Reconstructing Palaeogeography and Inter-island Visibility in the Wallacean Archipelago During

14 Pages·2017·1.6 MB·English

View metadata, citation and similar papers at core.ac.uk brought to you by CORE
provided by Griffith Research Online
ArchaeologicalProspection
Archaeol.Prospect. 9999,(2017)
Published onlinein Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/arp.1570
Reconstructing Palaeogeography and
Inter-island Visibility in the Wallacean
Archipelago During the Likely Period of Sahul
–
Colonization, 65 45000 Years Ago
SHIMONA KEALY* , JULIEN LOUYS AND SUE O’CONNOR
ArchaeologyandNaturalHistory,SchoolofCulture,HistoryandLanguage,CollegeofAsiaandthePacific,
AustralianNationalUniversity,HCCoombsBuilding,9FellowsRd,Acton2601,ACTAustralia
ABSTRACT The palaeogeography of the Wallacea Archipelago is a significant factor in understanding early modern human
colonization of Sahul (Australia and New Guinea), and models of colonization patterns, as well as archaeological
surveyandsiteinterpretation,areallheavilydependentonthespecificpalaeogeographicreconstructionemployed.
Herewepresentfivereconstructionsfortheperiods65,60,55,50,and45000yearsago,usingthelatestbathometric
chartandasea-levelmodelthatisadjustedtoaccountfortheaverageupliftrateknownfromWallacea.Usingthisdata
wealsoreconstructedislandarealextentaswellastopographyforeachtimeperiod.Thesereconstructionswerethen
used to estimate visibility for each island in the archipelago, and finally to model how intervisible each island was
duringtheperiodoflikelyhumancolonization.OurmodelsprovidethefirstevidenceforintervisibilitybetweenTimor
andAustraliaatca.65–62kaand47–12ka,thesecondofwhichisnotableforitsoverlapwiththeoldestradiocarbon
datesfromTimor-LesteandAustralia. Basedonintervisibility alone, however,ourstudy suggeststhat thenorthern
routeintoPapuaNewGuineawasthemostparsimoniousrouteforfirstmodernhumanentryintoSahul.Ourstudy
providesarchaeologistswithanimportantbaselinefromwhichtoconductphysicalsurveys,interpretarchaeological
data,andtheorizethecolonizationofWallaceaandSahul.©2017TheAuthors.ArchaeologicalProspectionPublished
byJohnWiley&SonsLtd.
Keywords: Wallacea;intervisible;Sahul;Sunda;migration;palaeogeography
Introduction archaeological prospection zone. An understanding of
the palaeogeography of Wallacea significantly impacts
Palaeogeographyconcernsthestudyandreconstruction ourinterpretationsoftheregion’sarchaeology,andisvi-
oftheEarth’scontinentsandoceansthroughoutitsgeo- talforreconstructingmovementsofanatomicalmodern
logical past. Understanding the palaeogeography of a humans (AMHs) through the region and their first ar-
particularregionisvitalforinterpretingtheecology,dis- rival on Sahul (Birdsell, 1977; Butlin, 1993; Kealy et al.,
tribution,diversity,andevolutionoforganismsinthose 2015). Four possible routes of dispersal from Sunda to
regions (Benton and Harper, 2009). The biogeographic Sahul have been suggested (Birdsell, 1977; Sondaar,
regionofWallacea,definedasthearchipelagolyingbe- 1989; Morwood and Van Oosterzee, 2007), largely
tweenthecontinentalshelvesofsoutheastAsia(Sunda) reflectingBirdsell’s(1977)‘northern’(NewGuinean)or
and Australia-New Guinea (Sahul) to the exclusion of ‘southern’ (Australian) routes (Birdsell, 1977; Kealy
the Philippines (and thus following the extent of et al., 2015; Figure 1). While Birdsell (1977) favoured
Wallace’s original line; see Kealy et al., 2015), is a key thenorthernroute(Route1)basedonintervisibilitybe-
tween these island chains, O’Connor et al. (2010) sug-
*Correspondenceto:ShimonaKealy,ArchaeologyandNaturalHis-
gested that current archaeological evidence from
tory,SchoolofCulture,HistoryandLanguage,CollegeofAsiaand
the Pacific, Australian National University, Acton, 2601, ACT, Wallacea could be used to support either possibility,
Australia.E-mail:[email protected] though with the oldest dates for AMH occupation re-
ThisisanopenaccessarticleunderthetermsoftheCreativeCom-
coveredfromsitesinTimor-Lestefavouredthesouthern
monsAttributionLicense,whichpermitsuse,distributionandrepro-
ductioninanymedium,providedtheoriginalworkisproperlycited. route(O’Connor,2007;Figure1).
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Received11September2016
Revised30November2016
Accepted2January2017
S. Kealy,J. Louys and S.O’Connor
Figure 1. Map of Wallacea and neighbours showing Birdsell’s (1977) potential Sahul colonization routes, and the various archaeological sites
mentionedinthetext.Calibrateddaterangesareincludedinbrackets,roundedto1ka.Theextentofthecontinentalshelfdowntothe(cid:1)50m
bathometriccontourisshadedindarkgrey.
Hypotheses of migration made on the basis of Sahul colonization, with most studies suggesting an
molecular data have also been used to infer the initial arrival date of about 50 ka, and a period of ca.
direction, pattern, and timings of Sahul colonization 5 to 30 ka for complete dispersal between Asia and
(e.g. Redd and Stoneking, 1999; Oppenheimer, 2003; Sahul(Hilletal.,2007;Hudjashovetal.,2007;McEvoy
van Holst Pellekaan, 2013). Unfortunately, they have et al., 2010; Oppenheimer, 2012; van Holst Pellekaan,
added little clarification to the different possibilities 2013; Fregel, 2015; Bergströmet al.,2016).
suggested by the archaeological data. In part this is Itshouldbenoted thatnotallmolecularstudies are
theresultofthe use ofarchaeological data to calibrate basedonunevenlysampledpopulations,andsomeare
molecular models, creating circularity between the quite data-rich for Wallacea and Sahul; however these
results of the two approaches (Allen and O’Connell, tend to focus on more recent migration patterns (e.g.
2014). This issue notwithstanding, most molecular Redd et al., 2002; Soares et al., 2008; Jinam et al., 2012;
models are examined at a broad geographical scale Tumonggor et al., 2013). A case in point is the
that obscures regional patterns, because the focus of mitochondrial study by Hill et al. (2007), where about
such models is on Out-of-Africa and global coloniza- 20% of the modern inhabitants sampled in Island
tion rather than how AMHs made it into Sahul (e.g. South East Asia (ISEA) had mtDNA haplotypes that
Underhill et al., 2001; Oppenheimer, 2003, 2009, 2012; can be traced to the first AMH to colonize the region.
Endicott et al. 2007). Thus, genetic sampling in The other 80% of inhabitants trace their origins back
Wallacea and Sahul is comparatively limited in to more recent Holocene migrations out of Indochina,
relation to other sampled populations in these studies Taiwan, and Near Oceania, and these provide the
(e.g. Redd and Stoneking, 1999; Fregel et al., 2015). majority of genetic data for ISEA (Hill et al., 2007). A
Despite these limitations, global models do provide more recent analysis by Gomes et al. (2015) focused
useful information concerning the timing and rate of on the AMH migration from Sunda to Sahul, and
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
Reconstructingpalaeogeography andintervisibility in Wallacea
compiled over 16 thousand samples from 33 different heightstoaccountfortheloweredsea-levels(aseleva-
geographic locations. Their analysis suggested that tion is measured in metres above sea-level). Birdsell
both the northern and southern routes were used in (1977) then used these reconstructed island heights,
the colonization of Sahul, thus supporting other widths, and the distances between islands to infer
molecular models (Ingman and Gyllensten, 2003; inter-island visibility and ease of prehistoric travel.
Oppenheimer, 2009; McEvoy et al., 2010; Rasmussen His model, however, neglected to account for the
et al.,2011;van Holst Pellekaan, 2013). region’s extensive tectonic uplift (Birdsell, 1977). Our
While excavations ontheislands alongthedifferent palaeogeographic reconstructions of Wallacea
proposed routes are ongoing, archaeologists have yet essentially represent an updated, quantitative, and
torecoverevidenceofAMHoccupationthatpre-dates digitized analysis of Birdsell’s (1977) island intervisi-
the colonization dates from Sahul, and thus provide bility hypotheses with the important addition of an
support for any particular route. Understanding the uplift variable.
route of AMH dispersal through Wallacea is not just ThestudyofislandupliftinWallaceaisnotcompre-
an academic pursuit. It can assist with interpretations hensive, rather, select studies on only specific islands
ofearlymarinetechnologiesandconsequentlyaspects has left much of the differential uplift rates in the
of tool development, communication and social/ region inquestion.Thus,inorder toaccountforuplift
community structures (Bednarik et al., 1999; Balme, in our palaeogeographic reconstructions, we calcu-
2013).Withover2000islandsinWallacea,understand- latedtheaverageupliftrateforWallaceafromdifferent
ing the most likely route(s) taken by AMHs is also ratesrecordedforislandsintheregion(Table1).While
advantageous when selecting particular islands for the use of an average will result in slight over- and
archaeological attention. Additionally, the choice of under-estimates of uplift for some islands, until more
‘landing site’ for Sahul colonization has wide ranging data is available, the average rate provides a
impacts on all models of subsequent colonization of reasonable approximation for this variable. We then
Greater Australia, and Island Melanesia (Birdsell, usedthisaverageupliftratetoreconstructanadjusted
1977; Irwin et al., 1990; Clark, 1991; Field and Lahr, sea-level curve for Wallacea (Figure 2) based on the
2005; Clark et al., 2008). Until more illuminating palaeoclimactic model of Lambeck and Chappell
archaeological evidence is found, computer simula- (2001). Because most studies posit colonization of
tions and mathematical models canbe used topredict Sahul occurred sometime around 45–65 ka (Hill et al.,
likely patterns of AMH movement, and in turn, these 2007; Hudjashov et al., 2007; Oppenheimer, 2012; van
models can inform subsequent archaeological survey Holst Pellekaan, 2013; Bergström et al., 2016), we
efforts (Kealy et al., 2015). A key variable in such produced reconstructions for five periods spanning
models is palaeogeography (Van Andel, 1989; this interval: 65 ka, 60 ka, 55 ka, 50 ka, and 45 ka.
Oppenheimer,2009;O’Connelletal.,2010).InWallacea The resultant difference between ancient and current
in particular, the method and extent of sea-levels for each of the five time periods were then
palaeogeographic reconstruction plays a significant applied to the most recent bathymetric chart of
role in dispersal model outcomes (Kealy et al., 2015), Wallacea, obtained from the General Bathymetric
and a detailed reconstruction of palaeogeography is Chart of the Oceans (GEBCO_14) dataset (Smith and
vital for testing the validity of the various proposed Sandwell, 1997 and downloaded from http://www.
models of AMH dispersal toSahul. gebco.net/), using ArcGIS [Environmental Systems
ResearchInstitute (ESRI), 2014].
We calculated the distance to the geometric horizon
Materials and methods according the formula d = √2r, where d is the distance
tothehorizon,andristheradiusoftheEarth.Ignoring
TheWallaceanarchipelagoisazoneofincrediblyhigh refraction, this resulted in a distance of 3.57 km to the
tectonicactivityandgeologicalcomplexitythatresults geometric horizon. We incorporated two different
fromthesimultaneouscollisionofthreetectonicplates heights extending above the Earth’s surface
(Eurasian, Indian-Australian and Pacific-Philippine (equivalent to a person’s eye height and the top of a
Sea; Hall, 2009). Consequently, Wallacea has experi- mountain) using thefollowingformula:
enced a significant degree of tectonic uplift since the (cid:3) pffiffiffiffiffi(cid:4) (cid:3) pffiffiffiffiffi(cid:4)
time of initial AMH colonization. Birdsell (1977) used VisibilityðkmÞ¼ 3:57(cid:3) h þ 3:57(cid:3) h
1 2
the palaeoclimatic model of Chappell (1976) to lower
sea-levels by 150 m (to account for Pleistocene glacial whereh istheeyeheightoftheviewer(inmetres)and
1
conditions), then added 150 m to present island h is the height of the island (inmetres).
2
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
S. Kealy,J. Louys and S.O’Connor
Table1. UpliftratescalculatedfordifferentislandsthroughoutWallacea.ModifiedfromPedojaetal.(2014,table1).
Island/region Sitelocation Proposedmaximum Apparentupliftrates Dating Reference
record (m/ka) method
EastTimor Atauro Plio-Quaternary 0.516 U/Th Chappell&Veeh,1978
EastTimor Baucau/Ponta Notknown 0.553 U/Th Cox,2009;Chappell&
Bondura Veeh,1978
EastTimor Hau Notknown 0.082 U/Th Chappell&Veeh,1978
EastTimor Lautem Notknown 0.41 U/Th Chappell&Veeh,1978
EastTimor Laga MIS7 0.5 U/Th Cox,2009
WestTimor Namosain MIS7 0.3 U/Th Jouannicetal.,1988
WestTimor/ CapeOeloimi >MIS5e 0.328 U/Th Jouannicetal.,1988
SemauIsland
RoteIsland PointDombo >MIS5e 1.4 U/Th Roosmawati&Harris,2009
SavuIsland WestSavu >MIS5e 0.7 ESR/U/Th Roosmawati&Harris,2009
SumbaIsland CapeLaundi 1Ma 0.49 ESR Nexeretal.,2015;
Pirazzolietal.,1993
AlorIsland Kabola >MIS15 1.2 14C/U/Th/ESR Hantoroetal.,1994
KisarIsland Kisar MIS9 0.5 U/Th Majoretal.,2013
EastSulawesi Luwuk MIS7orolder 0.18 U/Th/14C Sumosusastroetal.,1989
EastSulawesi Luwuk MIS9orolder 0.336 U/Th/14C Sumosusastroetal.,1989
The equation thus takes into account the increasing height. Again, this under-estimates visibility for most
visibility that results as the object being viewed (the individuals viewing from the shore; however, it
island) extends upwards from the Earth’s surface allows us to accommodate the likelihood that the
(h ) in addition to the degree to which visibility is viewer is not standing but rather seated in some type
2
increased as a result of eye height above sea level of craft such as a raft or canoe (Friedman et al., 2010).
(h ). Further corrections to account for atmospheric For example Alor Island has an estimated elevation
1
refraction (which increases the visibility distance) of 1717 m at 45 ka. Thus a person whose eye is
are possible; however as the equation also operates approximately 0.5 m above sea level would have
under the assumption of perfect weather conditions been able to see Alor up to 150 km away. If that
we chose to use the simpler equation and avoid person was to stand up, however, and their eye
over-estimation of visibility. Furthermore, we elected was then 1.8 m above sea-level, they could see the
to use a height of only 0.5 m for the viewer’s eye island up to 152.7 km away.
Figure2.Sea-levelcurveadjustedforanaverageupliftrateof0.5m/kaforWallaceaforthelast75000years.Sea-leveldatafromLambeckand
Chappell(2001),upliftratecalculatedfromTable1.
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
Reconstructingpalaeogeography andintervisibility in Wallacea
Aftercalculatingthevisibilityforeachislandaccord- directional and can thus cause low islands to look as
ing to their reconstructed heights at the five different though they are visible from the shore of a higher
time periods, we calculated the distances out to sea island (when in fact it is only the higher island that is
that the individual island could be seen using the visible from the shores of the lower one), a higher
‘buffer’ tool in ArcGIS and our reconstructed island island also offers the viewer with a higher lookout
coastlines. Areas where this ‘buffer zone’ overlapped point should they climb inland, thus effectively miti-
were thus areas of continuous island-to-island gating this bias. Furthermore, we consider absolute
visibility. In other-words, overlap indicated that one intervisibility to be less relevant to migrating people
could travel by canoe (or some other type of water than relative intervisibility. If AMHs in Wallacea had
craft) between two islands without ever going out of the marine technology and skill to travel notable dis-
sight of land. This is not an indicator that one island tances from their ‘home’ shore in pursuit of marine
can be seen from the shore of another (see later), but resources (O’Connor et al., 2011; Samper Carro et al.,
that whilst travelling, before someone is completely 2016), then the likelihood is that new islands were
out of sight of the starting island they are in sight of initially spotted at sea rather than from land. There-
the next. We term this ‘relative intervisibility’, as two fore, we argue that relative intervisibility provides a
islands are visible from a point at sea between their better indication of how the Wallacean archipelago
shores, relativeto the observer. might have been viewed by Pleistocene peoples.
Absoluteintervisibilityisusedtodefineislandsthat In addition to interpretations of intervisibility, we
arevisiblefromtheshoreofanother.Thecalculationof used our palaeogeographic reconstructions to investi-
awholescale, shore-to-shorevisibilitymodel(absolute gate the locations of the oldest sites known from
intervisibility) is exponentially more computationally islands with Pleistocene occupation dates. We
intensive than the relative intervisibility model: rather compiled a dataset, calculating the distance from the
than determining the distance at which an island can coast and elevation of the archaeological sites at their
be seen, absolute intervisibility requires knowledge of oldest known occupation, as well as the size of the
which of all the surrounding islands can be spotted island at that time (Table 2). Using a non-parametric
from the shores of each other. As the islands are of rank correlation analysis (Spearman’s rho), we tested
significantly varying heights, a precise analysis of whether any correlation between any of our
absolute intervisibility requires multiple and simulta- geographic measurements and the timing of first
neousislanddirectionalanalyses.Whilesuchanalyses occupation existed.
would no doubt provide precise indications of visibil-
ity relationships between islands, this degree of detail
was deemed unnecessary forour study. Results
Rather, we proceeded with the (computationally
simpler)assumptionthatislandswereofthesameap- We produced five separate maps representing the
proximate height, such that halving the islands’ palaeogeography present at 65 ka (Figure 3), 60 ka
relative intervisibility produced an area of overlap (Figure4),55ka(Figure5),50ka(Figure6),and45ka
indicating approximate absolute intervisibilty. This is (Figure7).Eachofthesemapsshowsthereconstructed
because if the relative intervisibility buffer of one coastlines, island topography, relative intervisibility
island, which highlights all areas from which that andapproximateabsoluteintervisibilityforthesetime
islandcanbeseen,touchesthecoastofaneighbouring periods.Ourresultsshowthegreatestvisibilityforthe
island, then that indicates that the first island can be periodsof65and45ka,withtheleastvisibilityoccur-
seen from the shores of its neighbour (absolute ringat60ka.ThenorthernislandsbetweenSundaand
intervisibility). Halving the visibility distance of two Sahul show continued relative intervisibility between
absolutely intervisible islands of the same height 65to45 ka, while visibility is less continuous towards
results in each ‘half buffer’ merging to form an the south. Relative intervisibility between Timor and
absolute intervisible buffer equal to the islands’ Australia is present at 65 and 45 ka through a small
relative intervisibility buffer – in other words indicat- emergent island chain to the north of Darwin and the
ing that each island falls within the other’s zone of Tiwi Islands, but absent in the other reconstructions.
visibility.Thismethodallowsforthestudyofthearchi- Interpolating from the five maps in light of the
pelago as a whole by displaying chains of continued uplift-adjusted sea-level curve (Figure 2), this
(approximate) absolute intervisibility everywhere Timor-Australia relative intervisibility was also likely
these ‘half buffers’ overlap. While this approach presentbetween65and62kaand47–12ka.According
introduces a ‘highland bias,’ as the analysis is not to estimates for absolute intervisibility, at no time
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
S. Kealy,J. Louys and S.O’Connor
Table2. Archaeologicalsiteswiththeoldestknownoccupationdates(maximumcalibratedageBP,95.4%probabilityrange)forWallacean
islandswithPleistocenedata.
Island Site Dateofinitialoccupation Reference Distancetocoast Elevation Islandsize
(maximumka) (km) (km) (km2)
Talaud LeangSarru 35 Onoetal.,2009; 0.3 93.3 1654
Kealyetal.,2015
Moroti Daeo2 17 Bellwoodetal.,1998; 2 N/A 39423
Kealyetal.,2015
Gebe GoloCave 36 Bellwoodetal.1998; 3 N/A 644
Kealyetal.,2015
Sulawesi LeangTimpuseng 42 Aubertetal.,2014 59 70 227360
Alor TronBonLei 21 SamperCarroetal.,2016 2 155.5 3862
Kisar HereSorotEntapa 15 WK43324 0.5 107.5 117
Timor Jerimalai 43 O’Connoretal.,2014; 3 128.5 31583
Kealyetal.,2015
Roti LuaMeko 27 Mahirta,2003; 7 N/A 2869
Kealyetal.,2015
— — — — — — —
Spearman’srho — — 0.56631 –0.2 0.35714
Probability — — 0.1504 0.68333 0.38938
Note:Measurementsforsitedistancetothecoast(inkilometres),siteelevation(inkilometres)andisland(islandgroup)size(inkm2)interpretedfromthe
palaeogeographicreconstructionsatthemaximumoccupationdateassociatedwitheachisland.Spearman’srhoforeachsetofvariablesisincluded.
N/A,datanotavailable.
Figure3.PalaeogeographyreconstructionoftheWallaceaArchipelago65kaago,showingreconstructedtopographyandvisibilitybuffers.The
lightbluebufferindicatesregionsofrelativeintervisibility,whilethedarkblueshowstheestimatedabsoluteintervisibility.
would land have been visible from the south coast of and Sahul, but absolute intervisibility was also likely
Timor. Conversely, not only did northern Wallacea present between each ‘stepping-stone’ island along
have continued relative intervisibility between Sunda theway.Whilethevisibilitybuffersmayappearlarge,
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
Reconstructingpalaeogeography andintervisibility in Wallacea
Figure4.PalaeogeographyreconstructionoftheWallaceaArchipelago60kaago,showingreconstructedtopographyandvisibilitybuffers.The
lightbluebufferindicatesregionsofrelativeintervisibility,whilethedarkblueshowstheestimatedabsoluteintervisibility.
we note that much of present day Indonesia already into Timor and down onto the Fantone Bank on the
experiences significant relative intervisibility, and exposed shelf of northwest Australia (Birdsell, 1977;
lowered sea-levels (and thus increased island heights) Figure 1). The intervisibility models constructed here
in the Pleistocene, as well as the emergence of small show the greatest support for a northern colonization
islands throughout the archipelago, adds significantly of Sahul. Birdsell’s Route 1B in particular is well
tothe extent ofthe buffers. supported by both relative intervisibility and the
Using the distance from the coast and elevation of approximated absolute intervisibility. In addition to a
the archaeological sites at their oldest known occupa- landing point on Misool, our models also suggest the
tion, as well as the size of the island at that time possibilityofalandingsitesomewherealongthecoast
(Table 2), we found no correlationbetween any of our of the West Papua Fakfak Regency. Not only is inter-
geographic measurements and the dates of apparent visibility continuous between Seram and the Fakfak
initial occupation. coast, but this part of Papua shows a high elevation
that would have made it easy to spot by early
colonizers. Future archaeological surveys both along
Discussion theFakfakcoastandinSeramcouldprovidetheneces-
sarydatatotestthishypothesis.Fortheroutethrough
Birdsell’s (1977) intervisibility study favoured two HalmaheraandontothePapuaBirdsHead(Route1A;
routes into Sahul (Figure 1). The first (Route 1B) ran Birdsell, 1977), relative intervisibility is continuous;
from present day Borneo/Kalimantan, through howeverourestimateforabsoluteintervisibilityisless
Sulawesi and the Peleng islands, down to Ambon connected than forthe routethrough Seram.
andSeramandupacrosstoMisool.Thesecond(Route The archaeological record, however, is less support-
2B) route passed through Java and Bali, across to ive of the northern route, with both the Wallacean
Lombok and through the Nusa Tenggara island chain and nearby Sahul (New Guinea) dates of occupation
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
S. Kealy,J. Louys and S.O’Connor
Figure5.PalaeogeographyreconstructionoftheWallaceaArchipelago55kaago,showingreconstructedtopographyandvisibilitybuffers.The
lightbluebufferindicatesregionsofrelativeintervisibility,whilethedarkblueshowstheestimatedabsoluteintervisibility.
lacking the antiquity of their southern route counter- Riwi; Figure 1) do not extend back to 65 ka (Balme,
parts. A possible exception is the Vilakuav dates from 2000; Hiscock et al., 2016; Wood et al., 2016). Archaeo-
the Ivane Valley in the highlands of Eastern New logicalTL/OSL(thermoluminescence/opticallystimu-
Guinea (Summerhayes et al., 2010). While the oldest lated luminescence) dates from Madjebebe do have
date recovered from this site has a range extending ranges that overlap with the 65–62 ka period of inter-
back to ca. 49 ka, it also has a minimum age of ca. visibility;howevertheysufferfromlargeerrormargins
43 ka, making any further interpretations of the site that make further interpretations unreliable (Roberts
in relation to the initial colonization of Sahul etal.,1990;Clarksonetal.,2015).WhileMadjebebe’slo-
dependant on refinement of this date. Should the cation in the central-north of Australia might suggest
maximum date prove to be correct, the site’s location greatest support foraFantone Bank landing site,with
in Eastern New Guinea (Figure 1) still leaves a large a lack of corresponding archaeological evidence, the
geographicandtemporaldearthofarchaeologicalsites possibility that the initial occupants did not arrive via
along this route. the Kimberly or the northern routes through New
Forthesouthernroute,onlythe65and45karecon- Guinea cannot be ruled out. With the exception of the
structionssupportBirdsell’s(1977)Route2B,however, TL/OSL dates from Madjebebe (Roberts et al., 1990;
they do suggest the same Fantone Bank landing Clarkson et al., 2015) the 65 ka reconstruction, whilst
location. The 65 ka reconstruction also shows relative showing the furthest extent of southern intervisibility
intervisibility to the west, south of Roti through the forour models and thus theoretically a better time for
present day Ashmore reef, to a landing point in the southern colonization, is also at the upper limit of
north Kimberly region. This south-western route models supported by archaeological and molecular
currently receives little support from Wallacean data (Hill et al., 2007; Hudjashov et al., 2007;
archaeology (e.g. Mahirta, 2003), and the oldest dates Oppenheimer, 2012; van Holst Pellekaan, 2013; Allen
from the Kimberly region (i.e. Carpenters Gap 1 and and O’Connell, 2014; Bergström et al., 2016).
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
Reconstructingpalaeogeography andintervisibility in Wallacea
Figure6.PalaeogeographyreconstructionoftheWallaceaArchipelago50kaago,showingreconstructedtopographyandvisibilitybuffers.The
lightbluebufferindicatesregionsofrelativeintervisibility,whilethedarkblueshowstheestimatedabsoluteintervisibility.
Conversely, radiocarbon dates from Australia and allourreconstructions,althoughourestimateforabso-
Timor-Leste do overlap, with the latter dating within lute intervisibility was absent at all times between 65
the range 47–12 ka when the south-eastern route and45ka.Therearearchaeologysitesalongthisroute,
between Timor and Sahul would have experienced in Kisar and Aru, however none of these have dates
continuous relative intervisibility (O’Connor, 2007; that extend to the dates associated with AMH coloni-
O’Connor et al., 2010; Clarkson et al., 2015; Hiscock zation(O’Connoretal.,2005a,2005b).Asintervisibility
et al., 2016). Most archaeological and molecular would have remained continuous between Timor and
studies, however,suggestaninitialSahulcolonization Aru until at least 12 ka, the later occupation dates of
event before 47 ka (Hudjashov et al., 2007; Kisar and Aru could be associated with later migra-
Summerhayes et al., 2010; van Holst Pellekaan, 2013; tions from Timor along this route. Interestingly, the
Clarkson et al., 2015; Bergström et al., 2016; Hamm molecular study by Hudjashov et al. (2007) suggests
etal.,2016;Hiscocketal.,2016),whichourintervisibil- colonization of Australia along the southern route,
ity study suggests was more likely from the north. with Aru and the Fantone Bank as their two landing
Nevertheless, later colonization events from Timor points. One of the haplogroup lineages to travel the
could have occurred once relative intervisibility was Aruroute(Route2A)hasamutationalseparationdate
established, supporting a multiple-colonization hy- that overlaps with the oldest dates from the Aru
pothesis (Birdsell, 1977; Ingman and Gyllensten, 2003; excavations (O’Connor et al., 2005a; Hudjashov et al.,
Oppenheimer, 2009; McEvoy et al., 2010; Rasmussen 2007).
et al.,2011;Balme, 2013; van Holst Pellekaan, 2013). ThealternativehypothesisthatSahulwascolonized
Birdsell’s (1977) Route 2A, from Timor through the between 65 and 62 ka from Timor either to the
Tanimbars and east to Aru and Sahul receives much Kimberly coast or Fantone Bank cannot be refuted on
greater support from our models than Route 2B. This the basis of our study, and is not entirely lacking in
latter route has continuous relative intervisibility in archaeological or molecular support (Ingman and
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp
S. Kealy,J. Louys and S.O’Connor
Figure7.PalaeogeographyreconstructionoftheWallaceaArchipelago45kaago,showingreconstructedtopographyandvisibilitybuffers.The
lightbluebufferindicatesregionsofrelativeintervisibility,whilethedarkblueshowstheestimatedabsoluteintervisibility.
Gyllensten, 2003; Oppenheimer, 2012; Clarkson et al., not only conducive to extensive and rapid island
2015). As it is not supported by the majority of data colonization,butalsolikelypromotesthedevelopment
available, however, we consider the possibility to be ofmorespecializedmaritimetechnologiesandamore
less likely than the later use of the northern route. maritime confident culture. The development of trade
Further archaeological investigations and dates from not only between neighbouring islands (e.g.
both Wallacea and Sahul will impact the likelihood of Reepmeyer et al., 2016) but also more extensively
this scenario. throughout the archipelago is rendered more likely
At a broader level, our reconstructions show a considering the extent of relative intervisibility
prehistoricislandarchipelagothatwas,visually,signif- throughout the region. Intermittent occupation
icantly inter-connected during the period of initial records, such as those of Leang Sarru in the Talauds
AMH colonization. Our models indicate that AMH (Onoetal.,2009),mightbeexplainedbyanarchipelago
exploration of the Wallacean Archipelago could have with such extensive inter-island connections that a
been far more extensive than previously suggested, remote colony could easily move back to a larger
as the vast majority of the Wallacean seascape was in island if conditions on the smaller island became less
sight of land between 65 and 45 ka. The dependency favourable. Similarly, initial landing by AMHs on the
of early Wallacean communities on marine resources smaller islands of Wallacea may have occurred much
andtheirexploitationofbothnear-shoreandoff-shore earlier than their occupation records suggest, as their
environments (O’Connor et al., 2002; O’Connor et al., proximity to larger, colonized islands removes the
2010,2011;SamperCarroetal.,2016)isconsistentwith necessity to occupy the islands permanently or semi-
these models, indicating that the vast majority of the permanently, and reduces the risk associated with
Wallacean Archipelago’s marine environments were settlementofnewandunfamiliarislandenvironments.
accessible to AMH hunter-gatherers without the risk In addition to the measures of relative and absolute
of travelling beyond sight of land. Such a situation is intervisibility, other factors would have also
©2017TheAuthors.ArchaeologicalProspectionPublishedbyJohnWiley&SonsLtd. Archaeol.Prospect.9999,(2017)
DOI:10.1002/arp