The MayaArch3D project: A 3D WebGIS for analyzing ancient architecture and landscapes

There is a need in the humanities for a 3D WebGIS with analytical tools that allow researchers to analyze 3D models linked to spatially referenced data. Geographic Information Systems (GIS) allow for complex spatial analysis of 2.5D data. For example, they offer bird’s eye views of landscapes with extruded building footprints, but one cannot ‘get on the ground’ and interact with true 3D models from a pedestrian perspective. Meanwhile, 3D models and virtual environments visualize data in 3D space, but analytical tools are simple rotation or lighting effects. The MayaArch3D Project is developing a 3D WebGIS—called QueryArch3D—to allow these two distinct approaches to ‘talk to each other’ for studies of architecture and landscapes—in this case, the eighth-century Maya kingdom of Copan, Honduras. With this tool, researchers can search and query, in real time via a virtual reality (VR) environment, segmented 3D models of multiple resolutions (as well as computer-assisted design and reality-based) that are linked to attribute data stored in a spatial database. Beta tests indicate that this tool can assist researchers in expanding questions and developing new analytical methods in humanities research. This article summarizes the results of a pilot project that started in 2009, with an art historian and an archaeologist’s collaborative research on the ancient Maya kingdom and UNESCO World Heritage site of Copan in Honduras—called MayaArch3D. The project researches inno736 digitalcommons.unl.edu T h e M a y a a r c h 3D p r o j e c T : 3D W e b GIS f o r a n c I e n T a r c h I T e c T u r e a n D l a n D S c a p e S 737 1 The Gap between GIS and 3D Modeling Systems 1.1 3D modeling Modern sensor and computing technologies are changing the practice of art history and archaeology because they offer innovative ways to document, reconstruct, and research the ancient world in 3D (El-Hakim et al., 2008; Reindel and Wagner, 2009). State-of-the-art imaging technologies allow researchers to document 3D objects to the level of the micron (e.g. Grün, 2008), whereas Virtual Reality (VR) simulation programs enable reconstructions of ancient buildings in their ancient environments and landscapes. However, as Frischer has noted (2008), the perception is that 3D models are purely illustrative—ideal for education or conservation— whereas how 3D models can assist with comparative research on architecture is an ongoing question. Since 1998, Jennifer von Schwerin has addressed this question for ancient Maya architecture when she began collaborating with Harvard University archaeologists to analyze the collapsed façade sculpture of an eighth-century temple at Copan, Honduras, called Temple 22 (Ahlfeldt 2004; Fash 2011b; von Schwerin 2011a). As an art historian, von Schwerin seeks to correlate political and social changes in ancient Maya kingdoms with developments in architectural form over space and time. But the first challenge is simply to bring together data on the temple that is spread around the world in various archives and museums and to determine how the building once appeared in the past. To test her reconstructions, von Schwerin turned to digital 3D tools. Different methods are possible for creating 3D models of ancient monuments—such as computer graphics, procedural modeling (models created from sets of rules), and reality-based modeling (models created from real-world data such as laser scanning)—and increasingly, these are being combined to create multi-resolution 3D reconstructions. Although this combination can expand research possibilities, it is critical to identify optional modeling techniques based on researcher needs and to define the workflow for dealing with multi-resolution models in a 3D WebGIS tool. The MayaArch3D project is addressing this by creating test data of multi-resolution 3D models from Copan, including various 3D simulations of Temple 22 (Remondino et al., 2009, von Schwerin et al., 2011b). The 3D models are being generated at different levels of detail (LoD) and resolutions ranging from individual buildings to archaeological complexes using methodologies based on image data acquired with passive sensors (e.g. digital cameras), range data acquired with active sensors (e.g. laser scanning), classical surveying, and procedural modeling using existing maps. The choice depends on the required accuracy, object dimensions and location, the surface characteristics, the team’s level of experience, the project’s budget, and the final goal. For example, computer-assisted design (CAD) models such as the 3D Studio Max model of Temple vative approaches to integrate GIS, 3D digital models, and VR environments online for teaching and research on ancient architecture and landscapes. It has grown into an international, interdisciplinary project that brings together art historians, archaeologists, and cultural resource managers with experts in remote sensing, photogrammetry, 3D modeling, and VR. The Start Up Phase was funded by two National Endowment for the Humanities, Digital Humanities Start-Up grants to the University of New Mexico (PI: Jennifer von Schwerin) and developed and beta tested a pipeline and prototype 3D WebGIS—called QueryArch3D. The prototype version is available at http://Mayaarch3d.org/project-history/). Project results indicate that it is possible to bridge the gap between 3D and GIS to create a resource for researchers of Maya architecture to compare and analyze 3D models and archaeological data in the context of a geographically referenced, VR landscape. 738 v o n S c h W e r I n e T a l . I n L i t e r a r y a n d L i n g u i s t i c c o m p u t i n g 28 (2013) 22 depicted in Figures 1 and 2 offers the ability to test hypothetical reconstructions and to analyze a building from multiple perspectives (e.g. bird’s eye, exterior versus interior view) with rotation or lighting effects (Figure 3).1 Reality-based models created using active and passive sensors allow for comparison against CAD reconstructions (Figs 4 and 5). VR such as this low-resolution SketchUp model of Copan’s landscape (Figure 6)—created using georeferenced building footprints—provides an urban context for high-resolution 3D models of individual structures and allows users to virtually navigate through ancient cities and landscapes and to increase their awareness of mass, space, and spatial relationships. This interaction facilitates a sense of embodiment and place (Forte and Bonini, 2010), and it also is useful for visualizing the results of archaeological research—for example, an affiliated project is working to display the results of archaeoastronomical studies at Copan (see Figure 10). These are just a few reasons that counter the common perception that 3D models are purely illustrative (e.g. Frischer and Dakouri-Hild, 2008). Increasingly, projects are demonstrating the value of 3D models for scientific analysis. Researchers developing tools for viewing and analyzing sophisticated 3D architectural models include the two big VR environment re-creation laboratories—the Experimental Technology Center at University of California, Los Angeles and the Institute for Advanced Technology in the Humanities at the University of Virginia, who have collaborated on the project ‘Rome Reborn’ (romereborn.frischerconsulting.com). In Europe, 3D models of architecture are used to analyze building plans and phases [for instance, the projects on Roman emperor palaces in Rome and Serbia (Weferling et al., 2001) and analyses of the Cologne cathedral (Schock-Werner et al., 2011)]. More recently, a few researchers have begun to explore how digital models might be used for comparative online research. One example is Stephen Murray’s “Mapping Gothic France” project—a collaborative project linking text, Quick Time VR, and 2D and 3D images to an interactive map of Gothic cathedrals. One promising opportunity—the approach taken by the MayaArch3D Project—is to use 3D models as visualization “containers” for different kinds of information (Manferdini et al., 2008). These recent advantages have initiated a broader interest in 3D modeling for archaeology and cultural heritage, which is evident at conferences such as CAA Figure 1. The 3D low-resolution CAD model of Temple 22 used for testing hypothetical reconstructions integrated with high-resolution reality-based 3D models of architectural sculpture (3D model created by F. Galezzi) Figure 2. Preliminary high-definition model of Temple 22 used to test the process of integrating various data sources into the reconstruction process. (3D model by R. Maqueda and J. von Schwerin) T h e M a y a a r c h 3D p r o j e c T : 3D W e b GIS f o r a n c I e n T a r c h I T e c T u r e a n D l a n D S c a p e S 739 (Computer Applications and Quantitative Methods in Archaeology), CIPA (International Committee for Documentation of Cultural Heritage), and the recently founded peer-reviewed journal Digital Applications in Archaeology and Cultural Heritage. 1.2 3D models in ancient American archaeology Most applications of 3D archaeology focus on archaeological sites in Europe or the Middle East; however, the acquisition of reality-based data for 3D models also is increasing for the archaeology of the ancient Americas (e.g. Reindel and Wagner, 2009; Lambers et al., 2007). As for current 3D projects that deal with the remains of the ancient Maya specifically, some are engaged with high-resolution scanning of individual sculptures for conservation and analysis and are considering ways to offer them online. These include Harvard University’s Corpus Project (Tokovinine and Fash, 2008; Fash 2011a, 2012), the MayaArch3D Project summarized here (see also Remondino et al., 2009), and the Mesoamerican Three-Dimensional Imaging Database (Doering and Collins, 2009) (http://www.famsi.org). Other web-based applications, like CyArk, use Google Earth and make point clouds available of whole Maya structures (http://archive.cyark.org). Meanwhile, some archaeological projects in the Maya area have published static maps on the web with links to still views of 3D reconstructions (http://www. papacweb.org/copan.html), whereas other projects such as the Palenque Map provide interactive maps with Quick Time VR panoramas as well as fly-throughs of 3D buildings (http://learningobjects.wesleyan.edu/palenque/explore/). The Maya Skies project has gone further to link 3D reconstructions and animations of buildings not only to a map but also to an archaeological database (http://Mayaskies.net). The La Blanca Project in Petén, Guatemala, an archaeological project carried out by the University of València, the Polytechnical University of València and the University of San Carlos in Guatemala since 2004 (http://www.uv.es/arsMaya), also has linked scanned data with simulated models of ancient buildings created in CAD programs such as 3DStudioMax or SketchUp and has built prototypes of online tools to analyze the 3D data and to display the results of excavations and hypothetical reconstructions. Such 3D visualizations obviously are effective ways to educate the public about Mayan cultural heritage. One can even now download apps of reconstructions of Maya temples (http:// www.Maya-3d.com) for use on mobile devices. In sum, 3D documentation is becoming a new standard for accurate, reality-based archaeological documentation, research, and visualization of results in Maya archaeology. Figure 3. Interior views of high-definition model of Temple 22 used to simulate lighting in the interior rooms. (3D model created by R. Maqueda) 740 v o n S c h W e r I n e T a l . I n L i t e r a r y a n d L i n g u i s t i c c o m p u t i n g 28 (2013) 1.3 Limitations of 3D models for archaeological and art historical inquiry The dissemination of these 3D data or products, however, still is limited due to developing countries’ limited access to hardware, software, and sufficient band-width. The 3D models therefore present challenges for enabling public access and longer term digital use/preservation (e.g. copyright issues or large files sizes that make them difficult to visualize via the web), and as a result they often only are published via 2D images in printed journals. Thus, most 3D models cannot be measured or compared with each other in any way, and it is difficult to share source models between users. Moreover, although powerful 3D visualization tools do exist, they implement either no or only limited query functionalities for data retrieval. Additionally, most 3D models themselves are not digitally linked to scientific data and not contextualized in their broader spatial and/or temporal context. These limitations become problematic when an art historian or archaeologist, for example, wants to analyze a temple within its urban context to understand its relationship to other temples, and changes in temple and urban design through time. To reveal spatial and temporal patterns, scholars need to be able to compare structures in both quantitative and qualitative ways and to analyze them within their larger spatial and temporal context, and along with their associated archaeological data (Robertson Figure 4. Results from unmanned helicopter flights over the East Court of Copan and Temple 22 to capture images for aerial photogrammetry. Right: Surface model with 5 cm resolution. Left: Orthophoto with 1 cm resolution. (Graphic: H. Eisenbeiss) Figure 5. Reality-based 3D model of Temple 22 generated from laserscan data. (Graphic: F. Remondino) T h e M a y a a r c h 3D p r o j e c T : 3D W e b GIS f o r a n c I e n T a r c h I T e c T u r e a n D l a n D S c a p e S 741 et al., 2006). For example, a research project comparing temples built >100 years at Copan and commissioned by three different rulers. These temples were part of an urban context, and surely their messages were intended to convey to a larger audience throughout the city; therefore, we need a tool that will examine the temples at multiple scales and perspectives and allow us to address questions such as: how did the messages change, or the intended audiences change, between the reigns’ of different rulers? How were temples, similar or different in their relationship to the natural landscape, or to the urban settlement at large? Specific methods of inquiry that such a tool could assist with would be: 1. Distribution (Figure 7): How did the distribution of freestanding monuments such as stelae in space and time between the reigns’ of different rulers? What were the spatial and temporal distribution of forms, symbolism, and texts? Do patterns exist between the content and spatial location (interior, exterior, lower story/upper stories, etc.) of motifs/glyphs on the temples that inform on message and audience? 2. Accessibility (Figure 8): Which residential groups had the easiest access to the temples? What were possible ritual procession routes between ceremonial sites, and what was their relationship to natural features in the landscape such as mountains or springs? What was the accessibility of ceremonial sites in comparison with residential sites and in relation to temples in the civic-ceremonial center as well as the natural landscape? 3. Visibility (Figure 9): What was the overall visibility of hypothesized civic-ceremonial sites (e.g. Group 8L-10), visual connections between civic-ceremonial sites as well as to which social groups they were most visible? Which temples were more visible from the elite residences to the East? How could visibility inform us about possible boundaries for ritual activities? 4. Orientation to the urban and natural landscape (Figure 10): What was the spatial alignment of temples in relation to (1) other ceremonial structures in the urban landscape and (2) mountain peaks and horizon markers in the natural landscape and what might this tell us about cosmological associations of space and place in ancient Copan? 1.4 Geographic information systems For these types of approaches, Geographic Information Systems (GIS)—linking map features to searchable databases—currently are better suited because they include queries as standard functions and allow for spatial and temporal analyses of relationships, patterns, and trends that are not evident when using traditional, non-spatial, databases (Lock 2000; Wheatley and Gillings, 2002; Conolly and Lake, 2006; Bodenhamer et al., 2010; Zerneke et al., 2006). Archaeologists began to use GIS in the 1980s to create, manage, and analyze geographically referenced information. For example, early archaeological research applications analyzed artifact distributions or predicted site locations. More recently, archaeologists have begun to perform visibility, accessibility, and network analyses in GIS to quantitatively explore the structure of ancient societies and the relationships between anthropogenic and natural phenomena. 1.5 GIS in Maya archaeology Maya archaeologists are using GIS in diverse ways. For example, to understand sites in a landscape context, archaeologists are combining remote sensing technologies (such as satellite imagery and airborne LIDAR) with GIS to discover new sites and offer new understandings of ancient Maya kingdoms such as at Caracol (Chase et al. 2011) and Figure 6. VR Environment—created using georeferenced building footprints in SketchUp—View facing northeast and overlooking Principal Group (SketchUp model by H. Richards-Rissetto) 742 v o n S c h W e r I n e T a l . I n L i t e r a r y a n d L i n g u i s t i c c o m p u t i n g 28 (2013) San Bartolo (Saturno et al. 2007). Researchers have applied GIS and aerial photos to predict site locations in the Yucatan peninsula (Podobnikar and Sprajc 2010). GIS also has been used for visibility studies to reconstruct site lines and identify intergroup connections and ancient political boundaries (Hammond and Tourtellot 1999; Richards-Rissetto 2010; Doyle et al. 2012). The only project that makes GIS data on Maya archaeology available online for research, however, is the Electronic Atlas on Ancient Maya Sites. This project uses GIS as a repository to store the locations of Maya archaeological sites and to create maps that overlay these sites on terrain, hydrology, or other features to illustrate polity size or political boundaries (http://Mayagis.smv.org/). One issue of concern is to what extent GIS data should be made available to the public, given the endemic looting that is significant at archaeological sites in Latin America. User management systems are useful in this way and can allow for password-protected access to sensitive data, particularly real-world coordinates or overlaying satellite imagery. The MayaArch3D Project has planned to institute five levels of user access ranging from most restricted access for the public to open access for internal researchers, and in this way can address concerns about looting. Figure 7. Distribution of Stelae for Rulers 12 and 13Copan GIS (Map by Heather Richards-Rissetto 2011) Figure 8. Accessibility: Western sacbe leading to Copan’s main civic-ceremonial complex (SketchUp model by H. Richards-Rissetto) T h e M a y a a r c h 3D p r o j e c T : 3D W e b GIS f o r a n c I e n T a r c h I T e c T u r e a n D l a n D S c a p e S 743 While Maya archaeologists currently use GIS, the ability to link GIS data to 3D models online would expand research possibilities dramatically. For example, Heather Richards-Rissetto created a GIS for Copan to study the visual and spatial relationships between built forms and natural landscape features (Figure 11). The GIS of Copan’s archaeological and topographical features covering over 24 km2 (Richards-Rissetto 2010, 2012) provides the data required to investigate the accessibility and visibility of different types of architecture at Copan and then relate these findings to possible levels of social interaction. Soon however, Richards-Rissetto realized that the 2D perspective of GIS maps limited her interpretations. For example, viewsheds calculated in GIS identified what could be seen from fixed vantage points at Copan, but it is not possible to ‘get on the ground’ to view the results from a pedestrian perspective (see Figure 8). Because GIS software was created to handle mainly terrain models (i.e. 2.5D data), it falls short when dealing with real 3D models (e.g. a building with interior). 1.6 State of the field in linking GIS and 3D models In the humanities, Geo-browsers, or ‘virtual globes’ (such as Google Earth, NASA’s World Wind, and ESRI’s ArcGIS Explorer) are the most common solution to ‘link’ GIS to 3D models. For example, the Digital Karnak Project, which traces the development of a temple precinct in Egypt from its origins as a local shrine to a powerful center, has a time slider that enables users to visualize changes (using 2D site plans) in the temple precinct throughFigure 9. Visibility: Viewshed of hypothesized civic-ceremonial group (8L-10) in Copan’s urban core, derived using GIS (Map by H. Richards-Rissetto) 744 v o n S c h W e r I n e T a l . I n L i t e r a r y a n d L i n g u i s t i c c o m p u t i n g 28 (2013) out its 3,000 year history (http://dlib.etc.ucla.edu/ projects/Karnak/google_earth). For users to track in 3D the construction phases of the Temple of Karnak through time, the system uses Google Earth’s time slider. However, more complex interactive queries are not implemented in Google Earth because this and other existing geo-browsers cannot query the 3D models against a database. It is not possible, for instance, to select all 3D models of structures in a city/site built between a certain time intervals, or planned by a certain architect/ruler. Finally, geo-browsers cannot visualize big and complex polygonal models. Because of these limitations, GIS and geo-browsers are not ideally suited to more recent approaches in archaeology and art history that are concerned with 3D space, such as performance studies, phenomenology and aesthetics, the relationship of architecture to the landscape, and archaeoastronomy. Some of the first experiments in 3D WebGIS include the Via Appia Antica Project—which developed a specific tool in Open Scene Graph (Forte et al., 2005) that integrated topographic landscapes with 3D architectural models in a VR environment to offer interactive virtual exploration and multiperspective experiences. There are some ‘3D GIS’ software products (such as ESRI’s CityEngine) that can rapidly build virtual cities, but they have two shortcomings: (1) they are based on procedural modeling; in other words, they create buildings with standard geometries and textures (not useful for studying aesthetics) (taking accurate measurements) and (2) they do not perform complex 3D spatial analyses. These systems cannot be used, for example, to model the aesthetic experience of ritual processions while simultaneously quantifying how far away a certain sculpture on a temple could be seen as people walked in this procession. This type of analysis still has to be done using separate GIS and 3D modeling systems. Given the state of the field summarized earlier in the text, our goal for the QueryArch3D tool is to combine the benefits of 3D visualizations with the analytical capabilities of a GIS to enable online realtime comparisons and analyses of multiple types of data. In other words, as one reviewer elegantly put it: “If these two distinct approaches to modeling reality could ‘talk to each other,’ one could do Figure 10. Orientation: View of eastern sky solar alignments in 8th Century, Copan—visualized using Stellarium Scenary3D Plugin and SketchUp (3D SketchUp model by H. Richards-Rissetto and G. Zotti) T h e M a y a a r c h 3D p r o j e c T : 3D W e b GIS f o r a n c I e n T a r c h I T e c T u r e a n D l a n D S c a p e S 745 research inside the 3D model (or its hosting environment).” To be able to perform interactive queries on high-resolution models and change parameters “on the fly” (e.g. restrict access to spaces based on gender or class) would significantly enhance research and education on ancient architecture and