1.79
0
25
1.95
0
N and longitudes 88
58.50
0
88
58.70
0
E (Fig. 1). Paharpur is situated in the monsoon region
with a summer dominant rainfall, lying just north of the Tropic of
Cancer. The climate of the area is generally warm and humid,
though based on rainfall, humidity, temperature and wind pres-
sure four seasons are recognized: (a) pre-monsoon, (b) monsoon,
(c) post-monsoon and (d) winter. The maximum daily tempera-
ture ranges from 37
C to 39
C and the minimum from 7
C to
10
C. Rainfall is very light from November to February, increases
somewhat in March and April and continues uniformly at about
250 mm/month during the monsoon months June to September.
In May and October the rainfall average decreases to approxi-
mately 130 mm/month. The total average rainfall for the year is
5080 mm.
The Paharpur region is well-drained by numerous small,
entrenched meandering streams and rivers. Approximately 4.5 km
to the west of Somapura, the Old Jamuna River ows from north to
south, maintaining a meandering course, forming many oxbow
lakes. The traces of relict watercourses are detectable from the air
and mostly run in a northsouth direction, although an eastwest
arm appears to have extended laterally just south of the monastery
(Chowdhury, 2003).
The Somapura Mahavihara site is currently surrounded by rich
alluvial farmland with small open elds that are intensively culti-
vated, though small pockets of natural vegetation persist in some
areas. Articial cultivation platforms and homestead areas are used
for growing vegetables and bananas, fruit trees and Betel leaf (pan).
Mixed forest, scrub and replanted Gajari (Assam Sal) and Sal
(Shorea robusta) occupy patchy areas of the higher ridges of the
Barind terraces.
3. Archaeological background
The archaeological ruins of the Somapura Mahavihara monas-
tery were discovered by Buchanan Hamilton who visited the site
between 1807 and 1812 whilst surveying for the East India
Company (Sanday et al., 1983) and the rst excavations were con-
ducted at the site later in the 19th century by Sir Alexander Cun-
ningham (1879). The ruins were placed on the Bangladesh List of
Protected Monuments in 1919, and inscribed to the World Heritage
List in 1985.
The site is extensive, covering approximately 10 km
2
(10 ha). It
occupies a quadrangular court measuring ca. 274 m on each side,
with high enclosure walls ca. 5 m thick and between 3.6 and 4.5 m
high. This gigantic establishment is dominated by a central shrine
with 177 surrounding monastic cells and is conspicuous by its lofty
height and unusual architectural design (Alam, 2003; see Fig. 2).
The site lies surrounded by a slightly elevated terrace, meaning the
area remained ood-free throughout historic times, a feature which
it is thought may have been an important factor in allowing the Pala
Empire to ourish.
The chronology of the occupation sequence at Sompura Maha-
vihara has largely been established on the basis of excavations
undertaken from 192728 and 193334 (see Dikshit, 1938). With
the support of UNESCO, the Department of Archaeology of the
Ministry of Cultural Affairs, Bangladesh subsequently completed
further excavations from198182, 198485, 198889, and 199091,
which have further contributed to our understanding of the
sequence of events at the site. Such excavations have revealed that
Somapura Mahavihara was occupied predominantly by the Pala
Dynasty from the middle of the 8th century through to the early
12th century. Dharmapala, the second Pala monarch who ruled
from about AD 770810, founded Somapura Mahavihara in the
second half of the 8th century. Towards the end of the 9th century
the Palas were defeated by the Gurjara-Pratiharas, but roughly
a century later the Pala Dynasty ourished once more. In the 11th
century, the Pala again suffered a setback and were devastated by
war, though towards the end of the century prosperity returned
once more. In the 12th century the Senas, who were followers of
Hinduism, replaced the Palas, and from this time onwards the
human use of Somapura Mahavihara gradually declined before the
monastery was permanently abandoned (Sanday et al., 1983).
The site sits on the PleistoceneHolocene aged Barind Clay
Residuum (Alam et al., 1990), which is the largest Quaternary
physiographic unit of the Bengal Basin and comprises slightly
elevated landform terraces within the alluvium. The underlying
sedimentsdknown as the PleistoceneHolocene Madhupur Clay/
Barind Claydare considered to be marineestuarine deposits
(Brammer, 1996). The Barind Clay Residuum is a strongly iron-
stained, clay-rich unit of weathered alluvial sediment that is
compact and resistant to erosion. In the study area it consists of two
zones: an upper zone comprising a grey to yellowish brown
mottled clayey silt/silty clay, with a lower zone of red and
brownish-yellow mottled clayey silt; the amount of sand increases
gradually with depth. Both zones feature abundant pottery frag-
ments, brick chips and plant roots, and are described as archaeo-
logical soils belonging to the Holocene epoch.
4. Materials and methods
4.1. Sampling
The archaeological soils of Somapura were sampled in May 2006
and later in August of the same year. For the collection of samples
two trenches (labelled Trench 1 and Trench 2), were excavated from
the surface to the rst occupation level of the monastery. Trench 1
is located in the vicinity of the northern outer wall of the monastery
A.K.M.M. Alam et al. / Journal of Archaeological Science 36 (2009) 504512 505
(near the main entrance of the monastery), has a maximum depth
of 1.50 m below surface, and the lower and upper zone of the
Barind Clay Residuum is clearly present. Trench 2 is located in the
vicinity of the eastern inner wall of the monastery and reached
a maximum depth below surface of 1.38 m (see Fig. 3), and reveals
the upper zone of the Barind Clay Residuum. Samples for phytolith
analysis were mostly collected from Trench 1 at regular intervals of
5 cm, while charcoal samples for radiocarbon dating were collected
from Trenches 1 and 2 at depths of 67 cm and 138 cm below
surface, and 10 cm and 95 cm depth, respectively.
4.2. Phytolith extraction techniques
Fifteen soil samples were processed using phytolith extraction
procedures as described by Wang and Lu , 1993 at the Key
Laboratory of Biogeology and Environmental Geology in China
University of Geosciences. Samples were air dried at room
temperature and then crushed with a mortar and pestle and
subsequently sieved through a 100 mm mesh (0.154 mm) to remove
root fragments and the gravel fraction. Approximately 10 g of each
soil sample was then placed into sterile beakers. Organic matter
was removed using 30% hydrogen peroxide, followed by a 10%
hydrochloric acid wash to remove any carbonates present. The
samples were then rinsed four times with distilled water in order to
remove clay particles and to return the suspension to a neutral pH
(pH 7).
Residual samples were transferred to test tubes and 5 ml of zinc
iodide heavy liquid (adjusted to a specic gravity of 2.35 to ensure
phytoliths would oat) was added. The samples were then
centrifuged to separate the phytoliths from the heavier mineral
Fig. 2. Photograph showing the archaeological ruins of the Somapura Buddhist monastery, Paharpur, Naogaon District, Bangladesh.
Fig. 1. Location map. (A) General map of Bangladesh (after Khan, 1991). (B) More detailed viewof the Paharpur study area showing the location of Trenches 1 and 2 fromwhich soil
samples were collected (based on an ASTER FCC image).
A.K.M.M. Alam et al. / Journal of Archaeological Science 36 (2009) 504512 506
fraction. The suspended phytoliths were poured into clean test
tubes and rinsed three times with distilled water. This was followed
by a 5% acetic acid wash to remove any remaining carbonates. The
phytolith fractions were then rinsed again with distilled water and
centrifuged, before being rinsed with absolute alcohol and trans-
ferred to vials for long-term storage. Phytoliths were mounted for
viewing on slides using glycerin, covered with cover slips and
sealed with parafn. The studies were then examined using a Zeiss
stereomicroscope at 400 magnication.
4.3. Phytolith identication and quantication
No modern phytolith reference database was available for the
study region and it was beyond the scope of this project to develop
one. In the absence of a regional phytolith database, the archaeo-
logical phytoliths were identied, described and classied
following Twiss (Twiss et al., 1969; Twiss, 1992, 2001) classication
scheme supplemented by published keys including Mulholland and
Rapp (1992), Pearsall and Elizabeth (1992), Fredlund and Tieszen
(1994), Kealhofer and Piperno (1998), Pearsall (2000), Lu and Liu,
2003, Wallis (2003) and Piperno (2006). Phytolith descriptions are
based on the new phytolith code of terminology (IPCN) (Madella
et al., 2005). The number of phytoliths counted ranged from445 up
to 595 per slide in order to identify at least 250 short-cell mor-
photypes per sample. Phytoliths with unrecognizable shapes were
counted and listed in the unidentied category. Sponge spicules
and microcharcoal were also recorded and are included in the
graphs as percentages of the overall assemblage.
Short-cell phytoliths were separated into two main categories:
those coming fromthe Pooideae (predominantly C3 grasses) versus
those derived from the Panicoideae (predominantly C4 grasses)
sub-families. The Pooideae-types are characterized by trapeziform
sinutes and trapeziform polylobates (Mulholland and Rapp, 1992;
Twiss, 1992; Wang and Lu , 1993; Fredlund and Tieszen, 1994;
Pearsall, 2000; Lu and Liu, 2003; Piperno, 2006). It should be noted
that these phytolith morphotypes have also been described from
the intercostal zones of the epidermis of Chloridoideae and Pan-
icoideae grasses, but are generally produced therein only in small
amounts and hence are considered relatively reliable indicators of
the Pooideae (Brown, 1984; Mulholland, 1989; Barboni et al., 1999).
The Panicoideae-type subfamily is indicated by the presence of
distinctive bilobate and polylobate morphotypes (Brown, 1984;
Mulholland and Rapp, 1992; Twiss, 1992; Pearsall, 2000; Piperno,
2006).
Other grass morphotypes identied in the phytolith assem-
blages included those from the epidermis long-cells, hair cell
mesophyll and bulliform cells, including the forms elongate
smooth, elongate echinate, scutiform hair cell (point-shaped type),
cuneiform (fan-shaped) bulliforms and parallelepipedal (square/
rectangular shaped) bulliforms (Twiss et al., 1969; Pearsall and
Elizabeth, 1992; Twiss, 1992, 2001; Wang and Lu , 1993; Lu and Liu,
2003; Piperno, 2006; Barboni et al., 2007).
4.4. Climate index
In Twiss (Twiss et al., 1969; Twiss, 1992, 2001) classication of
grass phytoliths, the division and classication of Festucoid (Pooid),
Panicoid and Chloridoid type phytoliths generally correspond with
different climates. For example, Panicoid-type phytoliths are
mainly produced by C4 grasses that predominantly grow in warm
and humid areas. C4 grasses also produce Chloridoid-type phyto-
liths, but these grasses are typical of warm semiarid or highly
seasonal regions well beyond the limits of the study area. Pooid-
type phytoliths are produced dominantly by C3 grasses that
ourish in cool seasons and climates in high latitudes or elevations.
Hence, changes in the relative abundance of Panicoid versus Pooid
phytolith types throughout a section can be used to infer past
climatic change (cf. Gu et al., 2007). For the purposes of this anal-
ysis, after comprehensively considering the different photosyn-
thetic pathways in plants phytolith morphotypes were divided into
two groups: The cool group includes Pooid-type morphologies (i.e.
trapeziform polylobates, trapeziform sinutes), elongates (both
smooth and echinate) and point-shaped (scutiform hair cell) phy-
toliths; in contrast, the warm group morphologies include Pani-
coid-type shapes (i.e. bilobates and polylobates), cuneiform
bulliforms and parallelepipedal bulliform. The ratio of cool to warm
group phytoliths is used here as a climate index, calculated as:
Climate=temperature index Ic