6 Calculating Late Classic Lowland Maya Population for the Upper Belize River Area
Total Page:16
File Type:pdf, Size:1020Kb
6 CALCULATING LATE CLASSIC LOWLAND MAYA POPULATION FOR THE UPPER BELIZE RIVER AREA Anabel Ford, Keith C. Clarke, and Sebastian Morlet The problem ofestimating ancient Maya populations has vexed archaeologists since the first temples were discovered, and existing best estimates range considerably. Using a probability map of Maya settlement patterns derived fro m predictive Bayesian modeling ofthe Upper Belize River Area, we have developed a means of estimating populations in the unknown areas.Using the ancient land use patterns ofresidential units revealed in surveyed areas, we expanded our model across a greater area, creating a probability map ofall Maya sites of the area, discovered or not. Based on a classification of the sites and demographic assumptions about the average fa mily, we derive estimates and their ranges ofpopulationfo r the Late Classic Maya that reveal an intensive land use system. Introduction Estimating Maya population numbers throughout their settlement history in .A.DziNnc:he Central America has been a preoccupation of Maya archaeologists for nearly a century. Contemplating the abandoned, forest covered temples and the numerous smaller platforms that appeared to be everywhere, gave rise to a rich variety of interpretations. Yet it was the focused attention on the major architecture and the abundance of the smaller structures that led researchers to envision a dense and widespread system of Maya settlement. As first noted by Bullard, as one looks beyond the temples and plazas variation in Maya settlement forms and configurations can be recognized (1960; see also Fedick 1995; Fedick and Ford 1990; Ford 1991; Iannone and Connell 2003; Isendahl 2002; Sabloff 1992; Smyth et al 1995; Webster Figure 1. The Central Maya Lowlands with Sites 2008). Still, the perception remains of vast indicated. cities surrounded by their sustaining rural hinterland habitations (Rice and Culbert obstructed by the density and isolation ofthe 1990 following Redfield 1967). Estimates Maya forest itself. Centers ranging in size for Late Classic settlement densities also from the large site of Tikal to the residential appear to reflect this division (Rice and community of Barton Ramie (Figure I) Culbert 1990:30-31). Yet this distinction is relied on the landscape for their subsistence possibly more structured by our own needs. In this research we sought answers to experiences of contemporary urbanization, questions of how the Maya used their land, and not by actual evidence from the ancient and consequently, how they chose to settle Maya landscape, where the perceived and farm the landscape. With answers to urban/rural dichotomy falls short of these questions, estimates of ancient describing the diversity of Maya settlement populations become feasible. (Ford 1991; Levi 2002). It is hard to The nature of settlement patterns and evaluate the efficacy of the dichotomy, densities suggests nuances that can only be however, since archaeological surveys are revealed by a more detailed examination of Research Reports in Belizean Archaeology, Vol. 8, 2011, pp. 75-87. Copyright © 2011 by the Institute of Archaeology, NICH, Belize. Late Classic Lowland Maya Population Chan b 8ana na Ban k. • 'Chorro "Ferm inReyea "Ca S~n ~ Lc;okout'Onlar1o~R~1e • • "Unltedvlle ,.! .I 1-- M eh Figure 2. The Upper Belize River Area with Study area indicated. Figure 3. Comparison of the Original Model (L) with the New Model (R). 76 Ana bel Ford et. al. the Maya landscape. In prior work, we to appreciate from the perspective offered created an assessment for Maya sites of the via a survey transect, but Geographic Upper Belize River Area with the Information Systems (GIS) offers a means to development of a predictive model using identify patterns and produce maps based on geographic variables (Ford and Clarke 2006 ; samples of identified settlements and Ford et al. 2009). This research involved patterns. This is how we constructed a extensive data correlation, building a model predictive model of Maya settlement for the based on Weight-of-Evidence (Bayesian) Upper Belize River Area (Ford and Clarke predictive methods , appl ication of the 2006, Ford et al 2009) . The results of the model, and its validation through field GIS research forms the basis of this further testing. The model proved highly successful effort to calculate Maya population density in predicting both the greater and lesser and distribution. presence of Maya settlements (Ford, Clarke, We begin with an assessment of the and Raines 2009). predictive model of Maya settlements as a Our work with Maya settlement patterns basis for the extrapolation of settlement and offers a new perspective that builds on the residential patterns. Refining our predictive constellation of environmental model to build the population estimates, we characteristics commonly associated with used the transect surveys for the Belize ancient patterns of settlement (Campbell et River Archaeological Settlement Survey aI. 2006). Put simply and in local terms, (BRASS), along with additional surveys of well-drained ridges within the tall canopy Barton Ramie , to characterize residential forests had high settlement densities (Ford configurations and settlement patterns across 1986:68-69, 88, 1991); poor-drained the wider Upper Belize River Area. The lowlands with short forest had low recognized patterns are then propagated settlement densities; and seasonally from the actual surveys to the broader map, inundated and perennial wetlands had no based on the predictive model using the GIS. settlement (Fedick and Ford 1990). Given Populations were estimated on the basis that the Maya civic centers are perched of defined residenti al units plotted with the within well-drained areas and that the GIS to create the broader settlement map. general non-central surveys incorporate the To calculate population, we develop a variety .of landforms, it may be that the strategy to first determine primary and proportions of varied environmental secondary residence derived from characteristics explain the differences in ethnohistoric and ethnographic cases. The densities. High densities around major results provide both a view of how centers such as Tikal were located in areas settlement and population vary across the with a high proportion of well-drained landscape and also present a picture of high terrain ; archaeological surveys that surround population levels and intense land use for such centers will therefore represent well the Late Classic Maya . drained ridges and high settlement densities. Broader estimates based on these densities if Predictive Model Map of Ancient Maya extended will overestimate populations. Settlement Patterns Intersite areas , such as those located Over the past several years, we have between Tikal · and Yaxha, incorporate been working with the results ofa predictive different' environmental zones ranging from model of Maya sites drawing on data ridges to wetlands (Ford 1986; 2003) and gathered in the GIS (Ford and Clarke 2006 , yield average settlement densities lower than Ford et aI. 2009; Merlet 2009, 2010). that ofTikal, for example. Focused on the data gathered by the BRASS When the environmental characteristics research , where three basic transects in the are taken into account, there is a wide western portion and targeted quadrants to variation. These variations may be difficult the east were designed to gain an 77 Late Classic Lowland Maya Population T~~i - - - - - -- rP-;-~a m ;te-;'- - - - Re~~it ;- __n _ --- - -- .---.------ ----- r -- -- - -- - Training sites for II Training points - I R ~ ~do~ Sa l~ p li ng-~fT r~ i ni ng ------ - ----- ------ -- - ----- -f-------- - -- ---- - - -------- ---POints- ------ -- --- --- --- -:.----- , ~. -;~: : :: : :;b~;SI ;t ;" -- t::~;~;-::;;,~~mt s J~;:~::~~:;~;: ~: ;: ':':i ; ::i~S ,-----.. ---- - -- -- --- ----.......-- - -------- .------ ----._-.--------.---------L--- --- ..,I ! ~~g.i~~ i c ~egE~~}<?!1 _ . .___ _R__a ~.~~~__<?f.~_e !g ht~ . .____ _ 1 .~~!£~Et~~d_ . J3.egr e. ~~<?~ Statistic . i L~ e i g h t s of EYigence (~o fE) Raster.of Weight.s. _ I Calcu !ated. ~e ig h!_s_of Ev i de!1.~e -J : f\g~e..~le..~g :...C heng__Ies~_ __ _ ~~!e!!<?!_~_ob ab l !I ty S~~n d~~ ~ ~_ Te3~ _~o_n~ !.~ !.O'!1 ~! _ l ~de p.e.. ~.d e f.lc ~_ j :_A.~~!.!::~~g ue'!9 _Tab 1 ~2.. no s ~e.. !io. !:.. ~ ro b ab i 1_~~x _ I!..~~ !l_~y~ints_Tabl~ _o. .f ~~.e.. __~ re d i ct i ye ~~~ ~J _ ___ Table L Geogra phic Information System Database for the Predictive Model of Maya Settlement (Merlet 2009). appreciation for the variation of settlements presence of sites but also the absence of in the alluvial valley, the rolling marl sites . foothills and the well-drained ridges to the In the final model refinements, north. Before the BRASS study, Maya experiment results were compared with and studies concentrated in the valley suggested without the coverage of rivers. We found that sett lement was a ribbon line pattern that the interaction of the river coverage was only along the river (Coe and Coe 1956; clouding patterns of land use, especially Thompson 1942; Willey et al. 1965). This where there was a low probabili ty of view is now altered and we have a more settlement (Ford et al. 2009: 12-13). The complex understanding of settlem ent relation between rivers and sites was found patterns today. Thu s we can compensate for to be compl ex, positive at some distances the sampling bias of prior sites selected too and negative at others. Evaluating the close together. relia bility of the models with and without Based on the BRASS research, we have the river coverage, we determined that we developed a predi cted map of the c. 1300 sq had better predictive results without the km of the Upper Belize River Area that rivers. predicts sites and their density at a 96% The validation of the predictive model confide nce level (Figure 2 and 3). Based on used successive fie ld tests based on the the known sites and their relationship to the initial mode ls (Monthus 2004; Ford and geographic variables of soil fertili ty, soil Clarke 2006; Ford et al.