Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 www.elsevier.com/locate/palaeo

Relationship of environmental changes in central to possible prehistoric land-use and climate changes ⁎ Rathnasiri Premathilake

Postgraduate Institute of Archaeology, University of , 407, Colombo 7, Sri Lanka and Physical Geography and Quaternary Geology, S-10691, Stockholm University, Sweden Received 1 October 2003; received in revised form 5 August 2005; accepted 2 March 2006

Abstract

A radiocarbon-dated pollen-analysed peat sequence from the Horton Plains (>2000 m a.s.l.), in central Sri Lanka, together with physical and chemical parameters (organic carbon, mineral magnetics, carbon isotopes and phytoliths), indicates major environmental changes during the last 24,000 years. The results suggest that a mobile life form, i.e. a hunter–forage culture, predominated in an open landscape, associated with xerophytic vegetation, e.g. Chenopodium spp. at ∼17.5 ka BP. Incipient management of cereal plants and slash-and-burn techniques seem to have prevailed between 17.5 and 13 ka BP, which was indigenous and associated with grazing. Evidence of systematic cereal cultivation in the form of oat and barley pollen grains is found from the late Pleistocene (∼13 ka BP). This is the earliest evidence of farming activities noted in Sri Lanka as well as in south . After 13 ka BP, cereal cultivation was associated with an increase in humidity. With a later abrupt increase in aridity, agricultural land-use decreased from ∼8to∼3.6 ka BP, when the area appears to have been almost deserted. After a severe middle Holocene arid phase (i.e. 5.4–3.6 ka BP), the agricultural activity with a limited extension was again initiated by ∼2.9 ka BP. During the next ∼900 years, cultivation ceased allowing the upper montane rain forest to dominate. Between 0.2 and 0.15 ka BP, new phases of agricultural activities were undertaken and potato cultivation took place lately, between 1950 and 1969 AD. © 2006 Elsevier B.V. All rights reserved.

Keywords: Sri Lanka; Horton Plains; Pollen; Land-use; Climate; Vegetation

1. Introduction fauna. Evidence of Mesolithic settlements on the island has been recorded in the form of stone artefacts, together Sri Lanka is a tropical island in the , with osteological and archaeo-botanical remains col- hosting prehistoric human settlements from ∼0.2 million lected from several sequences in cultural layers in cave years ago to the beginning of the historical period at sites, radiocarbon dated to 37 ka BP (, ∼2.5 ka BP (Deraniyagala, 1992; Kennedy, 1999). The 1992; Kennedy, 1993; Kennedy and Zahorsky, 1995; stratified archaeological data on prehistoric subsistence Kennedy, 1999). Possible indications of domestic dogs strategy in Sri Lanka indicate a hunting and gathering have been found from several Mesolithic camp sites economy based on resources from existing flora and (Deraniyagala, 1992). Tentative evidence of domesti- cated cereals is available from a cave at ∼7kaBP ⁎ Fax: +46 94 11 2 694151. (Deraniyagala, 2001). In the Indian sub-continent, this E-mail address: [email protected]. domestication is supported by finds of microlithic

0031-0182/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.palaeo.2006.03.001 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 469 industries associated with Mesolithic cultures of hunting tural land-use with barley (Hordeum sp.) cultivation people, fishermen and pastoralists or people practising dates back to the middle Holocene (Premathilake, some form of early agriculture (Allchin and Allchin, 1997). A 6 m long peat sequence, covering the last 1989). 24 ka years, provided evidence for a climatic amelio- Despite numerous attempts in the past, conclusive ration around 18.5 ka BP, favouring a gradual expansion evidence of the origins of has of tropical/equatorial forests and subsequently agricul- not yet been found (e.g. Deraniyagala, 1992; Kennedy ture (Premathilake and Risberg, 2003). and Zahorsky, 1995). Therefore, our knowledge regard- The horizons at Doravaka-lena shelter incorporate ing the early farming culture in Sri Lanka is incomplete, geometric microlithic tools and crude red pottery in mainly because of a lack of systematic and interdisci- association with finger millet (Eleusine coracana) plinary investigations. However, the location of Sri grains radiocarbon dated to ∼7 ka BP (Wijeyapala Lanka can be considered as critical for testing the pers. comm.). Black and Red Ware pottery from origins of agriculture due to a diverse landscape and overlying strata, dated to ∼5 ka BP, is possibly evidence resource availability (Abeywickrama, 1955; Harris, for a Neolithic culture in Sri Lanka (Deraniyagala, 1972; Hather, 1994; Manatunga, 1996; Yasuda, 2002). 2001). The agricultural history has been interpreted It is obvious that several of the basic premises for the from stratified archaeological data obtained from the origin of agriculture are available in Sri Lanka. These main site at Anuradapura, as representing a clear change comprise diversified natural habitats: i.e, the location in land-use between 3 and 2.5 ka BP. During this period, where these plants and animals occur in the wild, the intensification of agricultural land-use had already floristic richness, woodlands, diversified terrains and occurred in the lowland area. Evidence for the earliest special skills e.g. a very old tool making tradition. A farming activities in Sri Lanka date to the prehistoric mild climate with reversed monsoons giving abundant early iron using communities (i.e. 3–2.5 ka BP) engaged rains and dry periods also occur. in horse breeding, iron production and rice cultivation. Evidence of Neolithic settlements has been recorded This contradicts the traditional opinion that according to in the form of stone axes and associated domestic plant the 6th century AD chronicles (i.e. Mahavamsa, 1950), and animal remains, collected from cultural layers at Indian migrants have introduced the farming tradition to many archaeological sites in , but not in Sri the island ∼3kaBP(Gunawardana, 1984). Lanka. The origins of farming practices appear to have a It is suggested that a more comprehensive study is considerable antiquity in the Indian sub-continent, necessary to test the origin of agriculture in Sri Lanka, as starting from ∼10 ka BP (Allchin and Allchin, 1989; well as in south Asia. The present paper addresses how Gupta, 2004; Sharma et al., 2004; Singh, 2005). During new interdisciplinary research on prehistoric land-use the first three millennia thereafter, mud–brick architec- patterns from the Horton Plains, central Sri Lanka, can ture was developed and domestic wheat, barley, cattle be interpreted and integrated in the current discussion and goat have also been exploited in an aceramic about the history of agriculture. The following topics are Neolithic context. Evidence from Indo-Iranian border- considered: (1) when did agriculture start, (2) what lands suggested that domestic sheep, cattle and goat crops were cultivated, (3) how has climate affected were already husbanded by humans 10–7 ka ago. The agriculture and (4) what other human activities can be data on the origin of crop cultivation in the Indian sub- traced. Methods applied are high-resolution analyses of continent, however, are still too inadequate to lead to a pollen, charcoal, organic carbon, carbon and nitrogen coherent agricultural history (Chakrabarti, 2000). isotopes, and mineral magnetic properties. In south-eastern Asia the earliest known evidences show that nuts and tubers were domesticated ∼11 ka BP 2. The environment of the study area (Sundara, 1985). In other parts of the world the changes in subsistence patterns from foraging to farming have The Horton Plains are characterised by a rolling been described from several palaeoecological investiga- landscape with mires, plains, forested hilltops, grassy tions (Normile, 1997; Shen and Crawford, 1998; slopes, precipices, brooks and waterfalls. It is located Hillman et al., 2001; Toyama, 2001). Hillman et al. between 6°47′–6°50′ N and 80°46′–80°51′ E in the (2001) report that the earliest farming activities in south- Central Highlands, 20 km to the south from the well western Asia date to 13 ka BP and in south-eastern Asia known town “Nuwara Eliya” of Sri Lanka and was to 14 ka BP (Toyama, 2001). designated as a National Park in 1969 in order to A palynological study of a peat sequence obtained preserve the natural montane ecosystem and habitat from the elevated Horton Plains indicated that agricul- (Fig. 1). The area covers ∼3160 ha at 2100–2300 m a.s. 470 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

pers. comm.). This weathering product is overlain by sediments consisting of clastic particles i.e. sandy clay and silt mixed with organic matter (Table 1). Herba- ceous peat constitutes the main part of the core. The peat is dominated by partially decomposed plant remains (e.g. grasses) and humus, occasionally mixed with fine to coarse-grained sediments. Pollen and spore analyses, mineral magnetic, organic carbon and phytolith mea- surements were obtained from one of the cores, while chemical parameters were measured on the comple- mentary core.

3.1. Pollen and spores

The conventional acetolysis method was used for concentration of pollen grains (Berglund and Ralska- Jasiewiczowa, 1986). Analyses were carried out in intervals varying between 4 and 1 cm. Counting took place under a magnification of X500 (standard), with X1250 and phase contrast for critical identifications Fig. 1. Location map of the Horton Plains national park in central Sri (microscopes used were Leitz, Laborlux-S and Zeiss, Lanka (park boundary is marked by the dashed line). Thin lines mark Axioplane 2). On average, 750 pollen grains were the major drainage patterns. Thick lines indicate major roads. FI = counted in each sample. Identifications were based on Farr's Inn bungalow. Mountain peaks: To = Totupola, Ki = relevant literature in tropical/subtropical pollen flora Kirigalpotta. Insert map of southern and Sri Lanka. Archaeo- logical locations mentioned in the text are included. 1 = Mannar, 2 = (e.g. Huang, 1972; Zheng, 1982; Reille, 1998), together , 3 = and 4 = Doravaka-lena. with two reference slide collections, one made at the Palynological Laboratory, Museum of Natural History, l. altitude. The bedrock mainly consists of highly Stockholm, Sweden, and the other at the Postgraduate metamorphic rocks, e.g. garnetiferous gneisses, quartz and granulites (Cooray, 1984). The present day climate Table 1 of the Horton Plains is wet (Balasubramaniam et al., Lithological description of the material analysed 1993). The average rainfall is about 2200 mm/yr. The Depth (cm) Lithology Colour Macrofossils precipitation is to a great extent determined by the 000–139 Peat BL/VDB (10YR 2/1, 2/2) 1,2,3,4 Southwest Monsoon (SWM), which reaches peaks in 139–165 Sandy peat BL/VDB (10YR 2/1, 2/2) 1,2,3,4 June and August. The lower amount of precipitation in 165–172 Peat BL (10YR 2/1) 1,2,3,4 December is determined by the Northeast Monsoon 172–200 Sandy peat BL (10YR 2/1) 1,4 – (NEM). During the driest period, in February, the mean 200 216 Peat BL (10YR 2/1) 1,2,4 216–228 Sandy peat BL (10YR 2/1) 1,4 temperature is 12 °C and the night temperature drops to 228–353 Peaty sand BL/VDB (10YR 2/1, 2/2) 1,2,3,4 5 °C. Upper montane rain forest (UMRF) and grasslands 353–367 Peat BL (10YR 2/1) 1,3,4 dominate the recent vegetation of which ∼50% is 367–373 Sandy peat BL (10YR 2/1) 1,4 endemic to Sri Lanka. Stratigraphical studies and 373–405 Peat GA/BL (10YR 5/1,2/1) 1,2,4 405–432 Sandy peat GB (10YR 5/2) 1,2 sampling were performed in one of the major valley – – – 432 525 Peaty sand BL/VDB (10YR 2/1, 2/2) systems of Belihul Oya extending in north south 525–529 Silt + organic GB (10YR 5/2) 1,3,4 direction. The valley bottom is occupied by mire, 529–585 Silt + organic BL/YE/WH/OYE 3,4 mainly covered by dwarf bamboo (Arundinaria densi- (10YR 2/1, 5Y 8/6, 8/1, 6/8) folia) and grass vegetation. 585–600 Sand + clay, WH/OYE/YE 1,3,4 (5Y 8/1, 6/8, 8/6) 3. Materials and methods Colours are determined according to the Munsell Soil Colour Chart (1988). Colours: black (BL); very dark brown (VDB); dark brown (DB); Two parallel 6 m long cores were taken with a greyish brown (GB); yellow (YE); olive yellow (OYL); white (WH); Russian sampler (diameter 5 cm). The lowermost part greyish (GA). Macrofossils: 1 = grass; 2 = wood; 3 = charcoal; 4 = consists of kaolinite, i.e. weathered bedrock (Risberg non-identified. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 471

Institute of Archaeology, University of Kelaniya, Sri (1981), Fernando (1984), Berglund and Ralska-Jasie- Lanka. wiczowa (1986), Manatunga (1996), Saarse et al. (1999) The number of microscopic charcoal particles with a and Poska (2001). Palynological richness, i.e. the total diameter >25 μm was also counted on the same slides as number of taxa, was calculated as an additional indicator used for pollen analyses. Percentages were calculated as of human impact. Especially, the number of herb taxa is charcoal particles/charcoal particles + pollen sum. The expected to increase in connection with human activity. abundance of each taxon was calculated as a percentage Since the basic sum is varying, this technique gives only of the total pollen sum excluding aquatic pollen taxa. a rough estimate. Pollen concentration values were calculated as an additional parameter. These, together with microscopic 3.2. Identification of cereal pollen types charcoal and organic carbon curves, are also shown in the pollen diagrams. The local pollen assemblage zones The identification of pollen types from cultivated (LPAZ) were differentiated on the basis of fluctuations Hordeum, Avena and Triticum was based on measure- in tree pollen frequencies, by subjective boundary ments of size, annulus diameter (anl-D) and surface definitions. CONISS constrained cluster analysis was structures referring to Beug (1961), Andersen and used as complementary method (Grimm, 1987). Bertelsen (1972), Andersen (1979) and Kohler and In the human impact diagram, the identified herb Lange (1979). According to these references, cultivated pollen taxa were categorized in 16 groups following the cereal pollen types should be >37 μm in size and have ecological descriptions given by Bond (1953), Pema- an anl-D>8 μm. This literature refers to a temperate dasa (1984) and Dassanayake et al. (1980–2000). These environment making a direct comparison with tropical groups were summarized into five land-use categories conditions difficult. Measurement on pollen grains from (Table 2) following Holmes (1951), Bond (1953), Perera recent material collected from the Horton Plains during (1969), Muller-Dombois and Perera (1971), Behre the 1970's, deposited at the herbarium in Peradeniya,

Table 2 List of pollen taxa included in recognized land-use categories Groups Ecological groups Indicator taxa Land-use categories A Cultivated plants Hordeum sp., Avena sp., Triticum sp. Cultivated fields B Pastures Indigofera sp., Cerastium sp., Desmodium spp., Pastures Achyranthes spp., C Field and disturbed grounds Polygonum lapathifolium, Plantogo spp., Humulus Ruderal communities (major) sp. D Waste grounds Chenopodiaceae, Chenopodium spp., Gomphrena sp., Ruderal communities (major) Hewittia sp., Cruciferae, Cassia sp., Spermacoce sp., Malva sp. E Open ground/patanas Bupleurum sp. Ruderal communities (major) F Weed Centella sp., Coriandrum sp., Aster sp., Crepis sp., Ruderal communities (minor) Cleome sp., Stellaria sp., Evolvulu sp., Caryophyllaceae, Salvia sp., Ocimum sp., Labiatae, Polygonum caespitosum, Veronica sp. G Patanas Heracleum sp., Pimpinella sp., Emilia sp, Senecio sp., Patanas Campanula sp., Crotalaria sp., Smithia sp., Gentiana sp., Swertia sp., Agrimonia sp., Pedicularis sp. H Disturbed/damp grounds Hydrocotyle sp., Cardamine african, Dipsacus sp., Ruderal communities (minor) Exacum sp., Juncus sp., Liliaceae, Limnophila sp. I Open grounds Peucedanum sp., Sanicula sp., Galium sp. Ruderal communities (minor) J Shady grounds Arisaema leschenautii, Rhynchotechum sp., Ruderal communities (minor) Balanophora sp. K Shady/damp/waste grounds Asteraceae (echinolophate), Patrinia sp. Ruderal communities (minor) L Damp/shady grounds Impatiens spp., Commelinaceae, Neanotis spp., Ruderal communities (minor) Ranunculus spp. M Waste/patanas Hypericum japonicum Ruderal communities (minor) N Secondary forest/grasslands Euphorbia sp. Ruderal communities (major) O Secondary forest/shady grounds Coleus sp., Rubia sp. Ruderal communities (major) P Disturbed grounds Oenothera sp., Plumbago sp., Verbascum sp., Ruderal communities (minor) Verbena sp. 472 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 revealed that cultivated Hordeum sp. (n=25) had a mean and 8.6–10.5 μm, respectively, and indicates that pollen size of 46.8±2.3 μm, a mean anl-D of 10.1 cultivated Hordeum sp. common for Group II dom- ±0.5 μm and a scabrate pattern. Avena–Triticum spp. inates. Larger grains with bigger anl-D belong to Group (n=25) had a mean pollen size of 50.9±2.3 μm, a mean III, which mostly comprises cultivated species e.g. anl-D of 12.8±0.7 μm and a verrucate pattern. The Avena sp. and Triticum sp. A cereal type pollen grain differentiation between Avena sp. and Triticum sp. is from Avena sp. is shown in Fig. 4. The identification of based on the surface ornamentation and the shape of the fossil cereal type pollen in a montane tropical grains (Beug, 1961; Andersen, 1979). These measure- environment is ambiguous, due to the representation ments can be compared with data from the eight most of a large number of species and their uniformity. Since common wild Poaceae species growing on the Horton the identification of cereal pollen grains from temperate Plains (n=80), where mean pollen size ranges between areas (Andersen, 1979) is based on the morphological 40.1±1.5 and 26.8±4.4 μm and mean anl-D range is features, i.e. size, anl-D and surface pattern, the 8.5±1.0 and 6.3±0.4 μm. In this work, the positive comparison with cereal pollen found in LPAZ 3–6 identification of cereal pollen types is based mainly on might involve errors. It is likely that Poaceae pollen comparison with the recent material. Highest emphasis from tropical areas display both a larger size and a is put on anl-D followed by the surface pattern and grain bigger anl-D. The surface pattern, however, should not size. In order to describe the size and anl-D variations in be different. These errors might have been reduced in the fossil sequence, pollen grains>37.5 μm with an anl- LPAZ 1–2, because temperate climatic condition D>8.6 μm were measured and grouped into eleven (11) prevailed in the Horton Plains during the late Pleisto- and five (5) classes, respectively (Figs. 2 and 3). Total cene and early Holocene. Poaceae pollen counted at each level constitutes the base for percentage calculations. 3.4. Mineral magnetic parameters The accuracy of the measurements within the 11 size classes in Fig. 2 is in the order of ±1 μm being Mineral magnetic parameters were analysed in order influenced mainly by the degree of physical preserva- to describe concentrations and grain size variations of tion. The accuracy of the measurements within the five the magnetic minerogenic component in the sequence. anl-D classes in Fig. 3 is in the order of ±0.5 μm. The These variations are believed to depend on erosion of zones in Figs. 2 and 3 follow the LPAZ. the valley sides close to the sampling site. Since the pollen diagram indicates the presence of humans in the 3.3. Comments on the identification of cereal-type near vicinity, the degree of erosion can be interpreted in pollen grains term of anthropogenic activities (Lagerås and Sandgren, 1994; Eriksson and Sandgren, 1998). In LPAZ 1, a number of the observed Poaceae pollen Sub-sampling for mineral magnetic analysis was grains fulfilled the size requirement (i.e. >46.8 μm) to performed at 1 cm intervals over the entire 6 m length of be classified as domesticated cereal types. These grains, the core. Magnetic susceptibility (χ) can be related to the however, did not have large enough anl-D, i.e. concentration and presence of ferrimagnetic minerals in >10.1 μm, and no clear surface patterns, and so they a sample. Anhysteretic remanent magnetization (ARM) were not considered as belonging to a cereal type. The reflects the concentration and presence of finer magnetic data resemble Group II, which comprises wild and grains. Saturation isothermal remanent magnetization cultivated species of the Hordeum group (Andersen, (SIRM) is mainly a measure of ferrimagnetic minerals, 1979). It is likely that the pollen grains identified in generally magnetite (Fe3O4). This parameter depends on LPAZ 1 are close to the following wild species: magnetic grain size. The S-ratio was calculated as Hordeum jubatum (anl-D=9.6 μm), Hordeum nodosum IRM− 0.1 T/SIRM (Thompson and Oldfield, 1986). This (anl-D=9.3 μm), Hordeum murinum (anl-D=9.3 μm), ratio reflects the magnetic mineralogy in a sample. The Glyceria fluitans (anl-D=9.6 μm), and Glyceria plicata measurements were carried out at the Department of (anl-D=9.7 μm). Quaternary Geology, Lund University, Sweden. The In LPAZ 2–6, the size of the Poaceae pollen is 46.6– magnetic susceptibility (χ) was measured using an air- 67.5 μm and the anl-D values are 8.6–16 μm, which coiled susceptibility bridge in a peak alternating field of resemble Group II (Hordeum group) and Group III 0.1 mT. ARM was induced in an AC demagnetizer with (Avena–Triticum group) as suggested by Andersen a peak alternating field of 100 mT and the remanence (1979). In the sequence analysed, the predominant was measured on a Molspin spinner magnetometer. mode class for pollen size and anl-D size is 46.6–49.5 SIRM was achieved in a high magnetic field of 1 T .Peahlk aaoegah,Pleciaooy aaoclg 4 20)468 (2006) 240 Palaeoecology Palaeoclimatology, Palaeogeography, / Premathilake R.

Fig. 2. The distribution of the grass pollen (Poaceae) size classes in percentages from the sequence analysed. Less than 37.5 μm size class includes wild grass species. The zones are referring to the local – pollen assemblages. + = occurrences. 496 473 474 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

Fig. 3. The distribution of the grass pollen annulus diameter (anl-D) size classes in percentages from the sequence analysed. Less than 8.5 μm anl-D size class includes wild grass species. + = occurrences. produced by a Redcliff pulse magnetic charger and using the Molspin anhysteric remanent magnetizer and induced remanence was measured on the spinner Molspin Minispin fluxgate magnetometer. After the magnetometer. ARM and SIRM were determined saturation procedure, the samples were placed in a weak

Fig. 4. Fossil pollen grain of Avena sp. i.e. oat (A) natural grass type (B and C). Scale bar: 10 μm. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 475 negative field of 0.1 T (isothermal remanent magneti- susceptibility (χ) determinations of the lowermost part zation; IRM− 0.1 T) using a Molspin pulse magnetic of the cores remains still somewhat questionable. It charger and the remanence was measured on the spinner might be related to very local lithological changes of the magnetometer. After completion of the magnetic core positions. Relatively high concentration of clastic analyses, the samples were dried and weighed to allow particles (e.g. sand) due to bedrock erosion or soil- calculation of the mass specific concentration para- forming processes at local vicinity of the lowermost part meters (Thompson and Oldfield, 1986). Correlations of the MS1 core positions may indicate somewhat high between the master core (MS) and complementary core χ values. Locally increased erosion also causes much (MS1) collected for chemical parameters were carried higher concentrations of clastic particles which relate to out by means of lithostratigraphy and magnetic an abrupt increase in χ determinations throughout of the susceptibility and the depths were transferred to the MS1 core. MS (Fig. 5). Note: In general, lithological compositions of both 3.5. Chemical parameters (δ13C, TOC, TN and C/N cores are similar. However, the correlation of the ratio)

There are two principal carbon fixation pathways (C3 and C4) during the process of photosynthesis in plant communities (Sage, 1999). C3 and C4 plants have 13 12 13 widely different C/ C ratios (δ C). Notably, C4 plants are mainly tropical arid-climate grasses while C3 plant community includes trees, shrubs and cool- climate grasses. Therefore, δ13C can be used as a palaeoecological indicator. Fifty seven (57) samples were analysed at 10 cm intervals from the comple- mentary core. The sample material was put in an Ag- container and 2 M HCl was added to remove carbonates. Sample weights varied between 2.1 and 91.6 mg. The samples were allowed to dry in 60 °C overnight and then compressed to small spheres. For stable isotope analysis, the samples were combusted at 1020 °C with a Carlo Erba NC2500 analyser connected, via a split interface to reduce the gas volume, to a Finnigan MAT Delta plus mass spec- trometer. From these measurements the reproducibility of δ13C (vs PDB) was calculated to be better than 0.15‰. Total organic carbon (TOC) and total nitrogen (TN) values were determined simultaneously when measuring the isotope ratios. TOC was also determined as an additional indicator to understand changes in, for example, the productivity of source plants, hydrology and agricultural practices when measuring the mineral magnetic parameters using an ELTRA CS 500 simultaneous carbon sulphur analyser. The relative Fig. 5. Correlations as solid lines between the graphs in mineral errors were <1% for both measurements. Because of magnetic susceptibilities (χ) for the master core (MS) and compli- mentary core (MS1) collected for measurement of chemical para- the heterogeneous character of the samples, the actual meters. Correlations were made according to increasing and values may be somewhat uncertain. The general trends, decreasing trends in the both graphs. Both curves show relatively however, are considered reliable. good correlations, however, there are some features that occur only in one of the sequence. This can be explained by local variations in the 3.6. Phytoliths analysis amount of clastic particles mixed with the peat. The high concentra- tions of χ in the lower part of MS1 may be explained by varying mineralogical composition of sediment. The high concentrations in the Phytoliths were separated from the materials by upper part may be attributed to the presence of bacteria. standard methods (Battarbee, 1986). Organic material 476 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

was oxidised in 40% H2O2 in room temperature and on a 4. Results and interpretation water bath. Clay particles were removed after 2 h of sedimentation in 100 ml beakers. Sand and coarser 4.1. Chronology particles were removed after 5 s of sedimentation. The residue was mounted in Nephrax. In order to calculate accumulation rates, linear interpolations were performed between adjacent cali- 3.7. Radiocarbon dating brated radiocarbon dates (Fig. 6). The uppermost date (Ua-16393 at 80 cm) was extrapolated to the ground Fourteen (14) bulk samples were collected from the surface at the 0 point. This age–depth model was chosen core and dated (Table 3) at the Ångström Laboratory, within the 1-sigma uncertainty ranges of the 14 dates. In Uppsala University, Sweden, by the AMS technique general, accumulation rate was low throughout the late (Possnert, 1990). In all measurements except the Pleistocene and Holocene. According to the age–depth uppermost sample the soluble fraction was dated. relationship, high accumulation rates appear between Coarse plant remains e.g. rootlets were removed 10.5–9 and 3.1–2.6 ka BP. In general, the rates vary manually. Ages are stated with ±σ and a normalization between 0.05 and 1.60 mm/yr and may be explained by of δ13C=−25‰ against PDB was carried out. The half- fluctuation of the ground water table and/or the input of life (T1/2) is 5570 years. In order to facilitate comparison sand in various proportions. Another possibility is an with previous publications, the ages stated in the text input of nutrients from human activities within the refer to calibrated years BP (ka BP). The ages are drainage area of the valley. calibrated according to et al. (1998) (<18,000 Peat started to accumulate on the valley bottom 14C yrs BP) and Kitagawa and van der Plicht (1998) ∼17.1 ka BP. The accumulation rate was low, resulting (>18,000 14C yrs BP). in only about one metre to be formed until 10.2 ka BP.

Table 3 Radiocarbon dates of bulk peat obtained from the sampled core Depth (cm) Lab. nr Fraction 14C age±1σ (yr BP) δ13C(‰) (PDB) Bulk samples Calibrated age±1σ (yrs BP) 80 Ua-16393 INS 1025±70 −24.2 Peat 1060–1020 (8.1%); 1000–900 (43.8%); 870–790 (16.3%); 125 Ua-17188 SOL 3235±60 −23.2 Peat 3560–3510 (13.6%); 3.480–3380 (54.6%) 160 Ua-16394 SOL 2675±60 −23.0 Sandy peat 2850–2820 (16.3%); 2815–2745 (51.9%) 230 Ua-16395 SOL 3120±70 −25.7 Sandy peat 3450–3420 (3.9%); 3410–3240 (62.0%); 3230–3210 (2.3%) 250 Ua-17189 SOL 3060±50 −22.6 Sandy peat 3350–3210 (68.2%) 290 Ua-17190 SOL 5080±65 −24.1 Sandy peat 5910–5740 (68.2%) 305 Ua-16396 SOL 7935±80 −18.1 Sandy peat 8990–8880 (21.4%); 8870–8820 (9.7%); 8810–8800 (1.5%); 8790–8640 (35.6%) 350 Ua-17191 SOL 8805±70 −19.2 Sandy peat 10,150–10,130 (1.2%); 10,120–10,080 (6.9%); 10,070–10,060 (1.6%); 10,040–10,020 (1.9%); 10,010–9990 (2.3%); 9930–9690 (54.6%) 420 Ua-16397 SOL 9125±80 −18.6 Peat 10,400–10,210 (68.2%) 460 Ua-17192 SOL 9195±75 −18.9 Peat 10,480–10,460 (3.1%); 10,430–10,240 (65.1%) 487 Ua-16398 SOL 12,970±115 −25.5 Peaty sand 16,000–15,200 (68.2%) 499 Ua-17193 SOL 12,950±100 −25.6 Peaty sand 15,950–15,150 (68.2%) 540 Ua-16399 SOL 15,045±140 −25.5 Silt + org. 18,350–17,650 (68.2%) 580 Ua-17194 SOL 18,410±185 −22.3 Silt + org. 22,300–21,450 (68.2%) Date intervals in the sequence analysed represent ±0.5 cm. Calibrated ages are according to Stuiver et al. (1998). Probability percentages are shown in brackets. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 477

Fig. 6. Age/depth relationship according to calibrated radiocarbon dates for the core studied. Sedimentation rates are calculated by linear interpolation between calibrated ages. Zone boundaries for pollen and spores, mineral magnetic and chemical parameters are included. The detailed lithology can be found in Fig. 2 and Table 1.

The period 9–3.6 ka BP exhibits the lowest accumula- particular site due to insufficient number of radiocarbon tion rate throughout the sequence, including a palyno- dates. logical hiatus between 5.4 and 3.6 ka BP i.e. a total lack of pollen and spore concentration associated with the 4.2. Reliability of the chronology sequence, but no indication for a lithological hiatus (cf. Sukumar et al., 1993 and Gasse, 2000). These can be Several problems are associated with the radiocarbon explained by a deterioration in climate, i.e. an aridifica- dates. Developing a more accurate chronological model tion causing oxidization of the organic material. The for a particular site can be limited by an insufficient hiatus inferred from palynological changes between 284 number of radiocarbon dates, plateau positions in the and 256 cm prohibited a single linear regression through calibration curve, location of the cores collected, sub- all 14 dates. Therefore, the dates above and below the sampling procedures and nature of the sample. Radio- hiatus were treated separately. Between 10.6 and 15 ka carbon bulk dates of peat are problematic, yielding in BP another hiatus might be present; however, this is some case ages that are too young (cf. Olsson, 1974; very difficult to explain due to lack of data. During Possnert, 1990) or too old (cf. Nilsson et al., 2001). In periods with high accumulation rates several dates yield the present study, the source of the dated carbon is not similar ages, despite a stratigraphic difference of up to known; therefore, it has to be taken into consideration one metre. Around 3 ka BP, some of the dates are that the quality of the AMS dating on bulk samples in inverted (i.e. 2675±60 and 3235±60). The AMS date some cases might be regarded as less reliable. The 2675±60 yr BP has been regarded while the date 3235 absence of carbonate in the bedrock of the Horton Plains ±60 yr BP has been disregarded only after considering implies that one of the major sources (e.g. older carbon possible options when the construction of age–depth contamination) of error can be neglected. But, in one model was adopted (Fig. 6). The AMS date 12,970 case, the date appears to be too old (i.e. 3235±60) and ±115 yr BP at 487 cm depth is considered to fit within this may contain somewhat older detrital organic the age–depth model than the younger date, 12,950 material derived from a lower stratigraphic level. ±100 yr BP at 499 cm depth. This can be explained by Thus, the bulk dates obtained from the Horton Plains the contamination of younger organic materials, such as sequence may show somewhat younger ages due to the remaining root fragments. Note that some gaps may be following possible sources of error: (1) downward present in developing a chronological control for this penetrating humic acids, (2) ground water fluctuations, 478 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

(3) remaining rootlets/roots i.e. bioturbation (cf. Olsson, At 527 cm depth (corresponding to ∼17.5 ka BP) 1974; Possnert, 1990). The effect of the first two error domesticated Avena sp. (oat) pollen appears (Fig. 8A). sources (i.e. 1 and 2), however, is considered diminished This continues sporadically upwards within LPAZ 2. to a much lower level in most cases, partly due to the Pollen grains of Hordeum sp. (barley) also appear later high degree of compactions throughout the sequence. in the LPAZ 2. The sharp fluctuations (Fig. 7A) in the Thus, it is obvious that bioturbation (i.e. 3) may have curves of UMRF taxa indicate a succession in the been the most likely source of contamination. Therefore, vegetation. The taxa included in the ecological groups it is believed that possible contamination was mainly in (Figs. 8A and B) “G”, “C”, “F”, “secondary forest/shady the remaining rootlets. Because the rootlets were (O)” and “disturbed ground (P)” may indicate the manually removed and the fact that vegetation is continuing disturbance as do the occurrences of open dominated by Poaceae and Cyperaceae with short root land shrub taxa, e.g. Juniperus sp. and Rubus spp., systems in the vicinity of the coring points it is assumed Sarcococca zeylanica, and Rhamnus arnottianus. In the that this contamination is also negligible. The radiocar- middle part of the zone, minor occurrences of pollen bon dates were calibrated without correction (Table 3). grains from Buchanania (axillaris?), an edible plant, and the cultivated shrub Spiraea sp. occur. The high 4.3. Zonation of pollen data and interpretation percentages of microscopic charcoal particles may originate from forest burning. More than 200 pollen taxa have been identified most of them occurring at low frequencies (Figs. 7A and B). 4.3.3. LPAZ 3 (364–284 cm; 9.9–5.4 ka BP) Unknown pollen types make up 1–15% of the count at There is a high representation of rain forest and any given level analysed. Pollen preservation is aquatic taxa (Premathilake and Risberg, 2003). A generally good throughout the sequence except between domesticated pollen type of Poaceae appears sporadi- 284 and 256 cm depth at which a hiatus is suspected. cally while minor sporadic re-occurrences of S. The key features of the local pollen assemblage zones zeylanica, Spiraea sp., Rubus spp. and Indigofera sp. (LPAZ) are discussed in detail below. Their age are noted (Figs. 7A and 8A). Sporadic occurrences of intervals are deduced from the age–depth model (Fig. “O” and “disturbed ground/damp” (H) taxa, e.g. Rubus 6). An interpretation of possible climatic conditions for spp., Rubia cordifolia and Cardamine african can be each LPAZ can be found in Premathilake and Risberg identified. The few pollen finds of plant species (2003). representing the “cultivated” groups and/or “waste ground” (A/D), e.g. Malva parviflora, occur. Relatively 4.3.1. LPAZ 1 (600–546 cm; >24–18.5 ka BP) high values of microscopic charcoal particles (e.g. This zone is characterised by high values of average 50%) occur. Indications of human disturbances Chenopodium spp. and Gomphrena sp. (Fig. 8A) decrease upwards in the zone corresponding to an indicating a xerophytic vegetation community. Pollen increase in pollen from upper montane rain forest trees values in the following groups: “pastures” (G), e.g. up to 300 cm (Fig. 7A), indicating a recovery of the Achyranthes aspera and Cerastium sp., “field and vegetation and a rapid regeneration of the rain forest. disturbed ground” (C), e.g. Plantago spp., together Uppermost part of the zone, UMRF pollen values and with “weeds (F)”, e.g. Caryophyllaceae and Labiatae, the organic carbon content rapidly decrease. Appearance provide possible evidence of a disturbance that main- of cereal pollen grains also gradually declined upward in tained some pioneer vegetation (Fig. 8A). the zone.

4.3.2. LPAZ 2 (546–364 cm; 18.5–9.9 ka BP) 4.3.4. LPAZ 4 (284–256 cm; 5.4–3.6 ka BP) The pollen concentration values of this zone are This zone is characterised by extremely low pollen fairly high in relation to LPAZ 1. UMRF arboreal taxa, concentration values. Only a few (1–5) pollen grains of e.g. Syzygium sp. Elaeocarpus spp., Symplocos spp., tree taxa, e.g. Ilex spp., Elaeocarpus spp., and A. Ilex spp., Adinandra lasiopetala and Meliosema simpli- lasiopetala, together with pollen from shrubs, e.g. cifolia, Semicarpus sp., Rhododendron arboreum (Fig. Strobilanthus spp., Phylanthus spp. and herbs, Poaceae, 7A), start to increase and appearance of shrubs e.g., Asteraceae and Cyperaceae, were observed (Figs. 7A, B Juniperus sp. and Cupressus sp. and the herb taxa and 8B). The surface characters of the exine on most of Chenopodium sp. and Gomphrena sp. (Fig. 7B) values the pollen grains were not distinct (e.g. very faint decrease upwards suggesting an expansion of tropical/ colour), and also ruptured and bent. The lack of pollen equatorial forest as a result of climatic amelioration. grains in this zone can possibly be ascribed to aridity, .Peahlk aaoegah,Pleciaooy aaoclg 4 20)468 (2006) 240 Palaeoecology Palaeoclimatology, Palaeogeography, / Premathilake R. – 496

Fig. 7. A. Simplified percentage pollen diagram displaying selected upper montane rain forest (UMRF) tree and shrub pollen taxa. + = occurrences. B. Simplified percentage pollen diagram displaying selected upper montane rain forest (UMRF) herbs and aquatics/semi-aquatics pollen taxa. Spore flora (Pteridophytes and Moss), algal cysts, total concentration and curve for organic carbon are included. + = occurrences. 479 480 .Peahlk aaoegah,Pleciaooy aaoclg 4 20)468 (2006) 240 Palaeoecology Palaeoclimatology, Palaeogeography, / Premathilake R. – 496

Fig. 7 (continued). .Peahlk aaoegah,Pleciaooy aaoclg 4 20)468 (2006) 240 Palaeoecology Palaeoclimatology, Palaeogeography, / Premathilake R. – 496 Fig. 8. A. Human impact diagram for the sequence analysed. Individual curves for ecological groups (i.e. A = Cultivated lands, B = Pastures, C = Field and disturbed grounds, D = Waste ground, E = Open ground/patanas, F = Weeds). Palynological richness (PL. richness) is shown. + = occurrences. B. Human impact diagram (continue). Individual curves for G = Patanas, H = Disturbed/damp ground, I = Open ground, J = Shady ground, K = Shady/damp/waste ground, L = Damp/shady ground, M = Waste/patans, N = Secondary forest/montane grasslands, O = Secondary forest/shady ground, P = Disturbed ground and indifferent taxa. Microscopic charcoal particles are shown. Total sum of the ecological groups C, D, E, N, O and P are considered as representative of major ruderal communities, while sum of the other groups F, G, H, I, J, K, L, M and P are considered as representing minor ruderal communities. + = occurrences. 481 482 .Peahlk aaoegah,Pleciaooy aaoclg 4 20)468 (2006) 240 Palaeoecology Palaeoclimatology, Palaeogeography, / Premathilake R.

Fig. 8 (continued). – 496 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 483 causing unfavourable edaphic conditions for preserva- (around 35%). Observations of pollen from Berberis sp. tion. The organic carbon content is low due to more arid and Nandina domestica also support this interpretation. conditions. There is no conclusive pollen evidence for Sarcococca zellanica, Rubus spp., and Indigofera sp. human activity; however, microscopic charcoal particles give evidence of forest clearance and grazing. and grass phytoliths appear. 4.4. Mineral magnetic analyses 4.3.5. LPAZ 5 (256–108 cm; 3.6–2 ka BP) The pollen values UMRF, e.g. Syzygium spp., According to the variation in magnetic concentra- Elaeocarpus spp., and Symplocos spp., gradually tions and mineralogy reflected by χ, ARM and SIRM increase upwards suggesting a climatic amelioration. and S-ratio, the core can be divided into three main In the lowermost part of the zone, cultivated pollen magnetic units (Fig. 9). The origin of the clastic types and taxa representative of the ecological group component is mainly detrital, indicated by low ARM/ “C” occur. Similarly, pollen taxa included in the SIRM ratios, i.e. 0.025, for the entire core, except the ecological groups, e.g. “O” and “P” are also indicative uppermost 40 cm and some samples around 550 cm in of human induced environment. The sharp fluctuations depth. In general, magnetic susceptibility is extremely in the curve of UMRF trees indicate changes in the low in the peat and sediment sequence due to organic vegetation. and sandy composition. The low SIRM/χ ratios throughout the core imply that coarse multidomain 4.3.6. LPAZ 6 (108–14 cm, 2–0 ka BP) magnetic grains dominate in the magnetic stratigraphy. The pollen values for UMRF trees in the lower part of the zone are fairly low in relation to LPAZ 5. This 4.4.1. Unit 1 (600–529 cm; 24–17.5 ka BP) relative reduction of the UMRF (mainly Syzygium spp., The lowermost unit, which consists of sandy clay and Elaeocarpus spp., and Euonyumus revolutus) can be silt with organic material is characterised by low interpreted as a result of forest destruction. Agricultural magnetic concentrations. The lowest S-ratios, −0.8 activity is indicated by high occurrences of pollen from around 585 cm in depth, are indicative of magnetically Triticum sp. (6%) and microscopic charcoal particles soft minerals, e.g. magnetite. The detrital origin of

Fig. 9. Mineral magnetic concentrations and ratios for the sequence analysed. In magnetic unit 2, the periods with low magnetic concentrations are marked with a light grey shade, whereas periods with high values are white. Magnetic unit 2a corresponds to the incipient cereal plant management (LPAZ 2 in Fig. 8A) and unit 2b to the cultivation phase (LPAZ 3 in Fig. 8A). 484 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 magnetite may have derived from bedrock erosion or minerals (e.g. bacterial magnetite) might be produced soil-forming processes (Snowball, 1996; Walden et al., from bacterial activity in the very uppermost portion. 1999). Sand and clay accumulated during LGM are The increase in ARM may have been caused by sub- characterised by low S-ratios. The S-ratios increase to 0 recent cereal cultivation (cf. LPAZ 6). at the boundary between magnetic unit 1 and 2a, suggesting an increase in magnetically hard minerals 4.5. Chemical parameters (δ13C, TOC, TN and C/N (e.g. hematite). The magnetic unit 1 corresponds to ratio) and phytoliths LPAZ 1, defined as representing xerophytic vegetation associated with a semi-arid climate. The sequence can be divided into three zones according to the variations in δ13C, TOC, TN% and C/ 4.4.2. Unit 2 (529–87 cm; 17.5–1.1 ka BP) N ratios (Fig. 10). Phytolith assemblage zones follow This unit consists of a mixture of peat and sand, and the zones in Fig. 10. has overall higher magnetic concentrations than unit 1. High S-ratios indicate the presence of magnetically hard 4.5.1. Zone 1 (600–475 cm; >24–13.6 ka BP) minerals in the sequence. The unit is divided into three The δ13C values become lighter throughout the zone, sub-units (2a–2c, from older to younger) according to averaging around −26‰ PDB, which indicates the the variations in magnetic concentrations. Periods with domination of C3 plants (e.g. grasses). Low TOC values low, or relatively low, magnetic concentrations are (average 2%) reflect vegetation consisting of plants marked with a grey shade in the diagram, whereas producing low biomass and amount of detrital matter periods with high values, or relatively high values, are (e.g. silt and sand) or low preservation (cf. LPAZ 1). In white. the very uppermost part of the zone (<15 ka BP), few Higher and lower magnetic concentrations alternate fan-shaped grass opal phytoliths occur, consisting of over relatively long periods in sub-unit 2a compared to Oryza sp. (i.e. swamp grass). Consistently low values in sub-unit 2b. In sub-unit 2b, several prominent high TN may reflect poor nutrient conditions in the peaks can be identified, sandwiched in between periods environment. Increased C/N ratios from 10 to 22 with lower concentrations. In general, magnetic con- indicate that the local environment around the sampling centrations are stable in sub-unit 2c, except between 175 site was subject to slope processes, bringing reworked and 162 cm. terrestrial material into the site (cf. Aucour et al., 1999; In sub-unit 2a, the initial increase in χ, ARM and Lücke et al., 2003). SIRM is probably the result of an input of magnetic particles into the valley, possibly caused by erosion due 4.5.2. Zone 2 (475–300 cm; 13.6–8 ka BP) to reduced vegetation cover. This input was favoured by The δ13C values range between −15‰ and −26‰ the presence of an open landscape as indicated by the PDB, which indicates the presence of a mixed C3/C4 pollen data, which records major forest clearance. In the plant community in the vicinity of the sampling site. upper part of the zone, magnetic concentrations However, δ13C values become heavier at the middle decrease, probably as a result of more dense vegetation part of the zone (i.e. −15‰ PDB), which indicates cover binding the soil. In sub-unit 2b, the variations in that C4 grasses dominated as the major carbon source SIRM in particular, may be attributed to the shifting (e.g. swamp grasses) and aquatics were dominated e.g. pattern of cereal cultivation interpreted from the pollen Limnophyton sp., Rhynchoglossum sp. (cf. LPAZ 2). data. The uniform pattern of concentration values in This interpretation is supported by the frequent sub-unit 2c may be due to the lack of cereal cultivation occurrence of fan-shaped grass opals phytolith from activities (cf. LPAZ 5). Oryza sp. around 10 ka BP. The combined fan-shaped and papillae phytoliths evidence are indicative of 4.4.3. Unit 3 (87–0 cm; 1.1–0 ka BP) cereals. TOC values increase to about 15% then drop The magnetic concentrations are again higher than in to 5% and rise again to 20% maximum around 9 ka sub-unit 2c. The values in S-ratios decrease from 0 to BP. They very abruptly drop again to less than 1% −0.6, reflecting a change in magnetic mineralogy. χ, around 8 ka BP. The increasing trend in TN implies ARM and SIRM increase gradually indicating higher relatively nutrient rich environment. The almost concentrations of magnetic minerals. identical patterns of TOC and TN are indicative of The rise in ARM/SIRM ratios indicates an increase in climatic amelioration as indicated by pollen records. the portion of fine-grained, secondary, soil-derived C/N ratios increase ranging from 22 to 30, and magnetic minerals (Walden et al., 1999). Secondary gradually decrease to 20%. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 485

Fig. 10. The graphs for chemical parameters from the sequence sampled close to the Master Core. Values of TN (%), TOC (%), δ13C (vs PDB) and C/ N ratio are displayed. Zone 1 corresponds to the human induced environment and zone 2 to the systematic cultivation phase. Numbers of phytoliths (fan-shaped and papillae) found from the master sequence are included.

4.5.3. Zone 3 (300–0 cm; 8–0kaBP) studies have been undertaken. Therefore, a general The δ13C values reach c. −25‰ PDB suggesting a introduction to modern pollen–vegetation relationships dominance of C3 plants. There is a gradual increase in can be determined from the analysis of the very TOC and TN, reaching to 20% and 1.5% respectively, uppermost fossil pollen assemblages and distinct with the climatic amelioration. Stable C/N ratios remain contemporary vegetation communities of the Horton constant at 21 for the entire zone, apart from the Plains. A total of 54 woody species belonging to uppermost 80 cm. Very few papillae and fan-shaped 31 families have been encountered, of which 50% are phytoliths of cereals occur only in the uppermost part of endemic to Sri Lanka. Lauraceae, Symplocaceae, the zone. Myrtaceae, Clusiaceae are the most dominant families and Syzygium spp., Symplocos spp., Calophyllum spp. 5. Discussion and Cinnamomum sp. the most dominant genera. Small and large tree trunks and branches carry a thick coast of 5.1. Catchments of pollen and charcoal mosses, such as Frullania sp., Bazzania sp. and Sphagnum spp. large tree Cyathea cinuata is one The relationship between the spatial distribution of example of a large tree fern species and Lycopodium modern pollen and its representative vegetation (i.e. spp. and Psilotum sp are also present. Rhododendron pollen influx values) is very critical in palaeoecological arboretum ssp. zeylanicum is one of the most charac- research. Pollen influx values should be established in teristic species within the upper montane rain forest. various environments, from a closed forest to an open Undergrowth of the forest often consists of dense valley system of the Horton Plains, in order to evaluate masses of Strobilanthus spp. In addition, Coleus spp., the fossil pollen spectra. On this line, however, no Osbeckia spp. and Impatiens spp. are common. The 486 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 grasslands are dominated by Chrysopogon zeylanicum, show variations in total annual rainfall over South Asia Andropogon lividus Arundinella villosa and Garnotia during the late Pleistocene/Holocene (Sirocko et al., tectorum. Pollen taxa distinguished in the arboreal fossil 1993; Gupta et al., 2003; Staubwasser et al., 2003). It pollen spectra were probably from the species growing is obvious that impact of annual rainfall changes in upper montane rain forest and low montane forest in described by investigations in cave deposits, archaeo- Sri Lanka. However, very few pollen taxa from the low logical deposits and organic accumulations in the montane arboreal forest (<900 m) were distinguished. It interior of the island of Sri Lanka and in sand dune seems that fossil pollen spectra represent much more areas along the coast is coherent with twists and turns regional view of the composite vegetation of the Horton of early human settlements and civilization (cf. Plains. Burning and cutting have been the most Deraniyagala, 1992; Premathilake, 2003). However, important disturbances maintaining the successional this section shows the converging evidence for climate cycles in the upper montane rain forest and grasslands change during the late Quaternary and briefly outlines (Amarasinghe and Pemadasa, 1983; Deraniyagala, the environmental history of the Horton Plains, Central 1992). Climate change has also drawn attention to the Sri Lanka. The rainfall characteristics in different parts importance of those two types of vegetation communi- of Sri Lanka are fairly different (Suppiah, 1987). The ties. By burning of the communities, charcoal particles author showed a bimodal pattern in the annual principal preserved in the peat and sediment sequence may permit rainfall cycle. Except for the east coast region, all other evaluation of past fire activity. parts of Sri Lanka are coincident with the northward and southward migration of the Intertropical Conver- 5.2. Comments on swamp history (mire history) gence Zone (ITCZ). During the South West Monsoon, the ITCZ-induced rainfall prevails from April to June Pollen and spore spectra constructed from the and from September to November and westerly winds sequence were used to reconstruct effective changes in lead to the heavy rainfall on the western hill slopes of moisture, precipitation and to trace human activities. It the Island from June and September i.e. more than is suggested that the main causes behind the vegetation 2500 mm/yr (Zubair, 1999). The winter monsoon changes and human activities should be highlighted rainfall of the island is from October to the December separately. For this purpose, ecological groups including i.e. less than 1750 mm/yr. It is estimated that the pollen indicator species approach were used. However, western hill slopes of the island (i.e. 2500 m a.s.l.) have some ecological grouping contains species, which are the highest rainfall reaching 6000 mm/yr from May to growing on swamps (i.e. mire). Therefore, it is December (cf. Zubair, 2003, 2004). South west necessary to take account of the site environment, i.e. monsoon rains (SWMR) have very high impact upon local swamp history, from the pollen and spore analysis. the Horton Plains, Central Sri Lanka. The connection Pollen and spore spectra and siliceous microfossil between the SWMR and the environmental setting (e.g. analysis suggested eight periods of distinctive on-site local and regional vegetations types) has been recorded vegetation changes during the 24 ka years. 24–18.5 ka: in Table 4. The changes in vegetation types are, partly, aerial condition and grasses and herb-rich environment, the results of large variations in the SWMR in response 18.5–12 ka: semi-dry mire, 12–10.2 ka: shallow water to seasonal shifts in the position of the ITCZ and its body, 10.2–5.4 ka: mire, 5.4–3.6 ka dry mire, 3.6–2 ka: associated belt of convective activity (cf. Premathilake wet mire, 2–0.5 ka: mire and 0.5–0 ka: semi-dry mire and Risberg, 2003). In contrast, the maximum Upper (Premathilake and Risberg, 2003). Montane Rain Forest (UMRF) condition (i.e. under the hyper humid; Table 4) is reported during the heavy 5.3. Climate and palaeoenvironmental setting for regional rains (strong SWMR), when the ITCZ human activity migrates to its most northerly position, almost directly over the Horton Plains, central Sri Lanka, which is High-resolution palaeoclimatic records from differ- supported by comparison of the several Holocene ent archive (e.g. peat, lake sediment and deep-sea) records (cf. Neff et al., 2001; Gupta et al., 2003). The

Notes to Table 4: Vertical arrows indicate gradual changes, while solid horizontal lines denote more abrupt changes. Dotted lines indicate gradual transitions. UMRF = Upper Montane rain Forest. G/PG = Glacial/Postglacial. Capital letters A–G refer to marked peaks in the precipitation (i.e. high rainfall). Data in the column showing regional climate events are compiled from references mentioned in the text covering mainly the Arabian Sea and the Indian sub-continent. In the same column arrow heads indicate short-term intensified monsoonal climatic regime i.e. SWM (South West Monsoon), while the bars indicate more consistent patterns. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 487

Table 4 Compilation of the data obtained from the different parameters and interpretations of palaeoenvironment and climate 488 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

South West Monsoon downturn, particularly mid- (i.e. lack of rain forest) in the Horton Plains as well as Holocene aridity (i.e. severe decrease in mean annual in most parts of the island during the LGM rainfall) when the ITCZ seems to have gradually (Deraniyagala, 1992). It is thought that they settled reduced its presence in the region, could be special in small camps in different environments, ranging time periods in the continental tropic that have been from the damp and cold high plains (e.g. Horton previously underestimated. In addition, El Nino- Plains) to the lowlands (e.g. Mannar). Human bone southern oscillation (ENSO) phenomenon is now material found from several prehistoric cave sites of recognized as a primary mode of seasonal climatic the island, dated between 37 and 6.5 ka BP, has been variability, particularly in the tropics (Ropelewski and subject to detailed physical anthropological studies Halpert, 1987, 1989; Kane, 1998; Zubair, 2002, 2003). (Kennedy et al., 1987; Kennedy, 2000; Hawkey, Rasmusson and Carpenter (1983) and Kane (1998) 2000). These studies show that the prehistoric have studied the relationship between the ENSO and population constitutes close genetic affinities to the rainfall over India and Sri Lanka. The authors have recent Veddha aboriginal population in Sri Lanka. The shown that the ENSO warm conditions are often modern life styles between Veddha and hunter–forage associated with decreased of annual SWMR (i.e. populations could not have been too different in Sri increased droughts conditions) in Sri Lanka, but inter- Lanka and among indigenous populations in India, monsoon season rainfall shows a strong opposite Andaman Island and Malaysia (Deraniyagala, 1992; tendency (e.g. floods). Thus, interpretations of the Kennedy, 1993; Deraniyagala, 2001). Osteological proxy data found from this study reveal that the mean analyses indicate that the prehistoric human diet annual rainfall of Sri Lanka is diminished by variations comprised of a mixture of wide range of food-plant of the mean ITCZ position, with time and the ENSO and animals (Deraniyagala, 2001). phenomenon. It has been postulated that the hunter–foragers fed themselves using food-plants and animals from the 5.4. Late Pleistocene hunting and gathering island as well as from the Indian sub-continent (e.g. Deraniyagala, 1992). This includes canarium nuts, wild Long-term land-use history can be reconstructed breadfruit, bananas, grasses and fruits in the hilly areas using radiocarbon-dated pollen diagrams, interpreted and hunting animals on the open slopes and plains near in terms of vegetation changes (e.g. Vuorela, 1986; watering points. This interpretation is based on Delcourt et al., 1986; Kealhofer and Penny, 1998; stratigraphic cultural material studied from southern Poska, 2001). The approach of using indicator species Asia, i.e. India, Pakistan and Sri Lanka (Jarrige, 1985; in the pollen stratigraphy conduces to the understand- Kajale, 1990a; Kajale, 1996a,b). ing of past human economy and life styles. Therefore, Current results from the Horton Plains support this the main features of human life style in the Horton interpretation since herding (Bos indicus?) and the Plains are discussed here. Before 17.5 ka BP, the incipient management of cereals (i.e. oat and barley) landscape of the area was probably open with a occur to some extent from 17,500 cal yrs BP onwards. prevailing semi-arid climate as indicated by pollen (cf. This also coincides with a semi-humid event (17.6– LPAZ 1), δ13C and TOC (cf. Zone 1) records during 16 ka BP), suggesting favourable climatic conditions the Last Glacial Maximum (LGM). The occurrences (i.e. warmer/humid), influencing the plant management of the wild plant families, Plantaginaceae, Chenopo- system (cf. Singh, 1990; Premathilake and Risberg, diaceae, Polygonaceae, Amaranthaceae and Caryo- 2003). From 17.5 ka BP through the late glacial, the phyllaceae indicate dwelling patterns of hunters and cereal management was periodic in character. Rela- foragers (Kajale, 1996a). The pollen taxa of Cerastium tively cool and dry conditions followed in the early sp., Desmodium spp. and Achyranthes spp. appear, Holocene (Premathilake and Risberg, 2003). Several probably being favoured by grazing animals. Pollen other indications support this emergence of the plant from Plantago sp. indicates human induced environ- management system, which includes pronounced ment. Stands of wild relatives of Hordeum sp. and fluctuations in the arboreal and non-arboreal taxa (cf. Avena sp. were also present in the grasslands. The LPAZ 2), pastures, different types of disturbed fields, vegetation compositions deduced from the pollen data e.g. patanas and increase in mineral magnetic concen- and soil forming processes detected from S-ratios tration (e.g. χ, and SIRM). Frequent burning and forest suggest that the population consisted of mobile groups clearances (i.e. slash-and-burn) by the prehistoric of hunter–foragers. It is possible that this life form people may have contributed to the expansion of could have dominated due to the low carrying capacity patanas (Perera, 1967; Muller-Dombois and Perera, R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 489

1971). The high representation of pollen taxa within headwaters of streams, forested hills and grassy the group “patanas (G)” might be the result of a close vegetation (cf. Redman, 1988; Singh, 1991). The Horton relationship between prehistoric humans and the Plains, and the surrounding area, display all these basic prevailing vegetation. factors implying a particularly diverse landscape with a In the middle part of LPAZ 2, the presence of an wide range of resources. These highland areas are also edible plant (Buchanania axillaris) and a cultivated suitable for summer grazing and the streams could have shrub (Spiraea sp.), are further pieces of evidence of the been dammed with soil barriers in order to regulate the management system. The high percentages of micro- flow of water permitting incipient cereal plant manage- scopic charcoal particles may originate from domestic ment (Deraniyagala, 1992). fires in swamp (mire) burning, i.e. slash-and-burn It is believed that the hunter–foragers had some activity. This scenario was probable during the early experience with plant management since 17.5 ka BP, stages of the mobile life style. Therefore, it is logical to while they were still mainly engaged with hunting and assume that the hunter–foragers had acquired the foraging. The appearance of pollen from cereal type techniques and practices to convert the wild plants plants indicates that the mobile life style lost importance into domestic ones from ∼17.5 ka BP onwards. in the Horton Plains. This interpretation is supported by However, it does not indicate the start of agricultural arguments that the grasslands of the Horton Plains are a activities but, the beginning of plant management. result of major forest clearing during prehistoric time (de The process of domestication has played a central Rosayro, 1946a,b,c; Senaratne, 1956; Perera, 1967; role in the development of prehistoric social environ- Muller-Dombois and Perera, 1971; Deraniyagala, 1972; ment and the material culture. Archaeological and Amarasinghe and Pemadasa, 1983; Deraniyagala, historical data obtained from the Horton Plains suggest 1992). The forest clearance can be related to patch that the area was subject to fire, forest clearance and production for hunting or incipient cereal management grazing (Deraniyagala, 1992). Periodic fires seem to and subsequent cultivation. Some areas of the grass- have occurred since at least 28 ka BP in the surrounding lands were selected for campsites, contributing to the upland and montane grasslands, and in all likelihood, anthropogenic landscape of the Horton Plains (Dera- there was burning of the forests on the Horton Plains niyagala, 1992). Kennedy and Zahorsky (1995) have (Deraniyagala, 1992). These may have been caused by claimed, based on the analysis of human bone material humans during small-scale seasonal movements into the from several prehistoric sites located in lowlands that a montane region. Close to the site studied, there is relatively isolated population lived in restricted areas evidence of prehistoric humans in the form of undated from the late Pleistocene to the modern time (Deraniya- microlithic stone artefacts, which were assigned to the gala, 2001). Mesolithic culture in Sri Lanka (Deraniyagala, 1992). Palaeoecological evidence for the origin of agricul- The domestic cereal types may have derived from wild ture has been found from the Indian sub-continent as progenitors of the potentially domesticable Hordeum sp. well as in Southeast Asia (i.e. Pakistan, India, Papua and Avena sp. plants. These wild progenitors were New Guinea, Indonesia, Thailand and Taiwan, at a growing already before the LGM within the vegetation somewhat later stage as compared to this study (Allchin of the Horton Plains. At this early stage, such plants may and Allchin, 1989; Hather, 1994; Chakrabarti, 2000). It have been growing as weeds among other grasses. This is also proposed, but not established, that the mountain is supported by the presence of wild grass species region of Sri Lanka could have constituted another belonging to Aveneae (e.g. A. sterilis, A. fatua, A. barata “hearth” for the domestication of plants as suggested by and Helictotrichon sp.) and Triticeae tribes (e.g. Deraniyagala (1992). This would be the case if Brachypodium sp.) that are represented in the present prehistoric humans in the Horton Plains cultivated vegetation of the Horton Plains and its surroundings wild grains first, and subsequently, fully domesticated (Clayton et al., 1994). It is implied that the area could them during the end of late Pleistocene and beginning of have served as a primary source of genetic variability the early Holocene. The combined pollen, opal (wild species) for the domesticated Hordeum and phytoliths, mineral magnetic composition, and palaeo- Avena–Triticum-types, pollen of which appear later in climate evidence (shown by TOC, TN and δ13C) the core sequence. Several of the basic requirements for strongly implicates very abrupt climate change as a the origin of the domestication processes were available key factor leading to the agricultural community in the Horton Plains. These comprise e.g. a mild, dry between 13.6 and 9 ka BP (zone 2 in Fig. 10). In climate, burning of forest, presence of humans with a general, the palynological richness also increased during shifting of setting mode, wild progenitors, rolling plains, this period. This includes the occurrences of several 490 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 pollen taxa representative of “B”, “G” and “F” (e.g. SIRM (cf. sub-unit 2b, Fig. 9). This is supported by Achyranthes sp., Indigofera sp., Heracleum sp. and several other lines of evidence, e.g. the reduction in Veronica sp.). The crop cultivation probably took place pollen values of livestock farming, cultivated and/or close to the sampling site. These activities seem to have waste ground group, field and disturbed ground, caused an abrupt decrease in Chenopodium spp. and regeneration of rain forest. The decrease in cultivation increase in Pteridaceae (Pteridophytes). The initiation of can be explained by a gradual increase in aridity, which systematic cereal cultivation took place in the middle started ∼8.6 ka BP and lasted until the end of middle part of the climatically humid phase between 13.6 and Holocene i.e. 3.6 ka BP. 12 ka BP, i.e. the end of the late Pleistocene. To some The farming activity recorded from the Horton Plains extent, this also marks a change from mobile to and its age can be compared with corresponding results sedentary life style. obtained in South and Southeast Asia. Neolithic farming The results indicate that the incipient management of practices, including cultivation and cattle breeding, oat and barley continued for a period of 4.5 ka BP appear to have had considerable antiquity in Pakistan at (17.5–13 ka BP). During this period, prehistoric humans Mehrgarh and in India at Rajasthan, Vindhyan Hills, developed advanced techniques and practices for plant Chopani-Mando, Lahuradewa and Koldihwa between domestication. In the Horton Plains, environmental 10and7kaBP(Meadow, 1981; Jarrige, 1981; factors such as climate and soil forming processes Costantini, 1981; Allchin and Allchin, 1989; Chakra- coincide with the incipient and succeeding land-use. It is barti, 2000; Gupta, 2004; Singh, 2005). Sharma et al. also suggested that the time lag (i.e. 4.5 ka BP) is (2004) and Singh (2005) have described an early stage reasonably long to consider that the agriculture origins of farming at around 15 ka BP based on the occurrence followed a macro-evolutionary experiment as suggested of Cerealea and other cultural pollen and microcharcoal by Richerson and Boyd (2000). records found in Northern India. The fossil pollen evidence from the Papua New Guinea highlands 5.5. Farming activity indicates that complex agricultural systems were developed since 9 ka BP (Denham et al., 2003). In 5.5.1. Early Holocene farming addition, pollen and charcoal records achieved from a The origin of farming in different parts of the world is sedimentary sequence in the highlands of Indonesia complex (Normile, 1997; Toyama, 2001; Hillman et al., indicated anthropogenic disturbances ∼13 ka BP, which 2001). In the Jiangxi Province of China, South-Eastern may correlate with the spread of agricultural innovation Asia, rice plant (Oryza sativa) phytoliths have been in lowland Papua New Guinea (Golson, 1989; Hope and recovered from pottery at an excavation covering the Tulip, 1994). period 14–9kaBP(Toyama, 2001). Palaeobotanical data from China suggest that rice cultivation was the 5.5.2. Middle Holocene farming main subsistence for human culture. At the Abu Hureyra During the middle Holocene (8–3.6 ka BP), the site on the Euphrates in south-western Asia, the yielded overall percent decrease in cereal-type pollen (LPAZ 4), remains of cultivated rye (e.g. Secale cereale) indicate indicates a significant reduction of cultivation, which is systematic cereal cultivation, 13 ka BP (Hillman et al., also in good agreement with the very abrupt decrease in 2001), which is synchronous with the new data achieved the UMRF components implying unfavourable envi- from the Horton Plains. It is suggested that the ronmental conditions, e.g. persistent multi-century prehistoric cereal cultivation, which occurred in the aridity. Palynological hiatus appears to be coherent south Asian region, confirms the diversity of the with the most prominent climatic change (i.e. severely agriculture at the same time. increased aridity) between 5.4 and 3.6 ka BP. Thus, During the early Holocene, the main cultivated these data indicate a very significant reduction in South crops in the Horton Plains were Hordeum sp. and West Monsoonal rains (cf. Staubwasser et al., 2003). It Avena sp. The cultivation occurred continuously is suggested that repeated severe drought cycles may between 10 and 9.5 ka BP, associated with more have initiated depletion of habitats of agrarian commu- humid conditions around 8.7 ka BP. After this age, nity of the Horton Plains, probably re-locating to the towards the end of the early Holocene, the state of the lowland areas in Sri Lanka. This severe drought period cultivation was sporadic in the vicinity of the sampling is probably the result of climatic conditions that site, which reveals human influence occurred in limited prevented the ITCZ and its associated rainfall from extent. This may indicate temporary habitations and/or penetrating far north. In South and West Asia, shifting patterns of cultivation, as shown by variation in intensified aridity around 4 ka BP badly affected the R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 491

Old World societies e.g. Mesopotamia and India (Dalfes livestock farming. Thereafter, the farming activity was et al., 1997; Enzel et al., 1999). It is very obvious that completely abandoned for ∼800 years as indicated by the drought event at 4.2 ka BP traced from planktonic the lack of cereal type pollen. Uniform patterns in χ, oxygen isotope ratios, off the Indian delta is coherent ARM and SIRM (Fig. 9), in age corresponding to LPAZ with the termination of urban Harappan civilization in 5, are probably the result of an absence of disturbances the Indus valley (Staubwasser et al., 2003). around the site. Archaeological evidence of geometric microlithic The last stage of farming culture in the Horton Plains industry in association with cereals and Black and Red has been postulated for the period between 2 and 0.15 ka Ware from the Doravaka-lena shelter in Sri Lanka, were BP. The findings of cereal type pollen (Triticum–Avena dated to 7.25 and 5 ka BP, respectively (Wijayapala sp., Hordeum sp.) and the decrease in pollen values of pers. comm.). Furthermore, a geometric microlithic UMRF provide evidence of human impact. Pollen horizon dated to ∼3.75 ka BP revealing the knowledge grains from Berberis sp. and N. domestica in LPAZ 6 of copper industry, was found at Mantai in the North- indicate that the land area was mainly used for Triticum Western part of the island (cf. Seneviratne, 1995). sp. cultivation, which possibly started close to the Jarrige (1985) has suggested that a more diversified sampling site. These activities increased slightly during agricultural system for the Indus region and Peninsular the latter part of the period. It is believed that the small- India developed after 6 ka BP. This farming system scale cultivation of Hordeum vulgare, Avena sativa and included cultivation of naked wheat as the most Triticum aestivum occurred during the last century important crop within a diverse economic system within the upper montaneregionofSriLanka (Chakrabarti, 2000). Also, finds of domesticated (Senaratne, 1956). The increase in the values of mineral sheep, goat and wild boar attest to more diversified magnetic parameters in unit 3 (Fig. 9), corresponding to animal husbandry within this economic system. Well- LPAZ 6, is probably the result from erosion caused by developed farming practices, using a multi-cropping cereal cultivation. The rise in ARM at the very top of the system, existed in the Indian sub-continent between core is probably a result of potato cultivation between 4.55 and 3.65 ka BP. This system included e.g. sorghum, AD 1950 and 1969. Traces of these activities can be millet, finger millet, wheat, barley and an animal seen in the form of terraces in some places along the husbandry e.g. cattle, sheep and goat (Possehl, 1995; valley slopes. Weber, 1995; Kajale, 1996b; Weber, 1999). Most of the areas in the Indian sub-continent have In addition to western India and Pakistan, rice been subjected to a settled farming system as a basic cultivation amongst the cereal crops has been reported subsistence strategy during the late Holocene. Palaeo- from several sites in Rajasthan, Uttar Pradesh, botanical evidence for these systems includes finds of Maharashtra, Bihar, Kashmir, Swat Valley and at cereals and other crop cultivation from 3.2 ka BP to Pirak in the Kachi Plains dating from 4.5 to 4 ka BP modern time (Kajale, 1988, 1990b, 1994, 1996b, 1997). (Stacul, 1981; Allchin and Allchin, 1989; Chakrabarti, The earliest manifestation of the early iron using 2000). The micro-botanical data (pollen, spores, communities in Sri Lanka is radiocarbon dated from charcoal and phytoliths) recorded by Kealhofer and ∼3000 ka BP onward at sites located in lowland areas Piperno (1994) and Kealhofer and Penny (1998) (<900 m a.s.l.) e.g. Anuradhapura and Sigiriya suggest intensification of agricultural activities in (Seneviratne, 1984, 1988; Deraniyagala, 1992; Myr- Thailand as early as 7–6kaBP. dal-Runebjer, 1996; Mogren, 1999). The early iron using communities of Sri Lanka, 3–2.5 ka BP, are 5.5.3. Late Holocene farming referred to as the transition from prehistoric to historic There are only a few traces recorded of arable time. This transition is characterised by breeding of farming in LPAZ 5 suggesting that the area was horses, practising of iron production and rice cultivation. subjected to very limited human activities at the This period is obvious in the lowland areas of the island beginning of the late Holocene (i.e. 3.6–2.9 ka BP). and in much of the Indian sub-continent, suggesting an This is indicated by the few sporadic finds of cultivated expansion of agricultural activities with advanced cereal type pollen together with the pollen taxa included pottery, mortuary practices and irrigation techniques in the group “C”. Minor re-occurrence of the pollen taxa (Siriweera, 1983; Allchin and Allchin, 1989; Deraniya- included in the group “N/O” indicates small-scale gala, 1992; Myrdal-Runebjer, 1996). These data imply deforestation. The significant increase in the UMRF that the prehistoric farming settlements in the Horton components implies a prevailing humid climate and a Plains (i.e. highland) were gradually re-located at lower grazing pressure on the forest due to a decrease of ∼2.9 ka BP in the lowland areas, following the 492 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 monsoon regime. It is reasonable to assume that the involved. This diversification coincides with a climatic South West Monsoon (SWM) became stronger between amelioration, i.e. an increase in humidity, during the late 3.6 and 2 ka BP (Premathilake and Risberg, 2003). glacial. Several wild grass species, growing in semi-arid According to the geomorphology of the island, the climatic conditions during the LGM, served as primary increased water budget received from the SWM could sources of genetic variability for the domesticated cereal have been properly managed with the large-scale types. Thereafter, a certain degree of systematic irrigation works which started around 2.4 ka BP at cultivation started ∼13 ka BP probably within an Anuradhapura (Myrdal-Runebjer, 1996). This intensi- indigenous context, e.g. Veddha aboriginal population. fied the rain-fed agriculture (e.g. rice) on the lowland The cultivation might have started to decline ∼8kaBP, areas of the island. This surplus of water, however, and came to an end 3.6 ka BP possibly due to intensified could not be utilised in the highland areas due to aridity as indicated by the lack of UMRF pollen. Then, physiographic and technological barriers. Thousands of the agricultural activity was limited until 2.9 ka BP. tanks and canals were built all over the lowlands during From that age until 2 ka BP, no cultivation activities are the rules of several kings of the Anuradhapura and recorded, instead UMRF dominated. Soon after 2 ka BP Polonnaruva kingdoms. Between 1.6 and 0.8 ka BP, a a new phase of agricultural activities was re-established, most sophisticated hydraulic civilization existed, which terminating by 0.15 ka BP. Potato cultivation took place served as a great economic foundation for a diversified between 1950 and 1969. agriculture (Nicholas, 1960; Gunawardana, 1978; In short it is concluded that: Brohier, 1979; Fernando, 1982; Siriweera, 1990). This subsistence strategy includes several cereal • incipient cereal plant management started ∼17.5 ka crops, e.g. rice, millet and finger millet, but not barley, BP, oat and wheat, mainly due to a lack of temperate • slash-and-burn activity, forest clearances and grazing climatic conditions in the lowlands. The last stage of have been applied, cultivation, mainly of wheat, started in the Horton • cultivation of oat (Avena sp.) and barley (Hordeum Plains nearly synchronous with the time of the start of sp.) started ∼13,000 ka BP. Wheat (Triticum sp.) was irrigation techniques in Sri Lanka. introduced ∼2kaBP, These data indicate a complex subsistence strategy, • post-LGM and late glacial climatic ameliorations which was furthermore diversified by the introduction favoured the early development of cultivation while of wheat cultivation in the Horton Plains. Archaeobo- middle Holocene aridity caused a decrease. tanical evidences, i.e. seeds from cultivated wheat dated to ∼2.3 ka BP, were found in an excavation at the Acknowledgements ancient international port site in Mantai. It is likely that migrants introduced wheat to the Island from the Indian I thank Prof. Senake Bandaranayake, University of sub-continent as suggested by Kajale (1990c). Due to Kelaniya, Sri Lanka, Prof. Urve Miller, Department of the favourable climatic condition that prevailed in the Physical Geography and Quaternary Geology (DPG/ highlands, wheat was adapted at a later stage in the QG), Stockholm University, Dr. Eva Myrdal-Runebjer, Horton Plains. Department of Archaeology, Gothenburg University and Dr. Mats Mogren, Central Board of National 6. Conclusion Antiquities, Lund, Sweden and Dr. D. S. Epitawatta, University of Sri Jayawardenepura, Sri Lanka for Pollen records, plant opal phytoliths, TN, TOC, δ13C, initiating this project. The author is indebted to Ass. C/N ratios, and mineral magnetic measurements, Prof. Jan Risberg and Ass. Prof. Ann-Marie Roberts- together with radiocarbon dates on a peat sequence son, DPG/QG and Prof. Anandha Gunathilake for obtained from the Horton Plains, central Sri Lanka, commenting on the manuscript and Dr. Siwan Davies, provide a framework to study land-use changes during DPG/DQ Mr. S. Orugodawatta, Colombo for linguistic the last 24,000 years. The different proxies show that the correction. Microscopic facilities at the Palynological hunter–foragers in central Sri Lanka had a mobile life Laboratory, Swedish Museum of Natural History were style until ∼17.5 ka BP in an open landscape associated used under supervision of the Late Prof. Siwert with xerophytic vegetation. From this age onwards, the Nilsson. I thank Mr. Sirinimal Lakdusinghe, Director life style diversifies with the achievement of incipient of the Postgraduate Institute of Archaeology (PGIAR), cereal plant management together with slash-and-burn for his support and Mr. Asoka Perera, Cartographer, techniques. Grazing and forest clearances were also PGIAR, Colombo and Mr. Henrik Risberg who R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 493 assisted in the field. Sven Karlsson, Phil. Lic., DPG/ Costantini, L., 1981. The beginning of agriculture in the Kachi Plain: DQ, assisted in constructing the pollen diagrams. Ivan the evidence of Mehrgarh. In: Allchin, B., Allchin, R. (Eds.), South Asian Archaeology. Cambridge, pp. 29–33. Romero, DPG/DQ performed the organic carbon Dalfes, H.N., Kukla, G., Weiss, H. (Eds.), 1997. Third millenium BC measurements. Prof. Per Sandgren, Department of climate change and Old World collapse. NATO ASI Series, vol. 1 Quaternary Geology, Lund University, supervised the (49). 273 pp. mineral magnetic measurements and made Fig. 9. Dassanayake, M.D., Fosberg, F.R., Clayton, W.D. (Eds.), 1980–2000. – Financial support received from SIDA/SAREC/HRCE Flora of Ceylon, vol. 1 14. Oxford and IBH Publication, New Delhi. de Rosayro, R.A., 1946a. The montane grasslands (patanas) of Ceylon and the National Research Council, Sri Lanka, is II. Topical Agriculture 102, 4–16. gratefully acknowledged. de Rosayro, R.A., 1946b. The montane grasslands (patanas) of Ceylon III. Topical Agriculture 102, 81–94. de Rosayro, R.A., 1946c. The montane grasslands (patanas) of Ceylon References IV. Topical Agriculture 102, 139–148. Delcourt, P.A., Delcourt, H.R., Cridlebaugh, P.A., Chapman, J., 1986. Abeywickrama, B.A., 1955. The origin and affinities of the flora of Holocene ethnobotanical and paleoecological records of human Ceylon. Proc. 11th Annual Session, Ceylon Association for the impact on vegetation in the little Tennessee River Valley, Advancement of Science (Section D), Colombo, pp. 1–23. Tennessee. Quaternary Research 25, 330–349. Allchin, B., Allchin, R., 1989. The Rise of Civilization in India and Denham, T.P., Haberle, S.G., Lentfer, C., Fullagar, R., Field, J., Therin, Pakistan. Cambridge University Press, Cambridge. 379 pp. M., Porch, N., Winsborough, B., 2003. Origins of agriculture at Amarasinghe, L., Pemadasa, M.A., 1983. The ecology of montane Kuk Swamp in the highlands of New Guinea. Science 301, grasslands in Sri Lanka VI. Effects of burning and cutting. Ceylon 189–193. Journal of Science, Biological Sciences 16 (1 and 2), 1–14. Deraniyagala, S.U., 1972. Archaeological explorations in Ceylon: Part Andersen, S.T., 1979. Identification of wild grass and cereal pollen. 1. Horton Plains. Spolia Zeylanica 32 (1), 13–23. Danmarks Geologiske Undersøgelse. Årbog 1978, 68–92. Deraniyagala, S.U., 1992. The : an Ecological Andersen, S.T., Bertelsen, F., 1972. Scanning electron microscope Perspective: Part I and II. Department of Archaeological Survey, studies of pollen of cereal and other grasses. Grana 12, 79–86. Colombo. 813 pp. Aucour, A.M., Bonnefille, R., Marcel, C.H., 1999. Sources and Deraniyagala, S.U., 2001. Man before Vijaya in Sri Lanka: the accumulation rates of organic carbon in an equatorial peat bog archaeological evidence. In: Premathilake, L., Devendra, S., (Burundi, East Africa) during the Holocene: carbon isotope Devendra, T., Dambadeniarachchi, C. (Eds.), Men and Monu- constraints. Palaeogeography, Palaeoclimatology, Palaeoecology ments. Central Cultural Fund, Colombo, pp. 54–62. 150, 179–189. Enzel, Y., Ely, L.L., Mishra, S., Ramesh, R., Amit, R., Lazar, B., Balasubramaniam, S., White, R., Ratnayake, S., 1993. The montane Rajaguru, S.N., Baker, V.R., Sandler, A., 1999. High-resolution forest of the Horton Plains Nature Reserve of Sri Lanka. Holocene environmental changes in the Thar Desert, north-western In: Erdelen, W., Preu, C., Ishwaran, N., Maddumabandara, India. Science 284, 125–128. C.M. (Eds.), Ecology and Landscape Management in Sri Eriksson, M.G., Sandgren, P., 1998. Mineral magnetic analyses of Lanka. Markgraf Scientific Books, Weikersheim, pp. 95–102. sediment cores recording recent soil erosion history in central Battarbee, R.W., 1986. Diatom analysis. In: Berglund, E.E. (Ed.), Tanzania. Landscape and soil erosion history in central Tanzania: a Handbook of Holocene Palaeoecology and Palaeohydrology. study based on lacustrine, colluvial and alluvial deposits. Doctoral Wiley, Chichester, pp. 527–570. Thesis No. 12, Department of Physical Geography, Stockholm Behre, K.E., 1981. The interpretation of anthropogenic indicators in University, Sweden. 28 pp. pollen diagram. Pollen et Spores 23, 225–245. Fernando, A., 1982. The ancient hydraulic civilization of Sri Lanka in Berglund, B.E., Ralska-Jasiewiczowa, M., 1986. Pollen and pollen relation to its natural resources. Journal of Sri Lankan Branch of diagrams. In: Berglund, B.E. (Ed.), Handbook of Holocene the Royal Asiatic Society 27, 1–48. Palaeoecology and Palaeohydrology. John Wiley and Sons, Fernando, C.H., 1984. Ecology and Biogeography in Sri Lanka. W. Chichester, pp. 455–485. Junk Publishers, The Hague, The Netherlands, p. 550. Beug, H.J., 1961. Leitfaden der Pollen-bestimmung. Lieferung, vol. 1. Gasse, F., 2000. Hydrological changes in the African tropics since Gustav Fischer Verlag, Stuttgart. 63 pp. the last Glacial Maximum. Quaternary Science Review 19, Bond, T.E.T., 1953. Wild Flowers of the Ceylon Hills: Some 189–211. Familiar Plants of the Up-Country District. Oxford University Golson, J., 1989. The origins and development of New Guinea Press. 239 pp. agriculture. In: Harris, D.R., Hillman, G.C. (Eds.), Foraging and Brohier, R.L., 1979. Ancient Irrigation Work in Ceylon. Government Farming: the Evolution of Plant Exploitation. Unwin Hyman Ltd., Press, Colombo. 97 pp. London, pp. 678–687. Chakrabarti, D.K., 2000. The pattern and problems in the history of Grimm, E.C., 1987. CONISS; a Fortran 77 program for stratigraphi- crops. In: Lahini, N. (Ed.), The Decline and Fall of the Indian cally constrained cluster analysis by the method of incremental Civilization. Permanent Black, New Delhi, pp. 93–98. sum of sequences. Computer and Geosciences 13, 13–35. Clayton, W.D., Davides, G., Gould, F., Lazarides, M., Soderstrom, S. Gunawardana, R.A.L.H., 1978. Hydraulic engineering in ancient Sri T., 1994. Poaceae. In: Dassanayake, M.D., Fosberg, R.F., Clayton, Lanka: the cistern sluices. In: Premathilake, L., Indrapala, K., W.D. (Eds.), Flora of Ceylon, vol. 8. Oxford and IBH Publication, Leeuw, J.E.V.L. (Eds.), Senarat Paranavitana Commemoration, New Delhi, p. 458. vol. 7. E.J. Brill, Leiden, pp. 61–74. Cooray, P.G., 1984. The Geology of Sri Lanka. National Museum Gunawardana, R.A.L.H., 1984. Inter-societal transfer of hydraulic Department, Colombo, pp. 170–179. technology in pre-colonial South Asia: some reflections based on a 494 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

preliminary investigation. South Asian Studies 22 (2), 115–142 Kajale, M.D., 1997. Evidence of rice (Oryza cf. sativa Linn.) from (Kyoto). Koppa: a Megalithic site in Karnataka. Man and Environment 22 Gupta, A.K., 2004. Origin of agriculture and domestication of plant (1), 98–101. and animals linked to early Holocene climate amelioration. Current Kane, R.P., 1998. ENSO relationship to the rainfall of Sri Lanka. Science 87, 1, 10, 54–58. International Journal of Climatology 18 (8), 859–872. Gupta, A.K., Anderson, D.M., Overpeck, J.T., 2003. Abrupt change in Kealhofer, L., Penny, D., 1998. A combined pollen and phytolith the Asian southwest monsoon during the Holocene and their links record for fourteen thousand years of vegetation change in north- to the north Atlantic Ocean. Nature 421, 354–357. eastern Thailand. Review of Palaeobotany and Palynology 103, Harris, D.R., 1972. The origins of agriculture in the tropics. American 83–93. Scientist 60, 180–193. Kealhofer, L., Piperno, D.R., 1994. Early agriculture in southeast Asia: Hather, H.G. (Ed.), 1994. Tropical Archaeobotany: Application phytolith evidence from the Bang Pakong Valley, Thailand. and New Developments. Routledge, London and New York, Antiquity 88, 564–572. p. 270. Kennedy, K.A.R., 1993. Recent discoveries and studies of the human Hawkey, D.T., 2000. Out of Asia: dental evidence for affinities and skeletal remains of ancient Sri Lanka: a palaeoanthropological microevolution of early populations from India/Sri Lanka. update. In: Seth, P.K., Seth, S. (Eds.), New Perspectives in Doctoral Thesis, publication no. DA-9837698, Arizona State Archaeology. MD Publications Pvt. Ltd., New Delhi, pp. 300–341. University, Michigan. Kennedy, K.A.R., 1999. Paleoanthropology of South Asia. Evolu- Hillman, G., Hedges, R., Moore, A., Susan, Pettitt, P., 2001. New tionary Anthropology 8, 165–185. evidence of late glacial cereal cultivation at Abu Hureyar on the Kennedy, K.A.R., 2000. God–Apes and Fossil Men: Palaeoanthropol- Euphrates. The Holocene 11 (4), 383–393. ogy in South Asia. University of Michigan Press, Michigan. Holmes, C.H., 1951. The Grass, Fern and Savanna Lands of Ceylon, Kennedy, K.A.R., Zahorsky, J.L., 1995. Trends in prehistoric their Nature and Ecological Significance. Imperial Forestry technology and biological adaptations: new evidence from Institute, University of Oxford. Institute Paper 28. Pleistocene deposits at Fa Hien cave, Sri Lanka. In: Allchin, R., Hope, G., Tulip, J., 1994. A long vegetation history from lowland Irian Allchin, B. (Eds.), South Asian Archaeology, vol. 2. Science Jaya, Indonesia. Palaeogeography, Palaeoclimatology, Palaeoecol- Publishers, Inc., USA, pp. 839–853. ogy 109, 385–398. Kennedy, K.A.R., Deraniyagala, S.U., Roertjen, W.J., Chiment, J., Huang, T.C., 1972. Pollen Flora of Taiwan. Ching-Hwa Press Ltd., Disotell, T., 1987. Upper Pleistocene fossil hominids from Sri Taiwan. 276 pp. Lanka. American Journal of Physical Anthropology 72, Jarrige, J.F., 1981. Chronology of the earlier period of the Greater 441–461. Indus as seen from Mehrgarh, Pakistan. In: Allchin, B., Allchin, R. Kitagawa, H., van der Plicht, J., 1998. Atmospheric radiocarbon (Eds.), South Asian Archaeology. Cambridge University Press, calibration to 45,000 yr BP: late glacial fluctuations and Cambridge, pp. 21–28. cosmogenic isotope production. Science 279, 1187–1190. Jarrige, J.F., 1985. Continuing and change in the North Kachi Plain Kohler, E., Lange, E., 1979. A contribution to distinguishing cereal (Baluchistan, Pakistan). In: Schotsmans, J., Taddei, M. (Eds.), from wild grass pollen grains by LM and SEM. Grana 18, South Asian Archaeology, 1983. Istituto Universitario Orientale, 133–140. Naples, pp. 35–68. Lagerås, P., Sandgren, P., 1994. The use of mineral magnetic analyses Kajale, M.D., 1988. Plant economy. In: Dhavalikar, M.K., Sakalia, H. in identifying middle and late Holocene agriculture—a study of D., Ansari, Z.D. (Eds.), Excavation at Inamgaon, Deccan College. peat profiles in Småland, southern Sweden. Journal of Archaeo- Postgraduate and Research Institute, India, pp. 727–821. logical Science 21, 687–697. Kajale, M.D., 1990a. Some initial observation on palaeobotanical Lücke, A., Schleser, G.H., Zplitschka, B., Negendank, J.F.W., 2003. A evidence for Mesolithic plant economy from excavations at late glacial and Holocene organic carbon isotope record of Damdama, Pratapgarh Uttra Pradesh. In: Ghosh, N.C., Chakra- lacustrine palaeoproductivity and climatic change derived from barti, S. (Eds.), Adaptation and Other Essays. Visva-Bharati, varved lake sediments of Lake Holzmaar, Germany. Quaternary Santiniketan, India, pp. 8–102. Science Reviews 22, 569–580. Kajale, M.D., 1990b. Early village farming in western Maharashtra, Mahavamsa, 1950. 6th century AD ancient chronicle. In: Gieger, W. India: palaeobotanical evidence. Ancient Ceylon 11, 249–262. (Ed.), Ceylon Government Information Department. Kajale, M.D., 1990c. Ancient plant economy from excavation at Manatunga, A., 1996. Research potential for studies on the origin of Mantai, District Mannar, northwest Sri Lanka. Ancient Ceylon 11, agriculture in Sri Lanka: a feasibility survey. In: Fernando, S., 204–208. Amarasinghe, M., Bulankulame, I. (Eds.), Retrospect, vol. 1. Kajale, M.D., 1994. Archaeobotanical investigations on a multi- Department of Archaeology, University of Kelaniya, p. 20. cultural site at Adam, Maharashtra, with special reference to Meadow, R., 1981. Note on the faunal remains from Mehrgarh, with a the development of tropical agriculture in part of India. In: focus on cattle (Bos sp.). In: Allchin, B., Allchin, R. (Eds.), South Hather, J.G. (Ed.), Tropical Archaeology. Routledge, London, Asian Archaeology. Cambridge, pp. 34–40. pp. 34–50. Mogren, M., 1999. Seven years in sihagiri bim: an account of Sri Kajale, M.D., 1996a. Plant resources and diet among the Mesolithic Lankan–Swedish collaboration in settlement archaeology in hunters and foragers. Colloquium, vol. 33. International Sigiriya 1998–1995. Current Swedish Archaeology 7, Congress of Prehistoric and Protohistoric Science, Forli, Italy, 107–129. pp. 251–253. Muller-Dombois, D., Perera, M., 1971. Ecological differentiation and Kajale, M.D., 1996b. Neolithic plant economy in parts of lower soil fungal distribution in the montane grasslands of Ceylon. Deccan and South India. Abstract—the Section of the 13th Ceylon Journal of Science, Biological Sciences 9, 1–41. International Congress of Prehistoric and Protohistoric Science, Munsell Soil colour Chart, 1988. Munsell Colour. Macbeth Division of Forli, Italy, pp. 67–70. Kalmorgen Instruments, Baltimore, Maryland, USA. R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496 495

Myrdal-Runebjer, E., 1996. Rice and millet. An archaeological case Ropelewski, C.F., Halpert, M.S., 1987. Global and regional scale study of a Sri Lankan transbasin irrigation system. Doctoral Thesis precipitation patterns associated with El Nino/southern oscillation. No 6, Series B, Department of Archaeology, Göteborg University. Monthly Weather Review 115, 1606–1626. 224 pp. Ropelewski, C.F., Halpert, M.S., 1989. Precipitation pattern associated Neff, U., Burns, S.J., Mangini, A., Mudelsee, M., Fleitmann, Matter, with the high index phase of the southern oscillation. Journal of A., 2001. Strong coherence between solar variability and the Climate 2, 268–284. monsoon in Oman between 9 and 6 kay ago. Nature 411, 290–292. Saarse, L., Heinsalu, A., Poska, A., Veski, S., Rajamäe, R., 1999. Nicholas, C.W., 1960. A short account of the history of irrigation up to Palaeoecology and human impact in the vicinity of Lake 11th century. Journal of Royal Asiatic Society Ceylon Branch 7, Kahala, North Estonia. In: Miller, U., Hackens, T., Lang, V., 43–64. Raukas, A., Hicks, S. (Eds.), Environmental and Cultural Nilsson, M., Klarqvist, M., Bohlin, E., Possnert, G., 2001. Variation in History of the Eastern Baltic Region, PACT, vol. 57. Rixensart, 14C age of macrofossils and different fractions of minute peat Belgium, pp. 373–403.

samples dated by AMS. The Holocene 11 (5), 579–586. Sage, R.F., 1999. In: Sage, R.F., Monson, R.K. (Eds.), C4 Plant Normile, D., 1997. Yangtze seen as earliest rice site. Science 275, Biology. Academic Press, San Diego, CA, pp. 3–16. 309. Senaratne, S.D.J.E., 1956. Regional survey of the grasslands of Olsson, I.U., 1974. Some problems in connection with the evaluation Ceylon. Study of tropical vegetation. Proceedings of Kandy of 14C dates. Geologiska Föreningens i Stockholm Förhandlingar Symposium. UNESCO, Paris, pp. 175–180. 96, 311–320. Seneviratne, S., 1984. The archaeology of the Megalithic Black Pemadasa, M.A., 1984. Grasslands. In: Fernando, C.H. (Ed.), and Red Ware Complex in Sri Lanka. Ancient Ceylon 5, Ecology and Biogeography in Sri Lanka. Kluwer Academic 237–305. Publishers, pp. 99–129. Seneviratne, S., 1988. Mica Zone and Early Iron Age sites in Sri Perera, N.P., 1967. An evaluation of the human impact on the nature Lanka: an exploratory study. The Sri Lanka Journal of the and distribution of wild plant communities in the Ceylon Humanities 12, 121–132. highlands. Doctoral Thesis, University of Leeds. 198 pp. Seneviratne, S., 1995. The ecology and archaeology of the Seruwila Perera, N.P., 1969. The ecological status of the montane grasslands copper–magnetite prospect, northeast Sri Lanka. The Sri Lanka (patanas). Ceylon Forester 9, 27–44. Journal of the Humanities 21, 113–145. Poska, A., 2001. Human impact on vegetation of coastal Estonia Sharma, S., Joachimski, M., Sharma, M., Tobschall, H.J., Singh, I.B., during the Stone Age. Doctoral Thesis Summary, Faculty of Sharma, C., Chauhan, M.S., Morgenroth, G., 2004. Late glacial Science and Technology 652, Uppsala University, Sweden. and Holocene environmental changes in Ganga Plains, northern 48 pp. India. Quaternary Science Review 23, 145–159. Possehl, G.L., 1995. Seafaring merchants of Meluhha. In: Allchin, R., Shen, C., Crawford, G.W., 1998. The origin of rice agriculture: recent Allchin, B. (Eds.), South Asian Archaeology, vol. 1. Science progress in East Asia. Antiquity 72, 858–866. Publishers Inc., USA, pp. 87–97. Singh, G., 1990. Environmental changes in southern Asia during the Possnert, G., 1990. Radiocarbon dating by accelerator technique. Holocene. Journal of Palynology 91, 277–296. Norwegian Archaeological Review 23, 30–37. Singh, P., 1991. The Neolithic Origins. Agam Kala Prakashan, New Premathilake, T.R., 1997. Recent and fossil pollen of Horton Plains: Delhi. 169 pp. environmental changes as recorded by stratigraphical studies of Singh, I.B., 2005. Quaternary palaeoenvironments of the Ganga two mires in Horton Plains. M. Phil. Thesis, Department of Plains and anthropogenic activity. Man and Environment 30 (1), Botany, University of Sri Jayawardenapura, Nugegoda, Sri 1–35. Lanka. 184 pp. Siriweera, W.I., 1983. Settlement patterns and climatic fluctuations in Premathilake, T.R., 2003. Late Quaternary palaeoecological event Sri Lanka from the earliest time up to the beginning of the sixteenth stratigraphy in the Horton Plains, central Sri Lanka: with century. In: Yoskino, H.H., Kayane, I., Bandara, C.M.M. (Eds.), contributions to the recent pollen flora. Doctoral Thesis, Thesis Climate, Water and Agriculture in Sri Lanka. University of in Quaternary Geology No. 2, Department of Physical Tsukuba, pp. 51–60. Geography and Quaternary Geology, Stockholm University, Siriweera, I., 1990. Farming systems in ancient dry zone. In: Sweden. 66 pp. Bandaranayake, S., Mogren, M., Epitawatta, S. (Eds.), Settlement Premathilake, T.R., Risberg, J., 2003. Late Quaternary climate history Archaeology of the Sigiriya–Dambulla Region. PGIAR, Colombo, of the Horton Plains, central Sri Lanka. Quaternary Science pp. 141–148. Reviews 22, 1525–1541. Sirocko, F., Sarnthein, M., Erlenkeuser, H., Lange, H., Arnold, M., Rasmusson, E.M., Carpenter, T.H., 1983. The relationship between Duplessy, J.C., 1993. Century-scale events in monsoonal climate eastern equatorial Pacific Sea surface temperatures and rainfall over the past 24,000 yrs BP. Nature 364, 322–324. over India and Sri Lanka. Monthly Weather Review 111, Snowball, I.F., 1996. Holocene environmental changes in the Abisko 517–528. region of northern Sweden recorded by the mineral magnetic Redman, C.L., 1988. The origins of agriculture. The Rise of stratigraphy of lake sediments. GFF 188, 9–17. Civilization. W. H. Freeman and Company, San Francisco, Stacul, G., 1981. Cultural change in the Swat Valley and beyond, pp. 88–140. c. 3000–1400 BC. In: Allchin, B. (Ed.), South Asian Reille, M., 1998. Pollen et spores d' Europe et d' Afrique du Nord. Archaeology, vol. 34. Cambridge University Press, Cambridge, Supplements, vol. 2. Laboratory of Historical Botany, University pp. 205–212. Marseille. 521 pp. Staubwasser, M., Sirocko, F., Grootes, P.M., Segl, M., 2003. Climate Richerson, P.J., Boyd, R., 2000. Climate, culture, and the evolution of change at the 4.2 ka BP termination of the Indus valley civilization cognition. In: Heyes, C.M., Huber, L. (Eds.), Evolution of and Holocene south Asian monsoon variability. Geophysical Cognition. MIT Press, Cambridge, pp. 329–345. Research Letters 30 (8), 1425–1435. 496 R. Premathilake / Palaeogeography, Palaeoclimatology, Palaeoecology 240 (2006) 468–496

Stuiver, M., Reimer, P.J., Bard, E., Beck, J.W., Burr, G.S., Hughen, K. Weber, S.A., 1995. Harappa archaeobotany: a model for subsistence. A., Kromer, B., McCormack, G., van der Plicht, J., Spurk, M., In: Allchin, R., Allchin, B. (Eds.), South Asian Archaeology, vol. 1998. Intcal 98 radiocarbon age calibration, 24,000–0 cal BP. 1. Science Publishers Inc., USA, pp. 115–117. Radiocarbon 40 (3), 1041–1083. Weber, S., 1999. Seeds of urbanism: palaeoethnobotany and the Indus Sukumar, R., Ramesh, R., Pant, R.K., Rajagopalan, G., 1993. A 13C civilization. Antiquity 73, 813–826. record of Late Quaternary climate change from tropical peat in Yasuda, Y. (Ed.), 2002. The Origins of Pottery and Agriculture. southern India. Nature 364, 305–307. International Research Center for Japanese Studies, Oeyama-cho, Sundara, A., 1985. Cultural ecology of the Neolithic in India. In: p. 400. Deo, S.B., Paddayya, K. (Eds.), Recent Advance in Indian Zheng, R.Y., 1982. Angiosperm Pollen Flora of Tropic and Subtropic Archaeology. Deccan Collage, Post-Graduate and Research China. Institute of Botany and South China Institute of Botany, Institute, India, pp. 39–57. Academia Sinica. 453 pp. Suppiah, R., 1987. Atmospheric circulation variations and rainfall of Zubair, L., 1999. A dynamical model for orographic rainfall in Sri Sri Lanka. Science Reports of the Institute of Geoscience, Lanka. Proceedings of the TROPMET-99. Indian Meteorological University of Tsukuba. Section A 9, 75–142. Society, Chennai, pp. 269–273. Thompson, R., Oldfield, F., 1986. Environmental Magnetism. Allen Zubair, L., 2002. El Nino-southern oscillation influences on rice and Unwin, London. 227 pp. production in Sri Lanka. International Journal of Climatology 22, Toyama, S., 2001. The origin and spread of rice cultivation in China. 249–260. Monsoon 3, 105–106. Zubair, L., 2003. El Nino-southern oscillation influences on the Vuorela, I., 1986. Palynological and historical evidence of Mahaweli stream flow in Sri Lanka. International Journal of slash-and-burn cultivation in South Finland. In: Behre, K.-E. Climatology 23, 91–102. (Ed.), Anthropogenic Indicators in Pollen Diagrams, Rotter- Zubair, L., 2004. May 2003 disaster in Sri Lanka and cyclone 01-B in dam, pp. 53–64. the . Natural Hazards 00, 1–16. Walden, J., Oldfield, F., Smith, J., 1999. Environmental magnetic: a practical guide. Technical Guide, vol. 6. Quaternary Research Association, London. 243 pp.