{\rm{ \,+\, AC}}{{\rm{L}}_{{\rm{aC3}}}}{{\rm{f}}_{{\rm{aC3}}}}{\rm{ \,+\, AC}}{{\rm{L}}_{{\rm{C4}}}}{{\rm{f}}_{{\rm{C4}}}}{\rm{ = ACL}}$$where δ13CtC3 = −38.5 ‰ (−46.8 to −34.5 ‰, n = 10, Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007), δ13CaC3 = −24.8 ‰ (−26.5 to −22.8 ‰, n = 3, Chikaraishi et al. Reference Chikaraishi, Naraoka and Poulson2004), δ13CC4 = −21.1 ‰ (−22.6 to −19.7 ‰, n = 5, Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007), ACLtC3 = 28.5 (Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007), ACLaC3 = 26.6 (Ficken et al. Reference Ficken, Li, Swan and Eglington2000) and ACLC4 = 28.3 ACL (Chikaraishi & Naraoka, Reference Chikaraishi and Naraoka2007) as the end-members. The datasets of δ13C and ACL of end-members are not comprehensive and contain uncertainty. To estimate the influence of end-member δ13C values, the fractions ftC3, faC3 and fC4 were calculated from the average, minimum and maximum δ13C end-member values. The end-member δ13C value of each plant component changed from the minimum to maximum, and the fractions ftC3, faC3 and fC4 were calculated for each case (Fig. 6). We chose the values of C3 angiosperm trees rather than C3 angiosperm herbs and gymnosperms as a representative of terrestrial C3 plants because the values derived from C3 angiosperm trees can better explain the variations of the ACL and δ13C prior to 6.3 Ma (Fig. 5). The end-member calculation indicates drastic changes in the relative abundances of terrestrial C3 (ftC3), aquatic C3 plants (faC3) and C4 plants (fC4) during the last 10.8 Myr. Aquatic C3 plants were generally abundant most of the time, but terrestrial C3 plants dominated from 8.1 to 7.5 Ma, and C4 plants dominated intermittently after 6 Ma (Fig. 6).
Plots of the average chain length (ACL) against the δ13C of n-C30 and n-C32 fatty acids in Site U1457 samples. Yellow circles, green circles, blue triangles and red squares indicate samples from 10.8−8.3 Ma, 8.1−7.5 Ma, 7.4−6.3 Ma and 5.8−0 Ma, respectively. The diagram shows the δ13C ranges (horizontal bars) and average of literature values of terrestrial and aquatic C3 plants (tC3 and aC3) and C4 plants (C4).
The relative abundance of terrestrial and aquatic C3 plants (tC3 and aC3) and C4 plants in U1457 sediments during the last 10.6 million years. Vertical bars indicate the range of fractions calculated from the minimum and maximum δ13C end-member values.
In contrast to our findings, no significant variation in different homologues has been observed at Site 722 in the δ13C of C27−C33 n-alkanes from 10.6 to 6 Ma. All of homologues show similar values (Fig. 7; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). This suggests either a lesser influence of the contribution of aquatic plants to the δ13C of long-chain n-alkanes or a minimal contribution from aquatic C3 plants to the Site 722 sediments. The n-alkanes in aquatic plants are typically dominated by mid-chain homologues than are n-fatty acids (Ficken et al. Reference Ficken, Li, Swan and Eglington2000). The difference in the homologous distribution may have led to the lesser influence of the contribution of aquatic plants to the δ13C of long-chain n-alkanes, even with a significant contribution of aquatic C3 plants. Alternatively, Site 722 is located on Owen Ridge and is therefore unaffected by turbidite deposition on the adjacent Indus Fan. Aeolian transport is the principal pathway for terrestrial input to the site (Clemens et al. Reference Clemens, Murray and Prell1996). Because of this depositional setting, the leaf wax record of Site 722 does not reflect the contribution of aquatic plants in the Indus and western Indian river waters.
The relative abundance of C4 plants in terrestrial plants (C4/(tC3 + C4)) and the δ13C of n-C30 and n-C32 fatty acids at Site U1457 (this study); the δ13C of soil carbonate in the Siwalik palaeosol in Pakistan (Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007); and the δ13C of n-C27, n-C29, n-C31 and n-C33 alkanes and the δD of n-C31 and n-C33 alkanes at Site 722 in the Arabian Sea (Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007) during the last 10.6 million years.
4.b. Late Miocene vegetation changes in the Indus River basin and western India
The abundance of C4 plants in total terrestrial C3 and C4 plants, i.e. the C4/(tC3 + C4) ratio, suggests that the terrestrial C3 plants were replaced by C4 plants in the Indus River basin and western India from 9.7 to 6.3 Ma with a period of terrestrial C3 plant expansion around 8 Ma (Fig. 7). However, whether the onset of C4 plant expansion occurred at 9.7 Ma or 8.2 Ma is not clear because the contribution of aquatic C3 plants overprinted the δ13C signal of C4 plants. The high C4/(tC3 + C4) ratio at 8.5 Ma is a single peak, and the ratios of other samples from 9.7–8.2 Ma are not significantly higher than 0 (Fig. 7). The robust increase in the C4/(tC3 + C4) ratio started at 8.2 Ma, which was synchronous with the increase in the δ13C value of palaeosol carbonate in Siwalik palaeosol sequences (Fig. 7; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007; Behrensmeyer et al. Reference Behrensmeyer, Quade, Cerling, Kappelman, Khan, Copeland, Roe, Hicks, Stubblefield, Willis and Latorre2007), although the δ13C value of palaeosol carbonate (Sanyal et al. Reference Sanyal, Bhattacharya, Kumar, Ghosh and Sangode2004; Behrensmeyer et al. Reference Behrensmeyer, Quade, Cerling, Kappelman, Khan, Copeland, Roe, Hicks, Stubblefield, Willis and Latorre2007) and long-chain n-alkanes (Ghosh et al. Reference Ghosh, Sanyal and Kumar2017) in Siwalik palaeosol sequences showed various changing patterns due to differences in the sub-environments of the Siwalik alluvial plain (Behrensmeyer et al. Reference Behrensmeyer, Quade, Cerling, Kappelman, Khan, Copeland, Roe, Hicks, Stubblefield, Willis and Latorre2007). This trend is consistent with the occurrence of aridification in the Indus River basin and western India, which was indicated by positive shifts in the δ18O of soil carbonate in Siwalik palaeosols (Quade et al. Reference Quade, Cerling and Bowman1989; Quade & Cerling, Reference Quade and Cerling1995; Sanyal et al. Reference Sanyal, Bhattacharya, Kumar, Ghosh and Sangode2004; Behrensmeyer et al. Reference Behrensmeyer, Quade, Cerling, Kappelman, Khan, Copeland, Roe, Hicks, Stubblefield, Willis and Latorre2007; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007) and the δD of long-chain n-alkanes at ODP Site 722 in the Arabian Sea (Fig. 7; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). A modelling study with the boundary conditions of the late Miocene palaeogeography, orography and ice sheets and an atmospheric CO2 level of 395 ppm indicated that the Indus River basin and western India were covered by tropical forests and shrubs in the Tortonian Age (11.6–7.3 Ma) because precipitation levels were 1−2 mm/day higher than at present (Pound et al. Reference Pound, Haywood, Salzmann, Riding, Lunt and Hunter2011). An isotope study for river waters in the Indus River basin indicated that 64 to 72 % of the Indus waters are derived from moisture transport from the Mediterranean Sea, and the rest derives from the Indian summer monsoon at present (Karim & Veizer, Reference Karim and Veizer2002). Intensification of the Indian summer monsoon after 8 Ma, shown in marine foraminifera records (e.g. Kroon et al. Reference Kroon, Steen, Troelstra, Prell, Niitsuma, Emeis, Al-Sulaiman, Al-Tobbah, Anderson, Barnes, Bilak, Bloemendal, Bray, Busch, Clemens, de Menocal, Debrabant, Hayashida, Hermelin, Jarrard, Krissek, Kroon, Murray, Nigrini, Pedersen, Ricken, Shimmield, Spaulding, Takayama, ten Haven and Weedon1991; Prell et al. Reference Prell, Murray, Clemens, Anderson, Duncan, Rea, Kidd, von Rad and Weissel1992), is not consistent with the expansion of a C4 ecosystem in the Indus River basin. Instead, during the Messinian Age (7.3–5.3 Ma), the shrinkage of the Paratethys and Mediterranean seas (Ivanov et al. Reference Ivanov, Utescher, Mosbrugger, Syabryaj, Djorjevic-Milutinovic and Molchanoff2011) may have decreased moisture transport to the Indus River basin and western India. Decreased precipitation in the Indus River basin and western India during late Miocene time may have driven the replacement of terrestrial C3 plants by C4 plants in this region.
If the increase of the C4/(tC3 + C4) ratio from 9.7 Ma is real, the replacement of C3 plants by C4 plants in the Indus River basin and western India began before the positive shift in the δ13C value of palaeosol carbonate in Siwalik palaeosol sequences in the northern Indus River basin (Fig. 7; Quade et al. Reference Quade, Cerling and Bowman1989; Quade & Cerling, Reference Quade and Cerling1995; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). One possible interpretation is that C3/C4 replacement started earlier in the downstream area of the Indus River and western India than in the upstream area of the Indus River (Fig. 8). Higher levels of precipitation in the Himalaya Mountains may have supplied sufficient water to C3 vegetation in the upstream area. This interpretation is consistent with the distribution of modern and Holocene vegetation in the Indus River basin (Ivory & Lézine, Reference Ivory and Lézine2009).
Schematic views of vegetation in the Indus and western Indian river basins before and after 6.3 Ma.
An episodic negative excursion of the δ13C values of n-C30 and n-C32 fatty acids around 8 Ma was superimposed on an increasing trend of δ13C (Fig. 7). Because it is associated with the increase in terrestrial C3 plants and the decrease in aquatic C3 plants (Fig. 6), this excursion suggests that the contribution of terrestrial C3 plants was elevated compared with that of aquatic C3 plants in the Indus and western Indian river waters. A pollen study of Holocene sediments from core SO90-56KA recovered from the Makran coast (Indus margin) of the Arabian Sea indicates more abundant montane pollen taxa from the Himalayas when the Indian summer monsoon was stronger due to increased fluvial activity of the Indus River during those times (Ivory & Lézine, Reference Ivory and Lézine2009). In contrast, pollen taxa from shrubs in the lower reaches of the Indus River were relatively abundant when the Indian summer monsoon was weaker. This observation suggests that higher precipitation in the upper reaches of the Indus River may have increased the contribution of terrestrial C3 plants in the Indus Fan sediments.
The negative excursion at 8 Ma was also recorded at ODP Site 722 in the Arabian Sea (Fig. 7; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). Site 722 received terrestrial sediments transported from Pakistan, Iran, Afghanistan and the Arabian Peninsula by wind, suggesting that the excursion was a regional, rather than local, phenomenon (Clemens et al. Reference Clemens, Murray and Prell1996; Huang et al. Reference Huang, Clemens, Liu, Wang and Prell2007). The negative excursion around 8 Ma can be correlated with the second ‘washhouse’ event of elevated precipitation in Europe (Böhme et al. Reference Böhme, ILG and Winkhofer2008). Enhanced moisture transport from Europe to the Indus River basin increased precipitation. Increased precipitation may have allowed range expansion for terrestrial grassland and forest, increasing the proportion of terrestrial C3 plants to aquatic plants in the Indus River and causing the negative δ13C excursion.
5. Conclusions
A large difference in the δ13C values of mid-chain (C24 and C26) and long-chain (C30 and C32) fatty acids from 10.6 Ma to 6.3 Ma suggests the contribution of aquatic C3 plants. Before 6.3 Ma, the δ13C values reflected the relative abundance of terrestrial C3 versus aquatic C3 and C4 plants in the Indus River basin and western India. After 6.3 Ma, the δ13C values reflected heavier δ13C values of both C3 aquatic and C4 plants.
A three-end-member model calculation suggests that terrestrial C3 plants were replaced by C4 plants in the Indus River basin and western India from 9.7 or 8.2 to 6.3 Ma. Decreased precipitation in those areas during late Miocene time may have driven the replacement of terrestrial C3 plants by C4 plants. The shrinkage of the Paratethys and Mediterranean seas (Ivanov et al. Reference Ivanov, Utescher, Mosbrugger, Syabryaj, Djorjevic-Milutinovic and Molchanoff2011) may have decreased moisture transport to the Indus River basin and western India.
An episodic increase in terrestrial C3 and C4 plants around 8 Ma was superimposed on a decreasing trend of terrestrial C3 plants. The increase can be attributed to high precipitation caused by the regionally enhanced moisture transport from the west.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0016756819001109
Acknowledgements
This research used samples and/or data provided by the International Ocean Discovery Program (IODP). We thank IODP Expedition 355 shipboard scientists for valuable discussion, and Yuka Sazuka and Kaori Ono of Hokkaido University for analytical assistance. Yoshito Chikaraishi of Hokkaido University provided an electronic dataset of the homologous distribution of n-fatty acids in plants. Unpublished data courtesy of Peter Clift of Louisiana State University and Boo-Keun Khim of Pusan National University helped us to understand the provenance of sediments at Site 1457. Comments by Denise Kulhanek, editor, and two anonymous reviews improved this manuscript. The study is supported by grants-in-aid for JSPS (Program for Advancing Strategic International Networks to Accelerate the Circulation of Talented Researchers R2901) to MY and JAMSTEC (Exp. 355 post-expedition study) to KS.
