Imagining the Landscape of Maya Forest:

Lessons in Sustainability & Resilience

BRASS/EL PILAR PROGRAM

Report on the 2003 Field Season

 The Maya forest of Mesoamerica has been the landscape against which three different types of human-environment relations have played out. For about 3000 years, the Maya, supporting a high population density, cultivated the entire region. Then, over a short period of a few human lifetimes, this system decayed to a level supporting low human densities, and event commonly known as the classic Maya Collapse.  After the European era, emergent settlement along western agricultural lines has emerged to now place the region into a phase of unsustainable population growth.

Our 2003 field season was focused on the first phase on a long term program aimed at predictive modeling of the ancient Maya landscape.  We are initiating an integrated science project that will test hypotheses about how emergent sustainable and resilient human-environment systems evolved in the tropical Maya forest of Mesoamerica and we hope to examine models of changes in the system thresholds with collapse and regeneration.  We have begun this season to build a geographically referenced record of the Maya forest region that will ultimately incorporate properties of ancient Maya settlement, land use, soils, geology, water availability, topography, and flora and fauna. These data will be gathered together in the UCSB Maya Forest GIS for the use bay all scholars and with our intention to model and explain the past with scenarios that can be used in conservation designs for the future.

The project is based on our three-scale UCSB Maya Forest GIS.  This platform will benefit the incorporation existing field data from current projects, published records, as well as the new data that we collect in the field to share with Maya forest colleagues and to calibrate and refine our predictive Maya settlement model.  Our goal will be to distribute the integrated data and the resultant model in collaboration with the Belize Government through regional networks of educational and outreach programs.   We also envision GIS training programs to enhance local knowledge in the use of the model for reserve planning, conservation, and eco-archaeological tourism benefiting from the experiences at the El Pilar Archaeological Reserve for Maya Flora and Fauna in Belize and Guatemala.

Figure 1: Three Scales of data ~ regional, local, and site specific areas of concentration in the Maya forest

Objectives

Contemporary communities of the Maya forest are pioneering lands and adapting to environmental conditions that have a long and dynamic tradition stretching back millennia to the ancient Maya civilization. While recent community land use patterns have emerged under different conditions, similar natural and physical resources shaped links that depend on an intricate interwoven alliance between culture and nature. Regardless of contemporary political boundaries, this region shares a common past, is united by the kindred present, and stands threatened by an ominous future. Current land use strategies are demonstrably unsustainable, and the accelerated deterioration of cultural and natural resources could be creating a situation of irreversible damage across the region at every scale. Without a clear appreciation of alternatives, the situation will persist. Our long-term research design draws together fundamental interests in the integrated science approach to conservation of the Maya forest and integrates cultural, physical, and natural science data on the development of one of the world's most biodiverse regions: the Maya forest.

The goals of the research are focused on the creation of a regional model that accounts for the successful evolution of land use patterns of the ancient Maya at the local and site specific levels, models their demise, accounts for the creation of the contemporary Maya forest, and projects future scenarios based on community conservation models and population demography. Incorporated into our goals is a strong international education component for students from the US and Maya forest region to participate in the site-specific research, investigation, and conservation.

Spatial/Temporal Complexity of the Maya Forest~
 Indigenous Maya or Indigenous Forest?

Geographic Setting.  The Maya forest region is characterized by rolling limestone ridges (Turner 1978) covered by a deciduous hardwood forest. This verdant jungle thrives on an annual rainfall of 1000 to 3000 mm that falls mainly from June to January. A drought-like dry-season runs from January to June. Local production activities are impacted by this wet/dry sequence today, as they were in the Maya prehistory  (Ford 1986, Fedick and Ford 1990; Rice 1993, Scarborough 1996).

A composite mosaic of regional land resources (well-drained uplands, slow-drained and steep terrain, closed depression swamps, and riverine swamps) underwrites the foundation of Late Classic Period settlement distribution and intensity in the Maya forest. Settlement densities are the greatest in the well-drained ridges across the region (Fedick 1989, 1992; Ford 1991b, 1996; Puleston 1973; Rice 1976). Ridge lands are concentrated in the interior (Turner 1978) and are char­acterized by shallow, fertile, mollisol soils of excellent quality (Dunning 1996; Fedick 1988, 1995), representing only 1% of the world's tropics yet up to 50% of the Maya forest. These soils are superior for hand cultivation methods but are inappropriate for industrial methods (Fedick 19889; see Jenkin et al. 1976).  The ancient Maya cultivated without the metal, draft animals, and plow all initially introduced by the Spanish.

Well-drained zones preferred by Maya settlement (see Bullard 1960, 1962) are unevenly dis­tributed across the region, resulting in dispersed settlement patterns (Ford 1991a, Culbert and Rice 1990; Sanders 1981). There is a suggested relationship between the availability of well-drained ridges, settlement density, and the regional Maya hierarchy (Ford 1986, 1994, 1998) that needs comprehensive field validation and examination in greater detail at the site-specific scale.

Chronology. The ancient Maya occupation of the central lowland region can be traced back into the third mil­lennium BC (Pohl et al. 1996; Leyden et al. 1988; Matsuo and Deevey 1965; Deveey et al 1979; Dunning et al. 1998, 1999;Vaughn et al. 1985; Rice et al. 1985; Curtis, et al. 1996; Hodell et al. 1991; Hodell 1995; among others). The material archaeological record, however, starts in the Middle Preclassic around 1000 BC (Rice and Rice 1990; Pope et al. 1996; Awe et al. 1991; Powis et al. 1999). Steady settlement expansion typified the first millennium BC (see Puleston and Puleston 1972; Sabloff and Henderson 1989), based essentially on household farming decisions (cf. Wilken 1987). There is continuity in settlement location from these early times through the climax of the Late Classic period AD 600-900.

The center of Tikal dominated the region in the Late Clas­sic (Coe 1975; Culbert et al. 1990; Martin and Grube 1995; Haviland 1972; Mathews 1985; 1991; Marcus 1993, Jones 1991; Chase and Chase 1992; Schele and Mathews 1991). This is the time period when Maya settle­ment expansion and construction was at its maximum (Abrams 1994; Wernecke 1994). Densities in the well-drained ridges (15-49% of the area) averaged 200 residential structures/sq. km.  Data from sample secondary cultivation areas, those too steep or too wet, averages less than 30-structures/sq km.  Infertile swamps exhibit no settlement (Fedick 1989; Fedick and Ford 1990).  These patterns require verification.

The centralization process was sustained at least 300 years to AD 900, before the "collapse" (Bove 1981; Culbert 1988, Hodell et al 1995; Williams 1993; Pyburn 1996, Kvamme 1990; Redman 1999). Major administrative and political centers, such as Tikal, witness abrupt halts in public projects; while settlements persisted in the Terminal Classic Period (AD 900-1000) in many areas (Ford 1985,1986, 1990, 1991; Webster and Gonlin 1988; Pendergast 1981, Chase and Chase 1988). Further, testing at El Pilar suggest monument building continued through this period, to finally cease in the Postclassic (Ford and Wernecke 1996; Ford et al. 1998). The region never regained its grandeur (Turner 1990).

Civilizational Collapse.  The dramatic Classic Maya collapse exemplified by abandoned temples under the canopy of the forest dramatizes the failure of Maya civilization. This detracts from the obvious adaptive successes that supported the integrated human and natural systems of the region (Marcus 1993; Pyburn 1996 cf. Fletcher 1993). Expansion of occupation and development of social complexity by the Maya were based on population growth and concomitant land use diversification, scheduling, and intensification (Fletcher 1995; Boserup 1965, 1981; Cohen 1977; Stone 1996). Early investments endured over time supported by cultivation practices that maintained regenerative processes within cropping systems and forest management practices (Fedick 1996; Graham 1987; Gliessman 2000; Sanders 1977; Pyburn 1996). Environmental dimensions constrained subsistence strategies and cultural developments mediated those constraints. The proposed systems modeling of the Maya forest will create the landforms and quantify the dynamics of the human adaptations in the Maya forest.

It has been assumed that the forest of today bears little in common with land use in the past (Turner 1993, Rice 1990; Rice et al. 1985 Deveey et al. 1976, Redman 1999). Despite evidence of major settlement densities in the past based on long sustained land use strategies, little effort has been made to identify the complex balance attained by the ancient Maya pattern. Instead, views and strategies drawn from foreign contexts are imposed that bear no resemblance to the Maya practices.  This is even more dramatic when such strategies are vividly seen as destructive short-term trajectories in process (Sever 1999; Stuart 1992; Nations 1999; Turner 1990, 2000).

The Maya are mystified by the distinction of a tropical setting, defying western perceptions of urbanism (see Sanders and Price 1968; Service 1975).  Public centers emerged in the areas of greatest settlement, yet do not fit traditional notions of cities (Ashmore 1991; Arnold and Ford 1981; Puleston 1973, Rice 1976; Graham 1996; Haviland 1969; Marcus 1983; Voorhies 1982).  This suggests the importance of "green space" consistent with the contemporary forest garden concept. Even visual metaphors expressed values placed on nature (Peterson 1983, 1990, Towsend 1992). Predator Jaguars (Benson 1997;Houston and Stuart 1989; Saunders 1989), three omnivorous and herbivorous species of monkey including the Capuchin, locally extinct (Baker 1992), as well as cacao (Peterson 1990; Whitkus et al. 1998; Dilinger et al. 2000) figure prominently in Maya art and iconography of the Late Classic Period. These have habitat implications for the Maya forest that will be explored in the proposed project.

Throughout the entire Maya sequence, a series of environmental factors have been identified and interpreted through geological, paleoecological, tephrochonological, archaeological and historical sources (e.g. McNeish 1982, 1983; Deveey et al 1979; Andrews and Hammond 1990; Espindola et al 2000; Beach 1998; Jones 1998; Schwartz 1990). Volcanic activity in Mesoamerica has the ability to distribute large amounts of volcanic ash to stratospheric levels with local impacts (Sheets 1992; See Drexler et al 1980; Espindola et al. 2000; Rose et al 1999; Sarna et al 1981; Ford and Glicken 1987, Ford and Rose 1995; Voorhies and Thomasson 1979). Cyclical droughts have been identified over a long course of time (Hodell et al 2001), without correlated human consequences until the collapse.  Similarly, global sea level rise (see Pope et al. 1996; Pohl et al. 1996), storms, and hurricane events (CRED 1997; Dale et al. 1998; Vandermeer 2000) had impacts in the past as well as in the present. These phenomena require the integration of previously independent discipline-specific data sets into one comprehensive comparative base in our modeling of the complex human and natural processes.

We propose to test the hypothesis that the ancient Maya destroyed the forest in which they lived.  Clearly the Maya forest landscape fostered the development of the Maya civilization.  Increasing evidence from agro-ecological research (see Gleissman 2000, Nigh 2001; Levasseur 2000), studies of the folk ecology of the Maya (Atran 2001), and data from ethnological and ethnohistorical sources (Farriss 1984; Schwartz 1990) suggest a continuum of alternatives. The economic success of the Maya belies the intrinsic cultural ecological relationship, after all, the Maya cities were abandoned but the forest is our present conservation challenge.    

UCSB Maya Forest Geographic Information System

The UCSB Maya Forest Geographic Information System (GIS) project has focused on assembling available data to cover the greater Maya forest region of Belize, Guatemala, and Mexico. The initial efforts have been supported by an innovative UCSB program called Research Across Disciplines and has provided the basis for our research at the regional level. These data are now combined into a regional GIS destined to be an archived database in the Alexandria Digital Library. Our compilation is based on the GIS developed at 1:250,000 by the Paseo Pantera Consortium for the US Agency for International Development. We have integrated our digitized maps of topography and soils, included Sader's (1999) land use data for the Petén, geo-referenced 11 MSS satellite images, incorporated the local GIS data base developed by Fedick (1989) for the Belize River area (topography, soils, geology, settlement), and included a 1998 1:6,000 photo-mosaic we developed of the El Pilar Archaeological Reserve for Maya flora and Fauna (EPAR). This first version that we have compiled specifically for the UCSB Maya Forest GIS has been shared with our collaborators in Belize and Guatemala.

Data for the research derive from a variety of distinct resources, involve a significant effort in compilation, and selected efforts at fieldwork to develop the essential bases for the natural/human systems inquiry and analyses. Past environmental dimensions and archaeological resources are dispersed among research centers where common work is ongoing. Much of the contemporary data are gathered in the archives of the team investigators and collaborators, embedded in different GIS data sets across the Maya forest region, found within literatures of the diverse fields represented among the team, amassed in government agencies, university departments, and natural history museums and found in the gray literature from the development, management and community based non-government organizations that are found within the region collaborating with the Selva Maya Coalition. While most of the source materials are near at hand, the collection, incorporation, and development of a common shared GIS in consistent formats and common scales will require significant attention. We will be using experience of our to assist in this aspect of the project. The following outline lists the data sets that are designed for incorporation into a comprehensive GIS to be used in our research and analyses.

Data Out of the Past

Input-Scientists, their data and literature

                Paleo-environmental (human-environment relations)

                                -pollen records, sediment cores, and botanical remains

                                - volcanology, tephrochronology and associated ash deposits and signatures

                                - interpreted local, regional, and global climate history

                Archaeology

                                -architectural monuments

                                -site survey and settlement transects

                                - ceramics and stone artifacts

                                - stratigraphy and chronologies

                                -artistic motifs, inconographic representations of flora and fauna

                Ethnohistorical

                                Codices and interpretive accounts of the Maya

Military, Clerical and administrative accounts

Travelers and documentary accounts of the region

Data in the Present

GIS resources

Maya forest sources (mapping digitizing computer, existing GIS)

                Scientists (published/analyzed scientific data)

                                -brought by co-investigators

                                -incompatibilities and regularities

                Data sets-3 scales: Site Specific -GPS/transit/transects, Local -50K, Regional -1:250

-critical: topo, soil, vegetation, hydrology, archaeological sites, political boundaries

-additional: new satellite data, DEMs, land tenure, protected areas, vegetation change

Input-Scientists, their data and literature

                Ethnological, and agroforestry data

                Biological and botanical data

                Settlement size and composition

Literature and Sheet Maps (Mexico, Guatemala, Belize)

                Maya forest sources

Paleo-environmental reports

                Archaeological reports

                Geographic maps, volcanic ash flow distributions

Data for the Future

Weights of Evidence (WofE) and Agent Based (AB) Modeling

Site-specific data collection standardization

Local scale WofE and AB modeling: geographic/climatic inputs, human systems consumption

Regional scale geographic definition/predictive modeling for climate, natural and human systems

The Maya forest is considered a biodiversity hotspot, ranked second of 25 resources at risk by Population Action International (2000). There are more than 24,000 plants, 5,000 of which are endemic. Yet, data are accumulating that this same forest reveals a low alpha diversity 89-103 species of flowering plants greater than or equal to 1.5cm DBH per ha. and a low beta diversity index of similarity of 0.53-0.71. Given the high economic component, with up to 90% of botanicals listed as useful plants, we are asking what is the role of human systems in the development of the Maya forest, and what are the implications of this issue for the conservation of the Maya forest in the future? If human interventions prove to selectively high grade the species composition to favor their economic needs over 4000 years, what does this mean for the evaluation of the Maya forest? Flora and fauna now recognized to be at risk must have co-evolved with the ancient Maya. Revealing and modeling the complex biosystems relationships is the true mystery of the Maya forest.

We have developed the geographic database for the Maya forest founded on the UCSB Maya forest GIS. Using Weights-of-Evidence rankings of the independent geographic variables based on local scale research in the Belize River area, these geographic variables are used to predict settlement distribution in the Late Classic Period at the regional scale. Target field-testing areas combined with data from among the team have allowed for testing at the local scale. Future fieldwork at El Pilar combined with ongoing work by team scholars will provide the basis for the site-specific scale data. Succeeding steps will be to develop the spatial data from the Maya region and combine it with agent based modeling where archaeological data will be used for model validation. Paleo-environmental data will provide exploratory factors for explanation and calibration. The aim is to simulate the model dynamics of human and natural systems. In this way we hope to be able to create and test hypotheses on the complex human dynamic contribution to the Maya forest landscape in the past and today.

The contemporary model application will be configured with the Stella Maya forest landscape model for the contemporary land use change. We intend to examine what components of the model remain unchanged, which components work when modified, and which do not work at all. We will investigate whether sub-model components from the Maya forest model can be grafted into the contemporary model to permit “alternative futures” based on Maya land use practices. Given that the contemporary land use practice has put the Maya forest at risk, we will explore the possible long-term outcomes of a modified model.

Figure 2: The Prototype Stella Model for the Maya forest

Basic model systems modules are (1) a population dynamics for the human population (2) a land use module with the division by use intensity (3) a natural physical module with soil, surface geology, water, climate and vegetation interaction and (4) a biotic module with flora and fauna interact with the human and natural systems. Circles and arrows are factors and their influences. Blocks are stocks with pipe flows rates controlled by valves. The empirical relations in the model will be formulated using input from domain experts and the literature. Tests of the model will be from the field using the GIS. The model will be calibrated twice, once for the Maya period and once for the contemporary situation. With other data and model results, it will also be used for prediction of the Maya collapse, and of the next decades of d

Integrated Methods of Investigation

The project applies both traditional areas of geographic theory and new predictive methods using weights-of-evidence (WofE) and agent-based modeling to the Maya case. Our Maya forest GIS database is used as the basis for the weights-of-evidence (WofE) analysis and agent based models.. Environmental components of our settlement location model will include soils, geology, topographic variables, and surface hydrology as independent predictor layers with known archeological Maya settlement locations as both actual and predicted locations in the WofE analysis. The strength of associations provide ranks for environmental contributions to Maya settlement patterns, our first phase of work completed in 2003.  This new map basis will be the foundation for the agent based modeling.

The WofE origins are in mining geology and only recently have the essential tools been integrated into a GIS, ESRI's ArcView 3.2. Gary Raines, who helped build the ArcView extensions for WofE analysis has worked on the Columbia River Basin Ecosystem Management Project and in the development of GIS data standards (http://geology.usgs.gov/dm/). He has joined our group and is enthusiastically collaborating on this project. As developed, WofE analysis follows six steps: 1) select known points of some feature such as farming sites that are to be modeled, 1) select thematic maps that are suspected to contribute to the explanation of a distribution, 2) using the correlation analysis tools of WofE, convert selected map layers to binary or categorical form, 3) test for conditional independence comparing prior and posterior probabilities by class combinations, eliminating those maps which do not contribute explanatory power, 4) create a set of weights to use for each layer using Bayesian methods, and 5) develop posterior probability and the associated uncertainty maps using the weighted layers. The probabilities are then used as environmental weights in the agent based land-use model.

Additional analytical capabilities that complement WofE and deals with limitations of Bayesian assumptions of WofE are just publicly available. This new ArcView extension provides 1) logistic regression, eliminating the conditional dependency issues of WofE, 2) two neural network tools, providing the potential for dealing with nonlinear relationships, and 3) fuzzy logic, allowing for predictive modeling based on expert opinion without known training points in the area studied. This is a distinct advantage in terms of our application to the archaeology of the Maya forest.

The ranking of Maya forest environment based on land use intensity in the Late Classic Period provides a basis for envisioning the Maya forest at the population peak. Working along with our team, we have begun to calibrate environmental data across the region to understand the dynamics of cultural and natural systems and the changing forest over time.

We have succeeded in developing our first map for the settlement of the upper Belize River area.  This initial predictive model for the ancient Maya landscape uses digitized 1:50:000 topographic data, hydrographic data, soil fertility data and soil drainage data.  Together they present a first level view that was field verified.  The field tests were challenging and need more in depth data from 1) the archives of the Belize Institute of Archaeology to better resolve the location information on ancient Maya sites and 2) the hydrological data form the area.   We have refined the model with strategic field data collection focused on the archaeological sites as well as the landforms in general. 

Our first results are encouraging.  The model provides an excellent model for understanding the extremes of settlement patterns of the ancient Maya (Figure 3). Those areas devoid of settlement and those areas with high settlement density are predicted with the high level of confidence.  The issues arise in the lower settlement density areas.  These are harder to define and may relate to the general hydrology, distribution of water sources, and the Mayas own strategies for impounding water in reservoirs or aguadas.  This highlights the issues of water in the detection of Maya sites and supports the increasing identification of water issues are critical in the development and demise of the Classic Maya civilization. 

Our work with the GIS evidential layers of topography, fertility, drainage and water were useful in predicting Maya settlement.   Yet, this initial effort working with the predictive model demonstrates several key points.  First, we need to develop a more comprehensive base of sites in the various habitats of the Belize River area and develop a ranking to distinguish the range of settlements from major and minor centers to compounds and small isolated structures.  Second, we have found that chronological data are important in understanding the evolution of settlement. Initial examination of the BRASS settlement data over the predictive model suggests that the largest residential units have the longest sequences and they appear to be located in ecotones between several types of landforms. This may imply that residential locations \maximizes access to a variety of resources zones while occupying the best areas for agriculture.  Third, the GIS data on the hydrology of the area is weak and needs to be improved.  This is a challenging problem as it is not simply access to flowing water, the Maya were able to modify landscapes to impound water and this fact needs to be developed within the context of the Maya forest GIS.  We are presently working with the natural hydrology layer based on the Digital Elevation Models (DEM) for the area.  We will be able to identify water flow streams with levels of inclusion of 1, 2, 3, etc levels of capture.  This will begin to define the complexity of the water layer.  In sum, the initial model development and field tests have been illuminating, demonstrating the value of the model for predicting cultural resources on the landscape, identifying areas without evidence, and building a model to understand the ancient Maya landscape of the Belize River area. 

Figure 3:  Preliminary predictive model for Maya sites in the Upper Belize River area.

The Belize River area was one of the most important loci of settlement from the earliest times and Awe's and others work have demonstrated.  These early sites are at the base of the local area development and are often the location of Late Classic centers, suggesting a consistent incremental development of the initial resources across the prehistoric landscape. Assembling our collective data into one comprehensive data base on the settlement and environment will begin to expose the ancient Maya strategies of land use, the land use priorities in prehistory, and the foundation for the management in the future. 

We see an exciting series of field and laboratory activities to build the Maya predictive model, and collaboration is at the foundation of the next steps.  Clearly more field tests at the three scales of our Maya forest GIS are the very essence of the model development.  Based on the fieldwork at El Pilar, the expansion of the residential settlement survey and test excavations for the determination of settlement chronology will be important.    But equally important, we need the incorporation of the wide array of existing data that reside in a number of institutional arenas of Belize: Institute of Archaeology, Land Information Center, Department of Hydrology, University of Belize, MesoAmerican Biological Corridors among a few. These disparate data can be drawn together as an example of environmental integration for the research and protection of the Maya archaeological landscape. Specific data on the recorded sites in the Belize River area from the archives of the Institute of Archaeology can provide an important independent test and would begin the formal compilation of paper maps into the GIS for Maya sites of Belize.  

These are fundamental steps towards uniting the local Belize data base on the cultural environment, refining and calibrating of the predictive model of Maya sites, and facilitating the use of the model in scaling the major concerns of cultural resource management. Substantial data exist in the research archives and from the wide base of literature of the Belize River Area.  These too need to be targets of compilation.  With the integration of these data into a comprehensive foundation the Institute of Archaeology, as the national repository of the cultural resource inventory, would have a major base upon which to build education and planning for the future of these irreplaceable resources.