Vegetation and land-use at Angkor, Cambodia: a dated pollen sequence from the Bakong temple moat.

Penny, Dan ; Pottier, Christophe ; Fletcher, Roland 等

Introduction

Angkor was the location of the capital of the Khmer state for most of the period from the eighth to sometime in the fifteenth and sixteenth centuries AD. By the twelfth century Angkor had become a vast low-density urban complex covering about 1000 sq. km (Fletcher et al. 2003), stretching from the Tonle Sap lake in the south to the Kulen hills in the north. The urban complex incorporated numerous residential loci that were integrated by an immense network of canals and embankments (Figure 1). Within Angkor, successive rulers built major temples and palaces in several locations, shifting the focus of the urban complex until the late twelfth century AD when the central enclosure of Angkor Thorn was constructed. In the seventh to ninth centuries AD, there were concurrent residential loci around Ak Yum, and at Hariharalaya where Jayavarman II established his administration as the newly anointed cakravartin (universal king) after AD 802 (Jacques 1972). Hariharalaya (Roluos) was the site of several major constructions including the massive temple-mountain of the Bakong, the first model of the Angkorian capital (Stern 1954) and the archetype of the pyramidal temple.

[FIGURE 1 OMITTED]

One method of investigating this cultural sequence is through palynology, which was successfully applied in the present research project. Vegetation is sensitive to the activities of people and the composition of the flora will often indicate the nature of land use at any given point in time. For example, the abandonment of land or the attenuation of land use will be clearly manifest in the flora, with a predictable successional shift from a cultivated flora to invasive colonising plants, woody secondary forest and climax deciduous forest. Such changes are identifiable in pollen sequences, and can be dated absolutely. Palynological analysis at the site of Hariharalaya had significance for several key periods in Angkorian history, including the date for initial construction on the Bakong site in the eighth century AD, the impact of relocating the administrative centre to the vicinity of Phnom Bakeng in the late ninth century AD, and the conundrum of the demise of Angkor some time in the fifteenth and sixteenth centuries AD. This paper presents the first continuous palynological reconstruction of vegetation change at Angkor, encompassing the entire span of the Classic Angkorian period.

Site description (Figure 2)

[FIGURE 2 OMITTED]

Epigraphic evidence indicates that the Bakong temple was inaugurated in AD 881 by King Indravarman I (AD 877-889) (Coedes 1951: 31). It is a five-tiered pyramid, the first tier measuring 65 by 67m, built on an artificial mound, and topped by a central tower built in the twelfth century AD to replace the original, which is thought to have been destroyed some time earlier (Boisselier 1952). The pyramid and its numerous associated buildings are surrounded by two walled enclosures and two moats. The largest moat is 830 by 800m is size and 30-40m wide. The inner moat is 380 by 350m in size, 70-80m in width, and crossed to the east and west by a causeway (Glaize 1993; Pottier unpublished survey). The inner moat of the Bakong is deeply excavated into the regional alluvium, approximately 8m below the level of the temple platform, and 5m below the land surface between the two moats. Consequently, despite the marked seasonal fluctuation in the groundwater table the moat is likely to have carried water since its excavation, providing the conditions necessary for the preservation of organic materials. Groslier (1998: 40) hypothesised the existence of a hydraulic connection between the moats of the Bakong and Preah Ko temples and the Indratataka reservoir to the north (Figure 2), but no evidence of such a link has yet been discovered. Prior to September 2004 when it was cleared, the moat was entirely overgrown with a thick mat of floating vegetation that to our knowledge has not been disturbed throughout the twentieth century--and in all probability much longer--protecting the sediment accumulating on the bed of the moat from disturbance by people or livestock.

The vegetation surrounding the site today is a heavily disturbed mixed evergreen forest type that is species-poor relative to the primary forests of the region (Ashwell & Fitzwilliams 1993; see also Rollet 1972a, b, c; Boulbet 1979). Much of the surrounding area is under cultivation, including cultural features such as the seasonally dry outer moats of Preah K6 and Bakong, and the large trapeang (small rectangular reservoirs) to the north-east of the Bakong, outside the external moat enclosure. Extensive fresh-water swamp forest occurs approximately 9km to the south at the margins of the Tonle Sap lake (McDonald et al. 1997).

Materials and methods

Cores of sediment were extracted from the inner moat of the Bakong temple in July 2002, using a Kullenburg-type corer (Glew et al. 2001). The longest sequence recovered was 2.67m in length and 55mm in diameter (laboratory code BK/07/02/A), from a site approximately 60m south of the eastern causeway, in the approximate centre of the moat (Figure 3 no. 4).

[FIGURE 3 OMITTED]

Core tubes were cut longitudinally in the laboratory. After visual description of the sediments, sub-samples were taken for pollen analysis at 10cm intervals between 0 and 190cm depth, and at 5cm depth between 195 and 277cm depth. Laboratory protocols followed standard procedures (Berglund & Ralska-Jasiewiczowa 1986; Stockmarr 1971). Counting and identification were undertaken at 400-1000 times magnification using standard transmitted light microscopy. Pollen taxa are expressed as a percentage of either a primary (trees, shrubs and other woody taxa) or secondary (herbaceous plants and ferns) pollen group. Terminology follows Punt et al. (1994).

Core BK/07/02/A is composed of two horizons: a black (5Y 2.5/2) peat (0.00-2.10m depth) with an acute lower boundary, over a very firm pinkish-grey (7.5YR 6/2) clayey-sand (2.10-2.67m depth). The peat deposit, which is highly humified and has a crumbly rather than fibrous texture, we interpret as the autochthonous infill of the moat. The lack of mineral material in this horizon is expected, as there is no apparent fluvial input to the moat, and sources of mineral sediment are restricted to the banks of the moat and the fallout component. The underlying clayey sand is distinct from the regional substrate--typically heavy, pale clay with abundant redox features--and is interpreted as the initial infill of the moat immediately following its excavation, comprised largely of mineral clasts reworked from the banks of the newly excavated moat and the spoil from the excavation.

Chronology (Figure 4)

[FIGURE 4 OMITTED]

A chronology for the core was provided by AMS-radiocarbon dating of pollen concentrates. Eight samples were prepared following Brown et al. (1989). This procedure returned poor results in this case, with the pre-treated samples including sporopolleniferous algae, fungal hyphae, cellulose and charcoal fragments. The ages presented here are, therefore, based on a mixed assemblage of organic materials < 120[micro]m in size, rather than a pure pollen fraction. Radiocarbon results are shown in Table 1, and plotted in Figure 4. OZG481 and OZH177, based on samples taken immediately above the acute boundary that marks the initiation of organic sedimentation with the moat, return ages of 1290 [+ or -] 40 and 1250 [+ or -] 50 years BP, which are statistically indistinguishable at the 95 per cent confidence level (T = 0.39; Ward & Wilson 1978), combining to provide a weighted average age of 1274 [+ or -] 31 years BP (Stuiver & Reimer 1993, version 5.0). This pooled age calibrates to AD 680770 AD (1[sigma]). OZH174 returned a modern age, which is consistent with the slope of the age/depth curve (Figure 4), and is indicative of the thick fibrous mat of modern and recent vegetation at the top of the profile. OZG480 (10-11cm depth) is, therefore, erroneously old with respect to its neighbouring dates, possibly reflecting the in-wash of 'old' carbon from the banks of the moat as a result of some recent disturbance. The banks of the moat and the causeways that cross it are known to be unstable. Reports from the Angkor Conservation Office, for instance, indicate that the northern side of the eastern causeway collapsed in 1949, 1951 and 1952, and the distribution of fallen laterite blocks and sandstone pieces of the ornamental and monumental naga balustrade on the south side of the eastern causeway revealed much older episodes of bank collapse.

OZH176 (187cm depth) appears erroneously young with respect to its neighbouring dates. The reason for this inversion, and the significance of its coincidence with a major palynological boundary (between 180 and 190cm depth; see below), is unknown. Similar instances of age inversion have been reported from other Southeast Asian sites with floating vegetation mats (see Maloney 1984: 44), where younger material 'rafts' under the mat and is subsequently interred. Moreover, in instances where the sediment below the mat is highly unconsolidated, sloughing of younger material from the base of the floating mat may lead to substantial age inversions. Clearly, the sedimentation history of this site is more complex and subtle than the stratigraphy would imply. In any event, we consider that OZH176 and OZG480 represent inversions in the core stratigraphy, and are omitted here for the purposes of calculating the chronological model (Figure 4).

Palynology and sedimentary charcoal (Figure 5)

[FIGURE 5 OMITTED]

114 pollen and spore types were identified from 24 samples between 0 and 210cm depth. No specimens were observed at or below 215cm depth. The number of specimens counted per sample ranged between 116 and 1221, with a mean of 554 specimens. Stratigraphically constrained classification (Grimm 1987) of the pollen and spore assemblages indicates three sample groups representing successive vegetation phases.

At vegetation Phase 1 (see Figure 5) (210-190cm depth) the arboreal component is dominated by Celtis tetrandra-type and Macaranga, with palms (Arecaceae Sabal-sim. and Borassus flabellifer), Combretaceae/Melastomataceae, Pinus merkusii, Quercus, and Urticaceae/Moraceae commonly recorded. Poaceae and Cyperaceae dominate the nonarboreal pollen sum. Aquatic plants and ferns are rare and recorded sporadically. Charcoal particle concentrations are relatively high, with an average value of 1.8 x [10.sup.6] particles/[cm.sup.3].

At Vegetation Phase 2 (180-80cm depth) there is a fundamental change in the vegetation. Ilex cymosa (swamp holy) increases markedly to become the dominant tree pollen type, while those taxa dominant in the previous phase (Celtis tetrandra-type, Macaranga and Pinus merkusii, in particular) are severely reduced. The relative abundance of Uncaria-type increases steadily through the phase to become the sub-dominant arboreal pollen type. Eugenia-type is more abundant than in the previous phase, but its representation remains low and variable. Palms are less common, particularly Borassus flabellifer, which becomes rare. Poaceae and Cyperaceae values also fall markedly in this zone, while fern spores (psilate- and verrucate-monolete forms in particular, but also the climbing fern Steonchlaena palustris) are strongly represented. Charcoal particles are less abundant than in the preceding phase, with an average concentration of 0.9 x [10.sup.6] particles/[cm.sup.3]. Algae were rare or absent.

Vegetation Phase 3 (70-0cm depth) is characterised by a discrete fall in Ilex cymosa pollen values from 70 to 40cm depth and, while values increase again above 30cm depth, I. cymosa pollen values remain low relative to the preceding vegetation phase. This is coincident with higher values for Uncaria-type, Eugenia-type, and a slight increase in Pinus merkusii. Trema, Celtis and Macaranga all increase to reach maxima at 40-50cm depth, while Borassus flabellifer reappears in the pollen record at 40cm depth (absent from the record since 100cm depth). Calophyllum, which to that time had only been a minor component of the flora, increases markedly at 20cm depth. Fern spores are again very common in these sediments, while Poaceae and Cyperaceae pollen grains are relatively rare, excepting a discrete peak in Cyperaceae values at 40cm depth, and a slight increase in Poaceae above 20cm depth. Nymphoides remains the most common of the aquatic plants. Charcoal particle values are relatively low (an average of 0.4 x [10.sup.6] particles/[cm.sup.3]), but increase above 30cm.

Discussion

The historical date for the inauguration of the Bakong temple by Indravarman is AD 881, yet the inner moat of the temple, which defines its inner enclosure, was constructed some time between the late seventh and late eighth century AD, more than a century before the temple was complete, and most likely decades before Jayavarman II reputedly initiated the cult of devaraja and assumed the title cakravartin in AD 802. Prima facie, this date corroborates Pottier's (1996) analyses that scrutinised the chronology of Indravarman's capital, and the commonly accepted interpretation of epigraphic texts. In particular, these new radiocarbon dates confirm an earlier origin and a more complex sequence for the Bakong, opening the way for a redefinition of the current idealised Angkorian urban model (Stern 1954), and demonstrating the need for a detailed reappraisal of the earliest urban development in Angkor combining new concepts of urbanism with rigorous empirical research.

The palynological data presented here indicate an intensively used agricultural landscape around the site for more than a century, from the mid-late eighth to the late ninth century AD (Phase 1). The strong representation of grass pollen and absence of deciduous canopy trees is indicative of sparse forest cover, and the presence of Pinus merkusii pollen implies an open landscape in which a regional influence is significant. The high charcoal particle values imply frequent burning, although the relatively strong presence of Macaranga, a well-known indicator of disturbance and secondary forest development, suggests that return intervals were not sufficient to suppress regrowth. A number of useful plants are apparent, particularly Celtis tetrandra-type, typically a timber tree (Po 2003) but also with edible fruit, and the palmyra palm, Borassus flabellifer.

Borassus is particularly important for inferring agricultural land use. Stargardt (1983: 68, 128) considers it to be 'second ... only to rice' in terms of its economic importance, and Fox (1977: 200) describes Borassus as chief among those plants 'inextricably involved in human history'. The uses of Borassus are legion (lumber, thatch, rope, edible fruit, fodder, sweet juice from the inflorescence, which can be taken directly, reduced into a syrup or hard sugar, or distilled to make alcohol), and in Cambodia, as in other parts of Southeast Asia (Fox 1977), individual plants are owned by family groups or rented on a seasonal basis (Ebihara 1968: 292). The production and consumption of Borassus products are universally associated with the poor, and primarily service local markets (Fox 1977: 205-207). Zhou Daguan noted in the late thirteenth century, for example, that wine distilled from sugar--again, presumably from Borassus--was 'last in importance' of the fermented drinks (2001: 75).

Borassus is thought to have originated in India and Sri Lanka, though there is no compelling evidence to support this. Along with the African genus Hyphaene, Borassus has a distinctive pollen type (monosulcate, with sparse supratectal gemmate sculpture; Ferguson & Harley 1992) and the number of palynological studies in the region is now sufficient (Penny & Kealhofer 2005; White et al. 2004; Bishop et al. 2003; Maxwell 2001; Boyd & McGrath 2001; Penny 2001; Kealhofer & Penny 1998; Maloney 1991, 1999) to permit some preliminary comments on its palaeo-biogeography. The first occurrence of Borassus flabellifer pollen in the fossil record of mainland Southeast Asia occurs on the Malay peninsula, in sediments recovered from a sixth century AD canal at Satingpra (Stargardt 1983). At Angkor Borei, southern Cambodia, Borassus pollen is recorded at a depth of 117cm in a sediment core extracted from a small reservoir, dated to the mid-seventh century AD (Bishop et al. 2003). An inscription (K.9) records that orchards of the plant (tpal tern tunnot;, Coedes 1953: 35) were presented as gifts in the seventh century AD in Vietnam. These early occurrences of Borassus at entrepots located on the east-west trade route between China, India and Europe imply that Borassus had established a beachhead in mainland Southeast Asia some time between the sixth and seventh century AD. The pollen data presented here indicate that Borassus was growing in the Angkor region by the eighth century AD at least, and probably earlier.

The staccato-like occurrence of Borassus pollen in the fossil record renders its palaeo-biogeography 'disconcertingly circumstantial', according to Maloney (1994: 147). Its rarity outside cultural loci is due to its entomophily (Ramanujam & Kalpana 1995) and low pollen productivity, which is four times lower per flower than Cocos nucifera, another economically useful palm (Subba Reddi & Reddi 1986), seven times lower than Shorea (Bera 1990), a common forest tree in the area, and eleven and ninety times lower than the over-represented genera Pinus and Quercus, respectively (Erdtman 1969). This means that even a sparse occurrence in pollen assemblages is strongly indicative of its local presence in the landscape. Furthermore, Borassus is both slow growing and propagated from seed (Thorel 2001: 128), and is relatively shade-intolerant (Chandrashekara & Sankar 1998: 173), making it particularly susceptible to competition from more aggressive colonial plants. This suggests that Borassus is not viable unless tended and is gradually selected against as secondary forests regenerate, emphasising the 'close and complementary' (Stargardt 1983: 129), almost symbiotic relationship between agriculture and Borassus arboriculture.

In the Bakong pollen record Borassus pollen reaches a maximum representation in the early-mid ninth century AD, and declines thereafter. There are only two occurrences of the pollen between the early twelfth to early eighteenth century AD. The initial decline of Borassus is coincident with a dramatic and widespread change in the flora. The marked decrease in the abundance of grass pollen and falling charcoal particle concentrations suggest a decrease in the frequency of burning, although there is no apparent expansion of deciduous forests due, presumably, to the under-representation of the most common deciduous species (Bera 1990; Ashton 1982; Chan & Appanah 1980) and the over-representation of plants growing on or immediately adjacent to the core site.

These changes are coincident with the colonisation of the moat by hydrophytic plants. The sudden dominance of ferns suggests that vegetation had invaded the moat during the early-mid tenth century (180cm depth). Once an adequate substrate had formed from these initial colonisers, hydrophilic woody plants became established in the moat. Of these, Ilex cymosa was the most common. Ilex is a common component of peat swamp forest in Southeast Asia and, prior to c. 5600 years BP, occurred in floodplain vegetation around the Tonle Sap lake under a relatively stable inundation regime (Penny 2006), but became rare during the later Holocene as seasonality increased. Ilex is no longer present in the modern 'flooded forest' of the Tonle Sap (McDonald et al. 1997). The presence of I. cymosa and other typical swamp-forest trees (Calophyllum, Eugenia-type, Barringtonia) in the moat of the Bakong reflects the development of localised freshwater peat-swamp forest that owes its existence to the relatively shallow water table and a sheltered micro-environment afforded by the deeply incised moat. The gradual increase in the representation of the epiphytic fern Stenochlaena palustris and the liana Uncaria-type probably reflects development of larger woody shrubs and small trees within the moat.

For more than 150 years from the mid-late eighth century the moat had remained relatively clear of aquatic vegetation, which we presume indicates deliberate clearing of colonising swamp plants, as it would have taken only a few years in such a sheltered environment, perhaps less, for vegetation to close over the moat. It appears that such maintenance ceased in the early-mid tenth century. The colonisation of the moat is coincident with a decrease in the intensity of land use around the temple. The decrease in burning, the dramatic declines in grass (which, no doubt, includes rice), and the gradual exclusion of Borassus flabellifer from the flora, are all indicative of a marked decrease in the use of land for agriculture. These substantial changes in land use are coincident with the movement of Yasovarman I's administration from Hariharalaya to Yasodharapura at the centre of present day Angkor toward the end of the ninth century AD. One interpretation of these data is that with the relocation to the north-west, Hariharalaya ceased to be a central focus for settlement in the region and became peripheral to the urban complex.

An alternative interpretation is that layout of the temple was reformulated in the ninth century to include a second, larger moat. This may have precluded agricultural land use immediately surrounding the temple (between the first and second enclosures), but would have permitted agricultural life to continue unabated in the immediate hinterland, beyond the sacred enclosures. In that respect, it is also interesting to note that no specific changes in the flora in or surrounding the inner moat are apparent in the twelfth century, during a period of obvious architectural rejuvenation, with the building of several structures including the central shrine topping the pyramid. This architectural resurgence, it appears, did not extend to the clearance of the moat or the removal of vegetation from the surrounding land.

Further changes in the flora of the moat are apparent from the early-mid fifteenth to early sixteenth century (70-80cm depth). The marked decline in Ilex cymosa, and extremely subtle changes in the representation of other swamp plants (notably Eugenia-type and the fern family Polypodiaceae) indicate successional changes in the swamp-forest. The reasons for this are unknown, but may relate to excessive shading of the dominant woody shrubs by climbing plants such as Uncaria and Stenochlaena palustris, both of which have their strongest representations at this time. Further declines in charcoal particle concentrations are apparent between the mid fifteenth to mid eighteenth centuries, reaching a minimum in the mid-late seventeenth century. Regionally, land use for agriculture may have been at its most attenuated at that time.

There is no evidence of the supposed fifteenth century AD sack of Angkor. The dryland pollen assemblages appear stable, while those changes that are apparent are subtle, extremely localised and driven by successional processes rather than by changes in land use. However, it may be that the final sack of Angkor had little or no physical impact at Hariharalaya which may, at that time, have been already thinly occupied and, perhaps, of little strategic importance. Unequivocal evidence of disturbance in the dryland vegetation at Haraiharalaya occurs from the late seventeenth to mid eighteenth centuries (50-40cm), with increases in secondary forest (Trema, Macaranga), grasses, increases in charcoal particles and reappearance of Barassus flabellifer. Taken together, these data imply a limited re-activation of the landscape for agriculture, but the intensity of land use appears to be markedly lower than that apparent during Hariharalaya's zenith in the eighth and ninth centuries AD.

Conclusion

These data constitute the first palynological record of land-cover and land-use change at Angkor from the pre-Angkorian period to the present day. They indicate that the inner moat of the Bakong temple, which represents the main temple enclosure, was dug in the eighth century AD, prior to the historical date for the construction of the temple itself. Agricultural land-use around the Bakong temple was intense until the late ninth century AD, after which time the intensity of land-use declined. This is consistent with the movement of the capital from Hariharalaya to Yasodharapura (centred on Phnom Bakeng), and implies that from that time on Hariharalaya was on the periphery of Angkor. There is no evidence indicating widespread land-abandonment in the fifteenth century associated with the supposed Thai invasion. Rather there is a gradual attenuation of land use around the Bakong temple after the ninth century culminating in what appears to be near abandonment in the mid seventeenth century AD. Agricultural land use did not clearly reappear at Roluos until the eighteenth century.

Acknowledgements

This research is part of the Greater Angkor Project, a collaborative project between the University of Sydney, Ecole Francaise d'Extreme-Orient and APSARA, the Cambodian government body responsible for the management of Angkor and Siem Reap. Funding is provided by the Australian Research Council (DP0211012 and DP0211600), the University of Sydney U2000 Postdoctoral Research Fellowship scheme, and the Institute of Nuclear Science and Engineering (AINSE) (Grant 03/091P). Thanks to Mitch Hendrickson, Damian Evans and Eileen Lustig for their assistance in the preparation of the manuscript, and to Sam Player, Maital Dar and Tous Somaneath for assistance in the field.

Received: 1 November 2004; Revised: 23 August 2005; Accepted: 21 October 2005

References

ASHTON, P.S. 1982. Dipterocarpaceae. Flora Malesiana Ser. 1.9: 237-552.

ASHWELL, D. & J. FITZWILLIAMS. 1993. Zonation and Environmental Management Plan for Angkor: background report on the vegetation ecology of Angkor and environs. IUCN/UNESCO.

BERA, S.K. 1990. Palynology of Shorea robusta (Dipterocarpaceae) in relation to pollen production and dispersal. Grana 29:251-5.

BERGLUND, B.E. & M. RALSKA-JASIEWICZOWA. 1986. Pollen analysis and pollen diagrams, in B.E. Berglund (ed.) Handbook of Holocene palaeoecology and palaeohydrology: 455-84. New York (NY): John Wiley & Sons.

BISHOP, P., D. PENNY, M. STARK & M. SCOTT. 2003. A 3.5ka record of paleoenvironments and human occupation at Angkor Borei, Mekong delta, southern Cambodia. Geoarchaeology 18(3): 359-93.

BOISSELIER, J. 1952. Ben Mala et la chronologie des monuments du style d'Angkor Vat. BEFEO 46 (1): 187-238.

BOULBET, J. 1979. Le Phnom Kulen et sa region: carte et commentaire. Paris: Ecole Francaise d'Extreme-Orient.

BOYD, W.E. & R.J. McGRATH. 2001. The geoarchaeology of the prehistoric ditched sites of the upper Map Sam Mun Valley, NE Thailand, III: Late Holocene vegetation history. Palaeogeography, Palaeoclimatology, Palaeoecology 171: 307-28.

BROWN, T.A., D.E. NELSON, R.W. MATHEWES, J.S. VOGEL & J.R. SOUTHON. 1989. Radiocarbon dating of pollen by accelerator mass spectrometry. Quaternary Research 32: 205-12.

CHAN, H.T. & S. APVANAH. 1980. Reproductive biology of some Malaysian dipterocarps I: flowering biology. The Malaysian Forester 43(2): 132-43.

CHANDRASHEKARA, U.M. & S. SANKAR. 1998. Ecology and management of sacred groves in Kerala, India. Forest Ecology and Management 112: 165-77.

COEDES, G. 1951. Inscriptions de Cambodge. Vol. 1. Ecole Francaise d'Extreme-Orient. Paris: E. de Boccard.

--1953. Inscriptions de Cambodge. Vol. 5. Ecole Francaise d'Extreme-Orient. Paris: E. de Boccard.

EBIHARA, M.M. 1968. Svay, a Khmer village in Cambodia. Unpublished PhD Thesis, Faculty of Philosophy, Columbia University.

ERDTMAN, G. 1969. Handbook of Palynology--An Introduction to the Study of Pollen Grains and Spores. Copenhagen: Munksgaard.

FERGUSON, I.K. & M.M. HARLEY. 1992. The significance of new and recent work on pollen morphology in the Palmae. Kew Bulletin 48(2): 205-45.

FINK, D., M.A.C. HOTCHKIS, Q. HUA, G.E. JACOBSEN, A.M. SMITH, U. ZOPPI, D. CHILD, C. MIFSUD, H. VAN DER GAAST, A. WILLIAMS & M. WILLIAMS. 2004. The ANTARES facility at ANSTO. Nuclear Instruments and Methods in Physics Research B 223-4:109-15.

FLETCHER, R., D. EVANS, M. BARBETTI, D. PENNY, T. HENG, S. IM, C. KHIEU, S. Tous & C. POTTIER. 2003. Redefining Angkor: structure and environment in the largest low density urban complex of the pre-industrial world. Udaya 4: 107-25.

Fox, J.J. 1977. Harvest of the palm: ecological change in eastern Indonesia. Cambridge (MA): Harvard University Press.

GLAIZE, M. 1993. Les monuments du groupe d'Angkor. 4th edition. Paris: J. Maisonneuve.

GLEW, J.R., J.P. SMOL & W.M. LAST. 2001. Sediment core collection and extrusion, in W.M. Last & J.P. Smol (ed.) Tracking Environment Change Using Lake Sediments. Volume I: Basin Analysis, Coring and Chronological Techniques: 73-105. Dordrecht: Kluwer Academic.

GRIMM, E.C. 1987. CONISS: a FORTRAN 77 program for stratigraphically constrained cluster analysis by the method of incremental sum of squares. Computers and Geoscience 13(1): 13-35.

GROSLIER, B.P. 1998. Melanges sur l'archeologie du Cambodge ed. J. Dumarcay. Paris: EFEO.

HUA, Q., G.E. JACOBSEN, U. ZOPPI, E.M. LAWSON, A.A. WILLIAMS, A.M. SMITH & M.J. McGANN. 2001. Progress in radiocarbon target preparation at the ANTARES AMS Centre. Radiocarbon 43(2A): 275-82.

JACQUES, C. 1972. Etudes d'epigraphie cambodgienne-VIII. La carriere de Jayavarman II. BEFEO 59: 205-20.

KEALHOFER, L. & D. PENNY. 1998. A combined pollen and phytolith record for fourteen thousand years of vegetation change in northeast Thailand. Review of Palaeobotany and Palynology 103: 83-93.

MALONEY, B.K. 1984. Inverted and other anomalous radiocarbon dates from southeast Asian pollen sites: some comments. Modern Quaternary Research in SE Asia 8: 43-7.

--1991. Palaeoenvironments of Khok Phanom Di: the pollen, pteridophyte spore and microscopic charcoal record, in C.F.W. Higham & R. Bannanurag (ed.) The Excavation of Khok Phanom Di, a Prehistoric Site in Central Thailand. Volume II: The Biological Remains (part I): 7-19. London: Society of Antiquaries of London.

--1994. The prospects and problems of using palynology to trace the origins of tropical agriculture: the case of Southeast Asia, in J.G. Hather (ed.) Tropical Archaeobotany: applications and new developments: 139-71. London: Routledge.

--1999. A 10,600 year pollen record from Nong Thale Song Hong, Trang Province, south Thailand. Journal of the Siam Society 86: 201-17.

MAXWELL, A.L. 2001. Holocene monsoon changes inferred from lake sediment pollen and carbonate records, northeastern Cambodia. Quaternary Research 56: 390-400.

McDONALD, J.A., B. PECH, V. PHAUK & B. LEEU. 1997. Plant communities of the Tonle Sap floodplain. Phnom Penh: UNESCO/IUCN/WI.

PENNY, D. 2001. A 40,000 year palynological record from north-east Thailand; implications for biogeography and palaeo-environmental reconstruction. Palaeogeography, Palaeoclimatology, Palaeoecology 171: 97-128.

--2006. The Holocene history and development of the Tonle Sap, Cambodia. Quaternary Science Reviews 25: 310-22.

PENNY, D. & L. KEALHOFER. 2005. Microfossil Evidence of Land-Use Intensification in North Thailand. Journal of Archaeological Science 32: 69-82.

Po, S. 2003. Celtis, in Z. Wu & P.H. Raven (ed.) Flora of China. Vol. 5 (Ulmaceae through Basellaceae): 15-9. Beijing: Science Press/St Louis (MO): Missouri Botanical Garden Press.

POTTIER, C. 1996. Notes sur le Bakong et son implantation. BEFEO 83: 318-26.

PUNT, W., S. BLACKMORE, S. NILSSON & A. LE THOMAS. 1994. Glossary of Pollen and Spore Terminology. LPP Contribution Series No. 1. Utrecht: LPP Foundation.

RAMANUJAM, C.G.K. & T.P. KALAPANA. 1995. Microscopic analysis of honeys from a coastal district of Andbra Pradesh, India. Review of Palaeobotany and Palynology 89: 469-80.

REAMER, P.J., M.G.L. BAILLIE, E. BARD, A. BAYLISS, J.W. BECK, C. BERTRAND, P.G. BLACKWELL, C.E. BUCK, G. BURR, K.B. CUTLER, P.E. DAMON, R.L. EDWARDS, R.G. FAIRBANKS, M. FRIEDRICH, T.P. GUILDERSON, K.A. HUGHEN, B. KROMER, F.G. MCCORMAC, S. MANNING, C. BRONK RAMSEY, R.W. REIMER, S. REMMELE, J.R. SOUTHON, M. STUIVER, S. TALAMO, F.W. TAYLOR, J. VAN DER PLICHT & C.E. WEYHENMEYER. 2004. IntCal04 Terrestrial radiocarbon age calibration, 26-0 ka BP. Radiocarbon 46: 1029-58.

ROLLET, B. 1972a. La vegetation du Cambodge. Bois et Forets des Tropiques 144:3-15.

--1972b. La vegetation du Cambodge. Bois et Forets des Tropiques 145: 23-31.

--1972c. La vegetation du Cambodge. Bois et Forets des Tropiques 146: 3-20.

STARGARDT, J. 1983. Satingpra. I, The Environmental And Economicentury Archaeology Of South Thailand. BAR International Series 158. Oxford: Archaeopress.

STERN, P. 1954. Diversite et rythme des fondations royales khmeres. BEFEO 47 (2): 649-87.

STOCKMARR. J. 1971. Tablets with spores used in absolute pollen analysis. Pollen et Spores 13: 615-21.

STUIVER M. & P.J. REIMER. 1993. Extended 14C database and revised CALIB radiocarbon calibration program. Radiocarbon 35:215-30.

SUBBA REDDI, C. & N.S. REDDI. 1986. Pollen production in some anemophilous angiosperms. Grana 25: 55-61.

THOREL, C. 2001. Agriculture and Ethnobotany of the Mekong Basin. The Mekong Exploration Commission Report (1866-1868) Volume 4, trans W.E.J. Tips. Bangkok: White Lotus Press.

WARD, G.K. & S.R. WILSON. 1978. Procedures for comparing and combining radiocarbon age determinations: a critique. Archaeometry 20 (1): 19-31.

WHITE, J.C., D. PENNY, B. MALONEY, L. KEALHOFER. 2004. Vegetation changes from the Terminal Pleistocene through the Holocene from three areas of archaeological significance in Thailand. Quaternary International 113:111-32.

ZHOU, D. (CHOU, T.-K.) 2001. The customs of Cambodia. Bangkok: The Siam Society.

Dan Penny (1), Christophe Pottier (2), Roland Fletcher (3), Mike Barbetti (4), David Fink (5) & Quan Hua (5)

(1) School of Geosciences, University of Sydney, Sydney, NSW 2006, Australia (Email: d.penny@geosci.usyd.edu.au)

(2) Ecole Francaise d'Extreme Orient, P.O. Box 93300, Siem Reap, Cambodia

(3) Department of Archaeology, University of Sydney, Sydney, NSW 2006, Australia

(4) NWG Macintosh Centre, University of Sydney, Sydney, NSW2006, Australia

(5) Australian Nuclear Science and Technology Organisation, PMB No. 1, Menai Sydney, NSW 2234, Australia

Table 1. Results of AMS radiocarbon dating of pollen samples from
core BK/07/02/A, Bakong temple moat, Angkor, Cambodia. These ages
were measured by AMS radiocarbon using the ANTARES facility at
ANSTO (Hua et al. 2001; Fink et al. 2004). Radiocarbon dates are
calibrated (Stuiver & Reimer 1993 version 5.0) using the IntCal04
curve (Reimer et al. 2004). Ages are calibrated to years AD,
representing 68.2 per cent and 95.4 per cent confidence levels
respectively, and years BP (95.4 per cent confidence only). Figures
in square brackets indicate the proportion of the probability
range represented by each intersection of the probability curve
(thus, 1 is equal to 68.2 or 95.4 per cent). Ranges marked with
an asterisk (*) are suspect due to impingement on the end of the
calibration data set. Measured [[delta].sup.13] C values relate
solely to the graphite derived from the material used for
radiocarbon measurement.

 [delta] Percent modern
 Depth ([.sup.13]C) (pMC) [+ or -]
Lab. Code (cm) per mil 1 [sigma] error

OZG480 10-11 -28.3 94.89 [+ or -] 0.45
OZH174 20-21 -27.7 109.85 [+ or -] 0.50
OZG482 40-41 -28.7 97.63 [+ or -] 0.45
OZH175 85-86 -29.2 94.10 [+ or -] 0.42
OZG483 140-141 -26.8 90.28 [+ or -] 0.43
OZH176 187-188 -24.2 94.27 [+ or -] 0.45
OZG481 207-208 -22.9 85.21 [+ or -] 0.40
OZH177 208-210 -24.4 85.54 [+ or -] 0.53

 Radiocarbon age
 ([sup.14]C years Calibrated age
Lab. Code BP [+ or -] 1 range BP (2[sigma]
 [sigma]) 95.4% probability)

OZG480 420 [+ or -] 40 321-378 [0.175867]
 389-390 [0.001249]
 427-530 [0.822884]
OZH174 modern --
OZG482 195 [+ or -] 40 * -2-34 [0.176637]
 71-116 [0.059981]
 134-227 [0.513038]
 252-307 [0.250344]
OZH175 490 [+ or -] 35 496-553 [0.982617]
 611-620 [0.017383]
OZG483 820 [+ or -] 40 673-795 [0.986605]
 878-892 [0.013395]
OZH176 475 [+ or -] 40 466-554 [0.986221]
 610-621 [0.013779]
OZG481 1290 [+ or -] 40 1093-1106 [0.013874]
 1137-1162 [0.035908]
 1167-1297 [0.950218]
OZH177 1250 [+ or -] 50 1066-1282 [1.00]
 789-811 [0.140779]
 846-856 [0.05068]

 Calibrated age range Calibrated age range
 AD (1 [sigma] 68.2% AD (2 [sigma] 95.4%
Lab. Code probability) probability)

OZG480 1433-1491 [0.926329] 1420-1523 [0.822884]
 1603-1611 [0.073671] 1560-1561 [0.001249]
 1572-1629 [0.175867]
OZH174 -- --
OZG482 1659-1682 [0.222391] 1643-1698 [0.250344]
 1736-1804 [0.618262] 1723-1816 [0.513038]
 1936-1951 * [0.159347] 1834-1879 [0.059981]
 1916-1952 * [0.176637]
OZH175 1416-1441 [1.00] 1330-1339 [0.017383]
 1397-1454 [0.982617]
OZG483 1187-1199 [0.146346] 1058-1072 [0.013395]
 1206-1261 [0.853654] 1155-1277 [0.986605]
OZH176 1417-1447 [1.00] 1329-1340 [0.013779]
 1396-1484 [0.986221]
OZG481 670-721 [0.63193] 653-783 [0.950218]
 741-770 [0.36807] 788-813 [0.035908]
 844-857 [0.013874]
OZH177 682-782 [0.80854] 668-884 [1.00]