A mid- to late-Holocene record of vegetation decline and erosion triggered by monsoon weakening and human adaptations in the south-east Indian Peninsula

Abstract

The mid- to late-Holocene monsoon decline led to aridification of the Indian Peninsula impacting the early agricultural practices in the region. Our analysis of organic carbon, mineral magnetic properties and AMS ¹⁴C dating of a 54.2-m-long sediment core (CY) from the Godavari Delta, India, showed changes in the organic carbon source and sediment provenance, which are linked to the changes in vegetation and soil/rock erosion caused by widespread aridification and associated human adaptation in central India. Our results show a decline in the concentration of ferrimagnetic minerals, indicating reduced input from the basalts of the Deccan Plateau after ~6.0 cal. ka BP in response to the weaker Indian monsoon. δ¹³C values show a distinct increase from ~4.9 cal. ka BP, indicating an increase in C4 plant sources under the continued weak monsoon phase, whereas a higher ferrimagnetic mineral concentration in the sediment suggested an increased Deccan basalt source. Abrupt increase in δ¹³C values and decrease in TOC content accompanied with a significant increase in ferrimagnetic mineral concentration from ~3.2 to 3.1 cal. ka BP reflected a shift of organic carbon and sediment source and a severe decline in vegetation coverage. Such phenomena indicate intensified deforestation and soil/rock erosion in the Deccan Plateau producing higher ferrimagnetic mineral inputs, which is in agreement with significant expansion of agricultural activities in the Deccan Chalcolithic cultural period. In addition, C3 plants recovered and magnetic concentration declined during the wet events (4.6 and 4.0 cal. ka BP) of Neolithic time, while both C3 plants and magnetic parameters increased during the wet events (3.1–2.8 and 2.1 cal. ka BP) of the Chalcolithic cultural period. This implies increased agricultural activity and the onset of human modification of the ecosystem.

Keywords

C3/C4 plants Deccan Chalcolithic culture Deccan Plateau mineral magnetic property organic carbon sediment provenance

Introduction

The Indian monsoon, one of the most important tropical climate systems on earth, has a strong influence on agricultural production and thus society and the economy in South Asia. This is particularly the case in India, which accounts for about two-thirds of the population and over a half of the farmland. There have been many studies on the mid-Holocene decline of the Indian monsoon based on the analyses of sedimentary records from varied environments such as speleothems, lake deposits, marine and terrestrial sediments obtained from the Arabian Sea and the Bay of Bengal alongside the Indian Peninsula and from the Tibetan Plateau (Berkelhammer et al., 2012; Bird et al., 2014; Dixit et al., 2014; Fleitmann et al., 2003, 2007; Prasad et al., 2014; Sarkar et al., 2015; Thamban et al., 2007). Thamban et al. (2007) inferred a significantly weak monsoon period during 6–5.5 ka BP from the sediments of the eastern Arabian Sea. Lonar Lake sediments in central India showed the onset of aridity from ~6 cal. ka BP and a marked drought period during 4.8–4 cal. ka BP (Prasad et al., 2014; Sarkar et al., 2015). Carbon isotopes of sedimentary leaf waxes estimated from a marine core off the Godavari delta indicated an increased proportion of C4 plants since ~4 cal. ka BP in central India which was inferred to be a response to the mid-Holocene monsoon weakening (Ponton et al., 2012). Analyses of pollen and carbon isotopes from the sediments of two marine cores, one from off Godavari Delta and the other from off Mahanadi Delta, also implied aridification and expansion of drier vegetation from ~4.2 to 4.0 cal. ka BP in the core monsoon zone (such as in the drainage basins of the Godavari and Mahanadi Rivers; Figure 1a) of the Indian Peninsula (Ponton et al., 2012; Zorzi et al., 2015).

Figure 1.

(a) Locations of the Godavari drainage basin and core CY (red solid circle). Also indicated are the Indian monsoon wind directions (white arrows), rainfall in the monsoon season (cm, white dashed lines), previous research sites (black solid circles) of 723A (Gupta et al., 2003), Qunf Cave (Fleitmann et al., 2003) and NGHP-16A (Ponton et al., 2012) and the major geological formations of the Godavari drainage basin (Kulkarni et al., 2015). (b) Distribution of the ecological communities in the Godavari drainage basin (after Zorzi et al., 2015). (c) SRTM 90 m DEM showing the Godavari delta with the six delta growth lobes outlined (after Nageswara Rao et al., 2015). The red circle indicates the location of core CY.

Recent studies suggested that the mid-Holocene aridification led to the decline of Indus civilizations in northwest India at ~4.1 cal. ka BP (Dixit et al., 2014) and human migration from the Indus region towards the Ganga plain (Gupta et al., 2006). Asouti and Fuller (2008) and Ponton et al. (2012) suggested that adaptation of agriculture to the semi-arid climate and expansion of human settlements in central and south India occurred after 4 cal. ka BP when the Harappan civilization collapsed owing to the drought in the northwestern part of the Indian Peninsula.We propose such human adaptation to the climate change in the Deccan Plateau should have influenced the Plateau’s environment and the sediment flux to the Bay of Bengal, such as an increase in the Deccan Volcanic Province sediment source in the late Quaternary sediments of the Bengal Fan (Kulkarni et al., 2015; Sangode et al., 2001).

The Godavari River flows through a large part of the Deccan Plateau in central India. It encompasses a climatic transitional zone from semi-arid to wet (Figure 1a). There is also a long history of cultivation of millet, pulses, wheat, barley and so on in the Deccan Plateau since ~5000 years ago (Fuller, 2000, 2011; Fuller and Harvey, 2006; Fuller and Murphy, 2014; Misra, 2001; Shinde, 1990; Stevens et al., 2016). The vegetation, especially in the upper reaches of the Godavari River in the Deccan Plateau, is typically sensitive to the change in both monsoon precipitation and the widespread agriculture on the Deccan Plateau. Therefore, we propose that the changes in vegetation, such as its coverage, and deforestation and cultivation related to population expansion had induced changes in organic carbon and sediment yield from the Deccan Plateau, which is, however, less well documented.

In this study, we analysed organic elemental chemistry to reflect changes in organic carbon source and vegetation, which are the response to the monsoon precipitation and changes in sediment provenance that is linked to both the monsoon and human activity such as deforestation and cultivation in the Godavari Basin. Furthermore, we analysed mineral magnetism taking note of the zonal distribution of lithology in the Godavari drainage basin with the diagnostic basalt in the semi-arid western part, which on erosion yields higher quantities of ferrimagnetic minerals of specific composition and domain size than the granite/gneisses in the wetter eastern part (Kulkarni et al., 2015; Sangode et al., 2001). Thus, magnetic proxies are useful to decipher changes in sediment provenance, especially the contributions from the Deccan Plateau basalt province. We then presented a continuous high-resolution record of vegetation, sediment source and erosion vis-à-vis the monsoon decline and human activity for the period from 6.1 to 1.3 cal. ka BP, taking advantage of the high sedimentation rate and high sensitivity of proxies to the monsoon climate and sediment sources in sediment core CY from the Godavari Delta, India (Nageswara Rao et al., 2015). Our study will not only unravel the linkage between climate and human society but also reveal the ecosystem sensitivity in a semi-arid climate setting.

Study area

The Godavari River originates from the Western Ghats close to the west coast of India, flows south-easterly across the Indian Peninsula for over 1465 km and enters the Bay of Bengal on the east coast of India (Jain et al., 2007). Indian monsoon climate prevails over the drainage basin with a mean precipitation of 923 mm during the wet season, which accounts for 82% of the annual rainfall (Nageswara Rao, 1997). The rainfall in the monsoon season is <750 mm in the upper basin (the Deccan Plateau) and increases to >1000 mm in the lower basin (Figure 1a). Therefore, two broad vegetation communities – thorny vegetation (mainly C4 plants) and tropical dry/moist deciduous forest (mainly C3 plants; Figure 1b) – occur in the upper and mid-lower basin, respectively, controlled by the spatial distribution of monsoon precipitation (Zorzi et al., 2015).

The upper part of the drainage basin is located on the Tertiary Deccan Plateau basalts, while its middle and lower parts are underlain by acid igneous and metamorphic rocks (Ahmad et al., 2009). The Tertiary basalt covers ~50% of the Godavari drainage basin area and is a major sediment source for the surficial marine sediments in the Bay of Bengal, recognized by its high content of single-domain (SD) magnetic minerals (Kulkarni et al., 2015, 2014; Sangode et al., 2001).

Early human colonization, including Neolithic/Chalcolithic, Iron Age and early historic cultures, is mainly reported in the upper reach, that is, the Deccan Plateau, of the Godavari drainage basin (Fuller, 2011; Misra, 2001). The Neolithic landscape of the region was characterized by scrub and savanna woodlands, scattered shrubby areas and thorny thickets (Misra, 2001). From the early Holocene, as in other parts of the subcontinent, hunter-gatherer communities underwent a unique Neolithic transition with indigenous millet-pulse cultivation and pastoralism (Fuller and Murphy, 2014), instigating sedentism with food production based on domesticated plants and animals (Fuller et al., 2015). Cultivation of millets started from ~5 cal. ka BP and wheat and barley from ~4 cal. ka BP (Fuller, 2011). The number of settlements increased first at ~4 cal. ka BP and then at ~3 cal. ka BP (Ponton et al., 2012). Pit dwellings dug into the bedrock characterize the Savalda culture (2000–1800 BC) in the upper Godavari Basin (Shinde, 1990). In addition, urbanization in the first millennium BC, such as at Pratisthana in the upper Godavari Basin, promoted the development of irrigation systems, such as damming, for the cultivation of rice in central India (Shaw et al., 2007). In fact, Pratisthana, which flourished as the capital city of the Satavahana dynasty (2nd century BC to 2nd century AD) is now called Paithan, located near a major dam across the Godavari River and is known for its ancient shrines and silk trade.

The Godavari Delta developed during ~6 cal. ka BP through six stages (lobes 1–6; Figure 1c) of successive migration of the main depocenter (Nageswara Rao et al., 2015); the corresponding time intervals of each lobe are shown in Figure 2. The sediment accumulation rate in the delta had progressively increased from as low as ~12 Mt/yr during 6–4 cal. ka BP to ~27 Mt/yr during 4–3 cal. ka BP and 52–62 Mt/yr during 3–1 cal. ka BP and decreased to ~44 Mt/yr over the past 1000 years (Nageswara Rao et al., 2015).

Figure 2.

Lithology, sedimentary facies, age–depth model and the stages of six delta lobes including the 18 AMS ¹⁴C dates (solid dots, from Nageswara Rao et al., 2015) recalibrated using Calib v. 7.1 and the four new dates (open dots) in this study.

Materials and methods

The CY core that we analysed is from the beach ridge plain of the Godavari Delta close to the present shoreline (16°30′04.71″N and 82°08′51.46″E; Figure 1c). The total length of the CY core is 54.2 m with 85% core recovery. The upper 15-m sediment includes medium sand and sandy mud and the lower 39.2 m consists of dark grey shallow marine mud with small calcareous concretions, pebbles and shell fragments (Figure 2). Stratigraphy and sedimentary facies of the well-dated CY core were described in Nageswara Rao et al. (2015). We have obtained four more accelerator mass spectrometry (AMS) ¹⁴C dates for this core (Beta Analytic, Inc., US) in addition to recalibrating the 18 existing dates (from Nageswara Rao et al., 2015) using Calib v. 7.1 (Figure 2 and Table 1). Marine04 data set was used for all the shell samples with a ΔR value of 22 ± 37 years (Nageswara Rao et al., 2015).

Table 1.

AMS ¹⁴C ages including 18 dates from Nageswara Rao et al. (2015) and 4 new dates obtained in this study. All ages were calibrated using Calib v. 7.1 and the Marine04 data set was used for all shell samples by taking the reservoir factor ΔR as 22 ± 37 years (Nageswara Rao et al., 2015).

Depth (m)	Elevation (m a.s.l.)	Material	Conventional ¹⁴C age (yr BP)	Cal. BP (2σ)	Median probability	Probability	Laboratory code	Reference
3.29	−2.29	Shell	1250 ± 30	887−675	772	1	Beta-301749	Nageswara Rao et al. (2015)
5.71	−4.71	Plant material	1140 ± 30	1095−969	1040	0.829	Beta-301750	Nageswara Rao et al. (2015)
8.75	−7.75	Plant material	1440 ± 30	1382−1296	1333	1	Beta-301751	Nageswara Rao et al. (2015)
13.05	−12.05	Shell	1930 ± 30	1565−1333	1454	1	Beta-301753	Nageswara Rao et al. (2015)
16.81	−15.81	Shell	1810 ± 30	1461−1249	1333	1	Beta-301754	Nageswara Rao et al. (2015)
20.83	−19.83	Shell	2530 ± 30	2303−2041	2180	1	Beta-301755	Nageswara Rao et al. (2015)
22.06	−21.06	Shell	2570 ± 30	2329−2098	2220	1	Beta-301756	Nageswara Rao et al. (2015)
25.52	−24.52	Shell	2480 ± 30	2276−1979	2107	1	Beta-301757	Nageswara Rao et al. (2015)
27.74	−26.74	Shell	3000 ± 30	2863−2673	2755	1	Beta-301758	Nageswara Rao et al. (2015)
30.08	−29.08	Plant material	2850 ± 30	3059−2876	2961	1	Beta-301759	Nageswara Rao et al. (2015)
32.9	−31.9	Shell	3150 ± 30	3046−2775	2907	1	Beta-301760	Nageswara Rao et al. (2015)
34.34	−33.34	Plant material	2970 ± 30	3229−3056	3136	0.971	Beta-301761	Nageswara Rao et al. (2015)
40.32	−39.32	Shell	3410 ± 30	3374−3120	3255	1	Beta-429665	This study
41.94	−40.94	Shell	4030 ± 30	4159−3866	4019	1	Beta-301763	Nageswara Rao et al. (2015)
44.4	−43.4	Shell	4890 ± 30	5302−5019	5172	1	Beta-428994	This study
45.48	−44.48	Shell	4900 ± 30	5307−5027	5185	1	Beta-301764	Nageswara Rao et al. (2015)
48.2	−47.2	Shell	5110 ± 40	5566−5312	5448	1	Beta-301765	Nageswara Rao et al. (2015)
49.69	−48.69	Shell	5190 ± 30	5633−5431	5529	1	Beta-301766	Nageswara Rao et al. (2015)
50.64	−49.64	Shell	5380 ± 30	5857−5606	5722	1	Beta-428995	This study
51.79	−50.79	Shell	5570 ± 30	6101−5825	5936	1	Beta-428996	This study
52.56	−51.56	Shell	7440 ± 40	7989−7754	7880	1	Beta-301767	Nageswara Rao et al. (2015)
54.14	−53.14	Shell	8220 ± 40	8917−8547	8704	1	Beta-301768	Nageswara Rao et al. (2015)

AMS: accelerator mass spectrometry.

Radiocarbon ages showed a sedimentary record of ~9000 years in core CY. However, major sedimentation occurred only during the past 6000 years with rapid accumulation at a rate >9 mm/yr except during 3.2–5 cal. ka BP (Figure 2). Considering the effect of high sand content on organic matter preservation and the effect of changes in sedimentary facies on the interpretation of proxies, we used only the shallow marine facies of the sediment between 15.0 and 54.2 m depths in CY core for analysing the organic carbon content and mineral magnetic parameters and discussed the changes in vegetation, sediment provenance and soil/rock erosion during 1.3–6.1 cal. ka BP.

Organic carbon and stable isotope measurement

We analysed 167 samples for estimating the organic stable carbon isotope (δ¹³C), total carbon (TC), total organic carbon (TOC) and total nitrogen (TN) and computed the TOC/TN ratios. The samples were freeze-dried using a vacuum freeze dryer (SRK, Germany) in the State Key Laboratory of Estuarine and Coastal Research, East China Normal University. All samples were then ground to powder and split into two equal parts. The first part was used to measure TC and TN directly. The second part was treated with hydrochloric acid (HCl, 0.1 M) to remove the calcium carbonate content and then washed with deionized water to neutralize it. After drying the samples, TOC, TN and δ¹³C were estimated. The TC, TOC and TN were measured using a Vario Cube CN Elemental Analyzer (Elementar, Germany; reference material: acetanilide) in the State Key Laboratory of Marine Geology at Tongji University, China, to a mean precision of about 0.5%. The δ¹³C was measured using a mass spectrometer (Delta V Advantage, Germany; reference materials: urea and acetanilide) in the Third Institute of Oceanography, State Oceanic Administration, China, to a mean precision of about 0.2‰.

Calculation of proportions of C3/C4 plants

The value of TOC/TN (ratio by weight) fluctuated around the average of 17.28 in the whole sequence (Figure 3), indicating a predominantly terrestrial C3 and C4 plant source and a negligible contribution from algae (Figure 4; Lamb et al., 2006). We then used an end member model to estimate the C4 vegetation percentage following Pradhan et al. (2014):

δ^{13} C_{s} = f \cdot δ^{13} C_{4} + (1 - f) \cdot δ^{13} C_{3}, (0 < f < 100 %)

The equation can be written as

f = \frac{(δ^{13} C_{s} - δ^{13} C_{3})}{(δ^{13} C_{4} - δ^{13} C_{3})}

where f is the fraction of C4 plants; (1 − f) is the fraction of C3 plants; and δ¹³C_s, δ¹³C₄ and δ¹³C₃ are the values of δ¹³C of the sample and the end members of C4 and C3 plants, respectively. The values of C3 and C4 end members used here are −29.8‰ to −28.3‰ and −12.4‰ to −10.8‰ according to surficial samples from the Godavari Basin (Pradhan et al., 2014). The percentage of C4 plants shares the same curve as δ¹³C (but with an error bar ±4.4%) because the model includes only two end members, and it is a linear correlation between the two parameters (Figure 3).

Figure 3.

Variations in the organic elemental chemistry among the six sediment units (I–VI) in the muddy section (15–54.2 m depth) of core CY. Grain size (mud content) and lithology are indicated in the right panel (after Nageswara Rao et al., 2015).

Figure 4.

Plots of δ¹³C versus C/N to discriminate the organic carbon sources for the six units of CY (Lamb et al., 2006).

Mineral magnetic measurement

In total, 184 samples were analysed for room temperature magnetic parameters including χ_lf, χ_fd%, χ_ARM, saturation isothermal remanent magnetization (SIRM), soft IRM (Soft_IRM), hard isothermal remanent magnetization (HIRM) and S_−300mT (Liu et al., 2012) at Pune University, India. Low-frequency (465 Hz, χ_lf) and high-frequency (4650 Hz, χ_hf) magnetic susceptibility was measured using a Bartington Instruments MS2B sensor. Anhysteretic remanent magnetization (ARM) was imparted using a peak alternating field of 100 mT superimposed over 0.1 mT DC field, using an Agico alternating field demagnetizer and the remanence measured using a Minispin spinner magnetometer. Isothermal remanent magnetization (IRM) was imparted in the forward fields of 10, 20, 50, 100, 300, 500, 800 and 1000 mT and back fields of −5, −10, −20, −30, −50, −70, −100, −200 and −300 mT using the ASC impulse magnetizer and was measured on Minispin.

Magnetic parameters were then calculated as follows: frequency-dependent magnetic susceptibility (χ_fd% = ((χ_lf − χ_hf)/χ_lf) × 100); ARM expressed as χ_ARM by normalizing the mass-specific ARM with the bias field; SIRM as the mass normalized remanent magnetism at the field of 1000 mT; HIRM as (SIRM + IRM_−300mT)/2; Soft_IRM as (SIRM − IRM_−20mT)/2; and S_−300mT as ((SIRM − IRM_−300mT)/2/SIRM) × 100 (Bloemendal et al., 1992). Units for each parameter are as follows, χ_lf = 10⁻⁸ m³/kg, SIRM = 10⁻⁵ A m²/kg, χ_ARM = 10⁻⁵ m³/kg, Soft_IRM and HIRM = 10⁻² A m²/kg.

We further measured the temperature-dependent magnetization (J–T curve) for 12 samples, selected randomly from different depths, using a variable field translation balance (PETERSEN, Germany) in a 34-mT field at East China Normal University. The temperature was increased from room temperature to 700°C and decreased to room temperature again to determine the magnetic mineral species.

Results

Organic elemental chemistry, organic carbon flux and sources

Based on the variations in C-N elements, carbon isotope δ¹³C and the proportion of C4 plants (only the mean value is listed below), six major units are defined in the core (Figure 3). These units reflect the changes in the organic carbon flux and the proportions of C3 and C4 vegetation in the drainage basin. Details are given below.

Unit I (54.20–51.95 m, 8.7–6.5 cal. ka BP)

The values of TOC (average 1.56%, Table 2) and TN (0.10%) were relatively low but increased rapidly upwards (Figure 3). Low values of δ¹³C also occurred with an average of −22.92‰ (Table 2), and the proportion of C4 plants was ~35%. Carbonate was at a relatively high level (average 0.33%). Plots of δ¹³C versus TOC/TN suggest a predominantly C3 plant origin (Figure 4).

Table 2.

Mean values (left) and range (right) of proxies of organic elements and mineral magnetic parameters in the six sediment units (I–VI) in core CY.

Stage	I (8.7–6.5 ka BP) 51.95–53.95 m	II (6.5–4.9 ka BP) 43.75–51.95 m	III (4.9–3.2 ka BP) 39.20–43.75 m	IV (3.2–3.1 ka BP) 34.35–39.20 m	V (3.1–2.1 ka BP) 21.15–34.35 m	VI (2.1–1.3 ka BP) 15.00–21.15 m	Whole core
δ¹³C (‰)	−22.92; −23.00 ± 0.34	−23.02; −22.98 ± 0.52	−22; −22.58 ± 1.26	−20.56; −20.82 ± 0.98	−20.79; −20.82 ± 1.78	−20.23; −19.95 ± 1.28	−21.36; −21.26 ± 2.58
C/N	15.22; 15.15 ± 1.38	17.56; 17.82 ± 2.11	17.4; 17.75 ± 2.67	17.31; 17.42 ± 2.05	17.67; 16.61 ± 4.64	16.87; 16.89 ± 3.60	17.31; 16.61 ± 4.64
TOC (%)	1.56; 1.62 ± 0.17	2.07; 2.13 ± 0.41	1.99; 1.98 ± 0.53	2.07; 2.04 ± 0.42	1.4; 1.41 ± 0.88	1.31; 1.37 ± 0.52	1.68; 1.53 ± 1.01
TC (%)	1.89; 1.91 ± 0.17	2.12; 2.12 ± 0.21	2.12; 2.19 ± 0.47	2.25; 2.29 ± 0.29	1.67; 1.62 ± 0.66	1.61; 1.52 ± 0.52	1.88; 1.82 ± 0.85
TN (%)	0.1; 0.10 ± 0.01	0.12; 0.12 ± 0.02	0.11; 0.11 ± 0.02	0.12; 0.12 ± 0.02	0.08; 0.08 ± 0.03	0.08; 0.08 ± 0.02	0.1; 0.09 ± 0.05
Carbonate (%)	0.33; 0.33 ± 0.10	0.11; 0.13 ± 0.11	0.14; 0.16 ± 0.12	0.24; 0.27 ± 0.23	0.29; 0.32 ± 0.22	0.33; 0.52 ± 0.43	0.24; 0.48 ± 0.47
χ_lf (10⁻⁸ m³/kg)	8.94; 9.15 ± 1.42	8.18; 8.83 ± 2.40	11.49; 11.07 ± 3.53	12.22; 11.91 ± 3.34	17.37; 17.54 ± 6.21	15.75; 15.82 ± 4.94	13.54; 15.09 ± 8.66
χ_fd%	0.96; 1.30 ± 1.30	1.07; 2.02 ± 2.02	1.04; 1.43 ± 1.43	1.09; 1.30 ± 1.30	1.43; 1.20 ± 1.78	1.49; 1.32 ± 1.32	1.26; 2.02 ± 2.02
SIRM (10⁻⁵ A m²/kg)	1340; 1481 ± 339	1148; 1205 ± 374	1646; 1578 ± 603	1814; 1799 ± 512	2660; 2657 ± 958	2340; 2451 ± 766	2018; 2223 ± 1392
χ_ARM (10⁻⁵ m³/kg)	0.33; 0.32 ± 0.04	0.17; 0.19 ± 0.06	0.3; 0.31 ± 0.16	0.34; 0.33 ± 0.10	0.49; 0.55 ± 0.17	0.46; 0.46 ± 0.10	0.38; 0.42 ± 0.29
χ_ARM/χ_lf	37; 37 ± 5	21; 22 ± 4	25; 26 ± 8	28; 28 ± 4	29; 29 ± 5	29; 31 ± 6	27; 30 ± 12
χ_ARM/SIRM (10⁻⁴ m/A)	2.51; 2.41 ± 0.47	1.49; 1.57 ± 0.22	1.83; 1.86 ± 0.53	1.87; 1.90 ± 0.27	1.87; 1.88 ± 0.37	1.99; 2.04 ± 0.43	1.84; 2.11 ± 0.77
Soft_IRM (10⁻² A/m²)	450; 469 ± 68	422; 445 ± 143	570; 569 ± 204	626; 622 ± 188	936; 916 ± 371	820; 870 ± 316	709; 794 ± 492
HIRM (10⁻² A/m²)	33; 29.24 ± 8.55	35; 39.42 ± 17.59	51; 53.53 ± 30.28	81; 82.67 ± 32.69	81; 75.17 ± 61.09	66; 71.48 ± 35.42	63; 75.17 ± 61.09
S_−300mT(%)	97.5; 97.53 ± 0.84	96.9; 96.77 ± 1.28	96.9; 96.18 ± 1.43	95.7; 95.72 ± 1.39	96.9; 97.11 ± 2.09	97.2; 97.09 ± 1.09	96.9; 96.77 ± 2.43

TC: total carbon; TOC: total organic carbon; TN: total nitrogen; SIRM: saturation isothermal remanent magnetization; HIRM: hard isothermal remanent magnetization; Soft_IRM: soft IRM.

Unit II (51.95–43.75 m, 6.5–4.9 cal. ka BP)

Both δ¹³C and carbonate showed the lowest values (average −23.02‰; 0.11%) throughout the section, while TOC and TN contents were at their highest (average 2.07%; 0.12%). TOC and TN were higher in the middle part of this unit, and carbonate showed an inverse pattern. The maximum values of TOC and TN reached 2.54% and 0.14%, respectively, suggesting a peak flux of organic matter from the drainage basin at this stage. Plots of δ¹³C versus TOC/TN still indicate that the organic carbon was contributed predominantly by C3 plants (Figure 4). The proportion of C4 plants decreased slightly to ~34%.

Unit III (43.75–39.20 m, 4.9–3.2 cal. ka BP)

Values of δ¹³C distinctly increased (average −23.02‰) but with a large fluctuation in the lower part of the unit. The proportion of C4 plants varied between 30% and 44%. The TOC and TN values decreased slightly (average 1.99%; 0.11%), while carbonate showed a moderate increase (average 0.14%). Plots of δ¹³C versus TOC/TN reflect a migration towards the upper boundary of C3 plants (Figure 4), indicating an increase in C4 plants.

Unit IV (39.20–34.35 m, 3.2–3.1 cal. ka BP)

High values of δ¹³C, TOC and TN occurred in this unit (average −20.56‰; 2.07%; 0.12%). The proportion of C4 plants increased to 41–53%. TOC and TN declined slightly upwards, while δ¹³C and carbonate increased to −21.80‰ to 19.84‰ and 0.04–0.50%, respectively. The plots of δ¹³C versus TOC/TN showed that most of the samples are out of the range of C3 plants and are located between the ranges of C3 and C4 plants (Figure 4), indicating a mixed source of C3 and C4 plants.

Unit V (34.35–21.15 m, 3.1–2.1 cal. ka BP)

TOC and TN declined abruptly and showed a continuously decreasing trend upwards (2.29–0.53%; 0.11–0.04%), reflecting a sharp decline in the plant productivity in the catchment and therefore a decreased organic matter flux into the sea. The values of δ¹³C showed large fluctuations in a range of −22.6‰ to −19.04‰, and its average value decreased slightly to −20.79‰ (Table 2). By contrast, the carbonate content increased to 0.10–0.54% (average 0.29%). The range of δ¹³C versus TOC/TN correlation was slightly larger, and organic carbon was still contributed from both C3 and C4 plants. The proportion of C4 plants was calculated to be 37–54%.

Unit VI (21.15–15.00 m, 2.1–1.3 cal. ka BP)

The δ¹³C and carbonate values further increased (average −20.23‰; 0.33%; Table 2), while TOC and TN remained at low levels (average 1.40–1.31%; 0.08–0.08%). The range of δ¹³C versus TOC/TN was larger and deviated further from the C3 plant range (Figure 4). The proportion of C4 plants increased to 45–59%.

Mineral magnetic variations

The χ_fd % values are lower than 2% in almost all samples in the whole sequence, indicating negligible contribution from the fine-grained SP ferrimagnetic particles (Thompson and Oldfield, 1986); thus, they are not indicated in Figure 5. Significant linear correlations between χ_lf and SIRM (R²= 0.988), χ_lf and χ_ARM (R²= 0.8764) and SIRM and Soft_IRM (R²= 0.991), and low correlation between SIRM and HIRM (R²= 0.435; Figure 6), indicated that the variations of χ_lf and SIRM are mainly governed by fine-grained ferrimagnetic minerals (e.g. magnetite and maghemite), rather than the antiferromagnetic minerals (e.g. haematite). Based on the variations in the magnetic parameters, we divided the core section into six units, which coincide with the six units described above on the basis of the organic elements (Figures 3 and 5; Table 2) confirming these subdivisions as they are based on data from independent sources.

Figure 5.

Results of mineral magnetic properties of the six sediment units. Grain size (mud content) and lithology are indicated in the right panel (after Nageswara Rao et al., 2015).

Figure 6.

Linear correlations of χ_lf versus SIRM, χ_lf versus χ_ARM, SIRM versus Soft_IRM and SIRM versus HIRM from the sediments of CY core.

Unit I (54.20–51.95 m, 8.7–6.5 cal. ka BP)

All the χ_lf, SIRM, χ_ARM, Soft_IRM and HIRM values were relatively low with their averages at 8.94, 1339.65, 0.33, 450.43 and 33.13, respectively, while the ratios of both χ_ARM/χ_lf (37) and χ_ARM/SIRM (2.51 × 10⁻⁴) were the highest in the six units. The S_−300mT values were relatively high with an average of 97.47%. This indicates that the SD ferrimagnetic minerals (such as magnetite and maghemite) dominated the magnetic composition but were in low concentration.

Unit II (51.95–43.75 m, 6.5–4.9 cal. ka BP)

The χ_lf (average 8.18) and SIRM (average 1147.45) declined slightly and χ_ARM (0.17) and associated ratios (21, 1.49 × 10⁻⁴) declined sharply, while Soft_IRM, HIRM and S_−300mT remained similar to that in unit I. The decrease in these magnetic parameters indicates a lower contribution from ferrimagnetic minerals caused mainly by the decline in SD magnetic minerals.

Unit III (43.75–39.20 m, 4.9–3.2 cal. ka BP)

The χ_lf (average 11.49), SIRM (average 1645.53), χ_ARM (average 0.30), Soft_IRM (average 570.03) and HIRM (average 50.83) values increased slightly and ratios of χ_ARM/χ_lf (26) and χ_ARM/SIRM (1.83 × 10⁻⁴) distinctly increased, while that of S_−300mT remained high, indicating an increase in the absolute content of ferrimagnetic minerals.

Unit IV (39.20–34.35 m, 3.2–3.1 cal. ka BP)

The χ_lf, SIRM, χ_ARM, Soft_IRM and HIRM values and the ratios of χ_ARM/χ_lf (28) and χ_ARM/SIRM (1.87 × 10⁻⁴) decreased slightly at the bottom of this unit (39.20–38.3 m) but increased steadily upwards to the top, signifying the fall and rise in the contribution from both SD ferrimagnetic and antiferromagnetic minerals, respectively. The S_−300mT values remained high at 39.20–38.3 m depth but declined noticeably between 37.55 and 34.35 m, showing an increase in relative contribution of antiferromagnetic minerals in the upper section.

Unit V (34.35–21.15 m, 3.1–2.1 cal. ka BP)

The values of χ_lf, SIRM, χ_ARM and Soft_IRM increased to the maximum, with averages of 17.37, 2660.27, 0.49 and 936.31, respectively, and the ratios of χ_ARM/χ_lf (29) and χ_ARM/SIRM (1.87 × 10⁻⁴) remained high, reflecting the highest content of SD ferrimagnetic minerals, with an exception at 26.55–26.15 m where all these parameters showed lower values, indicating a sudden decrease in ferrimagnetic mineral content. HIRM values fluctuated frequently: trough values presenting at 34.35–32.55, 29.55–26.15, and 22.75–21.15 m depth ranges, interrupted by peak values from 32.55–29.55 and 26.15–22.75 m. The S_−300mT values fluctuated corresponding to HIRM but were all higher than 95.0%, signifying the dominance of ferrimagnetic minerals.

Unit VI (21.15–15 m, 2.1–1.3 cal. ka BP)

All the parameters, χ_lf (average 15.75), SIRM (average 2339.88), χ_ARM (average 0.46), Soft_IRM (average 820.03) and HIRM (average 65.97), declined, interrupted by a small increase at 17.75–16.55 m depth, indicating a decline and a small increase in both ferrimagnetic and antiferromagnetic minerals, respectively. The ratios of χ_ARM/χ_lf (29) and χ_ARM/SIRM (1.99 × 10⁻⁴) remained high, reflecting finer grain size of the dominant magnetic minerals. The S_−300mT values remained high, over 96.0% (average 97.2%), reflecting the predominance of ferrimagnetic minerals in this unit.

Thermomagnetic (J–T) variations

The thermomagnetization J–T curves of representative samples indicate a drop before 580°C (Figure 7), indicating clearly the presence of magnetite (Dunlop and Özdemir, 1997). A few samples showed the formation of magnetite during laboratory heating. Most of the samples, except at 47.2 m, showed a rapid decline during heating from 200°C (some even from 130°C) to 400°C, that is, giving a concave shape to the curve, with residual magnetization dropping at <680°C, suggesting the presence of maghemite and haematite; the latter could be transformed from maghemite during the heating or originally present in the samples. The increase in J between 350°C and 500°C for the samples at 40.0, 42.4 and 47.2 m reflects conversion of paramagnetic minerals such as iron-bearing silicate or clay minerals and/or pyrites into ferrimagnetic minerals (magnetite). Thus, J–T curves reflect the dominance of magnetite and maghemite, both being the diagnostic minerals in the erosion and weathering products of the Deccan basalts.

Figure 7.

Thermomagnetization curves for typical samples from six units in core CY. Only heating curves are indicated.

Summary of variations in organic carbon, magnetic minerals and sediment sources

The ratios of stable carbon isotopes and TOC/TN in the bulk organic matter directly reflected changes in organic carbon source and associated vegetation type, that is, C3 (high moisture) and C4 (low moisture) plants (Lamb et al., 2006), because the contribution from algal sources was negligible (Figure 4). The organic carbon was dominated by C3 plants before 4.9 cal. ka BP and joined by some C4 plants during 4.9–3.2 cal. ka BP. The C4 plant contribution abruptly increased from 3.2 cal. ka BP, but this was interrupted by some recovery of C3 plants during 3.1–2.1 cal. ka BP. The C4 plants increased again during 2.1–1.3 cal. ka BP (Figure 4). In addition, TOC supply declined from 3.1 cal. ka BP.

The fine-grained (SD) magnetic mineral concentration was high before 6 cal. ka BP, but declined during 6–4.9 cal. ka BP. The absolute concentration of SD magnetite and maghemite increased again from 4.9 cal. ka BP accompanied by an increase in antiferromagnetic minerals. There was another abrupt increase in SD ferrimagnetic minerals at 3.1 cal. ka BP. These changes in the content and types of magnetic minerals reflected changes in sediment supply from the Deccan Plateau since the SD magnetite and maghemite are the diagnostic magnetic minerals of the Deccan volcanic province (Kulkarni et al., 2014). We further infer from the high values of HIRM that antiferromagnetic minerals such as haematite became enriched in the soil owing to the aridification in the Deccan Plateau (Thompson and Oldfield, 1986). Therefore, the sediment supply originating from the Deccan basalt declined from 6 cal. ka BP and increased again from 4.9 cal. ka BP, followed by a further increase from 3.1 cal. ka BP.

Discussion

Vegetation and sediment provenance response to the declining Indian monsoon during the middle to late Holocene

High values of δ¹³C correspond to a larger proportion of C4 plants and low precipitation, which reflected a weak monsoon phase. However, the composition of organic carbon isotopes in both the CY core from the Godavari Delta plain (this study) and NGHP-16A, a marine core from off the Godavari Delta (Ponton et al., 2012), did not reflect a weak monsoon at 6 cal. ka BP unlike the other records from the Arabian Sea and Qunf Cave (Figure 8) and from Lonar Lake in the upper Godavari basin (Prasad et al., 2014; Sarkar et al., 2015). We infer, therefore, C3 plants were abundant in the upper drainage basin in the Deccan Plateau during the stronger monsoon phase (i.e. before 6 cal. ka BP), and as the monsoon weakened, the main contribution was from the middle and lower parts of the basin where C3 plants still persisted. Both scenarios resulted in a major organic carbon source from C3 plants, which explains the low values of δ¹³C before 4.9 cal. ka BP (Figure 8a and b). This inference is also supported by the change in magnetic properties in core CY (Figure 8c). The decline in the fine-grained magnetic minerals from ~6 cal. ka BP reflects a reduction in the sediment supply from the Deccan Plateau. We suggest such a decline can be induced by less atmospheric moisture being delivered to the Plateau and, therefore, less precipitation and low sediment contribution from the Plateau during the weak monsoon phase.

Figure 8.

(a) δ¹³C_org record in core CY from the Godavari delta plain (this study). (b) δ¹³C_wax record (weight-averaged for n-alkanoic acids C26–C32) in core NGHP-16A off the Godavari delta (Ponton et al., 2012). (c) χ_ARM record in core CY which reflects sediment contribution from the Deccan Plateau (this study). Values during 4.9–1.3 cal. ka BP are not presented because of a dominant information of human activity. (d) Percentage of foraminifera representative of upwelling, Globigerina bulloides, which reflects the strength of Indian monsoon from the Arabian Sea (Gupta et al., 2003). (e) Speleothem δ¹⁸O record from Qunf Cave, Oman (Fleitmann et al., 2003).

Then, the increases in δ¹³C values and the proportion of C4 plants from ~4.9 cal. ka BP at core CY (Figure 8a) suggested a gradual weakening of the Indian monsoon, which is consistent with the records of compound-specific carbon isotopes at core NGHP-16A (Figure 8b; Ponton et al., 2012), the foraminifera from the Arabian Sea representative of upwelling (Figure 8d; Gupta et al., 2003) and δ¹⁸O from a speleothem at Qunf Cave, Oman (Figure 8e; Fleitmann et al., 2003). The more or less stepwise changes in δ¹³C values and C4 plant proportions at ~4.5 cal. ka BP and an abrupt increase in C4 plants at ~3.2 cal. ka BP in core CY (Figure 8a) were also recorded at core NGHP-16A (Figure 8b; Ponton et al., 2012). By contrast, the speleothem and foraminifera recorded a continuing gradual weakening (Figure 8d and e). We infer such differences might reflect the nature of vegetation succession when responding to the decline in monsoon precipitation, that is, the C3 plant community may have resistance within a range of climate change and then be replaced by the C4 community when precipitation declines beyond that range. Furthermore, we suggest the shift of organic matter sources could also be linked to a change in sediment provenance, that is, an increase in sediment supply from the Deccan Plateau where only C4 plants grew due to the monsoon weakening. The stronger magnetic properties inferred the increased Deccan basalt source despite the monsoon weakening from ~4.9 cal. ka BP, which should be induced by human activity as we will discuss in the next section. This phenomenon indicates the limitation of organic elemental proxies for palaeo-monsoon reconstruction.

Of note, the aridification during 4.9–3.2 cal. ka BP was interrupted by two humid events as indicated by abrupt increases in C3 plants at ~4.6 and ~4.0 cal. ka BP (Figure 8a). Increases in C3 plants were also recorded at core NGHP-16A (Figure 8b; Ponton et al., 2012) but at somewhat different ages, which could be induced by the extremely low sedimentation rate in both the sediment cores (Figure 2; Ponton et al., 2012) and a relatively coarser sampling interval at core NGHP-16A than at CY (Figure 8a and b). At the same time, the foraminifera indicative of upwelling in the Arabian Sea (Figure 8d) and the speleothem in Qunf Cave, Oman (Figure 8e), also recorded stronger Indian summer monsoon and wetter conditions around 4.6 and 4.0 cal. ka BP. We infer that the recovery of C3 plants was much faster than its deterioration as indicated by the abrupt and large amplitude decline in δ¹³C values (Figure 8a). Similarly, there was again an abrupt increase in C3 plants during ~3.1–2.8 cal. ka BP and at ~2.1 cal. ka BP recorded in core CY, which was consistent with the wetter conditions in the Arabian Sea (Figure 8a and d).

Deforestation and erosion induced by human activity during late Holocene

The increase in ferrimagnetic concentration and sediment contribution from the Deccan Plateau since 4.9 cal. ka BP, despite the monsoon weakening, was because of increased soil erosion induced by human activity (deforestation and early agriculture) in the region (Figures 8 and 9). Increase in ferrimagnetic concentration and sediment contribution from the Deccan basalt has also been reported for the surficial sediments in the Bay of Bengal (Kulkarni et al., 2015; Sangode et al., 2001). Kulkarni et al. (2015) linked this phenomenon to the weaker monsoon, which induced a higher proportion of the sediment load of the Godavari River from the Deccan basaltic provenance. However, the sediment record in core CY reveals weaker magnetic properties during the weaker monsoon because of a decline in an overall supply of magnetic minerals from the Deccan Plateau from 6 cal. ka BP (Figure 8c). The sediment budget in the Godavari delta also witnessed a decline in sediment accumulation caused by the weaker monsoon during the third millennium BC (Figure 9f; Nageswara Rao et al., 2015). Subsequently, a significant increase in the amount of deltaic sedimentation that occurred during the second millennium BC (Figure 9f; Nageswara Rao et al., 2015) despite the continued weaker monsoon was apparently due to the increase in Neolithic settlements in the Deccan Plateau (Figure 9e; Fuller et al., 2004). The increasing number of human settlements (Figure 9e; Ponton et al., 2012) and, therefore, intensification of sedentary agricultural activity led to accelerated erosion in the Deccan Plateau despite the weaker monsoon from ~4 cal. ka BP, which is supported by the increase in magnetic parameters (Figures 5 and 9c and d).

Figure 9.

(a–d) TOC, δ¹³C, χ_ARM and HIRM in the sediments of core CY. (e) Number of settlements in Neolithic (solid grey) and Chalcolithic (solid black) cultures in the Deccan Plateau (after Ponton et al., 2012). (f) Total quantity of sediment accumulated in the Godavari delta on the millennial scale (Nageswara Rao et al., 2015).

Human activity was low key and not well established during the third millennium BC as indicated by the relatively low values of magnetic parameters except for a few spikes. We also suggest the impact of the Neolithic people was still limited in changing the natural vegetation during the second millennium BC as TOC remained high until ~3.1 cal. ka BP (Figure 9a), although the number of settlements in the drainage basin area and the sediment supply to the delta increased. Furthermore, the magnetic parameters decreased and TOC increased when the C3 plants recovered during the wetter events of ~4.6 and 4.0 cal. ka BP (Figure 9a–d), which suggested less erosion due to improved vegetation coverage. Thus, the Neolithic people in the Deccan Plateau mainly depended on or utilized the natural resources without resorting to any significant deforestation during 4.9–3.1 cal. ka BP.

The abrupt decrease in both the TOC and TN contents in the core since ~3.1 cal. ka BP (Figure 9a) suggested a severe decline in organic matter supply from the drainage basin. We suggest such a decline was induced by deforestation and cultivation in the Deccan Plateau during the Chalcolithic cultural period because it was accompanied by a sharp increase in the magnetic mineral concentration (Figures 5 and 9c); the latter represents a significant increase in erosion from the Plateau (Kulkarni et al., 2015; Sangode et al., 2001). Ponton et al. (2012) reported that the number of settlements increased from 255 (Neolithic) to 1056 (Chalcolithic) (Figure 9e) because of human adaptation in the semi-arid Deccan Plateau at ~3.2 cal. ka BP. The significant increase in both the sedimentation rate in core CY (Figure 2) and the sediment accumulation in the delta region (Figure 9f), in contrast to the continuous monsoon weakening since ~3.0 cal. ka BP (Figure 8d), also suggests that the substantial increase in sediment yield was induced by human activity in the drainage basin. Therefore, we infer that human activity led to significant vegetation deterioration in the semi-arid Deccan Plateau during the Chalcolithic cultural period. Indeed, the records of magnetic parameters in core CY give more detailed information of the human impact: the abrupt increase in HIRM occurred at ~3.2 cal. ka BP, which is ~100 years earlier than the abrupt increase in χ_lf (or χ_ARM) and decline in TOC (Figures 5 and 9). This suggests that soil erosion predated the beginning of the Deccan Chalcolithic culture, followed by deforestation and rock erosion a hundred years later.

Therefore, it seems that during the relatively wet periods such as 3.1–2.8 and 2.3–2.1 cal. ka BP, as reflected by the recovery of C3 plants in the Godavari drainage basin and foraminifera in the Arabian Sea (Figure 8), soil/rock erosion intensified as indicated by the rapid increase in HIRM and χ_ARM (Figure 9c and d) and deforestation and sedentary agricultural activities expanded as indicated by further decline in TOC values (Figure 9a). This pattern is reversed as compared with that at ~4.6 and 4.0 cal. ka BP in the Neolithic time (Figure 9), which implies a significant increase in the human impact on the natural environment since the start of the Chalcolithic cultural period owing to both the increase in human population and the improvements in agricultural practices. There have been reports about ancient dams in central India built before the Christian era, which were used for providing water for irrigation, most likely for rice cultivation as a response to the increased population (Shaw et al., 2007; Shaw and Sutcliffe, 2003). There were also historic cities (e.g. Pratisthana) and centralized states in the late centuries BC with the growing need for cultivating food crops. Thus, the increase in C3 plants could be partially related to cultivation of rice enhanced by the irrigation practices (Shaw et al., 2007) in a beneficial climate setting during 2.3–2.1 cal. ka BP. This implies a sensitive linkage between the human society and monsoon precipitation in central India.

Conclusion

This study, based on the proxies of organic matter and mineral magnetism from a well-dated sediment core CY from the Godavari delta plain, revealed vegetation response and associated soil/rock erosion driven by the Indian monsoon weakening along with enhanced human activity in the Godavari drainage basin during the middle to late Holocene. The variation of proxies can be divided into five stages characterized by (1) high organic carbon content, low values of δ¹³C and slightly high values of concentration-dependent magnetic parameters before 6 cal. ka BP; (2) high organic carbon content and low values of δ¹³C and concentration of ferrimagnetic minerals during 6–4.9 cal. ka BP; (3) increase in δ¹³C and concentration of ferrimagnetic minerals during 4.9–3.2 cal. ka BP; (4) high values of δ¹³C and magnetic parameters during 3.2–3.1 cal. ka BP; and (5) a decrease in organic carbon content but high values of δ¹³C and further increase in magnetic concentration during 3.1–1.3 cal. ka BP.

These changes in organic elemental and magnetic proxies indicate two stages of increase in organic carbon contribution from C4 plants from 4.9 and 3.2 cal. ka BP, respectively, and a significant decline in organic carbon supply from 3.1 cal. ka BP along with enhanced sediment yield from the Deccan Plateau in the upper Godavari drainage basin. The change in organic carbon source not only reflects the vegetation succession from C3 to C4 plants responding to the aridification in the Godavari drainage basin but also as a result of increased sediment supply from the C4-dominated Deccan Plateau induced by the intensified erosion driven by the expansion of agricultural activities since the start of the Neolithic period. The decline in organic carbon supply versus increase in sediment supply from the Deccan Plateau from ~3.1 cal. ka BP reflects severe deforestation and soil/rock erosion induced by the rapid development of the Deccan Chalcolithic culture. In addition, soil/rock erosion decreased during the wet events in the Neolithic period but increased during the wet events in the Chalcolithic period, indicating variations in the human–environment relationships during the two cultural periods and a rapid response of human agricultural activities in a beneficial climate setting in the Chalcolithic period.

Footnotes

Acknowledgements

The authors are grateful to Dr Brian Finlayson for his valuable suggestions for improving the manuscript.

Funding

ZW acknowledges the financial support by National Key Research Programme (grant no. 2016YFC0402600). KNR thanks Space Applications Centre, Department of Space, Government of India for funding the core drilling in the Krishna–Godavari delta region through PRACRITI project (SAC/RESA/PRACRITI/ESHD/SLR/WP02/2010) and DMSP project (SAC/EPSA/MPSG/DMSP/CV/WP/02/2012). SJS acknowledges Department of Science and Technology, New Delhi (grant SR/S4/ES-409/2009), for funding, which enabled mineral magnetic analysis.

References

Ahmad

Padmakumari

Babu

(2009) Strontium and neodymium isotopic compositions in sediments from Godavari, Krishna and Pennar rivers. Currentence 97: 1766–1769.

Asouti

Fuller

(2008) Trees and Woodlands of South India: Archaeological Perspectives. Walnut Creek, CA: Left Coast Press.

Berkelhammer

Sinha

Stott

et al . (2012) An abrupt shift in the Indian monsoon 4000 years ago. In: Giosan

Fuller

Nicoll

et al . (eds) Climates, Landscapes, and Civilizations: Geophysical Monograph, vol. 198. Washington, DC: American Geophysical Union, pp. 75–88.

Bird

Polisar

Lei

et al . (2014) A Tibetan lake sediment record of Holocene Indian summer monsoon variability. Earth and Planetary Science Letters 399: 92–102.

Bloemendal

King

Hall

et al . (1992) Rock magnetism of Late Neogene and Pleistocene deep-sea sediments: Relationship to sediment source, diagenetic processes, and sediment lithology. Journal of Geophysical Research Atmospheres 97: 4361–4375.

Dixit

Hodell

Petrie

(2014) Abrupt weakening of the summer monsoon in northwest India 4100 yr ago. Geology 42: 339–342.

Dunlop

Özdemir

(1997) Rock Magnetism: Fundamentals and Frontiers. Cambridge: Cambridge University Press.

Fleitmann

Burns

Mangini

et al . (2007) Holocene ITCZ and Indian monsoon dynamics recorded in stalagmites from Oman and Yemen (Socotra). Quaternary Science Reviews 26: 170–188.

Fleitmann

Burns

Mudelsee

et al . (2003) Holocene forcing of the Indian monsoon recorded in a stalagmite from southern Oman. Science 300: 1737–1739.

10.

Fuller

Korisettar

Venkatasubbaiah

et al . (2004) Early plant domestications in southern India: Some preliminary archaeobotanical results. Vegetation History and Archaeobotany 13: 115–129.

11.

Fuller

(2000) Ashmounds and hilltop villages: the search for early agriculture in southern India. Archaeology International 4: 43–46.

12.

Fuller

(2011) Finding Plant Domestication in the Indian Subcontinent. Current Anthropology 52: S347–S362.

13.

Fuller

Harvey

(2006) The archaeobotany of Indian pulses: Identification, processing and evidence for cultivation. Environmental Archaeology 11: 219–246.

14.

Fuller

Murphy

(2014) Overlooked but not forgotten: India as a center for agricultural domestication. General Anthropology 21(2): 1–8.

15.

Fuller

Kingwell-Banham

Lucas

et al . (2015) Comparing pathways to Agriculture. Archaeology International 18: 61–66.

16.

Gupta

Anderson

Overpeck

(2003) Abrupt changes in the Asian southwest monsoon during the Holocene and their links to the North Atlantic Ocean. Nature 421: 354–357.

17.

Gupta

Anderson

Pandey

et al . (2006) Adaptation and human migration, and evidence of agriculture coincident with changes in the Indian summer monsoon during the Holocene. Current Science 90: 1082–1090.

18.

Jain

Pushpendra Agarwal

Singh

(2007) Hydrology and Water Resources of India. Dordrecht: Springer.

19.

Kulkarni

Sangode

Bloemendal

et al . (2015) Mineral magnetic characterization of the Godavari River and Western Bay of Bengal Sediments: Implications to source to sink relations. Journal of the Geological Society of India 85: 71–78.

20.

Kulkarni

Sangode

Meshram

et al . (2014) Mineral magnetic characterization of the Godavari River sediments: Implications to Deccan basalt weathering. Journal Geological Society of India 83: 376–384.

21.

Lamb

Wilson

Leng

(2006) A review of coastal palaeoclimate and relative sea-level reconstructions using δ13C and C/N ratios in organic material. Earth-Science Reviews 75: 29–57.

22.

Liu

Roberts

Larrasoana

et al . (2012) Environmental magnetism: Principles and applications. Reviews of Geophysics 50: RG4002.

23.

Misra

(2001) Prehistoric human colonization of India. Journal of Biosciences 26: 491–531.

24.

Nageswara Rao

(1997) Interannual variations of Monsoon Rainfall in Godavari River Basin – connections with the Southern Oscillation. Journal of Climate 11: 768–771.

25.

Nageswara Rao

Saito

Nagakumar

KCV

et al . (2015) Palaeogeography and evolution of the Godavari delta, east coast of India during the Holocene: An example of wave-dominated and fan-delta settings. Palaeogeography, Palaeoclimatology, Palaeoecology 440: 213–233.

26.

Ponton

Giosan

Eglinton

et al . (2012) Holocene aridification of India. Geophysical Research Letters 39: 419–420.

27.

Pradhan

Shirodkar

et al . (2014) Multi-proxy evidence for compositional change of organic matter in the largest tropical (peninsular) river basin of India. Journal of Hydrology 519: 999–1009.

28.

Prasad

Anoop

Riedel

et al . (2014) Prolonged monsoon droughts and links to Indo-Pacific warm pool: A Holocene record from Lonar Lake, central India. Earth and Planetary Science Letters 391: 171–182.

29.

Sangode

Suresh

Bagati

(2001) Godavari source in the Bengal fan sediments: Results from magnetic susceptibility dispersal pattern. Current Science 80: 660–664.

30.

Sarkar

Prasad

Wilkes

et al . (2015) Monsoon source shifts during the drying mid-Holocene: Biomarker isotope based evidence from the ‘core monsoon zone’ (CMZ) of India. Quaternary Science Reviews 123: 144–157.

31.

Shaw

Sutcliffe

(2003) Ancient dams, settlement archaeology and Buddhist propagation in central India: The hydrological background. Hydrological Sciences Journal 48: 277–291.

32.

Shaw

Sutcliffe

Lloyd-Smith

et al . (2007) Ancient irrigation and Buddhist history in central India: Optically stimulated luminescence dates and pollen sequences from the Sanchi Dams. Asian Perspectives 46: 166–201.

33.

Shinde

(1990) Settlement pattern of the Savalda culture: The first farming community of Maharashtra. Bulletin of the Deccan College Research Institute 49: 417–426.

34.

Stevens

Murphy

Roberts

et al . (2016) Between China and South Asia: A middle Asian corridor of crop dispersal and agricultural innovation in the Bronze Age. The Holocene 26: 1541–1555.

35.

Thamban

Kawahata

Rao

(2007) Indian summer monsoon variability during the Holocene as recorded in sediments of the Arabian Sea: Timing and implications. Journal of Oceanography 63: 1009–1020.

36.

Thompson

Oldfield

(1986) Environmental Magnetism. London: Allen & Unwin.

37.

Zorzi

Sanchez Goñi

Anupama

et al . (2015) Indian monsoon variations during three contrasting climatic periods: The Holocene, Heinrich Stadial 2 and the last interglacial–glacial transition. Quaternary Science Reviews 125: 50–60.