<<

AUTHENTICITY IN DIGITAL ARCHAEOLOGICAL RECONSTRUCTIONS: A WORKFLOW PIPELINE AND DATA CLASSIFICATION SYSTEM TO INFORM AND VALIDATE THE DIGITAL RECONSTRUCTION PROCESS

Lazaros Kastanis Ba, MSc ()

Submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

School of Electrical Engineering and Computer Science Science and Engineering Faculty Queensland University of Technology 2019 Keywords

Anastylosis; ; Archaeological Reconstruction; Authenticity; Classification; Digital Reconstruction; Evidence; Extant Remains, Ground Truthing, Paradata; Transparency; Uncertainty; Virtual Environments; ; Visualisation; Workflow

ii Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process Abstract

Virtual reality (VR) and 3D modelling have been employed in archaeological reconstruction dating back to the early 1990’s. These technologies offer archaeologists the ability to objectively study reconstruction scenarios and alternative hypotheses without the need to physically interact with extant remains. With the advent of affordable computing hardware, VR modelling is increasingly being used in archaeological reconstructions to visualise lost ancient structures in highly realistic and evocative detail. This growth in digital reconstruction has been paralleled with a growing consensus recognising the dangers inherent in creating highly realistic outputs that are based on limited information and lack transparency of process. This thesis presents a workflow pipeline and data classification system for the digital reconstruction of archaeological structures, particularly, buildings. The workflow pipeline provides a framework for the digital reconstruction process. The classification system orders input data into classes of certainty. This thesis premises that the certainty of the inputs used in a digital reconstruction can be used as indicators of reconstruction authenticity. The workflow pipeline was applied to two complete reconstruction case studies. The classification system was applied to the two digital reconstructions and to two third- party reconstructions. The workflow pipeline has been demonstrated to be a robust and flexible framework for digital reconstruction allowing testing of unknowns through iteration. The classification system, when used with the workflow pipeline, has demonstrated its effectiveness in the reconstruction process as a data organisation tool and as a measure of reconstruction authenticity. Application of the classification system to third-party archaeological reconstructions highlights limitations which suggest that it has general applicability but may not be suitable in all cases. Table of Contents

Keywords ...... ii Abstract ...... iii Table of Contents ...... iv List of Figures ...... vi List of Tables ...... xi List of Abbreviations ...... xii Statement of Original Authorship ...... xiii Acknowledgements ...... xiv Chapter 1: Introduction ...... 1 1.1 Background ...... 1 1.2 Research problem ...... 3 1.3 Aims and Objectives ...... 5 1.4 Significance and Scope ...... 6 1.5 Thesis Outline ...... 7 2 Literature Review ...... 9 2.1 Introduction ...... 9 2.2 Digital Reconstruction of Heritage Architecture ...... 10 2.3 Uncertainty...... 12 2.4 The need for digital reconstruction and visualisation for Archaeology ...... 13 2.5 Case studies of digital reconstruction in Archaeology ...... 21 2.6 Problems associated with Digital Reconstruction in Archaeology ...... 43 2.7 Visualising Uncertainty ...... 47 2.8 Current Classification Systems for Visualising Uncertainty ...... 53 2.9 Summary ...... 59 3 Research Design ...... 65 This chapter is comprised of four sections. These are:...... 65 3.1 Research Design ...... 65 3.2 Classification System ...... 72 3.3 The Workflow Pipeline ...... 81 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case ...... 87 4.1 Application of the classification system to the Rose Theatre digital reconstruction ... 87 4.2 Digital Reconstruction of the Rose Theatre based on the Workflow Pipeline and Decision Tree ...... 96

iv Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case ...... 116 5.1 The (Komediehuset) Bergen Theatre Reconstruction ...... 116 6 Results - Application of the Visual Classification - Case Studies...... 131 6.1 Rose Theatre – Application of the Visual Classification ...... 131 6.2 Bergen Theatre – Application of the Visual Classification ...... 139 7 Results - Application of the visual classification system to external case studies ...... 152 7.1 School of gladiators, Carnuntum, Austria ...... 152 7.2 Temple of Apollo (Temple C), Kos, ...... 163 7.3 Discussion ...... 172 8 Discussion and Conclusions ...... 175 8.1 The Workflow Pipeline...... 175 8.2 The Classification System ...... 178 8.3 Contributions and Application ...... 188 8.4 Recommendations...... 188 8.5 Conclusion ...... 190 Bibliography ...... 193 Appendices ...... 203 Data Source acknowledgement ...... 203 Appendices ...... 204 Appendix A Tables ...... 204 Appendices ...... 215 Appendix B Figures ...... 215

List of Figures

Figure 2.1: The extent of archaeological material remains of the Carnuntum site. From Neubauer et al., 2014 (Wolfgang Neubauer et al. 2014, p. 177)...... 22 Figure 2.2: Virtual reconstruction of the Catnuntum ludus, viewed from the south, © Michael Klein. From (W. Neubauer et al. 2014, p. 2)...... 23 Figure 2.3: Examples of contemporary churches used as data sources for the reconstruction. (Gasparini et al. 2012, p.125) ...... 26 Figure 2.4: Examples of photography, urban documentation and archaeology records used in the reconstruction process. (Gasparini et al. 2012, p. 217)...... 26 Figure 2.5: Roof reconstruction based on similar architecture examples. (Gasparini et al. 2012, p. 222) ...... 27 Figure 2.6: Final reconstructed virtual model of the San Jacinto Church, Caracas, Venezuela. (Gasparini et al. 2012, p. 223)...... 28 Figure 2.7: Fragment restoration from the destroyed church. (Petrova et al. 2011, p. 391) ...... 29 Figure 2.8: Monochrome image of the wall, completed with pictures from archival photographs. (Petrova et al. 2011, p.392)...... 30 Figure 2.9: The reconstructed frescoes of the western wall based on colours extracted from the original fragments.(Petrova et al. 2011, p. 393)...... 30 Figure 2.10: Wireframe and rendered 3D model of the Yanqing Section of the Great Wall. (Issini 2012, p. 525)...... 32 Figure 2.11: The visualisation 'Home Page' The 3D model is displayed in association with information access tools. (Issini 2012, p. 526)...... 33 Figure 2.12: Examples of metadata available through the information visualisation system. (Issini 2012, p. 526)...... 33 Figure 2.13: Perspective view of the model of the Villa of the Papyri, note the use of colour.(Zarmakoupi, 2010) ...... 35 Figure 2.14: A schematic diagram of the 3D modelling stages and influences for the Villa of the Papyri as presented by Hermon 2008. (Hermon 2008, p. 38) ...... 35 Figure 2.15: A view from the south-west of a room within the Villa of the Papyri showing the, (a) existing state and (b) with a embedded restoration proposal. (Zarmakoupi, 2010) ...... 37 Figure 2.16: Pavilions in the 1x1 & 16x16 configurations (Das & Garg 2011, p. 11) ...... 39 Figure 2.17: A Preliminary Reconstruction of the lost central structure of Sompur Mahavihara, (Rashid et al. 2010, p. 32) ...... 40

vi Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process Figure 2.18: The framework of knowledge for the reconstruction, (Rashid et al. 2010, p. 32) ...... 41 Figure 2.19: High-level overview of the steps in the virtual reconstruction of the Octogon. (Thuswaldner et al. 2009, p. 6) ...... 42 Figure 2.20: High quality render of two different scenarios for a possible physical anastylosis of the Octagon. White indicates infered fill in stonework. (Thuswaldner et al. 2009, p. 22)...... 42 Figure 2.21: A study of the visual impact of the Octagon anastylosis at its prominent spot in the Curetes street in Ephesos. (Thuswaldner et al. 2009, p. 22) ...... 43 Figure 2.22: The Pool scene at the start of the Palace Midas level, Tomb Raider 1996, (Kelly, 2013) ...... 44 Figure 2.23: Reborn, an aerial view of the city centre. (Frisher Consulting, 2013) ...... 45 Figure 2.24: Rome had many small private bathing establishments. (Frisher Consulting, 2013) ...... 45 Figure 2.25: An aerial view of the Flavian Amphitheater ("Colosseum"), (Frisher Consulting, 2013) ...... 46 Figure 2.26: The plaza on the west side of the Flavian Amphitheater, (Frisher Consulting, 2013) ...... 46 Figure 2.27: This visualisation pipeline shows the introduction of data uncertainty from models and measurements. (Pang et al. 1997, p. 372) ..... 48 Figure 2.28: Archaeological reconstruction interface showing the animated time line (1), and Uncertainty cues. From a to c: no cues, rising/sinking cue, wireframe, and transparency. (Zuk et al. 2005, p. 104) ...... 50 Figure 2.29: Rendering of the reconstructed 1790 historical centre of the city of Cracow (Poland) showing earlier and later time period buildings (Dudek & Blaise, 2004, p. 14) ...... 51 Figure 2.30: Rendering of the reconstructed 1570 historical centre of the city of Cracow (Poland) showing input sources not yet analysed (circled) and translucent features resulting derived from interpretation (Dudek & Blaise, 2004, p. 15) ...... 51 Figure 2.31: Interactive ‘site plan’. Use of a time-slider to view reconstruction by era and ‘clickable’ objects allowing access to the ‘evidence options’ available for each reconstruction (Bonde et al., 2009, p. 367) ...... 52 Figure 2.32: Gershon's High-level taxonomy of the causes of imperfect knowledge of the information state (imperfection). (Gershon 1998, p .43) ...... 56 Figure 2.33: Credibility and Disagreement uncertainty span levels and sometimes disagreement results in credibility uncertainty. (Skeels et al. 2010, p. 75) ...... 57 Figure 2.34: Level of uncertainty awareness: (1) Known Knowns: what you know; (2) Unknown Knowns: what you need to find out; and (3) Unidentified Unknowns: what you don't even know you need to find out. Participants described the unidentified unknowns, or unrecognized information needs, as worst case. (Skeels et al. 2010, p. 76) ...... 58 Figure 3.1: Summary Stages - Research Design ...... 68 Figure 3.2: Decision tree detailing the process of assigning data a class of uncertainty...... 74 Figure 3.3: Modified Decision Tree ...... 80 Figure 3.4: Flowchart detailing the reconstruction process and identifying areas of inputs and associated uncertainty classes ...... 83 Figure 3.5: Revised flowchart detailing the reconstruction process and identifying areas of inputs and associated uncertainty classes ...... 86 Figure 4.1: Modified Decision Tree ...... 93 Figure 4.2: Figure 2: Base modelling shown with mapped archaeology and interpreted building footprint (Bowsher & Miller, 2009, p. 56,125)...... 97 Figure 4.3: Reconstruction of building foundations based on archaeological data and using the proposed workflow pipeline ...... 98 Figure 4.4: Derived building framework showing Sill beams, Joists, and main supporting pillars ...... 100 Figure 4.5: Derived building framework showing roofing framework, flooring and thatch roofing material ...... 102 Figure 4.6: The Rose (with circular roof, lower right area), ca 1600 From Norden's Panorama “Civitas Londini” (Source: http://www.luminarium.org/encyclopedia/nordenbankside.jpg) ...... 103 Figure 4.7: View of London, Claes Visscher 1616 (Source: https://en.wikipedia.org/wiki/File:London_panorama,_1616b.jpg) ...... 103 Figure 4.8: Proposed building entrance showing the narrow egress ...... 104 Figure 4.9 Postulated building entrance position (Bowsher & Miller, 2009, p. 125)...... 104 Figure 4.10: Exterior of the Rose Theatre showing Lath infill walls ...... 105 Figure 4.11: View of the Rose Theatre excavations in 1989. (Source Bowsher & Miller, 2009) ...... 106 Figure 4.12: Reconstructed Stage and underlying framework ...... 106 Figure 4.13: Top view showing the stage offset ...... 107 Figure 4.14: Position of Columns relative to the stage shape ...... 107 Figure 4.15: The stage canopy roof showing the interpreted covering and underlying framework ...... 108 Figure 4.16: The Tiring House bays ...... 109

viii Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process Figure 4.17: The Tiring House interiors showing stage access and the side walls...... 109 Figure 4.18: Proposed rear 'Tiring house' stairs and internal general access stairways relative to stage and yard ...... 111 Figure 4.19: Reconstructed railing and balusters. Unearthed baluster from the Rose excavation (Bowsher & Miller 2009)...... 112 Figure 4.20: Typical Doorways used in the reconstruction ...... 113 Figure 5.1: Original elevation (left) and plan (right) Blix drawings of 1870 renovation in the Bergen Theatre. Courtesy of Holledge et.al...... 118 Figure 5.2: Example original 1938 of the Bergen Theatre showing elevation and decorative details. Courtesy of Holledge et.al...... 118 Figure 5.3: Interior and exterior walls reconstructed based on 1938 plan survey drawing...... 120 Figure 5.4: Auditorium floor and gallery with supporting columns ...... 121 Figure 5.5: Detail of roof truss system...... 122 Figure 5.6: Completed external features of the Bergen Theatre: ...... 122 Figure 5.7: Example comparison of door designs between 1938 official survey drawing (left) and Blix's renovation drawings of 1870 (right). Courtesy of Holledge et.al...... 123 Figure 5.8: View of the auditorium showing internal doors...... 123 Figure 5.9: Comparison of doorways, the pre-1870 seating plan (Top), a Peter Blix renovation drawing (Middle) and the official 1938 Survey drawing (Bottom). Doorways outlined in red. Courtesy of Holledge et.al...... 124 Figure 5.10: Internal stair detail...... 125 Figure 5.11: View showing reconstructed Orchestra Pit and Stage ...... 125 Figure 5.12: The reconstructed Gallery and Wings...... 127 Figure 5.13: Side Elevation showing the chandelier recess in the auditorium ceiling. Courtesy of Holledge et.al...... 128 Figure 5.14: Reconstructed ceiling showing chandelier recess and low detail ceiling relief...... 128 Figure 5.15: Detail of the Orchestra Pit Reconstruction ...... 129 Figure 5.16: Render showing internal fittings...... 130 Figure 6.1: The exterior of the reconstructed Rose Theatre textured (top) and classified (bottom) ...... 134 Figure 6.2: The exterior entrance of the reconstructed Rose Theatre textured (top) and classified (bottom) ...... 135 Figure 6.3: The Rose Theatre Stage textured (Top) and classified (Bottom) ...... 136 Figure 6.4: Rose Theatre classified Stage frame and Foundations exposed ...... 137 Figure 6.5: The Rose Theatre Stair system textured (Top) and classified (Bottom) ...... 138 Figure 6.6: Bergen Theatre external reconstruction showing the unclassified (top) and classified (bottom) views...... 140 Figure 6.7: Bergen Theatre internal stairway reconstruction showing the unclassified (top) and classified (bottom) views ...... 141 Figure 6.8: Bergen Theatre orchestra pit reconstruction showing the unclassified (top) and classified (bottom) views ...... 142 Figure 6.9: Bergen Theatre stage reconstruction showing the unclassified (top) and classified (bottom) views ...... 143 Figure 6.10: Bergen Theatre gallery reconstruction showing the unclassified (top) and classified (bottom) views ...... 145 Figure 6.11: Bergen Theatre auditorium reconstruction showing the unclassified (top) and classified (bottom) views ...... 147 Figure 6.12: Bergen Theatre internal ceiling reconstruction showing the unclassified (top) and classified (bottom) views ...... 148 Figure 6.13: Bergen Theatre seating reconstruction showing the unclassified (top) and classified (bottom) views ...... 149 Figure 6.14: Bergen Theatre internal fixtures reconstruction showing the unclassified (top) and classified (bottom) views ...... 150 Figure 7.1: Perspective view of the reconstructed Gladiator school (ludus) showing the interpreted layout. Source: (W. Neubauer et al., 2014, p. 2) ...... 155 Figure 7.2: Figure 2: Application of the classification to the Gladiator School image as presented. Original image source: (W. Neubauer et al., 2014, p. 2) ...... 156 Figure 7.3: Proposed position of extant architectural elements of Temple C as interpreted by De Mattia 2009. Source: De Mattia, 2009, P. 79 ...... 166 Figure 7.4: Partial reconstruction and fill in pieces (dark grey) of the south-east and north-east corners of Temple C as proposed by De Mattia. Source: De Mattia, 2009, P. 79 ...... 167 Figure 7.5: Reconstruction of the north-east corner of Temple C showing position of columns, entablature and pediment components. Source: De Mattia, 2009, P. 77 ...... 168 Figure 7.6: Final reconstruction of Temple C as proposed by De Mattia 2009. Source: De Mattia 2009, p. 74 ...... 168 Figure 7.7: Application of the classification to Temple C images as presented. Original image source: (D. De Mattia, 2009, p. 74, P. 79) ...... 170

x Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process List of Tables

Table 2.1: Key guidelines from the London Charter for the Computer-Based Visualisation of , (Beacham et al., 2009) ...... 19 Table 2.2: Methodology for the Digital Reconstruction of theoretical pavilions.(Das and Garg, 2011), p.4 ...... 38 Table 2.3: The Uncertainty Matrix proposed by (Walker et al. 2003, p. 15) ...... 55 Table 3.1: Uncertainty classification matrix ...... 76 Table 6.1: Summary of the colour classification system ...... 132

List of Abbreviations

3D: Three Dimensional

AIS: Airborne Imaging Spectroscopy

ALS: Airborne Laser Spectroscopy

EMI: Electromagnetic Induction

GPR: Ground Penetrating Radar

VR: Virtual Reality

xii Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process Statement of Original Authorship

The work contained in this thesis has not been previously submitted to meet requirements for an award at this or any other higher institution. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made.

Signature: QUT Verified Signature

Date: 14/08/2019 Acknowledgements

Firstly, I would like to acknowledge the support and direction provided to me by my Principal Supervisor Daniel Johnson. His guidance and encouragement were crucial in completing this body of work. Daniel was the beacon of calm logic I needed to get to the end. I will be forever grateful for your steadfast support.

I would also like to acknowledge Michael Docherty, my first principal supervisor, who was there at the beginning and provided me with the support, direction and critical review that the early stages of study required.

I would like to express my deep gratitude to Joanne Tompkins who has supported me throughout this study. Her knowledge of theatre and theatre practise has been invaluable and instrumental in the completion of this work.

I am forever grateful to the team at Ortelia Interactive and particularly Darren Pack who have been so understanding and supportive of my research. Your contributions to my research and material support will always be remembered.

Finally, I would like to thank my family for their undying support. To my wife Laura, thank you for hanging in there with me. To my girls Pelagia and Melina, thanks for putting up with a sometimes frustrated and distracted dad.

xiv Authenticity in Digital Archaeological Reconstructions: A workflow pipeline and data classification system to inform and validate the digital reconstruction process

Chapter 1: Introduction

1.1 BACKGROUND Virtual Reality modelling (hereafter VR) and 3D modelling techniques are increasingly being employed as tools for the digital reconstruction and investigation of archaeological structures and sites. Researchers are turning to these technologies to test theories of past built form of lost buildings. The ability to digitally recreate a structure and explore it, facilitates the examination of more complex concepts such as the social conditions of the occupants of these past structures. Using VR and 3D modelling techniques in the context of archaeological reconstructions raises ‘the possibility of rediscovering not only the appearance of historic displays, but also some of their multiple, shifting, and contingent meanings’ (Roach, 2016, p. 2). Pierdicca et al. point out that detailed archaeological reconstruction models offer additional information to archaeological research and improve the dissemination of ‘cultural heritage artefacts’ to the wider public (Pierdicca, Frontoni, Malinverni, Colosi, & Orazi, 2016). Addison et al. note that the advent of VR technologies also means that there now exists the ability to create historical at a level not previously possible. (Addison, 2000)

Although there is a rapidly growing use of VR modelling and visualisation techniques in the field of archaeological reconstructions, as evidenced by the plethora of ' virtual Pompeii’s', many questions still remain with regards the authenticity of these virtual reconstructions (Addison, 2000, p. 25). The validity of the final reconstructions and transparency of the reconstruction process are being questioned. Researchers such as Beacham and Pletinckx et al. identify a significant issue of credibility in virtual archaeology within the scientific community (R. C. Beacham, 2011; Pletinckx & Tartessos, 2011). Roach states, when discussing digital reconstructions that they ‘contain great potential for distortion’ (Roach, 2016, p. 2). There are currently no formalised methodologies or guidelines for archaeological virtual reconstruction and, in general, no consistent approach to the use of such tools. This is of significance when considering that virtual reconstructions are often based on sparse and highly variable data sources to arrive at a final solution. Methodological rigour has been called into question when employing these data inputs (Pletinckx &

Chapter 1: Introduction 1

Tartessos, 2011). The sparseness of the data and the level of assumption required to completely reconstruct material remains adds a high level of uncertainty to the resulting reconstructions. This uncertainty is not visible to the viewer of these visualisations. There is also a generally held belief among researchers in the field that there is an tendency to employ 3D digital reconstructions to 'establish a photorealistic re-creation of physical while ignoring more fundamental issues of spatiality and use of space.' (Dunn & Woolford, 2012, p. 172). Davis et al. supports this assertion stating that many cultural heritage practitioners are in ‘awe’ of technology and the ‘lure’ of the ‘ease with which it can bring archaeology to life’ (Davis, Belton, Helmholz, Bourke, & McDonald, 2017, p. 1)

To overcome these perceived deficiencies, there have been efforts to standardise digital reconstruction methodologies and rigour. The London Charter for the Computer-based Visualisation of Cultural Heritage 2006 is one such attempt. This charter, and others, provide non-binding guidelines for the authentic and transparent virtual reconstruction of archaeological material remains. The guidelines do not go as far as to propose methodologies, techniques, and ontology with which to adhere to the guidelines. A 2013 survey of 686 publications or reports by Cerato and Pescarin detailing a virtual archaeological reconstruction revealed that only 1% addressed the issues of validating outputs (Cerato & Pescarin, 2013). This survey demonstrates the inherent problem that the guidelines remain only partially effective and are inconsistently applied.

There has been significant research carried out in recent times into the visual representation of uncertainty in scientific visualisation as evidenced in work by Pang et al. (1997) and Skeels et al. (2011). This research has been primarily concentrated on domains such as geovisualisation, fluid dynamics, intelligence gathering and the like. There has been no significant research into the display of uncertainty associated with an archaeological reconstruction. In addition to the lack of visual representation of uncertainty, there has been no attempt to form a classification system to deal with the disparate and sparse data available to the archaeologist when reconstructing material remains. The high level of uncertainty in archaeological data therefore makes the accurate prediction of how a structure looked, based on material remains alone, very difficult and subject to conjecture. There has been very little empirical research into

2 Chapter 1: Introduction

how to transfer the outcomes of these classifications to the visualisation environments to the field of virtual archaeological reconstructions.

1.2 RESEARCH PROBLEM This program of research has its basis in practical application. In 2007 the author of this thesis was engaged to digitally reconstruct the 1596 version of the Rose Theatre originally located on the West Bank of the river Thames in London (Kastanis, Pack, & Tompkins, 2007). The purpose of the reconstruction was for use in the research of theatre practices of the period and their relationship to space (Tompkins & Kastanis, 2017). In order to reconstruct the theatre for the specific time period a vast, and complex, body of data was collected. This data ranged from accurate archaeological records of the excavated Rose Theatre site, to 1600’s era sketches of the general area, analogous structure examples and literary texts. It became quickly apparent that the input data required classification to facilitate a logical and rigorous approach to its use in the reconstruction. The reconstruction process became highly iterative as less uncertain data inputs where introduced and which required extensive testing. This iterative approach and in-depth consideration of the inputs has spawned a workflow pipeline that is tailored to virtual archaeological reconstructions.

Many decisions in the reconstruction were based on the interpretation of evidence that varied greatly in its certainty. Data inputs that lacked a high degree of certainty was effectively ‘hidden’ from the user. In this case the users were the researchers of theatre practise. It became necessary to show the researchers what inputs were used to arrive at a building feature which facilitated decisions regarding the veracity of the element in question, and subsequently, the impact on the research. This study aims to address this lack of transparency and to provide a method whereby users can assess the veracity of a reconstruction by access to the underlying inputs. The study extends to two types of user, Primary and Secondary. Primary users are professional users such as researchers’, archaeologists, and specialists to name a few. Secondary users are lay users such as the general public who may have an interest in an archaeological reconstruction.

The following quotation from Collingwood aptly summarises the inherent underlying problem facing researchers when creating virtual reconstructions of cultural material remains. Although this statement was made in 1926 the issue of uncertainty and conjecture in the historical context has remained, as relevant, to the

Chapter 1: Introduction 3

present day. A digital reconstruction must still deal with uncertain data and the inherent conjecture that accompanies it.

'I have found in my historical inquiries that I can never determine the exact truth about any historical fact, but have to be content with an account containing a large and unverifiable amount of what I know to be conjecture."(Collingwood, 1926, p. 146).

Many studies have been undertaken to isolate and classify uncertainty in the visualisation pipeline. In fields such as fluid dynamics and geovisualisation, the identification of uncertainty in data is regularly addressed. The types of uncertainties being dealt with in these fields are more generally able to be statistically ascertained and therefore the resultant values can be transferred to a visual metaphor (Skeels, Lee, Smith, & Robertson, 2010; Wittenbrink, Pang, & Lodha, 1996). In order to quantify uncertainty, a classification scheme is required which allows users to create classes based on the source of the uncertainty in question (Wittenbrink et al., 1996). To date, apart from early work (Hynst, Gervautz, Grabner, & Schindler, 2001; Niccolucci & Hermon, 2004) there has been no significant research into the classification and visualisation of uncertainty in the field of virtual archaeology. Sorin (2008) argues that the application of VR technologies has not ' been fully adapted to the archaeological reasoning process' (Sorin. Hermon, 2008, p. 42). There is a clear need to make these visualisations more transparent to facilitate the viewer’s ability to make an informed assessment of the veracity of what is being presented in a virtual reconstruction.

The primary problem with the creation of a classification system and subsequent uncertainty visualisation is the complexity of data and observations gathered in the process of researching and uncovering an archaeological site (Pletinckx & Tartessos, 2011). Almeida et al. highlight that material remains are documented based on their past functions when much of the information relating to technical properties and past function is not available, or lost in . (Almeida & Barceló, 2012, p. 78).

In the case of an archaeological excavation there are established methodologies for the unearthing and recording of material remains. (Renfrew & Bahn, 2012) Current archaeological practise employs highly procedural approaches to recording data. In particular, positional data and associated metadata is usually of a very high accuracy (Renfrew & Bahn, 2012). In addition to highly accurate site data, a variety of other sources are also used in the reconstruction process. These sources may include, but are not limited to, information derived from ancient texts, examples of other unearthed

4 Chapter 1: Introduction

remains and analysis of historic events (S. Hermon, 2011). In all cases the researcher is relying on information with varying levels of uncertainty ranging from little to very high. The reconstruction of material remains is, in essence, a theoretical construct. In the case of most material remains, different interpretations of the available data will lead to different reconstruction outcomes. (S. Hermon, 2011) The remains of material culture are interpreted from three basic perspectives, material, social and ideational (S. Hermon, 2011). These perspectives raise the possibility of multiple possible interpretations of the data. The data sources used to arrive at a reconstruction of a lost structure ‘has always required conjecture, which is inevitably coloured by the translator’s worldview.’(Roach, 2016, p. 2).

Given that the data employed in the digital reconstruction of material remains is subject to varying degrees of uncertainty and that this uncertainty has a profound impact on the final reconstruction there is a defined need to:

• Isolate the type of uncertainty inherent in digital reconstruction

• Assign a value to uncertainty

• Visualise uncertainty within the virtual reconstruction

• Improve transparency in the archaeological reconstruction process

1.3 AIMS AND OBJECTIVES This study builds on recent efforts in both data uncertainty classification and, 3D modelling pipelines. The objective is to develop an uncertainty classification system that will allow researchers to identify and display the influences that lead to a final 3D reconstruction. The conversion of these ‘influences’ into a visual paradigm facilitates a viewer’s assessment of authenticity by providing access to the inputs used in the virtual reconstruction. It is the aim of this research to attempt to address the issues of authenticity and transparency in digital archaeological reconstruction, or, digital anastylosis, through the development of a framework that facilitates methodological rigour and promotes serious interpretive discussion. The research is divided into two distinct sections.

I. The development of a workflow pipeline for digital archaeological reconstruction II. The development of an uncertainty classification and visual display system specific to digital archaeological reconstruction.

Chapter 1: Introduction 5

Workflow Pipeline A key objective of this thesis is the development of a workflow pipeline for digital archaeological reconstruction. A workflow pipeline offers the possibility of addressing the issues of methodological rigour in virtual reconstructions identified in the literature. This study aims to build on the work of Hermon who identified the modelling stages of the Villa of Papyri reconstruction and the ‘influences’ on these stages (Sorin. Hermon, 2008). A key outcome of this thesis is to develop a repeatable and robust methodology for the reconstruction of archaeological structures.

Classification System This study aims to develop and test a data input classification system tailored to digital archaeological reconstruction of mass structures such as buildings. The classification system design aims to sort data inputs into classes of uncertainty. A basic premise of this approach is that classes of input data type and their corresponding levels of uncertainty can be used as measures or indicators of reconstruction authenticity. The objective of this classification system is to display an archaeological reconstruction as a classified output which facilitates a viewer’s ability to assess reconstruction authenticity.

1.4 SIGNIFICANCE AND SCOPE As discussed briefly above, and in more detail in chapter two, VR modelling tools and practises are increasingly being adopted for archaeological reconstruction. This increase has raised serious concerns with regards transparency of the reconstruction process, reconstruction authenticity and the ‘potential for distortion’ (Roach, 2016, p. 2). Virtual reality modelling and associated virtual environments represent powerful tools for communication, research and discussion in the field of archaeology. As these tools and the hardware that supports them becomes more ubiquitous, more virtual reconstructions will be produced and made available to the public. Currently viewers of a virtual reconstruction have no way to assess the accuracy or authenticity. The high production values associated with some detailed digital reconstructions may incorrectly give viewers an assumption of certainty regarding the reconstruction that has no obvious basis. The viewer relies solely on what is presented to them. Several attempts have been made to provide guidelines to reinforce the underpinning accuracy for virtual reconstruction as evidenced by the London charter.

6 Chapter 1: Introduction

This study seeks to take the first steps in developing a repeatable workflow pipeline and classification system for archaeological reconstruction. It is restricted to the investigation of mass structures such as buildings, as such, the scope does not extend to other areas of archaeological reconstruction. It aims to produce a visual output that can be users for interpretation but does not extend to the testing of the outputs with user groups. The primary focus of the study is the development of an integrated classification system and workflow pipeline that can be implemented for the entire reconstruction process.

1.5 THESIS OUTLINE Chapter 2: Literature Review provides a background to the relevant research in the disciplines of digital archaeological reconstruction and data uncertainty classification. The literature review begins with an in-depth review of virtual archaeological reconstructions and then delves into the current approaches to data uncertainty classification systems. Together the two form the theoretical approach to this study and inform the direction of the research Chapter 3: Methodology details the research progression adopted for this thesis. The chapter begins with research theory and scope of the study. The main research questions are posed, the research aims, and stages are identified. The initial approaches to the workflow pipeline and classification system are described and identified issues highlighted. Modifications to the workflow pipeline and classification system addressing these issues are briefly discussed. Chapter 4: Application of the Classification System and Workflow Pipeline details the initial application of the developed workflow pipeline and classification system to the Rose Theatre reconstruction. The reconstruction of the theatre using the workflow pipeline and classified data is discussed in detail. The results of the initial system are discussed in details and the issues highlighted. A revised classification system and workflow pipeline are discussed in detail. Significant issues were identified with the classification system. Modifications are proposed, and implemented with the Rose Theatre reconstruction, to address the issues identified. The results are discussed Chapter 5: Application of the modified Workflow Pipeline - Komediehuset, Bergen Theatre reconstruction details the reconstruction of the Komediehuset, Bergen Theatre employing the modified workflow pipeline. The reconstruction process is

Chapter 1: Introduction 7

described in detail and the resultant output discussed. Minor revisions are proposed for the workflow pipeline. Chapter 6: Application of the Visual Classification - Case Studies explores the application of the updated classification system to a series of case studies. The classification is applied to the Rose and Bergen Theatre reconstructions and the results discussed. The classification system is applied in the context of the workflow pipeline. The results and limitations of the application of the classification system to the two case studies are discussed. Chapter 7: Results - Application of the visual classification system - External Case Studies details and discusses the application of the classification system to virtual archaeological reconstructions by third-party researchers. The classification system is applied to two third-party reconstructions to test the efficacy of the system separate to the workflow pipeline. The results of the two case studies and discussed and the issues identified. Chapter 8: Discussion and Conclusions provides a detailed summary of the research. The results of the workflow pipeline and classification in the context of the case studies are discussed at length. The main discussion addresses the applicability of the research to the discipline as well as the limitations of the developed methodology. Recommendations for future research are proposed to address the issues identified and the contribution of this work to archaeological reconstruction is discussed.

8 Chapter 1: Introduction

2 Literature Review

2.1 INTRODUCTION

Virtual Reality (hereafter VR) and associated modelling techniques have been in existence since the 1960s and are also referred to as "virtual environments," "synthetic environments," "virtual worlds," and "artificial reality" (Bowen Loftin, R., Chen, J.X., Rosenblum & Loftin, 2005). Virtual Reality has proven to be a powerful tool that allows users to interact with three-dimensional data in both a static and dynamic manner, that is, "it inherently provides those who use it with the means to navigate and manipulate the information that is displayed” (Bowen Loftin, R., Chen, J.X., Rosenblum & Loftin, 2005, p. 479). VR can be used to facilitate a visualisation process that "primarily aims to amplify human cognition with different options in order to facilitate data meaning associations and to extend the interpretative or de-codifying skills of users" (Hermon & Kalisperis 2011, p. 60). This process can be summarised as "transforming data, information, and knowledge into visual form" that assist users in "getting an insight of abstract data values" (Hermon & Kalisperis 2011, p. 60). Hermon et al., conclude that ''the better the visual tool, the better the explanation and the interception' (Hermon & Kalisperis 2011, p60). Therefore, VR provides a visual framework in which objects, spaces and data are displayed and manipulated (S. Hermon & Kalisperis, 2011).

Virtual reality and 3D modelling techniques are increasingly being employed in archaeological virtual reconstruction and have a history of doing so dating back to the early 1990’s. The term ‘Virtual Reconstruction’ is often confused with the more simplistic process of mesh reconstruction arising from digitally acquired data. Virtual reconstruction is a set of steps which includes ‘the documenting, interpretation and visualisation of lost archaeological contexts.’(Demetrescu, 2015, p. 43). For this research, the term 'reconstruction' or 'Anastylosis' only applies to mass structures such as buildings and settlements. Addison et al. state, “Digital tools and techniques now emerging from academic, government, and industry labs offer new hope to the often painstakingly 3D documentation (everything from site surveys to epigraphy), complex tasks of archaeology, surveying, historic research, conservation, and education.” (Addison, 2000, p. 22). Addison is not alone in this view. Pollefeys et al., take the

2 Literature Review 9

view that "Virtual reality is a technology that offers promising perspectives for archaeologists."(Pollefeys et al., 1998, p. 1).

Researches such as Almagro point out that 'virtual reconstruction of buildings that have been destroyed or severely changed, carried out using computer graphics, enables the analysis of such important aspects as visual perception, character given to space by colour and material textures, light effects' (Almagro, 2007, p. 1). Furthermore, using VR reconstruction techniques allows the objective study of ‘different alternatives or hypotheses without having to make physical alterations to the original elements.' (Almagro 2007, p. 1). VR reconstruction can therefore also be viewed as ‘knowledge representations, that is, visual representations of the state of scholarly knowledge.’(Johanson 2009, p. 413).

2.2 DIGITAL RECONSTRUCTION OF HERITAGE ARCHITECTURE The digital reconstruction of archaeological evidence is now becoming common practise and is evidenced by many examples. These range from the physical restoration projects such as Villa of the Papyri project (Zarmakoupi 2010), to digital reconstruction of an 16th century theatre on the London South Bank (Tompkins & Kastanis, 2017). Digital reconstruction is being employed in archaeological reconstructions from many time periods. In Malaysia a significant project has been undertaken to recreate the 'A Formosa Fortress' constructed in 1511 and demolished in 1824, (M. I. Zainal Abidina, A. Bridges, 2008). Conversely, a virtual reconstruction of material remains of the Octagon in Ephesos, , is assisting researchers to piece together the fragments for physical reconstruction (Anastolysis) of the Octagon (Thuswaldner et al. 2009). The Qumran Digital Model project used VR reconstruction to question the reconstructions of the 'Qumran-Essene’ hypothesis as it was proposed in 1956 by de Vaux (Cargill, 2009). De Vaux argued that the original Khirbet Qumran inhabitants where responsible for the Dead Sea Scrolls which were discovered in the surrounding hinterland. De Vaux was criticized for relying primarily on the texts for his archaeological interpretation of the site (Cargill, 2009). Cargill’s virtual reconstructions of the Khirbet Qumran identified ‘flaws’ in de Vaux hypothesis whereby a number of his ‘archaeological assumptions did not hold up in the modelling process’ (Cargill, 2009, p. 34).

Along with the growth of digital reconstruction there is a growing consensus recognising the dangers inherent in any reconstruction project with limited information

10 2 Literature Review

and the subsequent presentation. The early adopters of virtual reality ‘were in awe of their ability to build large and accurate models’ but tended to ignore that these three- dimensional reconstructions were, in fact, ‘an instance of scholarship’ and this is generally not conveyed in the final outputs (Johanson 2009, p. 406). Denard, when examining the issue of digital reconstruction states, that if a virtual reconstruction 'lacks an account of the evaluation of sources or of the process of interpretation, does not, in itself, render the research process visible to the visitor and thus fails to allow the viewer to assess it as part of an argument.' (Denard 2011, p.60). Denard’s point is at the core of this thesis. A viewer of a virtual reconstruction (be they an archaeologist or a lay user) should have the ability, and accessibility, to assess the inputs used to arrive at the final reconstruction. In doing so the user may arrive at their own assessment of the authenticity of a reconstruction.

The issues faced in physical reconstruction of archaeological ruins, are, in many cases, similar and transferable to digital reconstructions. The 2013 UNESCO World Heritage Operational Guidelines state, reconstruction is ‘acceptable only on the basis of complete and detailed documentation and to no extent on conjecture' (UNESCO 2013, p. 22). In the case of archaeology, excavated material remains alone cannot provide 'complete and detailed documentation' (Stanley-price, 2009, p. 37). As has been previously discussed a level of conjecture in inevitably employed and therefore uncertainty introduced to the reconstruction process. As Stanley-Price et al. state 'Because reconstructions do involve conjecture to a greater or less degree, the tendency will be for their architects to be unconsciously prone to other influences.' (Stanley-price et al. 2009, p. 37). The use of archaeological remains in conjunction with evidence from similar other locations to reconstruct a building will, by default, create new knowledge from old evidence, and is therefore, the ‘digital act of recontextualization’ (Johanson 2009, p. 406). Almagro succinctly summarises the issue when stating that ' it is not easy to establish clear-cut limits regarding our capacity for ‘‘invention’’ when we are re- creating heritage which has been altered, destroyed and in many cases has disappeared' (Almagro 2007, p. 165). This "invention" or uncertainty has been a source of debate in digital archaeological reconstruction. There are many factors to consider, such as conjecture, interpretation, and lack of transparency of sources, and methodology that result in uncertainty when questioning the veracity of the digital

2 Literature Review 11

reconstruction. It is widely accepted that reconstruction of an object or building that is largely absent ‘is usually a matter of interpretation’ (Bonde et al. 2009, p. 365). This is clearly summed up in the following statement, 'It should always be possible to make a clear distinction between what is original and what is not, and between what is plausible and what is merely hypothesis' (Almagro 2007, p. 165). The process of digitally reconstructing a set of archaeological remains is informed and guided by the ‘information used and the design constraints applied’ by the creator (Johanson 2009, p. 406). As such there is a view that digital archaeological reconstructions ‘contain an element of deception’ and that the ‘ways in which we admit and confront the risk of deception are important.’ (Bonde et al. 2009, p. 363). As a result there are growing calls for standards and methodologies that fully inform the viewer. (Denard 2011).

2.3 UNCERTAINTY As stated, a key issue in physical and digital archaeological reconstruction is the limited, and varying degree of accuracy of information available as a starting point. This inherently leads the architect of the reconstruction to make inferences and deductions and is therefore, an inevitable source of uncertainty (Denard, 2011). Uncertainty is not only restricted to archaeological reconstruction but to a wide variety of other fields. In several fields using VR technologies such as, fluid dynamics, weather modelling, and decision support systems, uncertainty visualisation has become a topic of more focussed research. Skeels et al conducted a review of existing work on uncertainty from a number domains and proposed a refined classification for visualising uncertainty (Skeels et al., 2010). Pang, et al. surveyed uncertainty visualisation techniques in order to demonstrate ways that users can be more aware of 'locations and degree of uncertainties in their data' (Pang et al. 1997, p. 1). Others to have investigated the issue of lack of uncertainty visualisation in 3D visual representations include Johnson et al., 2003, who viewed the importance of visualising is such that, 'the visualisation research community needs to make visually representing errors and uncertainties the norm rather than the exception' (Johnson & Sanderson 2003, p. 9). The authors detail areas where uncertainty can be identified and propose categories of data quality using an uncertainty typology.

In order to visualise uncertainty there is first a need to categorise and classify not only what uncertainty is, but the sources of uncertainty. Uncertainty visualisation has recently become a focused topic of research as the use of the visualisation

12 2 Literature Review

technology increases. Very little research has been carried out in the application of uncertainty visualisation in archaeological reconstruction. The bulk of research has been in the fields of physics, geographic information systems and intelligence gathering. Thomson et al., discuss and propose an expanded typology for uncertainty’ in geovisualisation of national security data (Thomson, Hetzler, Maceachren, & Gahegan, 2005, p. 1). They base their typology on 'past frameworks targeted at scientific computing' (Thomson et al., 2005, p. 1). Skeels et al., review several classification systems and developed their own classification system whereby they identify five types of uncertainty. They attempt to create a classification that can be applied across a variety of domains (Skeels et al., 2010). Other researchers have developed a visualisation pipeline in order to demonstrate how uncertainty is introduced and then proposed 'a classification with five characteristics: value, location, data extent, visualisation extent, and axes mapping’ (Pang et al., 1997, p. 374).

The bulk of work in this area has not been conducted in the domain of . It is the aim of this research to draw on these past works and investigate a classification system more applicable to the subject. The final challenge is to develop a visual methodology for displaying uncertainty that can work as an additional information layers augmenting the visual output and quality of a digital reconstruction.

2.4 THE NEED FOR DIGITAL RECONSTRUCTION AND VISUALISATION FOR ARCHAEOLOGY As stated in the introduction of this document there are strong arguments for the use of VR technologies in the digital reconstruction of archaeological finds. Traditional methods of dissemination for archaeological reconstructions such as text, although usually exhaustive in description, fail to convey the knowledge to a wider audience effectively. Denard eloquently summarises this when describing the use of text for communication, pointing out that the authors of these texts 'force a synchronic perception into a diachronic narrative; that is, must strip the object down into a sequence of layers, to be presented sequentially rather than simultaneously, thereby negating the very cognitive experience that the author ultimately wishes to evoke.' (Denard 2011, p.60). The author goes on to argue that mediums such as static images, video presentations and real-time models 'might well become the expressive medium

2 Literature Review 13

of choice, conveying, all at once, the complete, synthetic image of the author’s idea.' (Denard 2011, p.60)

The possibilities of VR technology for the reconstruction of archaeological material were seen early in its infancy by Jablonka et al. They asked the question, how could these technologies 'help archaeology to meet its twofold challenge: To make scientific results accessible for the researcher, and to communicate them to the public' (Jablonka et al. 2002, p. 1). They recognised that VR reconstruction can be a powerful research tool used as 'authentic, accurate, up-to-date, and well-documented content.' (Jablonka et al. 2002, p. 1). The researchers state that VR tools can be used to 'arrange data at its actual position in space', and as a 'intuitive way of handling large amounts of archaeological information.' (Jablonka et al. 2002, p. 2). Perhaps their most compelling argument is that in creating digital virtual reconstructions of material remains it 'forces archaeologists to re-evaluate and discuss excavation results in previously unexpected ways.'(Jablonka et al. 2002, p. 4). This method of reconstruction allows archaeologists to ' try out different variants and thus arrive at a satisfactory result by an iterative process.'(Jablonka et al. 2002, p. 4). This view is supported by Olsen et al. who found that 3D VR technologies can 'bring archaeology to a new era where it not only can assist archaeologists for scientific discovery but it can also enrich the presentation of findings and enable the distributed scientific communication and discovery.' (Olsen et al. 2004, p.3).

There are many architectural material remains and relics throughout the world and the physical restoration/reconstruction of these remains is unrealisable (Petrova et al. 2011, p. 389). Not only are there many and strict guidelines around physical reconstruction of archaeological remains such as the (2008) ICOMOS Charter for the Interpretation and Presentation of Cultural Heritage Sites (Silberman, 2008) but also the complexity of the work and lack of information make such endeavours unfeasible (Petrova et al. 2011). For this reason Petrova et al. believe that VR reconstructions offer a means to 'achieve a new level of preservation and transmission of cultural heritage.' (Petrova et al. 2011, p.389). The concepts described by Petrova are echoed and expanded upon by Almargo 2007, who points out that due to legislative restrictions around reconstruction, the bulk of work carried out is of a 'conservative nature with due respect to historical values' which hinders the 'recovery of the image and perception of the original space' (Almagro 2007, pp. 161-162). Almagro champions

14 2 Literature Review

the use of VR reconstruction of lost buildings as a method to visualise 'such fundamental aspects as their visual perception, the importance to the space of aspects such as the colour and texture of the materials, the effects of light and even the scale of the building.' (Almagro 2007, p.162).

VR reconstructive techniques offer a strong visual medium in which to communicate, debate, and share information (Cargill, 2009). In using VR modelling to test the Qumran-Essene hypothesis as it was first proposed, Cargill concludes that the technology allowed researchers to analyse archaeological data and derive conclusions. The conclusions were drawn from the information gathered both through the process of 3D reconstruction, and from viewing the site in a VR setting (Cargill, 2009). Cargill concludes "this research highlights the value of digital modelling as a new approach to archaeological reconstruction" (Cargill, 2009, p. 41). Almagro, 2007, see multiple applications for VR reconstruction in archaeology that can be categorised into two general groups 'to facilitate reflection and research into the lost architectural heritage ' and 'the transmission of information' (Almagro 2007; Palombini et al. 2012, p. 164). He argues that the traditional, methods of communicating information about a lost building such as plans, elevations and sections ' have always proved to be rather unintelligible for those who have neither knowledge nor experience of the systems of representation.'(Almagro 2007, p. 164). Almagro argues that perspective views are more intuitive and easier to understand. (Almagro 2007)

In further support for VR in archaeological and historical reconstruction of material remains, Barceló compares the outcomes of traditional 2D artistic representations of reconstructed archaeology with that of computer-generated reconstructions. He points out that artists have collaborated with archaeologists for years to '“reconstruct” all those wonderful things not preserved in the archaeological record' (Barceló, 2000, p. 9). Barceló goes on to point out that the resulting image is not a reconstruction of the past but rather a subjective way of “seeing” it.' (Barceló, 2000). In contrast, a virtual 3D model is a visualisation of some, if not all, features of the object under concideration. The purpose of the 3D model is to assist users to understand the structure or behaviour of the entity under consideration, and to provide an environment that will support experimentation (Barceló, 2000). Daniels states that using 3D VR techniques in archaeological reconstruction and visualisation allows us

2 Literature Review 15

to more efficiently represent and record data (Daniels, 1997). Daniels points out that the traditional methods of data collation and integration are inefficient for accurately recording, archiving, and relating large data sets which are architecturally complex, and multi-dimensional in nature (Daniels, 1997).

Hermon and Kalisperis reviewed the role of VR in the cultural heritage domain. They concluded that information visualisation process transforms 'data, information, and knowledge into visual form.' (Hermon & Kalisperis 2011, p. 60). They conclude that VR environments enhance visual processing and allow an insight into 'abstract data values' (Hermon & Kalisperis 2011, p. 60). The authors believe that VR affords a superior visual tool to traditional methods for the delivery of information in the archaeological setting and conclude that 'the better the visual tool, the better the explanation and the interception.' (Hermon & Kalisperis 2011, p. 60). As an extension of this assertion they state that 3D visualisation of spaces, structures or objects ' gives a visual framework in which data is displayed.'(Hermon & Kalisperis 2011, p. 60). In addition to the obvious benefits of 3D Visualisation and reconstruction highlighted by the researchers mentioned above, there is also a strong belief by some researchers that VR ' complements perfectly documentation and conservation efforts and even can act as an integration activity to bring all information together in a structured way that allows long term preservation.' (Pletinckx & Tartessos 2011, p. 1). Pletinckx and Tartessos point out that the use of VR in archaeology has provided ' a large set of useful tools' and that past experiences show that ' it is much more than building 3D models only.'(Pletinckx & Tartessos 2011, p. 1).

Hermon argues that a great deal of 'intellectual effort has been dedicated in the past to the definition of theoretical frameworks of archaeological data interpretation.' but that far less importance has been placed on how we 'look at the data.'(Hermon 2011, p. 14). The author argues that the process of data interpretation and interrogation could be 'furthered' by the use of VR techniques thereby significantly contributing to a ' better understanding of the archaeological remains under investigation.' (Hermon 2011, p. 14).

Although Pletinckx, et al. expound the virtues of VR in archaeology as a method of integrating and visualising multiple disparate data sources they state that "Virtual archaeology has a problem of credibility and scientific rigour, as it lacks a widely supported methodology on how to turn its sparse sources into 3D models." (Pletinckx

16 2 Literature Review

& Tartessos 2011, p. 33). This is further supported by Demetrescu 2015 whom states that virtual reconstruction in the field of archaeological research is ‘as yet undefined discipline, one that is still largely fragmented when it comes to methodology, both in terms of data transparency and common standards’ (Demetrescu, 2015, p. 43). In addition to the perceptions of lack of scientific rigour there is also a generally held belief among researchers in the field that there is an emphasis on 3D digital reconstructions to be photorealistic while ignoring the more fundamental issues of spatiality and use of space (Dunn & Woolford, 2012). Pletinckx, et al., argue that VR archaeology has been primarily viewed as a communication medium with less emphasis on the scientific background and research. They point out that an analysis of specific archaeology projects shows that most are 'ephemeral' (Pletinckx & Tartessos, 2011).

Researchers are beginning to address the issue of these perceived deficiencies. It is now being recognised that virtual archaeology 'is teamwork, in which interdisciplinarity is the crucial success factor.' (Pletinckx & Tartessos 2011, p. 35). In recent times, there has been a concerted effort to standardise digital reconstruction methodologies and rigour. This is being partially addressed by the London charter for the Computer-based Visualisation of Cultural Heritage in 2006 (Denard, 2011), the Seville Principals (an extension of the London Charter) (Carrillo Gea, Toval, Fernández A., & Flores, 2015), and to some extent, the ICOMOS Interpretation and Presentation of Cultural Heritage Sites (Silberman, 2008). The London Charter was conceived to 'develop methodological rigour of computer-based visualisation as a means of researching and communicating cultural heritage.' (Denard, 2011, p. 57). In addition, the charter recognises that a 'lack of transparency had been identified, along with the epistemological problems posed by hyperrealism,' (Denard, 2011, p. 57). Both the London Carter and the Seville Principles support and promote ‘interpretation and simulation based on a theoretical and multidisciplinary scientific approach ‘ (Petriaggi et al., 2018, p. 2)

The objectives of the charter are:

1. Provide a benchmark that is recognised amongst users 2. Promote intellectual and technical rigour in the field of digital heritage visualisation in general

2 Literature Review 17

3. Ensure that computer-based visualisation processes and outcomes can be properly understood and evaluated by users 4. Enable computer-based visualisation authoritatively to contribute to the study, interpretation and management of cultural heritage sites 5. Ensure access and sustainability strategies are determined and applied. 6. Offer a robust foundation upon which communities of practice can build detailed London Charter Implementation Guidelines.

(Beacham et al. 2009, p. 4) For the purposes of the proposed research, objectives 1, 2, 3, and 5, are of interest. This study aims to contribute to the development of a easily understood benchmark methodology which focuses on the critical metadata associated with a virtual reconstruction by aiding transparency and fostering scientific rigour. It is clear objectives 4 to 6 cannot be fully realised without the implementation of the first three.

Although the London Charter is a significant step forward in the standardisation of 3D reconstruction in the cultural heritage domain it is yet to establish formalised guidelines or methodologies surrounding the use of VR in digital archaeological reconstruction. There are five broad guidelines presented in the charter for computer based visualisation of heritage sites, these are; Aims and Methods; Research Sources; Documentation; Sustainability and Access (R. Beacham et al., 2009). Of these, three are of primary interest in the context of the proposed research; Research Sources, Documentation, Sustainability and Access. The guidelines are summarised in Table 2.1 below with extracts from the charter and key wording highlighted. It will be demonstrated via case studies that, in general, research sources used in reconstructions are usually limited to research publications and seldom appear within the visualisations in a meaningful manner. Similarly, the outcomes of digital research are generally only discussed within the reporting papers and no serious attempts have been made to transfer this knowledge in some way to the digital reconstructions themselves. Finally, it is rare within reconstruction visualisations to have a clear indication of the 'hypothetical dependency relationships between elements,' (Beacham et al. 2009, p. 9). It follows that if this is the case, the maximum possible benefits for study, understanding, and interpretation are not being realised through such visualisations.

18 2 Literature Review

Research Sources Relevant research sources should be identified and evaluated in a structured and documented way, with particular attention 'to the way in which visual sources may be affected by ideological, historical, social, religious and aesthetic and other such factors.' p.7 Documentation Sufficient information should be documented and disseminated to allow computer-based visualisation methods and outcomes to be understood and evaluated in relation to the contexts and purposes for which they are deployed. p.8.

It should be made clear to users what a computer-based visualisation seeks to represent, for example the existing state, an evidence-based restoration or a hypothetical reconstruction of a cultural heritage object or site, and the extent and nature of any factual uncertainty.p.8

Computer-based visualisation outcomes should be disseminated in such a way that the nature and importance of significant, hypothetical dependency relationships between elements can be clearly identified by users and the reasoning underlying such hypotheses understood. p.9 Access The creation and dissemination of computer-based visualisation should be planned in such a way as to ensure that maximum possible benefits are achieved for the study, understanding, interpretation, preservation and management of cultural heritage. p.11

Table 2.1: Key guidelines from the London Charter for the Computer-Based Visualisation of Cultural Heritage, (Beacham et al., 2009)

The charter relies on voluntary implementation of the guidelines. This is, in part, due to the complexity of data and observations gathered in the process of researching and uncovering an archaeological site and the inherent gap between archaeologist/academic and the VR practitioner.

‘We now find ourselves, however, gravitating towards a, perhaps, wider, recognition of interconnectedness: an observation that computer-based visualisations are valuable precisely because they collapse boundaries between the mysteries of rarefied academics and popular understanding. A visualisation may at once embody deep and complex knowledge and at the same time make the contours of that knowledge intuitively accessible to a non-expert audience in a way that a text-based publication never could.’ (Denard 2011, p. 70). It is precisely because VR is powerful communication medium that emphasis must be placed on reconstruction rigour and transparency of data inputs.

Plentinckx, et al., recognise that, traditionally, archaeologists have not had the requisite knowledge of VR processes to provide 3D modellers with appropriate data

2 Literature Review 19

for modelling. 3D modellers, equally, did not have the requisite knowledge to ask the correct questions of the archaeologists in order to create a 3D model that represents a correct interpretation of the data (Pletinckx & Tartessos, 2011). To ask the right questions, the questioner must understand the process of archaeological research to some degree. Archaeological research analyses the remains of human activity in the past, these remains are discovered during excavations, collected during surveys or other remote sensing techniques (S. Hermon, 2011). The data collected during the excavation and research phase is subjected to an "interpretation of the findings within a given theoretical framework and according to the aims of research" (Hermon 2011, p. 11). Usually, to interpret findings and data, researchers and 3D modellers need to consider a number of factors beyond the physical structure being reconstructed. These could include:

"understanding the of an archaeological culture, the socio- economic organization of a human group, the human occupation of an area and its relationships with the environment, as well as an analysis of historic events, settlement patterns and movements of population, social behaviour, etc." (Hermon 2011, p. 13).

Hermon states that the remains of material culture can be analysed from three basic perspectives. (S. Hermon, 2011)

I. Material (the relationships with the environment)

II. Social (within the frame of the human society)

III. Ideational (how and what human ideas and beliefs they embed)

When interpreting multiple sources of information, and comparisons with existing similar material remains, there always exists the possibilities of 'several possible interpretations' (Hermon 2011, p. 18). It follows therefore, that there exists a need to be able to not only distinguish between various interpretations, but also the paradata that was used to arrive at those interpretations. This is supported by Beacham 2011 when discussing the quality of paradata and its role in the reconstruction/interpretation process. The author points out that ' one is concerned to make transparent, where does 'realism' or 'magic' come in, and what might be the relationship between them?' (Beacham 2011, p. 7). Beacham is making the point that there is a strong need for transparency in archaeological visualisation considering the quality of modern rendering systems and possible outputs. Beacham quotes Arthur C.

20 2 Literature Review

Clarkes famous 'Third ' and it is worth repeating here considering the argument made above.

' Any sufficiently advanced technology is indistinguishable from magic.' (Clarke, 1962).

Hermon and Kalisperis (2011), take the point above further by suggesting that when creating a visualisation it is generally aimed at a target audience, which can only be defined in general terms (S. Hermon & Kalisperis, 2011). They point out that because of this they face ' an apparent tough choice when creating a VR, balancing between "an objective representation of the whole", and a "selected visualisation", choosing only particular aspects of the real object to be virtually represented.' (Hermon & Kalisperis 2011, p. 61). Hermon points out that visualisation tools can be divided into two main categories, these being, 'Interpretive and Expressive' (Hermon 2008, p. 40). From most visualisation case studies examined it is clear the distinction between an interpretive or expressive visualisation is seldom made. The onus is on the viewer to research and investigate the paradata (often requiring extensive research) used to arrive at the reconstruction to come to an informed decision. The required information is rarely inherent in the visualisation itself.

Addison succinctly summarised the main issue facing the successful adoption of VR for archaeological reconstruction and interpretation as early as 2000 and remains the case although some small advances have been made.

Virtual heritage can be an invaluable tool, but if not applied wisely has the potential to do as much harm as good. Dozens of virtual Pompeii’s now exist, yet few are historically accurate or reliable enough to be truly useful (and may in fact mislead people). Historical accuracy, thorough documentation of sources, and care while at sites is perhaps more important than ever in the new digital landscape. (Addison 2000, p. 25).

2.5 CASE STUDIES OF DIGITAL RECONSTRUCTION IN ARCHAEOLOGY A significant body of work using virtual reality modelling and 3D visualisation techniques to recreate archaeological sites and monuments has already been carried out. The bulk of this work is based around the use of procedural techniques and technologies used to carry out these tasks. This is exemplified by researchers such as

2 Literature Review 21

(Andrés, Pozuelo, Marimón, & Gisbert, 2012; B Frischer, 2009; Rajapakse, Tokuyama, & Somadeva, 2011). It is clear with these works that rigorous scientific approaches are employed in the capture (and at times reconstruction) of archaeological data. The following examples are a selection of the most notable attempts that step beyond procedural acquisition to make use of VR in reconstruction and analysis of archaeological material remains. Denard make critical reference to the veracity of the resulting VR as to its ability to ”render the research process visible to the visitor … to allow the viewer to assess it as part of an argument” Denard, 2011, p60.

The most recent work in digital archaeology is the virtual reconstruction of the Roman Gladiator School at Carnuntum, Austria, carried out in 2014. The extensive 650 ha Carnuntum site (Figure 2.1) is located approximately 40km south-east of Vienna, Austria on the southern bank of the Danube river and has been excavated since the late nineteenth century (W. Neubauer et al. 2014).

Figure 2.1: The extent of archaeological material remains of the Carnuntum site. From Neubauer et al., 2014 (Wolfgang Neubauer et al. 2014, p. 177).

The Gladiator school was discovered in 2011 'using non-invasive survey methods' (Wolfgang Neubauer et al., 2014). The site dates back to around AD 124 during the time of the Roman emperor Hadrian (W. Neubauer et al. 2014, p. 174). The researchers used modern forms of archaeological survey such as ' airborne imaging spectroscopy, electromagnetic induction and ground-penetrating radar' to map and measure, as yet, unexcavated material remains of buildings (Wolfgang Neubauer et al. 2014, p. 1). These techniques uncovered the foundations of a building complex which

22 2 Literature Review

included training areas, assembly spaces, a bath complex and other infrastructure (W. Neubauer et al. 2014, p. 174). Although the paper describing the reconstruction is primarily a description of the data capture and processing phase a key statement is made in the introduction.

'The combination of techniques has led to the recording and visualisation of the buried remains in astonishing detail, and the impact of the discovery is made all the greater by the stunning reconstruction images that the project has generated.’ (W. Neubauer et al. 2014, p. 173).

The concluding words in this statement ' the impact of the discovery is made all the greater by the stunning reconstruction images' are of significance in the context of this review and will be elaborated on in more detail. It should, at this point, be noted in the context of a later statement where the authors present a 'best archaeological interpretation of the newly discovered monuments.' (W. Neubauer et al. 2014, p. 174).

As stated, the core of this report is the description of the techniques and methods used to map the material remains of the Gladiator School. The resulting 3D mapped data of the material remains were used to reconstruct the Gladiator School as a 3D visualisation, see Figure 2.2. A section of the paper is dedicated to the archaeological interpretation of the data and subsequent digital reconstruction.

Figure 2.2: Virtual reconstruction of the Catnuntum ludus, viewed from the south, © Michael Klein. From (W. Neubauer et al. 2014, p. 2). It is this section which is of interest in the context of the broader research proposed for this thesis. The digital reconstruction is primarily based on the data from the non-invasive scanning and from similar examples of Roman architecture unearthed

2 Literature Review 23

in other parts of Europe. It is clear from the descriptions provided, that assumptions have been made to reconstruct the Gladiator School. The actual plan layout of the reconstruction itself is essentially beyond question. The researchers used scientifically rigorous methodologies to determine the physical extents of the building walls and other remaining features. This is exemplified when the authors describe the structures and layout of the complex, 'A single and easily controlled entrance to the complex can be seen on its eastern side facing the amphitheatre' and the 'distance from the gateway to the amphitheatre is approximately 80m.' (W. Neubauer et al. 2014, p. 183).

The point here is that such descriptions can be based solely on the physical material remains of the complex. Assumptions such as the gateway controlling access are reasonable and supported by material evidence. Beyond the actual floor plans the researchers have drawn on past knowledge of Roman architecture as well as formulating hypothesis for the structure and use of the buildings. This is exemplified when the authors propose that the gate described above would have been a 'structure like a triumphal arch clearly discernible as two 4.4m x 3.8 wide foundations.' (W. Neubauer et al. 2014, p.184). At this point an inference has been made based on the limited material remains as well as other examples elsewhere. When describing the inner training arena of the School the authors suggest that the 19m diameter free standing structure was 'surrounded by wooden spectator stands' (W. Neubauer et al. 2014, p. 184).

It is not the role of this review to call these assumptions into to dispute but rather to highlight that assumptions are used in the reconstruction of the space. The use of words and phrases such as 'might', 'assumed', and 'most likely' are common not only in this paper but many others when discussing the final interpretation or visualisation of features such as buildings and other structures. No material evidence of the textures and colours of the walls of the building remain, as such, these features are usually assumed. In addition, heights of walls and roof types can only be inferred from foundation dimensions and existing examples of similar architecture. There is no doubt that the final visualised reconstruction is very impressive visually, and, quite probably, is as accurate as can be given the available information. Unfortunately, as is common with most reconstructions of this type, a high degree of uncertainty (resulting from assumptions and lack of data) is essentially embedded in the final output. This uncertainty remains hidden from the viewer and therefore contains, by default, ‘an

24 2 Literature Review

element of deception’ (Bonde et al., 2009). In terms of the desirable goals of the London charter the final visualisation does not provide a recognised benchmark in terms of an overall reconstruction methodology nor does it ensure that the outcomes can be properly understood or evaluated by a majority of users. It does, however, contribute to sustainable strategies.

Clearly the use of digital reconstruction of archaeological material remains is not restricted to any particular historical time period and VR techniques have been used effectively elsewhere in the historical context. Gasparini, et al, undertook the digital restoration of the missing San Jacinto Church in Caracas (1595 -1821). The 16th Century San Jacinto Church in Caracas Venezuela was badly damaged during an earthquake in 1812 and fell into disuse by 1821. The church was considered 'one of the two most important spiritual centers from the colonial times and first University of Venezuela.' (Gasparini et al. 2012, p. 216). Due to the lack of physical material remains, much of the data drawn on for the digital reconstruction came from a variety of sources. These sources included 'bibliographic sources, archaeologist excavations, urban planimetric, paintings, church photography, contemporary “retables”' (Gasparini et al. 2012, p. 214). Examples of the types of data sources can be seen in Figures 2.3 and 2.4.

2 Literature Review 25

Figure 2.3: Examples of contemporary churches used as data sources for the reconstruction. (Gasparini et al. 2012, p.125)

Figure 2.4: Examples of photography, urban documentation and archaeology records used in the reconstruction process. (Gasparini et al. 2012, p. 217).

It is clear from the examples above that the authors have a strong foundation in Venezuelan religious architectonic heritage and have drawn on this extensively in the development of the reconstruction. The authors developed a four-stage approach to the digital reconstruction and suggest this to be a method for similar such reconstructive situations. These stages were:

26 2 Literature Review

• Collation of bibliographic, archaeological, historic and urban documentation. The collation of church plans for reference from Canarian and Venezuelan churches.

• Revision of bibliographic, archaeological, historic and urban documentation from the site where the San Jacinto Church stood and from the church itself.

• The creation of a virtual model based on stage 1 and 2

• Develop a 2D plan model and a 3D model textured model of the church

The creation of the 3D model of the church was heavily based on analogous examples of similar architecture in Venezuela and the Canary Islands. Throughout the reconstruction process the authors have drawn not only on analogous examples, their own expertise, and available documentation but also took into consideration anecdotal accounts of visits to the church. This is exemplified when describing one of the three iteration the church went through, 'The third is described in the visit from Bishop Marti, presenting the same façade and one entry door plus the Terceros nave.' (Gasparini et al. 2012, p. 219). A date is not given for the Bishop's visit. Similarly, the reconstruction of the roof of the church is based on analogous examples, this is exemplified in Figure 2.5. The resulting 3D reconstruction of the San Jacinto church can be seen in Figure 2.6.

Figure 2.5: Roof reconstruction based on similar architecture examples. (Gasparini et al. 2012, p. 222)

2 Literature Review 27

Figure 2.6: Final reconstructed virtual model of the San Jacinto Church, Caracas, Venezuela. (Gasparini et al. 2012, p. 223).

This 3D reconstruction of the church is a simple and, visually well-defined reconstruction. Some effort has gone into adding a certain level of realism with the application of textures and the use of rendered lighting, (see Figure 2.6). The authors have reported their sources and assumptions faithfully, it is thus quite probable that based on the authors knowledge of Venezuelan religious architecture that the resulting 3D reconstructions are an accurate representation of the church as it stood in one of its iterations. Unfortunately, as with the previous case study examined, the critical information regarding assumptions and data sources is lost within the visualisation. It is not possible for the visualisation processes and outcomes to be properly understood and evaluated by users. A viewer will struggle to understand from the visualisation how the final reconstruction was arrived at. The embedded uncertainty is not accessible. This reconstruction does add to the body of sustainable strategies that can be employed in the virtual reconstruction process.

In a similar project to the reconstruction of the Church of San Jacinto, Petrova et al, 2011, undertook the digital reconstruction of the fresco paintings of the Church of the Transfiguration of Our Saviour on Nereditsa Hill in St. Petersburg, Russia. The church itself was destroyed during the Second World War, the rare 17th Century frescoes adorning the walls and ceiling were also lost. Fragments of the frescoes numbering 325,000 pieces remain. Figure 2.7 shows the placement of the pieces with relation to the original dome of the church (Petrova et al., 2011).

28 2 Literature Review

Figure 2.7: Fragment restoration from the destroyed church. (Petrova et al. 2011, p. 391)

The digital reconstruction process consisted of collating archival material in the form of photographs, written documentation and plan drawings as well as an investigation of the fragments themselves for colour retrieval. The authors compiled photography which was then merged to form a grey-scale image of the interior frescoes, (see Figure 2.8) These images were coloured using information gathered from original fragments (see Figure 2.9). This is a detailed analysis of procedure and sources but again no method of indicating uncertainty or assumptions is present in the final 3D reconstruction. The user is unable to distinguish between the known parameters and the assumed.

2 Literature Review 29

Figure 2.8: Monochrome image of the wall, completed with pictures from archival photographs. (Petrova et al. 2011, p.392).

Figure 2.9: The reconstructed frescoes of the western wall based on colours extracted from the original fragments.(Petrova et al. 2011, p. 393).

The methodology employed to arrive at the visualisation contributes to the intellectual and technical rigour of digital heritage visualisation. It does not however

30 2 Literature Review

provide a viable way for users to understand the data, the assumption and methods used to arrive at the final visual output.

The 3D reconstruction of archaeological material remains can be part of a wider information visualisation system of a large size heritage work. (Issini, 2012) describes a large-scale project examining the physical restoration of the Yanqin Section of the Chinese Great Wall. The researchers aimed to use VR technologies to assist in the interpretive process prior to any physical work being undertaken. The research team were tasked with collating and organising 'many different types of information, acquired across different times and with many methods and techniques' (Issini 2012, p. 523). The information needed to be organised in such a manner that it would allow access for analytical reference for design strategies in the reconstruction phase. It is important to note that the researchers aims were to collect the disparate data 'in a unique, interactive and immersive environment.' (Issini 2012, p. 523). In effect, the digital reconstruction aims to be the 'store house' of the data used. Although the element of digital reconstruction is a minor part of a greater project, the greater interest is the method the author chooses to display ancillary information or metadata in the visualisation environment.

The research describes a method of superimposing 'specialist themes and contents' within the 3D visual environment ' 'that can be activated on the basis of competence level and user needs.' (Issini 2012, p. 526). The author associates spatial locations within the virtual environment with the data contained in the database thereby allowing real-time display of metadata (Issini 2012, p. 526). This enhances the user experience by allowing them to gain a greater understanding of the visual element. The visual reconstruction of the wall section was in itself, a relatively straight forward task. The majority of the wall, and associated towers, still exist although in a state of disrepair. Acquisition of topography similarly was also straight forward as this data existed. The results of the initial reconstruction are seen in Figure 2 .10.

2 Literature Review 31

Figure 2.10: Wireframe and rendered 3D model of the Yanqing Section of the Great Wall. (Issini 2012, p. 525).

Along with survey data the research team also collected extensive textures of the wall from many locations allowing for a series of representative textures of sections of wall to be applied. This improved the visual display of the final reconstruction. The uniqueness of this project stems from the method of integrating the finalised reconstructed model into an interactive virtual environment that allows users to access information in real-time. This is exemplified in Figures 2.11 and 2.12. The significance of this approach is that the authors employed a navigable 3D environment as the base for an interactive information retrieval system. This system provides users with additional information beyond the visual, allowing them to gain a greater depth of knowledge of the reconstruction they are interacting with.

32 2 Literature Review

Figure 2.11: The visualisation 'Home Page' The 3D model is displayed in association with information access tools. (Issini 2012, p. 526).

Figure 2.12: Examples of metadata available through the information visualisation system. (Issini 2012, p. 526).

The system is coded in such a manner that a user can quickly understand what information is available using coded text. These text symbols and their significance are:

• VR - indicates that there is a navigable 3D model of the object;

• SURVEY - refers to metric survey technical tables;

• MA - indicates a materials report;

• DA - indicates an analysis of the deterioration related to collapse;

(Issini 2012, p. 526).

2 Literature Review 33

It is interesting to note that although this paper does not go into detail with regards any assumptions made in the digital reconstruction, these assumptions are individually detailed in the metadata. This information itself is inherently available (or deducible) through the system. This project represents the development of an application rather than a visualisation aimed at presenting transparency in the reconstruction process thereby facilitating viewers to better evaluate the outcomes of the reconstruction.

In a similar project the Villa of the Papyri Project (Zarmakoupi, 2010) makes extensive use of VR modelling and reconstruction techniques in order to inform a wider physical reconstruction project. The Villa of the Papyri was a private residence in the Ancient Roman city of Herculaneum situated near the modern town of Ercolano in the province of Naples in Southern . Since its discovery in the 18th century the villa has undergone a series of archaeological excavations and graphical 'reconstructions' as more of the site was unearthed (Zarmakoupi, 2010). The villa has also undergone physical reconstruction in the form of a copy known as the Getty Villa in Malibu, USA. Excavations continue today. An overriding reason for the digital reconstruction can be summed up by the following statement.

'Part of the flexibility of this virtual reality reconstruction is the ability to select among existing state and different restoration proposals.' (Zarmakoupi 2010, p. 185).

In addition to the above statement the author points out that VR models, may be employed to 'document and investigate archaeological sites as well as to present hypothetical reconstructions that may serve as virtual restoration proposal, of architectural monuments.' (Zarmakoupi 2010, p.181). A unique aspect of this study is the author’s attempts to differentiate source information in the actual 3D model. The methodology employed to do this involves the use of colour coding and layer overlaying. The author differentiates between the type of information that is visualised within the reconstructed model and provides users with the ability to ' switch between several architectural and wall painting restoration proposals and their existing state,' (Zarmakoupi 2010, p. 192). This is exemplified by the use of two different colours to ' differentiate the parts of the Villa that are known from Weber's plan and still lie underground from the recently excavated parts of the Villa.' (Zarmakoupi 2010, p. 185). This can be seen in Figure 2.13.

34 2 Literature Review

Figure 2.13: Perspective view of the model of the Villa of the Papyri, note the use of colour.(Zarmakoupi, 2010) The author acknowledges that a virtual reconstruction of archaeological material remains of the villa is based not only on excavation data but also 'comparative studies, as well as the modeller's informed hypotheses' (Zarmakoupi 2010, p. 185). The author presents a schematic diagram of the 3D modelling stages as proposed by Hermon 2008, Figure 2.14.

Figure 2.14: A schematic diagram of the 3D modelling stages and influences for the Villa of the Papyri as presented by Hermon 2008. (Hermon 2008, p. 38)

2 Literature Review 35

The model reconstructs and distinguishes the following areas, of the Villa of the Papyri,

1. Areas known from the 18th-century plan, 2. Areas revealed during the new excavations by Infratecna and the Archaeological Superintendency of Pompeii that are accessible today and 3. Restoration proposals, (fig. I5). These are known from the 18th-century plan are indicated by the yellow-beige colour. A brown-beige colour is used for the area, revealed during the new excavation,

(Zarmakoupi 2010, p. 187-188)

The colour-coding choices mentioned above were made to achieve two primary goals. These were:

• To create a reconstruction that makes as clear as possible what is reconstructed from the archaeological evidence and what is projected from the evidence in the form of restoration proposal and,

• To offer a reconstruction that is comprehensible as a three-dimensional building and it is not overly schematic.

(Zarmakoupi 2010, p. 188)

As a final note with regards the use of colour as an information indicator the author has experimented with a methodology whereby wall colour, or more precisely, wall texture is only applied in the cases where archaeological evidence indicates its true placement. A representative section is shown in Figures 2.15. It is interesting to note that from an authenticity perspective this approach has merit in that no assumptions are made with regards an assumed texture for the walls. Although this method allows the user the ability to discern information, the real from the assumed, the overall visual quality of the visualisation is of a poor quality and does not meet the output expected of modern visualisation systems and hardware. This raises an issue pertaining to user experience and the description of space.

36 2 Literature Review

Figure 2.15: A view from the south-west of a room within the Villa of the Papyri showing the, (a) existing state and (b) with a embedded restoration proposal. (Zarmakoupi, 2010)

In the case of this study the author has gone to lengths to maintain the authenticity of the reconstruction by concentrating on differentiating between actual archaeological evidence and inferred although this seems to relate primarily to the known wall paintings. The final 3D model is navigable, and locations of material remains can be identified and interrogated. The remaining reconstructed structure such as columns, wall types, roofs, etc., are based not only on material remains (where they exist) but also 18th Century architectural plans (created by the archaeologist Weber) and 'what is projected from the evidence in the form of restoration proposal' (Zarmakoupi 2010, p. 188). The author states that 'very little is known about the architectural detail, of the Villa.' (Zarmakoupi 2010, p. 189). In addition, the author has drawn on comparative studies of other luxury villas within the bay of Naples and 'architectural detail, surviving from other buildings of Herculaneum.' (Zarmakoupi 2010, p. 189). In some cases, information is missing completely such as the order of the columns of porticoes. The key point here is that although many unknowns (uncertainty) exist they are not treated with the same colour coding or classification system used for the wall frescoes. This visualisation represents a concerted attempt to address aspects of the London charter such as enabling users to understand the outcomes of the digital reconstruction with regards some of the data sources. It also contributes to body of sustainable strategies that could be employed in virtual reconstructions.

VR reconstruction methodologies in archaeology also extend into the theoretical domain where they are used to visually test hypothesis. Das and Garg 2011 used digital reconstructive techniques to 'present the visual interpretations of the pavilions described in the Mayamatam, a traditional ancient Indian architectural treatise. ' (Das

2 Literature Review 37

& Garg 2011, p. 2). The research is a theoretical exercise encompassing built heritage, Computer Aided Design, architectural proportions and aesthetics (Das & Garg 2011, p. 2). The paper is primarily a procedural study of rapid reconstruction using literary sources. The researchers developed a six-step method for the creation of the pavilions based on Sanskrit literature and other historical architecture publications. These six steps can be summarised as.

Step 1 Studying chapter 25 (Pavilions and Halls); reading the Sanskrit text and its English translation. Step 2 Tabulating the word-to-word meaning of the Sanskrit text and English translation. Step 3 Adding notes wherever alternative interpretation is possible, or some additional translation is required. Step 4 Based on the understanding, cross-referencing within the treatise and literature review (Archaeology Survey of India publications, by various authors), drawing first draft of the sketches on a square grid sheet along with comments. Step 5 Drawing final sketches on a square grid sheet. Step 6 Preparing CAD drawings and three-dimensional models for each sketch.

Table 2.2: Methodology for the Digital Reconstruction of theoretical pavilions.(Das and Garg, 2011), p.4

The authors have undertaken to use VR techniques as they present the opportunity to not only, effectively visualise the proposed structures, but also provide a database whereby a 'library of pavilion components, profiles, plans, etc. can also be prepared that will be useful for future researches.' (Das & Garg 2011, p. 2). In addition 3D modelling and associated 2D CAD modelling provides a method of 'easy storage, retrieval, manipulation, and editing of the objects created.' (Das & Garg 2011, p. 2). The resulting pavilion reconstructions can be seen in Figure 2.16. As with other case studies examined, the resulting reconstructions do not present a method of information dissemination which allows users access to ancillary metadata relevant to the authenticity of the reconstructions. This is not to say that the resulting reconstructions

38 2 Literature Review

are not accurate but rather that users have no way of intuitively assessing the authenticity.

Figure 2.16: Pavilions in the 1x1 & 16x16 configurations (Das & Garg 2011, p. 11)

In a similar project to that of the reconstruction of pavilions, Rashid et al. 2008 develop an interactive virtual model of the lost architectural heritage of the 8th century Buddhist Monastery of Sompur Mahaviahara in Bengal. This structure is of significance as it represents a type of Buddhist architecture (central cruciform structure) that is unique. The bulk of the superstructure no longer exists. The authors justify the use of VR modelling techniques as they 'can be used as a useful tool for multiple verification and criticism.'(Rashid et al. 2010, p. 31). The aim of this reconstruction was not to create a photorealist visualisation but rather to develop a 'method of evaluation and synthesis to conceptualize the formal expression of the structure.' (Rashid et al. 2010, p. 31). The virtual reconstruction of the monument was set in two stages. The first stage was the exact reconstruction of the existing remains and the second stage a 'process of evaluation and verification.' (Rashid et al. 2010, p. 31). The purpose of the second stage of the visualisation was to develop a methodology for reconstructing the structure based on all available information (Rashid et al., 2010). The researchers adopt a layered information system to the reconstruction efforts whereby all 'the threads all the available resources will be put together in a scientific manner to construct the bigger scenario. We kept the other end of the thread for inflow of the future resources so that the model or the proposal can be modified when newer resources would be available.' (Rashid et al. 2010, p. 31)

The authors point out that the main aim of the reconstruction was not the architectural form as it may have been originally. The research focuses on the 'collation

2 Literature Review 39

and the examination of all the available resources that may have had some impact on the architectural form', and fix this information in a framework that allows users to gain a better understanding of the possible 3D reconstructed outputs. (Rashid et al. 2010, p. 32). The authors state that the most important aspect of the proposed framework is that it relies on cross disciplinary approaches and exchanges of information (Rashid et al., 2010). The significance of this paper is that it attempts to create a framework for incorporating multidisciplinary data in the reconstruction process. The authors acknowledge the need for accurate representation of the structure as well as embedding information within the 3D reconstruction to enhance knowledge. The paper does not present any detail on the actual reconstruction methodology but does present a preliminary model of the reconstruction, (Figure 2.17), and the framework used, see Figure 2.18. They do not suggest ways to show uncertainty or assumptions and it is not possible for users to intuitively evaluate how the final built form was arrived at.

Figure 2.17: A Preliminary Reconstruction of the lost central structure of Sompur Mahavihara, (Rashid et al. 2010, p. 32) The workflow for the development of the temple described in Figure 2.18 is interesting in that it clearly shows the inputs used to develop the final model. Although the result is clear and shows how the model was created, there is no indication of the

40 2 Literature Review

level of uncertainty in any of the information. However, the framework does lend itself to being able to incorporate such elements and as such it contributes to strategies for developing virtual reconstruction methodologies.

Figure 2.18: The framework of knowledge for the reconstruction, (Rashid et al. 2010, p. 32)

Following on from the concept of using VR tools for reconstruction of archaeological material remains, a project by Thuswaldner et al., the Digital Anastylosis of the Octagon in Ephesos takes these reconstructive techniques one step further. The project presents the results of research in the digital Anastylosis (reconstruction) of cultural heritage monuments. The authors present a of building artefacts of the Octagon monument in Ephesos, Turkey that have been digitally reconstructed and reassembled. The primary focus of the paper are methods pertaining to geometry processing. The core of this project is to digitally reconstruct a structure for testing before an actual physical reconstruction is attempted (Thuswaldner et al., 2009). Figure 2.19 shows the methodology for the Anastylosis process, it is worth noting that four of the six steps in this procedure are entirely digital showing the importance of the technology in the reconstruction process.

2 Literature Review 41

Figure 2.19: High-level overview of the anastylosis steps in the virtual reconstruction of the Octogon. (Thuswaldner et al. 2009, p. 6)

Although this paper deals primarily with the technicalities of the data acquisition stage, the final reconstruction is of interest in that it shows a method of distinguishing the real from the assumed. The final method of display has validity in terms of conveying uncertainty and follows the principals of display employed by Zarmakoupi (2010). Figures 2.20 and 2.21 clearly demonstrates this principal.

Figure 2.20: High quality render of two different scenarios for a possible physical anastylosis of the Octagon. White indicates infered fill in stonework. (Thuswaldner et al. 2009, p. 22)

42 2 Literature Review

Figure 2.21: A study of the visual impact of the Octagon anastylosis at its prominent spot in the Curetes street in Ephesos. (Thuswaldner et al. 2009, p. 22)

2.6 PROBLEMS ASSOCIATED WITH DIGITAL RECONSTRUCTION IN ARCHAEOLOGY A virtual reconstruction of a lost structure is usually the result of fragmentary physical evidence and literary and epigraphic sources (Pollini, Swartz Dodd, Kensek, & Cipolla, 2005). Despite the growing use of VR modelling and visualisation techniques in the field of archaeological digital reconstruction and anastylosis, many questions remain with regards the authenticity of a virtual reconstruction; the accessibility of the associated paradata; the validity and transparency of the reconstruction process. Beacham 2011 succinctly summarises the issue in the following statement.

"…One is aware how easily - and how often - some practitioners (occasionally even established and reputable scholars) have been tempted by the publicity and hype of virtual reality as an element of popular culture to slip into what might be called the ‘P. T. Barnum syndrome', in which scholarship takes second place to showmanship.

2 Literature Review 43

Such visualisations are produced and launched with media hype, articles in the press and the like, and in the process questions of accuracy and the scholarly basis for such visualisations are too often displaced by their undeniably compelling 'magic" (Beacham 2011, p. 10).

This technology has been influenced primarily by the consumer gaming and movie market permitting users to be immersed in “texturally rich, fast, navigable worlds on a standard PC in the home” (Addison, 2000). As such, there is a tendency to create virtual reconstructions that are being driven by these consumer demands. This is exemplified by the use of Minoan Dolphin Fresco from the Palace of Knossos on in the Palace of Midas level of the 1996 video game 'Tomb Raider', see Figure 2.22. (Kelly, 2013)

Figure 2.22: The Pool scene at the start of the Palace Midas level, Tomb Raider 1996, (Kelly, 2013) Although the subject of discussion here is a video game, an important point should be noted, that is, this is a melding of factual history with a fantasy game environment. More factual based examples of game style rendering can be seen in the Rome Reborn project whose goal it is 'to build 3D digital models that illustrate the urban development of ancient Rome from the first settlement in the late Bronze Age'

44 2 Literature Review

(Wells et al. 2009, p. 373). The project has produced high quality visualisations of Rome in A.D. 320. The building and city models (see Figures 2.23 to 2.26) have been generated from a digital scan of a plaster cast, the '“Plastico di Roma antica,” a large plaster-of-Paris model of imperial Rome (16x17 meters) created in the last century' (Guidi et al. 2005, p. 119).

Figure 2.23: Rome Reborn, an aerial view of the city centre. (Frisher Consulting, 2013)

Figure 2.24: Rome had many small private bathing establishments. (Frisher Consulting, 2013)

2 Literature Review 45

Figure 2.25: An aerial view of the Flavian Amphitheater ("Colosseum"), (Frisher Consulting, 2013)

Figure 2.26: The plaza on the west side of the Flavian Amphitheatre, (Frisher Consulting, 2013) Again, the point being made is that although the resulting visualisations may be correct there is no access to paradata within the visualisation allowing viewers to evaluate the models. Combined with this is the fact that the models are rendered in great detail which can lead viewers to accept that the model are a true and correct view of history rather than an approximation. Regardless of factual accuracy it should be

46 2 Literature Review

remembered that a digital reconstruction of a now lost building is effectively a knowledge representation or hypothesis that is open to questioning (Bonde et al., 2009). The examples detailed above highlight the need to integrate paradata, the information used to arrive at the final reconstruction, within the final visualisation.

2.7 VISUALISING UNCERTAINTY To begin to identify the uncertainty in archaeological visualisations one must be able to visualise this uncertainty in a coherent and effective manner. Pollini, et al. point out that the task of visualising uncertainty within a virtual reconstruction is problematic as “…the immediacy of the experience is diminished because we are constantly compelled to look away from the visual reconstruction to the textual evidence.” Grigoryan et al. report that 'there is a rich and active body of research addressing the challenge of showing data values in the context of their certainties in an organic and effective manner.' (Grigoryan & Rheingans 2002, p. 147).

Johnson and Sanderson state that to have an effective visualisation ' complete and accurate visual representation of data and models' must be available for user interpretation (Johnson & Sanderson 2003, p. 8). They go on to point out that such visual representations should adhere to standard scientific visualisation techniques and include representations of uncertainty and error (Johnson & Sanderson, 2003) .

It is crucial therefore, to define the concept of “uncertainty” with clarity. Pang et al. broadly define uncertainty to be any data that includes ' statistical variations or spread, errors and differences, minimum-maximum range values, and noisy or missing data.' (Pang et al. 1997, p.371). This definition although statistical in origin and defined for fields other than archaeology such as physics, is valid in the context of reconstructive archaeological visualisations. Pang et al. make the compelling argument that to visualise uncertainty it first must be identified. They identify ' three major blocks in a visualisation pipeline' (Pang et al. 1997, p.371).

1. Data Acquisition (for example, material remains, literary works, analogous structures etc.) 2. Data transformation (data recording, data collation, data interpretation) 3. Data visualisation (3D modelling, video output, real-time output etc.)

2 Literature Review 47

Figure 2.27, the visualisation pipeline as described by Pang et al. describes the points at which uncertainty can be introduced. Different forms of uncertainty are introduced into the pipeline as data are acquired, transformed, and visualized.

Figure 2.27: This visualisation pipeline shows the introduction of data uncertainty from models and measurements. (Pang et al. 1997, p. 372) The introduction of uncertainty (imperfect data) into the visualisation pipeline leads to misleading or inaccurate visualisations (Gershon, 1998). Gershon proposes a high-level taxonomy of sources of imperfection which broadly align with the three proposed by Pang et al (1997).

These are: • Corrupt data - errors of measurement, errors of location etc.

• Incomplete data - very common in the real world

• Inconsistency - information or data not consistent with each other or what is known.

• Difficulty in understanding - overly complex data sets that are difficult to understand

• Uncertainty - The credibility of the sources and veracity of the data are unknown

(Gershon 1998, p. 43) Thomson et al. point out that certain sources of uncertainty such as the credibility of a source or the completeness of a set of information is an abstract uncertainty and that as ' the uncertainty becomes more abstract, it is more difficult to quantify, represent, and understand.' (Thomson et al. 2005, p. 147). Research in the field of visualisation of uncertainty has been primarily in the domains of, geo-visualisation,

48 2 Literature Review

meteorological visualisation, physics and mathematics. It is acknowledged as a complex area of research with many variables. Pang et al., in a review of uncertainty visualisation state that:

‘With few exceptions, most visualisation research has ignored or separated the presentation of uncertainty from data. Part of the reason is the inherent difficulty in defining, characterizing, and controlling the uncertainty in the visualisation process’ (Pang et al. 1997, p. 370). This may be a contributing reason as to why there has been no significant work in uncertainty visualisation relating to the field of archaeological reconstructive visualisation.

Pang et al. acknowledge that research in the field of uncertainty visualisation has 'great potential in a wide variety of applications' (Pang et al. 1997, p. 370). Examples of applications include, but are not limited to, simulations data; quantitative and visual analysis; finite element analysis; and volumetric data sets to name a few. The authors point out though that the ' common underlying problem in these areas is visually mapping data and uncertainty together into a holistic view.' (Pang et al. 1997, p. 370). The authors review several established uncertainty visualisation methods and exemplify them in the application areas of Radiosity, Meteorology, and Geovisualisation. The methods reviewed are:

• The use of glyphs,

• The addition of geometry,

• Geometry modification

• Attribute modification

• Animation

• Sonification

Despite the development of uncertainty visualisation methods as described by Pang et al. very few of these methods have been applied in the field of virtual reconstruction. Demetrescu 2015 identifies a ‘major problem’ with current virtual reconstruction methodologies being the ‘difficulty of representing and dealing with uncertainty.’ (Demetrescu, 2015, p. 43). Demetrescu argues that in archaeological virtual reconstructions the 3D model is considered a ‘tool with which to synthesize and convey different elements, each with varying degrees of reliability’ (Demetrescu, 2015, p. 44).

2 Literature Review 49

This effectively masks the processes behind the reconstruction and lead to a ‘black box’ effect (Demetrescu, 2015, p. 44). A number of researchers in the field of archaeological virtual reconstructions have attempted to represent uncertainty in the field virtual archaeological reconstructions by implementing model validation processes (Demetrescu, 2015, p. 44). Few of these methods draw on some of the visualisation principals detailed above.

Zuk et al. employed animation techniques to 'visualise temporal uncertainty of multiple 3D archaeological data sets with different dating.' (Zuk et al. 2005, p. 99). The authors created an interface that allowed users to control the 'temporal position of the time window.’ (Zuk et al., 2005, p. 101). The user moves a slider through time periods controlling animated visual cues for uncertainty data (Zuk et al. 2005, p. 101). The authors used several visual cues in association with animation such as, shading, haze, wireframe views and transparency. These techniques are exemplified in figure 2.36.

Figure 2.28: Archaeological reconstruction interface showing the animated time line (1), and Uncertainty cues. From a to c: no cues, rising/sinking cue, wireframe, and transparency. (Zuk et al. 2005, p. 104) Dudek and Blaise proposed visualising uncertainty in digital archaeological visualisations through ‘graphic variables’ to describe the typlogical details of reconstructed objects (Dudek & Blaise, 2004, p. 2). These graphic variables included Colour, Emissivness, Tranlucency, Highlighting and Symbols to denote uncertainty in

50 2 Literature Review

an reconstructed object. Colouring of an object was used to denote the temporal differences within a reconstruction. Two colours were employed, (figure 2.29) Red and Blue. Red denotes the object is of a later time period to the sourrounding reconstruction and blue an earlier period.

Figure 2.29: Rendering of the reconstructed 1790 historical centre of the city of Cracow (Poland) showing earlier and later time period buildings (Dudek & Blaise, 2004, p. 14) Emissive colour and translucency (Figure 2.30) were used to highlight reconstructed elements of the virtual reconstruction whose ‘documentation’ (input sources) were not yet analysed or the reconstruction was based on ‘interpretation’(Dudek & Blaise, 2004, p. 15)

Figure 2.30: Rendering of the reconstructed 1570 historical centre of the city of Cracow (Poland) showing input sources not yet analysed (circled) and translucent features resulting derived from interpretation (Dudek & Blaise, 2004, p. 15)

2 Literature Review 51

Bonde et al., proposed an interactive visualisation (Figure 2.31) to describe the chronological uncertainties of the monastery of Saint-Jean-des-Vignes (Soissons, ) reconstruction. The monastery evolved from a Romanesque church to a Gothic and then late 17th Century structure. During this period many buildings have been destroyed and others erected. The aim of the reconstruction was to provide an 3D digital environment that could be used for ‘…understanding and reconstructing monastic life at the abbey of Saint-Jean-des-Vignes, Soissons, during the medieval and early modern periods’ (Bonde et al., 2009, p. 364). It was the intention of the authors to ‘…present alternatives and levels of uncertainty in order to lead to critical analysis and debate’(Bonde et al., 2009, p. 365). An interactive ‘site plan’ was developed that allowed the use of a time-slider to view reconstruction by era as well ‘clickable’ objects allowing access to the ‘evidence options’ available for each reconstruction (Bonde et al., 2009, p. 367).

Figure 2.31: Interactive ‘site plan’. Use of a time-slider to view reconstruction by era and ‘clickable’ objects allowing access to the ‘evidence options’ available for each reconstruction (Bonde et al., 2009, p. 367)

52 2 Literature Review

2.8 CURRENT CLASSIFICATION SYSTEMS FOR VISUALISING UNCERTAINTY There are many sources of uncertainty in data as identified by authors such as Pang et al 1997 and Gershon 1998. In many cases the sources and types of uncertainty are abstract and complex (Thomson et al., 2005). It is necessary therefore to try and classify uncertainty as best as possible to be able to meaningfully include it in any virtual reconstruction. The classification of uncertainty allows users to order to information such that there is a cohort method display. In effect classification of uncertainty allows users ' more accurate depictions of critical data sets so that people can make more informed decisions.' (Skeels et al. 2010, p. 70). There has been a growing body of work in the classification of uncertainty in recent times, the more notable are discussed. MacEachren et al identify seven challenges facing uncertainty visualisation in the Geographic Information Systems (GIS) Sciences (MacEachren et al., 2005). The challenges identified by MacEachren are directly applicable in the context of virtual reconstruction.

These challenges are: 1. Understanding the components of uncertainty and their relationships to domains, users, and information needs 2. Understanding how knowledge of information uncertainty influences information analysis, decision making, and decision outcomes 3. Understanding how (or whether) uncertainty visualisation aids exploratory analysis 4. Developing methods for capturing and encoding analysts’ or decision makers’ uncertainty 5. Developing representation methods for depicting multiple kinds of uncertainty 6. Developing methods and tools for interacting with uncertainty depictions 7. Assessing the usability and utility of uncertainty capture, representation, and interaction methods and tools

(MacEachren et al., 2005) Thomson et al. present a typology that provides a general categorization for the types of uncertainty present in geospatially referenced information. The typology describes the specific uncertainties faced by intelligence analysts when tasked with creating an analytic product. Graphical components of the visualisation that are related

2 Literature Review 53

to information uncertainty can be linked to specific aspects of the typology. This allows the typology to combine and propagate uncertainties as well as information about the types of visual components that are most suited for representing specific types of uncertainty (Thomson et al., 2005). The authors present a general typology for 'geospatially referenced information' that attempts to address the ' types of uncertainty intelligence analysts face.' (Thomson et al. 2005, p. 152). Table A-1 of the Appendix presents the uncertainty typology for the goal of creating an intelligence analytic product. Although this typology is designed around intelligence gathering the generality of it make it applicable in other fields of visualisation such virtual archaeology. The typology is comprised of nine categories (sources) of uncertainty in data.

The authors conclude that 'a visualisation that maps each type of uncertainty into a different visual dimension will permit the analyst to express the implications of each dimension separately and to perform cognitively the fusion of the different type of uncertainties.' (Thomson et al. 2005, p. 155). The research did not present an evaluation of the applicability of the typology to practical visualisation. There are numerous elements in the typology which have applicability in the virtual reconstruction process when attempting to isolate sources of uncertainty. This can be exemplified with the categories of ‘Precision’ and ‘Completeness’. The precision and completeness of the data gathering of an archaeological excavation will have a direct influence on a resulting virtual reconstruction. There are many data sources used however that are difficult to quantify in terms of relative uncertainty using the typology proposed.

Walker et al. provided a conceptual framework for the 'systematic treatment of uncertainty in model-based decision support activities such as policy analysis, integrated assessment and risk assessment.' (Walker et al. 2003, p. 5). The authors distinguish two types of uncertainty. Uncertainty arising from lack of knowledge and 'uncertainty due to variability inherent to the system under consideration.' (Walker et al. 2003, p. 8). An 'uncertainty matrix' is proposed to provide a 'tool' that gives a graphical overview of uncertainty features in the models being used for decision support (Walker et al. 2003, p. 14). Table 2.3 shows the uncertainty matrix proposed by the authors. The vertical axis identifies the location of uncertainty within the decision support model while the horizontal axis identify the level and nature of the

54 2 Literature Review

uncertainty. The three levels of the uncertainty utilise determinism as a ‘…limiting characteristic at one end of the spectrum’ and indeterminacy or irreducible ignorance at the other end. Put simply, at one end of the scale a statistical value can be placed on uncertainty while at the other end ‘…neither research nor development can provide sufficient knowledge about the essential relationships.’ (Walker et al., 2003). The uncertainty matrix makes attempts to include uncertainty that cannot be directly defined using statistical methods. The approach is strongly geared to a very specific domain and outcome. No evaluation is made regarding the applicability of the matrix to practical decision support applications.

Location Level Nature Statistical Scenario Recognised Epistemic Variability uncertainty uncertainty ignorance uncertainty uncertainty Context Natural, technological Economic, social and political, representation Model Model structure Inputs System Data Parameters Model Outcomes

Table 2.3: The Uncertainty Matrix proposed by (Walker et al. 2003, p. 15)

Gershon attempts to describe uncertainty away from the constraints of a specific domain. The author proposes a taxonomy of 'imperfect knowledge' (Gershon 1998, p. 43). Unlike Thomson et al, the proposed high-level taxonomy includes 'corrupt data', ‘incomplete data', 'inconsistency' of data, 'too complicated information', and 'imperfect presentation' (Gershon 1998, p. 43). Figure 2.37 demonstrates Gershon's high-level taxonomy of 'imperfect knowledge of the information state' (Gershon 1998, p. 43).

2 Literature Review 55

Figure 2.32: Gershon's High-level taxonomy of the causes of imperfect knowledge of the information state (imperfection). (Gershon 1998, p .43) As with many earlier classifications of uncertainty this one does not discuss or include commonalities that may exist with other domains. Skeels et al. note that this point is endemic to many endeavours in the field of uncertainty classification stating that 'epistemological differences between fields add to the difficulty of creating a unified classification of uncertainty' (Skeels et al. 2010, p. 72). They go on to point out that many such classifications do not account for the 'additive' nature of uncertainty (Skeels et al. 2010, p. 72). A final point that the authors make is that most models of uncertainty have not been evaluated empirically and therefore it is difficult to choose between classifications as well as possibly integrate them (Skeels et al., 2010).

In light of previous research Skeels et al. propose a classification that draws on previous studies in order to ' describe uncertainty for information visualisation.' (Skeels et al. 2010, p. 72). They base their classification on consistencies and similarities found across previous classifications and identify five types of uncertainty.

1. Approximation 2. Prediction 3. Disagreement or inconsistency 4. Completeness 5. Credibility

A classification system was developed that incorporates ' multiple kinds of certainty and different levels of uncertainty in a data set or process.' (Skeels et al. 2010, p. 74). The classification development was driven by responses from questionnaires and interviews conducted with 'knowledge workers, scientists and

56 2 Literature Review

researchers' from'a variety of domains.' (Skeels et al. 2010, p. 73). Figure 2.38 below illustrates the resulting classification.

Figure 2.33: Credibility and Disagreement uncertainty span levels and sometimes disagreement results in credibility uncertainty. (Skeels et al. 2010, p. 75)

Measurement precision applies to the very beginning of the visualisation process. This class covers any ' variation, imperfection or theoretical precision limitations in measurement techniques that produce quantitative data.' (Skeels et al. 2010, p. 75). In the case of archaeological reconstruction, both virtual and physical, this is can be the recording of material remains and the collation of analogous data. The level of measurement precision of surviving remains will determine what level of uncertainty would be inherited at this stage. The physical proximity and dating of an analogous structure will also be a point of introduction of uncertainty. The imprecision discussed here can be represented by a range of values such as confidence intervals (Skeels et al., 2010).

Completeness of data and data sampling techniques will impact the level of uncertainty being inherited in a 3D reconstruction of archaeological material remains. Missing values represent incompleteness and therefore uncertainty but are different to the issue of sampling. The authors have determined a level of uncertainty awareness, see figure 2.39, in order to separate out 'Known Known, Unknown Knowns', and Unidentified Unknowns' (Skeels et al. 2010, p. 76).

2 Literature Review 57

Figure 2.34: Level of uncertainty awareness: (1) Known Knowns: what you know; (2) Unknown Knowns: what you need to find out; and (3) Unidentified Unknowns: what you don't even know you need to find out. Participants described the unidentified unknowns, or unrecognized information needs, as worst case. (Skeels et al. 2010, p. 76)

There is an inherent level of incompleteness (missing values) with archaeological sites. Most virtual reconstructions are of structures, objects or buildings that are largely absent (Bonde et al., 2009). Sampling can imply 'deliberate extrapolation from a few specific measurements to cover a larger set of possible measurements' (Skeels et al. 2010, p. 75) Once again, in the area of 3D reconstruction of archaeological material remains there is an inherent uncertainty in sampling. In the traditional sense sampling can be described as multiple measurements of the same process, phenomena, or item to ultimately achieve a consensus result. This may not be the case when recording material remains such as buildings. However, it is possible to apply this idea by quantifying how much of a site has been uncovered and analysed. Other discoveries within the same archaeological site may shed light on a reconstruction or add more data to the process.

The aggregation of data can have a profound influence on uncertainty levels. Aggregation is essentially the summarising of data, this inevitably leads to the loss of data and therefore the 'data are no longer complete.' (Skeels et al. 2010, p. 75). In the case of reconstructing archaeological material remains it is often the case that data needs to be collated from a variety of sources and then aggregated into a presumed coherent physical outcome. It is impossible to avoid the inclusion of uncertainty during this process. Inference is a broad class that covers ' prediction and extrapolation.'

58 2 Literature Review

(Skeels et al. 2010, p. 77). There is clearly a strong link between inference and the decision-making process. It is clear from the proposed classification scheme that Level 2 of the classification has a profound impact on Level 3. In relation to the reconstruction of archaeological material remains there is a high degree of inference. This inference can be informed by experience, other examples, past records, practicality etc., but, in general, all these are an inference.

Both Disagreement and Credibility span all three levels in varying degrees. Disagreement, in and of itself, will lead to uncertainty. This can be caused simply by measurements of the same object by differing people giving different results, or as in the case of archaeological reconstruction, two separate researchers arguing the possible original form of a building from existing floor remains and their individual knowledge base. Credibility is a type of uncertainty that is very difficult to measure (Skeels et al., 2010). For example, a 'human source may be considered untrustworthy based on past behaviour or associations', or the veracity of sources of information being used cannot be adequately verified (Skeels et al. 2010, p. 76). In the case of virtual archaeological reconstruction this is very common as researchers may rely on experiential knowledge, personal suppositions, texts from ancient writers or stylised images of a structure from antiquity etc.

The primary aim of the development of this classification was the ability to span multiple domains in its applicability. Table A-2 shows examples of uncertainty from three separate domains, Ecology, Computational Biology, and Medicine. Skeels et al. aimed to create a classification that spans multiple domains and in doing so provide a medium by which to ' to discuss and understand uncertainty so that we can find good ways to display it to a wide range of users.' (Skeels et al. 2010, p. 80). The authors acknowledge that the next step is to uncover ways to visualise this uncertainty. Suggestions are put forward as to how this may be achieved, for example the use of graphical layers, but do not go as far as implementation. This is an overriding theme that is apparent from the research into uncertainty classification, significant work is being invested in classification systems, but this has not yet transferred to methods of visually depicting the classifications in a meaningful manner.

2.9 SUMMARY We have seen from the evidence presented that there has been significant effort invested in the visualisation of digitally reconstructed material remains of buildings

2 Literature Review 59

and other archaeological artefacts. There is a clearly defined need for technologies such as VR in the domain of archaeology.

Denard compared VR technologies to traditional methods of communicating the reconstruction of material remains and concluded that static images, real-time environments and video may well be 'the expressive medium of choice, conveying, all at once, the complete, synthetic image of the author’s idea.' (Denard 2011, p. 60). Jablonka et al. concluded that VR technologies are an effective medium to make scientific data and results available to researchers while at the same time making them more accessible to the public (Jablonka et al., 2002, p. 1).

Virtual reconstructions have been termed ‘knowledge representations’ and can be used as hypothesis ‘to be questioned and tested.’ (Bonde et al., 2009, p. 363). This is also supported by Olsen et al. who state that the technologies discussed ‘can not only assist archaeologists with scientific discovery but also enrich the presentation of findings and enable the distributed scientific communication and discovery.' (Olsen et al. 2004, p.3).

Cargill also believe that VR reconstructive techniques offer a strong visual medium in which to communicate, debate, and share information (Cargill, 2009). Petrova et al. see the use of VR technologies as a means to effectively achieve preservation and transmission of cultural heritage (Petrova et al., 2011). Following the point of cultural preservation, Almagro points out that due to legislative and ethical restrictions the physical reconstruction or anastylosis of structures is highly problematic. VR technologies help to significantly overcome these issues while still allowing researchers to investigate such aspects as ‘visual perception, the importance to the space of aspects such as the colour and texture of the materials, the effects of light and even the scale of the building.' (Almagro 2007, p. 162) Almagro states that there are two major advantages to the use of VR technology in archaeology, it facilitates reflection and transmits information. (Almagro, 2007)

With the identified need for VR reconstruction and visualisation techniques there has been a plethora of VR visualisations in the archaeological domain. The case studies show that there is a wide use of VR across a range of archaeological applications. Neubauer et al. used VR techniques to recreate the Roman Gladiator School of Carnuntum, Austria. The authors used a combination of digital techniques to record buried remains of the structures and then visualise these remains. The authors state

60 2 Literature Review

that 'the impact of the discovery is made all the greater by the stunning reconstruction images that the project has generated.' (Wolfgang Neubauer et al., 2014, p. 173).

VR techniques were used in the digital restoration of the missing San Jacinto Church in Caracas (1595 -1821) by Gasparini et al. The reconstruction assisted the authors 'to define an educational project and implement a combination of new digital low cost methodologies applicable to teaching,' (Gasparini et al. 2012, p. 11). In a similar project Petrova et al. (2011) used digital reconstructive techniques to reconstruction of the fresco paintings of the Church of the Transfiguration of Our Saviour on Nereditsa Hill in St. Petersburg, Russia. (Petrova et al., 2011)

Issini used VR reconstruction techniques to recreate the damaged Yanqin Section of the Chinese Great Wall in order to assist in the physical anastylosis of the section. The 3D model was then further used as a base for an interactive information retrieval system. (Issini, 2012) Zarmakoupi (2010) makes extensive use of VR modelling and reconstruction techniques to inform a wider physical reconstruction project. The author concluded that 'part of the flexibility of this virtual reality reconstruction is the ability to select among existing state and different restoration proposals.' (Zarmakoupi 2010, p. 185). The anastylosis of the Octagon in takes these reconstructive techniques one step further. The project tested the digital reassembly of existing fragments of the Octagon monument. Digital reconstructive techniques were used by Das & Garg to develop visual interpretations of the pavilions described in a traditional ancient Indian architectural treatise, the Mayamatam (Das & Garg, 2011). In a similar project Rashid et al. (2008) develop an interactive virtual model of the lost architectural heritage of the 8th century Buddhist Monastery of Sompur Mahaviahara in Bengal.

Along with the growth of VR visualisation for archaeological reconstruction there is increasing concern for the transparency and veracity of visualisations being presented to the public. In order to be able to present effective visualisations of archaeological material remains there is a need to present complete and accurate representations of the data used to derive the final 3D reconstructions (Johnson & Sanderson, 2003). The issue of uncertainty in the visualisation process is now being addressed with increased vigour. Detailed investigations into of sources of uncertainty in data have been undertaken by researchers such as Gershon (1998), Thomson et al. (2005), and Gahegan, and Ehlers (2000). These researchers identify areas such as

2 Literature Review 61

corrupt data, incomplete data, credibility, data acquisition, data transformation, and data visualisation as sources of uncertainty.

Much work has been done in visualising uncertainty in data as can be seen by Pang et al. (1997), Gershon (1998), and Thomson et al. (2005) to name a few. Most of the research conducted into visualising uncertainty has been in the fields of Geovisualisation, Physics, Fluid Dynamics, Intelligence Gathering, and Meteorology, as discussed earlier.

It is interesting to note that, with the exception of, Zuk et al. (2005), Hynst et al. (2001), Blaise and Dudek, (2009), Bonde et al., (2009), and Zarmakoupi (2010), the above accepted methods of visual representation of uncertainty have not been applied in digital archaeological investigations and visualisations. Demetrescu 2015 notes that application of these uncertainty visualisation approaches ‘lack a common and agreed- upon standard’ and therefore do not present a structured and unified approach to uncertainty visualisation in virtual reconstructions (Demetrescu, 2015, p. 44).

Although there are accepted techniques of visualising uncertainty, the application of these techniques is often hampered by how to classify uncertainty. Thomson et al. point out that in many cases the types of uncertainty can be complex and abstract in nature (Thomson et al., 2005). This situation has led researchers to try and classify uncertainty in order to visualise it. Skeels et al. note that in order to understand data and make informed decisions, information needs to ordered and categorised (Skeels et al., 2010). There has been an increased focus on the classification of uncertainty as evidenced by MacEachren et al (2005), Thomson et al (2005), Walker et al (2003), Gershon (1998), and Skeels et al. (2010). In all cases, but Skeels et al., the authors propose classification systems that are specific to their domain interests. Skeels et al. propose a classification system for uncertainty that attempts cross domains in its application. At issue with all classification systems is that most models of uncertainty have not been evaluated in a practical application and therefore it is difficult to determine their effectiveness (Skeels et al., 2010). Part of the reason may be that the appropriate approach to evaluation of a visualisation of uncertainty is not clear. Despite this issue perhaps the most important point is that, in all cases of classification, there has been no attempt to transfer the classification system to a visual domain. Skeels et al. state that the next step in the development of uncertainty

62 2 Literature Review

classifications needs to be identifying methods to visualise the uncertainties that have been classified (Skeels et al., 2010).

2 Literature Review 63

3 Research Design

This chapter is comprised of four sections. These are: 3.1 - Research Design

o Research Theory o Research scope o Research stages

3.2 - The data classification system

o The modified data classification system

3.3 - The Workflow Pipeline

o The modified Workflow Pipeline

A workflow pipeline is proposed for the reconstruction of archaeological structures that is informed by classified data inputs. The classification of inputs is dealt with in chapter 4. The workflow pipeline is applied to a virtual reconstruction of the 17th century Rose Theatre discovered in London’s ‘south bank’ area in the United Kingdom. The results of this application are analysed, and issues associated with the workflow pipeline are identified. Changes are proposed for testing in subsequent digital reconstructions.

3.1 RESEARCH DESIGN

3.1.1 Research Methodology This research is based on Practice-Led research methods. The foundation principal of this research is that practice facilitates ‘research insights’ (Smith & Dean, 2009, p. 5). Additionally, it is the intent of this research to form ‘new understandings’ about the ‘practice’ and processes of digital archaeological reconstruction (Candy & Edmonds, 2018, p. 64). A core tenet of Practice-Led Research is that the practice itself is ‘embedded in the research process’ and that ‘research questions arise from the process of practice’(Candy & Edmonds, 2018, p. 63). This research aims to lead to ‘new knowledge that has operational significance’ for the practice digital archaeological reconstruction (Candy & Edmonds, 2018, p. 65). The nature of this research is that it depends heavily on ‘an iterative cycle of practice-led research and research-led practice’ that is representative of the processes of digital archaeological reconstruction (Candy & Edmonds, 2018, p. 64). The research methodology employed

3 Research Design 65

involves iterative testing cycles within each stage in which ‘results from tasks’ determine the usefulness and applicability of the process under investigation (Candy & Edmonds, 2018, p. 64).

As previously covered, much of the data used in archaeological reconstructions is primarily qualitative. The exception to this is the available quantitative data such as that sourced from archaeological reports. By their very nature archaeological excavations are limited in the amount of available quantitative data that can be gathered. Archaeological sites are incomplete often only yielding fragments of a structure’s original form. Much of the information used to recreate a lost structure is based on inference and examples from other sites. It is this very inference that this research aims to clarify and present as a qualitative measure of authenticity. Practice-Led Research methods offer a way of sorting and assessing all the available data used in the decision-making process of a virtual archaeological reconstruction. Data used in a reconstruction can be assessed through the iterative process of testing options. In the initial stages of the research data can be assessed to determine veracity, usefulness, and pertinence. The second phase is focused on application where the data selected for use in a digital reconstruction can be ‘tested’ in the actual reconstructive environment.

3.1.2 Research Scope Archaeological reconstruction, whether it be digital or physical ‘Anastylosis’, covers a broad area of activities. The reconstruction of an ancient amphora from surviving extant remains and existing examples is equally as challenging as the reconstruction of an ancient building or mechanical device. Although there may be similar challenges and approaches across the many forms of archaeological reconstruction this research focuses only on the reconstruction of buildings. The research is not constrained by a time-period, era or building type. The primary focus of this research is the data available for use in a reconstruction and the impact of that data on the final reconstruction in terms of accuracy and therefore authenticity.

Primary Research Questions

1. RQ1: What methodological pipeline is appropriate for archaeological digital reconstruction? 2. RQ2: How is the input data used for digital reconstruction assessed and classified? 3. RQ3: How is the authenticity of the reconstruction assessed? 4. RQ4: Can input data be assigned classes that reflect the authenticity of a digital reconstruction? 5. RQ5: How can a measure of authenticity be meaningfully presented in a digital reconstruction?

66 3 Research Design

3.1.2 Research Stages Figure 3.1 presents an overview of the five research stages including, the aims of each stage and, the research questions addressed.

3.1.3 Stage 1 – Assessing methodological construction pipelines, data classification methods and authenticity measures for archaeological digital reconstruction

Research Aim: Research current methodologies and systems available to digital archaeological reconstructions. A review of existing research literature revealed that the issue of methodological construction pipelines for archaeological reconstruction has not been addressed in any meaningful manner. Extensive discussion has centred on the possibilities and pitfalls presented by digital tools. Among researchers there is a growing consensus recognising the dangers inherent in any reconstruction project with limited information and the subsequent presentation. It is acknowledged that the issues faced in physical reconstruction of archaeological ruins, are, in many cases, transferable to digital reconstructions. There has been a concerted effort to standardise digital reconstruction methodologies and rigour. This is being partially addressed by the London charter for the Computer-based Visualisation of Cultural Heritage in 2006 (Denard 2011), and to some extent, the ICOMOS Interpretation and Presentation of Cultural Heritage Sites (Silberman 2008). Despite these efforts only limited work has been carried to identify an effective and robust construction pipeline. Agile software development methodologies which encourages iteration in discrete phases are reviewed for applicability to a construction pipeline methodology.

Extensive research has been undertaken in data classification. The application of data classification in the field of archaeological data has seen far less attention. Although a number of researchers have explored the issue of authenticity and data veracity in archaeological reconstruction, to date there has been no dedicated efforts to implicitly integrate classified data with a construction pipeline.

Stage 1 identifies current digital reconstruction and data classification systems and draws relevant material to formulate an input data classification system and reconstruction pipeline.

3 Research Design 67

• Assess current methodological pipelines (if any) employed for STAGE 1 archaeological digital reconstruction • Assess current methods of data classification and authenticity measures o RQ1, RQ2, RQ3, RQ4, RQ5

• Assess and classify the input data used in an archaeological STAGE 2 reconstruction o RQ2, RQ3, RQ4

• Develop and implement a methodological pipeline for the digital STAGE 3 reconstruction of a building from classified input data o RQ1, RQ2

• Review modify the data classification system and methodological STAGE 4 pipelines based on stage 3 outputs o RQ1, RQ2, RQ3, RQ4

• Present measures of authenticity and data classes in a meaningful visual STAGE 5 medium o RQ4, RQ5

Figure 3.1: Summary Stages - Research Design

68 3 Research Design

3.1.4 Stage 2 – Assessing and classifying the input data used in an archaeological reconstruction Research Aim: Create an input data classification system. Many disparate data inputs are employed in the reconstruction of an ancient mass structure such as a building. These data range from quantitative data such as extant remains and archaeological reports to qualitative sources such as literarily reference, drawings and analogous structures. In general, a reconstruction, be it digital or physical will employ as many data sources as possible to arrive at a final reconstruction. The value of the available data and the subsequent assessment of authenticity relies on the establishing the accuracy and veracity of the inputs so that it is clear how they influence the decision-making process. Classification offers a pathway to arranging the input data in clear levels of accuracy and veracity.

This stage explores the use of a classification system and data decision tree to assign classes to data for use not only in the decision-making process of the construction pipeline, but to also as a measure of authenticity of the final reconstruction. Input data are assigned numerical classes. The numerical classes, in turn, are used to calculate a final generalised authenticity score. The premise is that input data impacts the decision-making process and therefore by extension, will impact the accuracy and authenticity of the final reconstruction. There is a distinct lack of transparency with regards how this data is used in a reconstruction. The classification of data into logical and easily understood classes serves to increase the transparency of a virtual reconstruction by making input of that data in the reconstruction process more visible.

The classified data is applied in Stage 3 with the implementation of the reconstruction pipeline.

3.1.5 Stage 3 – Developing and implementing a methodological pipeline for the digital reconstruction of a building from classified input data Research Aim: Develop and implement a methodological pipeline for the digital reconstruction of a building from classified input data Much work has been published on the subject of virtual reconstruction of lost buildings as has been covered previously in this study. However, very little research has been devoted to the development of a pipeline specifically adapted to virtual reconstruction. Each reconstruction is approached based on the individual needs and desired outcomes of the author. This is not to say that these specific pipelines are incorrect but that they generally lack a standardized approach. The London charter for the Computer-based Visualisation of Cultural

3 Research Design 69

Heritage attempts to address this issue with broad non-compulsory guidelines. This study attempts to address the issue through the development of a repeatable pipeline that is directed by classified data inputs.

The pipeline is initially applied to the reconstruction of the 16th Century Rose Theatre in London. Classified data inputs are used in each stage of the pipeline to arrive at a final reconstruction.

3.1.6 Stage 4 – Reviewing and modifying the data classification system and methodological pipelines based on stage 3 outputs Research Aim: Review and modify the data classification system and methodological pipelines. An initial procedural pipeline and data classification system is established and applied in Stage 3 of the study. The results of the final reconstruction of the Rose Theatre are analysed to identify any issues with both the procedural pipeline and the classification system. Numerous deficiencies are identified in the classification system with regards data classes and authenticity measures and are addressed in this stage. These deficiencies are discussed in detail in subsequent sections of this study. Stage 4 revisits research questions 1, 2, 3 and 4

• RQ1: What methodological pipeline is appropriate for archaeological digital reconstruction? • RQ2: How is the input data used for digital reconstruction assessed and classified? • RQ3: How is the authenticity of the reconstruction assessed? • RQ4: Can input data be assigned classes that reflect the authenticity of a digital reconstruction? The procedural pipeline for reconstruction is analysed and found to be appropriate as it is assisted by discrete stages of construction with embedded feedback loops. The pipeline is demonstrated to be robust although this is influenced by the classified data inputs. The initial classification system is found to be flawed. The approach proposed to calculate the average score was overly simplistic and did not account for data variation appropriately. The assigning of data inputs to classes of uncertainty is a subjective process and numerous subjective decisions have a demonstrated impact on the accuracy of the computed values. The computed value of uncertainty/authenticity for the entire reconstruction did not contribute anything meaningful or of value from an interpretation standpoint. The deficiencies of the original classification system are dealt with in detail in subsequent sections.

70 3 Research Design

A modified classification system employing Fuzzy sets is developed. Fuzzy sets offer a way to order data classes that are somewhat ‘fuzzy’ with a continuum of grades and which reflect the nature of the input data used (Zadeh, 1965, p. 338). As already detailed, input data used in archaeological reconstructions is often qualitative. Inputs that include literary texts, drawings and analogous objects to name a few are not measurable in a quantitative sense. The classification of such inputs using a numerical value alone is difficult and the use of Fuzzy sets allows more flexibility in assigning inputs to logical classes. The use of numerical classes is discontinued and replaced with a colour coded system of classes.

3.1.7 Stage 5 – Presenting measures of authenticity and data classes in a meaningful visual medium Research Aim: Display measures of authenticity and data classes in a meaningful visual medium In stage 5 the application of the modified classification system is applied in this to several other examples of archaeological reconstructions. These specific reconstructions are discussed in detail in subsequent chapters.

This stage of the research addresses research questions 4 and 5

• RQ4: Can input data be assigned classes that reflect the authenticity of a digital reconstruction? • RQ5: How can a measure of authenticity be meaningfully presented in a digital reconstruction? In this stage two visual methods of displaying the authenticity of a reconstruction are presented. These are: 1. Class Colour per Element Display 2. Shading of Known Data Entities Class colour per element display involves assigning data class colours (derived from Stage 4) to individual elements or group of elements of the final reconstruction. Elements of the reconstruction are displayed as classes allowing the interpreter to view the element in relation to the data input source. This proves to be a viable method of improving transparency of data input and as a viewer assessment measure of authenticity.

Shading of known data entities involves the texturing of known elements of a reconstruction. That is, applying real-life textures to elements of the reconstruction that are known whilst leaving all other construction elements grey. This method allows the user to

3 Research Design 71

determine actual data inputs from calculated or assumed inputs. Neither display method is proposed over the other. A merging of the two methods offers future possibilities.

3.2 CLASSIFICATION SYSTEM In previous sections of this document attention has been drawn to the issues surrounding data uncertainty and the implications for resulting reconstructions, see Johanson 2009 and Denard 2011. The classification system developed for this study is based on values of uncertainty assigned to input data. Input data used in a virtual reconstruction such as extant archaeological remains can generally be assumed to have a low level of uncertainty with very clear influences on the outcomes in the reconstruction. Anecdotal evidence, on the other hand, can generally be assumed to have a high level of uncertainty but not necessarily a low level of accuracy. Researches such as Beacham and Kalisperis point out that the outcomes with uncertain data used in the reconstruction can, in themselves, be profound and may lead to assumptions that are incorrect or, an interpretation presented as fact. (R. C. Beacham, 2011) (Sorin Hermon & Kalisperis, 2011)

A Decision Tree has been developed which allows input data to assigned to uncertainty classes, see figure 3.2. The decision tree is used to separate data into two decision trunks. These are, data comprised of ‘Physical Evidence’ and data comprised of ‘Non-physical Evidence’ (anecdotal evidence, written descriptions etc). In the case of ‘Physical Evidence’, the tree branches at the first level into physical extant remains and documentary evidence. Both these branches are then split into fuzzy classes of uncertainty. Three fuzzy classes (levels) of uncertainty are derived at the final level of branching.

• Low Uncertainty – (Numeric value – 1) • Medium Uncertainty - (Numeric value – 3) • High Uncertainty - (Numeric value – 5) The classes are informed using a numerical tabular matrix which is detailed below in Table 3.1.

For data comprised of physical evidence uncertainty classes are assigned using qualitative evaluation to determine the amount and quality of physical evidence. The extent of the remaining archaeology, the accuracy of the recordings and the reputation of the recorder for example, all play a role in the assigning of classes. Data drawn from documentary evidence (photography, sketches, reports etc.) are assigned uncertainty classes that are determined by quality and veracity. For example, photographic evidence of an original building coupled with

72 3 Research Design

detailed plan drawings can be considered low uncertainty, while a ‘not to scale’ hand drawn sketch of unknown would represent high uncertainty. A photograph of unknown origin and date has an increased uncertainty as the context of the image cannot be determined.

The anecdotal evidence section of the decision tree branches, at the first level, into construction techniques and analogous examples. Construction techniques uses known building techniques of the time coupled with engineering principals and extant remains to determine structural elements of the reconstruction. Where there exist significant extant remains that support the placement and form of the reconstructed element it can be assigned a low uncertainty. Conversely, where little or no extant remains exist to support the proposed construction element a high uncertainty class is assigned. For data sourced from analogous structures the uncertainty classes assigned are determined by the degree of analogy, geographical location, era of build and the like.

3 Research Design 73

Figure 3.2: Decision tree detailing the process of assigning data a class of uncertainty.

74 3 Research Design

Another similar structure existing onsite and clearly of the same period can be considered data with low uncertainty. In contrast, a structure from another site but exhibiting similar features to the structure of interest will be considered a higher uncertainty data class.

To better quantify the concept of a level of data uncertainty, numerical values based on classification criteria detailed in the matrix below (Table 3.1) are used. The classification system proposed, determines levels of authenticity based on four broad uncertainty classes.

The four classification criteria are:

I. The level of uncertainty associated with the archaeological evidence II. The level of uncertainty associated with the anecdotal, literary or cultural evidence present III. The level of uncertainty associated with the similar analogous structures or methods of construction IV. The degree to which the reconstruction is supported by known construction methods/materials of the time-period and the of physics.

The four broad classes as defined by the decision tree, allow input data to be sorted into discrete classes. The degree of uncertainty for each broad class is determined by the value of the sub-classes. This means that input data is identified as a discrete piece of information that has properties describing how reliable it is. Each piece of input data used is classified using the decision tree into one of four broad classes, the data is further allocated to subclasses based on the uncertainty classification matrix below.

3 Research Design 75

Uncertainty Class Description

A very low level of uncertainty associated with: • Complete or extensive archaeological evidence – I

• Very detailed documented evidence, literary or cultural evidence – II

• Many similar examples of analogous structures or methods of construction available on 1 site or geographically close – III The digital reconstruction is supported by known construction methods/materials of the time-period and the laws of physics. – IV

A low level of uncertainty associated with:

• Substantial significant archaeological evidence – I

• Detailed documentary, anecdotal, literary or cultural evidence – II 2 • Similar examples of analogous structures or methods of construction exist – III

• The digital reconstruction is supported by known construction methods/materials of the time-period and the laws of physics. – IV

A moderate level of uncertainty associated with:

• Significant archaeological evidence – I

• Documentary, anecdotal, literary or cultural evidence exists although not detailed – II 3 • Somewhat similar analogous structures or methods of construction – III

• The reconstruction is supported by the laws of physics and known construction methods/materials of the time period – IV

A high level of uncertainty associated with:

• Limited archaeological evidence – I

• Limited documentary, anecdotal, literary or cultural evidence – II 4 • Few analogous structures or exemplified methods of construction – III

• The components of the reconstruction support the laws of physics and some known construction methods/materials of the time period – IV

A very high level of uncertainty associated with:

• Very limited archaeological evidence – I

• Very limited documentary, anecdotal, literary or cultural evidence – II 5 • Lack of analogous structures or exemplified methods of construction – III

• The reconstruction is not supported by the laws of physics and known construction methods/materials of the time-period. – IV

Table 3.1: Uncertainty classification matrix

76 3 Research Design

The classification draws on fuzzy sets as defined by Zadeh and has been created with numerical value classes ranging from one to five. The scale has been deliberately restricted to 5 values with half values possible (eg. 3.5) to maintain a level of operational simplicity. The classes themselves are somewhat ‘fuzzy’ with a continuum of grades that reflect the nature of the input data used (Zadeh, 1965) . Zadeh states that ‘More often than not, the classes of objects encountered in the real physical world do not have precisely defined criteria of membership.’ (Zadeh, 1965, p. 338). The classes used here are not considered classes in the ’usual mathematical sense’(Zadeh, 1965, p. 338). The classification represents a framework which provides a ‘natural way’ to deal with the ‘imprecision’ inherent in the high variation input data (Zadeh, 1965, p. 339). An aim of this methodology is to create a straight-forward analytical system which allows rapid assignation of data to classes whilst avoiding complicated statistical routines which are inappropriate in the context of this study. It is not the aim of this work to create a quantitative analytical system. The adoption of fuzzy set classification is ideal for this purpose as ‘the notion of a fuzzy set is completely nonstatistical in nature.’(Zadeh, 1965, p.340). Fuzzy sets are ’a better way to describe the uncertainty’ of data sources as the application of ‘Boolean true or false (black and white) values’ cannot account for ‘all the shades of uncertainty (grey) between them’(Sorin Hermon & Nikodem, 2007, p. 2).

Using the matrix described above, input data was assigned a fuzzy numerical class of uncertainty. The classes were used in the final reconstruction output as possible indicators of overall authenticity of the virtual reconstruction. The above decision tree and classification system was applied to the Rose Theatre virtual reconstruction and is discussed in subsequent sections.

3.2.1 Review of the Classification Methodology and Decision Tree

Results of the initial classification test with the Rose Theatre that the classification system showed promise with regards the organisation of data inputs and particularly as a method of informing the reconstruction process. However, significant questions were raised as to the rigour of the approach. The approach proposed to calculate the average score was demonstrated to be somewhat overly simplistic and it could be successfully argued that not every part (input) is equal. This is exemplified by the overall impact of three elements in the Rose reconstruction (Stairwells, Building Entrance and Windows) on the overall uncertainty level. The combined impact of these components on the overall reconstruction authenticity

3 Research Design 77

assessment was disproportionate. Similarly, the calculation of a single number as a final measure of authenticity did not reflect possible extremes that may be assigned to individual reconstruction elements.

As the assigning of data inputs to classes of uncertainty is a subjective process, the many subjective decisions have an impact over the accuracy of the computed values. Therefore, the computed value of uncertainty for the entire reconstruction is somewhat flawed. The process of assigning classes for numerical computation is costly in terms of time. Assigning a single numerical value of uncertainty impacts the generalizability of the overall reconstruction and data classification. Finally, the use of a single numerical value as an indicator of authenticity or uncertainty does not afford the viewer any opportunity to evaluate the evidence used to achieve the reconstruction.

The individual classes of uncertainty for data inputs can however be beneficial from a visualisation standpoint and this is discussed in subsequent sections. The process of calculating an overall authenticity score was removed from future iterations of the classification system and the focus shifted to the data input uncertainty classes and how these could be conveyed in a meaningful visual form.

The decision tree has proven to be a very useful tool for rapidly assigning input data to a class of uncertainty. The ability to sort inputs based on evidence type allows for a clearer understanding of the volume and types of data and greatly assisted data organisation. The decision tree allows a user to sort the low uncertainty data inputs and use these to form the basis of reasoning for the subsequent interpretive phases. The decision tree does however have deficiencies. In its original form the decision tree generated numerical values which were shown to be of limited value. The assigning of data inputs to classes is valuable for the reasons stated above, however, the class is poorly represented by a simple numerical value which cannot be translated effectively to a visual form in a 3D digital model and do not represent the variations possible within a class.

A simpler and more effective decision tree was developed. The revised decision tree is presented in figure 3.3. The most obvious modification to the original tree is the combining of the ‘Analogous’ evidence branches into a single. The concept of classifying inputs based on construction techniques is somewhat redundant when using existing examples or analogous structures and building techniques. It added an unnecessary element of complexity to data organisation and evaluation. The initial tree branches were renamed ‘Related Evidence’ and

78 3 Research Design

“Associated Evidence’. Related evidence is effectively any data inputs derived from information directly related to the reconstruction in question. These inputs can range from extensive extant archaeological remains to stories and myths. Associated evidence is data inputs derived from sources not directly related to the reconstruction. These inputs can range from onsite analogous structures to off-site structures of similar era and style. This modification to the decision tree facilitated a much clearer distinction and simpler method of rapidly assigning input data to a class of uncertainty. Additionally, this modification affords a rapid and effective method of distinguishing and sorting data based on evidence type. Data inputs have been categorised into three basic types. These are:

I. Archaeological evidence II. Documentary evidence III. Analogous structures

At the first level of data classification, inputs can be rapidly assigned to one of the three top level categories. Each class of data is then sorted into the final sub-classes of uncertainty based on the volume and veracity of the inputs. The final major change to the decision tree is the addition of colour elements associated with levels of uncertainty.

3 Research Design 79

Figure 3.3: Modified Decision Tree

80 3 Research Design

The addition of colour levels for each major uncertainty class facilitates a method of visualising reconstructed elements based on levels of uncertainty. It also offers a method whereby a viewer can make informed decisions about the overall authenticity of a reconstruction based on the data used. This is explored in detail in following discussions.

The final uncertainty classes continue to retain their numerical values although only as a point of reference. The colours chosen for each data category are arbitrary. A colour was chosen for each category and then segmented into three shades to represent each broad uncertainty class in that category. It is possible that more appropriate colours could be chosen to define categories although this would require significant exploration of colour theory and human perception to determine.

The refined decision tree and colour classification system is applied to the Rose Theatre and other test cases in subsequent chapters and the results discussed.

3.3 THE WORKFLOW PIPELINE A workflow pipeline is proposed that standardises the virtual reconstruction process. The proposed workflow draws from, and expands on, Hermon’s process of 3D modelling for reconstructions in archaeology. The purpose of following this method is that it allows the evaluation of ‘the model’s reliability and allow its (virtual) deconstruction’ (Sorin Hermon, 2008a, p. 40). The pipeline facilitates the identification of inputs that affect the authenticity of the reconstruction. Additionally the workflow draws on agile software development methodologies which encourages iteration in discrete phases (Coleman, 2016). The iterative procedure facilitates the identification of issues within the reconstruction that result from uncertain data inputs and seeks to address them.

The workflow pipeline (See Figure 3.4) is based on Decision nodes (squares), Chance nodes (circles) and End nodes (triangles). The workflow details the reconstruction process while identifying inputs and their associated uncertainty classes. The following phases are identified within the framework of the flowchart.

Phase 1: Digital reconstruction of extant physical remains; this represents low uncertainty values in the data and high authenticity in the reconstruction.

3 Research Design 81

Phase 2: Digital reconstruction incorporating mixed uncertainty data and reasoned inference. This phase represents an iterative process

Phase 3: Final reconstruction based on phases 1 and 2

Decision nodes are driven by available data. Where decision nodes are presented with rounded corners this represents the input of higher uncertainty data. In the case of phase 1, extant remains are reproduced, and this represents a low uncertainty value assigned to available data. Typically, extant remains will incorporate significant elements such as building foundations. Foundations, for example, provide a high level of certainty of the overall floorplan or extents of a building as well as possible wall thicknesses. They do not however, provide information on wall heights or roof type for example. Phase 2 in comparison is the merging of variable uncertainty value data (chance) with the results of phase 1. In this phase, typically other evidence such as analogous structures, archaeological artefacts, descriptive texts etc. are used to supplement the existing high certainty data inputs to arrive at a more complete reconstruction. This process is highly iterative and yet constrained by the known low uncertainty data inputs such as building foundations. This phase requires extensive testing which may well incorporate aspects such as the presumed social use of the reconstructed space to determine if, overall, the reconstruction approaches an acceptable level of authenticity. Phase 3 represents the end node in the decision process where iterations in Phase 2 lead to a final output. This phase is a final ‘logical’ incorporation of all the available data inputs to realise a final reconstruction. This reconstruction is a ‘knowledge representation’ or a visual representation of the state of knowledge of the reconstructed item (Johanson, 2009).

82 3 Research Design

Figure 3.4: Flowchart detailing the reconstruction process and identifying areas of inputs and associated uncertainty classes

3 Research Design 83

3.3.1 Review of the workflow pipeline Overall the workflow pipeline worked well. It was clear from the initial test case of the Rose Theatre that, for the workflow pipeline to be effective, it must be applied at the individual building component level. It is clear that the applying the workflow in the same manner as a physical structure would be erected is very effective. At each phase of the building process the workflow pipeline model allowed for decision making. The workflow pipeline facilitated an iterative decision-making process. This is detailed in subsequent sections.

The second phase of the workflow pipeline facilitated the construction of the building framework and the classification of input data. The iterative nature of Phase 2 allowed for the testing of possible alternative building frame layouts. The first phase of the workflow pipeline is aimed at utilising data inputs with low uncertainty values such as quantified archaeological data. This provides a starting point for a reconstruction where the actual extant remains form an integral part of the reconstruction.

The second phase of the workflow is characterised by the introduction of less certain data in the decision-making process. The iterative nature allowed various options to be tested based on the reconstruction from the first phase of the workflow pipeline. The second phase of the workflow pipeline demonstrated its usefulness with reconstructing the building components that used highly uncertain data inputs.

The workflow pipeline is designed to build upon higher certainty data to arrive a solution. That is, the workflow pipeline relies on first utilizing all the high certainty data available in the first phase. The outcomes of the first phase become in themselves data inputs that influence the second phase of the workflow where more uncertain data is usually introduced. The initial test case of the Rose Theatre suggests that the workflow pipeline required minor refinements to more accurately reflect the actual reconstruction and data assimilation pipelines when put into practice. It is clear from the initial construction of the Rose Theatre that broad sorting of input data occurs at the initial stages of the workflow. ‘Hard’ data such as extant remains, archaeological reports, architectural plans etc. are easily classified and separated from lower fidelity data. Broad groupings occur in Phase 1 to allow for the construction of extant remains whilst creating a grouping of data which requires broader classification. An updated pipeline (Figure 3.5) is proposed which reflects this initial data sorting stage.

84 3 Research Design

In addition to methodology pipeline phase 1 changes, minor revisions have been made to phase 2 resulting from subsequent digital reconstructions detailed later in this document. Phase 2 is a highly iterative process whereby known data is combined with data of higher uncertainty. The process of incorporation and experimentation requires extensive iteration in some cases to arrive at a logical solution. Revisions in Phase 2 reflect this iterative process.

3 Research Design 85

Figure 3.5: Revised flowchart detailing the reconstruction process and identifying areas of inputs and associated uncertainty classes

86 3 Research Design

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

4.1 APPLICATION OF THE CLASSIFICATION SYSTEM TO THE ROSE THEATRE DIGITAL RECONSTRUCTION

Many disparate data sources are used in each component of the reconstruction. In some cases, scant archaeological evidence is supported with extensive solid anecdotal evidence and analogous structures. In other cases, the reverse is true. The uncertainty classification matrix (See chapter 3) was applied to the digital reconstruction of the 1596 Rose Theatre. The theatre existed from 1587 until 1606 when it was demolished. During this period there were various renovations of the building. This study only considers the theatre as it existed in the 1596 iteration. Each data element used to inform the reconstruction of the Rose Theatre was assigned a class using the decision tree. The decision tree allowed input data to be sorted into classes and contributed to a better overall understanding of the reconstruction elements. The use of a decision tree and uncertainty matrix requires that a subjective judgement is made in assigning uncertainty classes. This subjective judgement is evidence based and represents a ‘balancing’ of all the available evidence. The use of fuzzy classes accommodates the subjective nature of these decisions. This classification process assisted in understanding individual discreet data inputs as part of a greater set.

The data input classes derived from the decision tree were applied to the classification matrix for each building component. In many cases several pieces of evidence were used in the reconstructing of a building component. A value of uncertainty for each building element based on classes was calculated from the matrix. The values generated from the uncertainty matrix ranged from 1 to 5 with the number 1 representing very low uncertainty and 5 representing very high uncertainty. The results of the classification are presented in Table A-3.

The thatch roof of the Rose Theatre is a good example of multiple data inputs used to arrive at a solution. Some extant remains of Thatch roofing were unearthed and documented during the excavations of the Rose Theatre (Bowsher & Miller, 2009). These extant remains provided strong evidence of the type of roof material and represented very low uncertainty

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 87

(Uncertainty value = 1.0) input data. In addition to the archaeological evidence, existing documentary evidence (Norden's Panorama) depict a pitched thatch roof (Uncertainty value = 2.0). Evidence such as Norden's Panorama of London circa 1600 sketch, “Civitas Londini” (Figure 4.6) supports the presence of a thatched roof. Although the image shows a thatched roof, dimensions and scale of the Rose building are not correct. This adds an element of uncertainty and therefore has been assigned a value of ‘2.0’. Known building techniques and surviving examples of thatch roofing of the era provide a low degree of uncertainty (Uncertainty value = 1.5) for the final reconstructed roof element. We can be confident that the reconstructed thatch roof, based on existing examples, has a high degree of certainty. However, existing examples are only analogous and do not conform the exact shape, size and texture of the Rose roof. A final score of 1.5 was calculated for the thatched roof based on the average of the totalled scores. This process was used for each building element to determine an authenticity score.

The value assigned is a median score of all the available evidence used to assess individual components of the reconstruction. Individual aspects of the reconstructions are detailed later in this document. The uncertainty value of the input data can be inversely viewed as a possible measure of authenticity of the final output. That is, a low overall uncertainty value can be inferred to represent a higher authenticity in the final reconstruction. Conversely, a high overall uncertainty value attributed to the input data could represent a lower authenticity score for the final reconstruction elements.

The resultant uncertainty values for each building component were combined to give a total score. The average of the total score has been used to assign a final indicative numerical uncertainty rating for the entire digital reconstruction. The average score has been determined by dividing the total score (41) by the number of individual elements of the reconstruction (15). The overall uncertainty value of input data was calculated to be 2.73. According to the classification matrix described, the overall reconstruction has a moderate level of uncertainty. Conversely the authenticity of the reconstruction can therefore be assumed to be moderate when considering the overall reconstruction elements and their relationship to the input data classes. This authenticity analysis does not extend to include the final visual rendering. The textures applied to this model are not considered in the analysis.

4.1.1 Analysis of the data classification system and authenticity measure Significant questions are raised regarding the effectiveness of a traditional numerical classification approach as a measure of authenticity. Calculation of an average score has

88 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

demonstrated itself to be somewhat overly simplistic. Not every component derived from multiple data inputs is equal in its contribution, or impact, on a reconstruction. An average score does not, and cannot, account for such variations in influence of components on the overall reconstruction authenticity. In effect, an average score may result in hiding important information about the variation of the uncertainty between elements.

This is clearly exemplified by the impact of three reconstruction components (Stairwells, Building Entrance and Windows) on the overall uncertainty level. Although all three elements are important parts of the building, their combined impact on the overall reconstruction is disproportionate. It is demonstrated in subsequent sections that the positions of the internal stairs were arrived at through a range of deductions that considered the lack of archaeology, building structure and known theatre practices of the time. However, all the information employed to arrive at the stairs lacks any hard evidence to support it and is effectively qualitative. It can be argued that the exact position of the stairs is not crucial to the overall authenticity of the reconstruction providing the stairs follow rules of construction and the requirements of the space. Similarly, the exact size and position of the entrance cannot be determined from the archaeology. Literary evidence suggests the position of the entrance was located on the South side of the building (Figure 4.7).

Theatre practices of the time suggest a narrow entrance to the theatre to control patron flow and collect entry fees. As with the stairs, the supporting evidence is qualitative and must be used in combination to arrive at a solution that cannot be verified with quantitative data. The impact of the windows is significant in terms of overall authenticity. It is highly likely that windows existed. Norden’s Panorama (Figure 4.6) suggest the presence of windows in the structure. Glass found during the archaeological excavations may have been related to window glass (Bowsher & Miller, 2009). No quantitative evidence exists to suggest window size, placement, number or shape. It is questionable if the current placement of windows in the reconstruction impacts the overall authenticity of the reconstruction to the extent of that resulting from the numerical classification approach. It is acknowledged that the exact size, number, shape and placement of the windows cannot be correct, but this may not necessarily mean the ‘overall’ reconstruction authenticity is only ‘Moderate’.

Authenticity could also be considered a function of utility. That is, the reconstruction facilitates the activities the structure was intended for. The position of the stairs, the entrance, and the windows do not affect the utility of the building. The current reconstruction allows the controlled flow of patrons between all tiers. Correct ‘choke’ points exist where patrons can be

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 89

monitored, and the correct fare extracted to move to a higher paying section. The narrow entrance would allow for the controlled flow of paying patrons. The windows would allow for some light egress to the tiers. The argument here is that although these elements cannot be verified, and, are highly uncertain, they do facilitate the utility of the structure as intended. If we consider that the reconstruction faithfully fulfils the utility requirements it was constructed for originally, then it is possible that the authenticity of the output is higher than determined numerically.

The classification of data inputs into classes of uncertainty is clearly a subjective process. The user must sort data inputs based on the decision tree. Although the decision tree provides broad categories and a hierarchical structure for sorting the user is compelled to ‘decide’ what level of uncertainty class data belongs in. The many subjective decisions must have a determining impact on accuracy of the computed values. A computed value of uncertainty for the entire reconstruction is flawed in that it does not treat inputs individually and in context of the decision-making process. As such it does not contribute anything meaningful or of value from a user interpretation standpoint or as a measure of authenticity.

The process of assigning classes for numerical computation is time consuming and was shown to be unwieldy in a large reconstruction project such as the Rose Theatre. The decision tree allows a class to be assigned quickly to input data. The subsequent calculation of individual components, although simplistic, is time consuming. In a large-scale reconstruction where there may be hundreds or thousands of elements to consider the current approach is unrealistic. It is also highly likely that with a large-scale project the errors in calculating uncertainty and authenticity in a numerical fashion will be exacerbated leading to incorrect assessments.

Assigning a single numerical value of uncertainty impacts the generalizability of the overall reconstruction and data classification methodology and its application to other types of archaeological reconstructions. It is doubtful for example, that a 3D reconstruction of a Neolithic site can be treated the same way as the reconstruction of a late 19th century church recently destroyed in a modern war. In the case of the Neolithic site the bulk of the data inputs would be inferential at best with material culture extant remains to infer from whilst the 19th century church would most likely be well documented and evidenced. Finally, and most importantly, the use of a single numerical value as an indicator of authenticity or uncertainty does not afford the viewer any opportunity to evaluate the evidence used to achieve the reconstruction. The viewer is forced to accept a numerical value which is, at best, an estimation. A single numerical value offers no level of transparency and a viewer cannot make an informed

90 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

assessment of the reconstruction. It is timely to reiterate that a reconstruction may indeed be accurate and have a high level of authenticity, but the current numerical classification approach cannot support such an assessment in the eyes of the viewers. In effect a number may well be meaningless to the viewer.

Individual classes of uncertainty for data inputs can be beneficial from a visualisation standpoint. The process of calculating an overall authenticity score, as discussed above, most likely is not representative of a final reconstruction. The numerical approach has been discontinued in favour of a more generalised classification approach based on visual outputs. The focus of the modified classification system shifts to the data input uncertainty classes and how these can be conveyed in a meaningful visual form. This shift in classification approach required the modification of the decision tree to simplify the classification process and assign visual attributes (colour) to classes of data uncertainty. Building components can then be assigned colour values as a method of visually classifying the reconstruction

The decision tree facilitates rapid assignation of input data to uncertainty classes. The ability to sort inputs based on evidence type and veracity allows for a clearer understanding of the data and assists data organisation. The decision tree allows a user to clearly separate data inputs and use these to form the basis of reasoning for the subsequent interpretive phases. The inputs are not weighted with a rigid numerical value that may not be representative of their impact on the reconstruction. This was evident during the Rose Theatre reconstruction where it was the volume of literary records and known practices of the time that supported the position and layout of the ‘Lords’ above the stage. Although the archaeological recorded suggested a sturdier construction in that area of the building it was the supporting evidence that was most instrumental in the decision-making process. In this instance, the qualitative evidence was as important, if not more so, than the limited archaeological record.

A simplified and more effective decision tree has been developed (Figure 4.1). The ‘Analogous’ evidence branches have been combined. Classifying input data based on construction techniques is redundant when using existing examples or analogous structures and building techniques. The fact that these structures exist and are erect testifies to the adherence to building techniques and the laws of physics. The primary tree branches have been renamed to ‘Related Evidence’ and “Associated Evidence’. Related evidence is any data inputs that can be directly related to the building in question. Inputs can range from extensive extant archaeological remains, to stories, drawings, descriptive texts, and the like. In the case of the Rose Theatre these inputs include the archaeology of the site, drawings of the South Bank

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 91

which include the Rose Theatre and Henslowe’s extensive diary of the Rose which details building expenses. Associated evidence is data inputs derived from sources that are not directly related to the reconstruction but are applicable in the context of the reconstruction. These inputs can range from onsite analogous structures to off-site structures of similar era and style. This is exemplified in the Rose reconstruction with evidence such as existing era buildings, documented building techniques of the era and known theatre practices of the time. The modified decision tree allows a much clearer distinction and simpler method of rapidly assigning input data to a class of uncertainty. The modification affords a rapid and logical method of sorting data based on evidence type. Data inputs have been categorised into three major classes. These are:

I. Archaeological evidence

II. Documentary evidence

III. Analogous structures

92 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

Figure 4.1: Modified Decision Tree

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 93

At the primary level of data classification inputs can be assigned one of the three top level categories. Each class of data is then sorted into sub-classes of uncertainty based on the volume (repeatability) and veracity (trustworthiness) of the inputs. In the case of the Rose Theatre reconstruction evidence such as extant remains from archaeology were clearly sorted into primary class I (Archaeological Evidence) and then into secondary classes of certainty. Extant foundations were assigned a very high degree of certainty whilst limited green glass fragments that possibly were associated with windows were assigned a low level of certainty. The identical process was followed for the Documentary Evidence and Analogous Evidence primary classes. Norden’s hand drawn sketch of London (Figure 4.6) was clearly documentary evidence. The image is not drawn to scale, lacks building detail, but does indicate some features of the Rose, however minor. This documentary evidence is of limited value in terms of detailed reconstruction information but does suggest that the structure was a circular half-timbered design with windows in the central tier. This evidence was assigned a low level of certainty but was valuable nonetheless in the decision-making process.

Arguably the most significant change to the decision tree is the addition of colour elements as classes of uncertainty. The addition of colour as a representation of class effectively transforms the decision tree into a classification and data sorting tool. Assigning each uncertainty class a colour, facilitates visualising reconstructed elements based on the inputs used to arrive at a solution. The uncertainty classes continue to retain their numerical values although only as a point of reference. A colour was arbitrarily chosen for each category and then segmented into three shades to represent each broad uncertainty class in that category. It is possible that more effective colours could be chosen to define categories. This would require significant exploration of colour theory and human perception and outside the scope of this study. The modified decision tree offers a method whereby viewers can make informed decisions about the overall authenticity of a reconstruction based on the data inputs used. By physically assigning classified colours to individual elements or components of the final reconstruction, viewers, in effect, classify the resulting reconstructions authenticity.

The refined decision tree and colour classification system is applied to the Rose Theatre and other test cases in subsequent chapters and the results discussed.

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 95

4.2 DIGITAL RECONSTRUCTION OF THE ROSE THEATRE BASED ON THE WORKFLOW PIPELINE AND DECISION TREE 4.2.1 Inputs, Tools and Scope The digital reconstruction of the Rose Theatre was based on the following data inputs:

• Archaeological evidence (Bowsher & Miller, 2009) • Historical descriptive texts and drawings • Known theatre practise and operations of the era • Surviving buildings of the era • Known building techniques of the era • Henslowe's diary of construction

The entire 3D modelling/reconstruction was conducted using Autodesk Maya 2011 and Adobe Photoshop for the creation of textures. The material properties of extant remains recorded during the excavation process have been used to inform the final visual output of the reconstruction where applicable. For example, it is known that two differing mortar types were used in the foundations of the 1594 iteration of building. Given that the foundations have been buried since the early 1600's it is doubtful that the current mortar colour is the same as when originally laid. The textures created for the foundations reflect the use of two mortar types but do not attempt to be photorealistic. Where no extant remains exist, texture is based on the evidence used to recreate the feature in question. This rationale is employed for all final texturing of the resulting structure. This description covers the main components of the building itself. The components that are discussed are:

I. The building structural framework II. The building Tiers III. The flooring and Roofing IV. The Tiring House V. The Stage VI. The stairs VII. The railings and balusters

96 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

VIII. The doorways Additional minor elements of the building such as the exterior assets and the bridge joining the building to Maiden Lane are not dealt with here but can be accessed via the online development journal (Kastanis, 2015). Throughout the reconstruction that is detailed below the workflow pipeline was applied at each stage of the reconstruction and reflects the iterative nature of the process. All input data used was assessed using the Decision Tree detailed above prior to inclusion in the final reconstruction.

4.2.2 The Rose Reconstruction 4.2.2.1 Tier 1 (Ground Level) The first step in the reconstruction was to create a mapping plane on to which the existing archaeological plan map created by Bowsher et al 2009 was applied. The plan was scaled to real-world size and treated as an information layer. (Figure 4.2)

Figure 4.2: Figure 2: Base modelling shown with mapped archaeology and interpreted building footprint (Bowsher & Miller, 2009, p. 56,125). A second layer was created containing the interpreted building base plan (Figure 4.2.) This was used to compare the existing archaeology with the proposed building layout. The interpretation, as expected, closely matched the extant remains and therefore was the most probable physical extent and shape of the building. This data input represented low uncertainty as it was derived from reliable quantitative sources. This stage of the reconstruction is the only section that contains relatively extensive archaeological data. Extant remains included foundation walls, wooden drains, stage

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 97

remains, etc. There is enough evidence in the archaeology to confidently determine not only the spatial extents of the foundations but also the overall footprint of the building. Two tiers of foundation were created reflecting the archaeology. The reconstruction of extant remains was modelled in gross form (Figure 4.3) with no consideration given to visual appearance or micro topography.

The physical dimensions and spacing of the inner and outer wall circuits in X (Width) and Z (Height) axis can be safely assumed to be as accurate. Two distinct levels of brick coursework were created as per the archaeology. The accuracy of the walls in terms of the Z axis is somewhat less accurate. The extant remains of the foundations rise to a height of approximately 98cm above datum but leave no apparent indication of the final height of the foundation walls. The final derived height of the foundation walls of 1.15m above datum is based on foundation heights postulated by Bowsher et al. but cannot be verified as the true original height (Bowsher & Miller, 2009). The resulting final reconstructed foundations can be seen in Figure 4.3.

Reconstructed extant foundation Base course of foundation The reconstructed remains (Phase 1) overlaid on the brickwork overlaying foundations. (Phase 3) excavation maps and associated extant remains of foundations. data (Phase 2)

Figure 4.3: Reconstruction of building foundations based on archaeological data and using the proposed workflow pipeline

4.2.2.2 The building structural framework The procedure of assessing and integrating the available evidence described above is repeated in the reconstruction of the building frame. This aspect of the reconstruction relied on data from sources other than the physical archaeology. These included analogous examples and documentary evidence. It is assumed that building

98 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

framework would have adhered to existing building practices of the day as well as obey the basic laws of physics to be an erect rigid structure.

It can be reliably assumed that the Rose Theatre was of half-timbered design as this was the standard form of timber-framed building of the era (Schofield, 1991). This assumption is supported by the many sketches of London architecture at the time and from the material list for renovations in Henslowe's diary which indicates substantial timber purchases (Henslowe, 2002). It should be noted that no identifiable remains of the actual building framework were unearthed in the archaeological excavation (Bowsher & Miller, 2009). The following description of reconstructing the building framework is presented here in an abbreviated form. A detailed discussion of each decision, material type, and stage of re-construction can be found in the online journal (Kastanis, 2015).

Main bearers (Sill beams) were laid on the foundations (Figure 4.5) of the building as per standard practice in buildings of the era (Schofield, 1991). The placement of the three main supporting bearers on top of the brick foundations are based on living examples of Elizabethan half-timbered buildings and the vast repository of documented building techniques of the era. Floor joists were placed evenly across the span between the inner and outer foundations (Figure 4.4). No extant remains of Sill beams, supporting pillars or floor joists have survived. The actual dimensions and spacing of joists are based on existing examples of period buildings. From a structural stability perspective, it is highly likely that size and placement of joists is indicative of the original.

Following the placement of Sill beams and Joists came the support pillars on both the inner and outer circuits (Figure 4.5). The intersection points of each segment of the floorplan are the logical points to position the supporting pillars and this is exemplified in surviving half-timbered buildings of the era.

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 99

Supporting Pillars

Sill Joists Beams

Figure 4.4: Derived building framework showing Sill beams, Joists, and main supporting pillars

Two additional tiers were added to the structure. The structural elements of tiers 2 and 3 are exact copies of tier 1 (Figure 4.4). This is a reasonable assumption as it reflects standard practices, would simplify the building process, and be the most economical solution. The only modification to the floor frames required would be to accommodate intrusions such as stairs. The height of the tiers is based on the measurements for the Fortune theatre, built in 1600, for which a building contract (and thus more precise measurements) remains (Gurr, 1992). It is assumed that the Rose Theatre tiers would have been similarly sized, even if the overall building size was smaller than the Globe (Greenfield & Gurr, 2003).

Given the physical constraints of the building shape and positioning of the foundations, the number of available options in terms of placement of structural elements is extremely limited. It is highly probable that the supporting pillars had to be placed on the intersection of each building segment. It is very likely that supporting Sill beams where placed on top of the foundations as this was standard practice in half- timbered buildings. To have floorboards, joists are required. Evidence of supporting posts for the joists were discovered in the form of holes aligned in the direction of the joists during excavations. It is postulated these holes originally had small posts in them supporting the mid-point of the ground-floor joists. This would have added an

100 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

additional level of rigidity to the ground floor. The point here is that although the building framework is effectively conjecture, it has been based on sound reasoning and analogous evidence. The main point of uncertainty in this stage of the reconstruction is the assumed height for each tier.

4.2.2.3 Flooring and Roofing Flooring was added after the main structural framework was completed (Figure 4.5). The Floorboard layout is effectively the same on each tier with the exception that Stairway access has been added to Tier 2 and 3. Once again, the absence of physical evidence introduces uncertainty to this aspect of the reconstruction. However, there is no doubt that flooring existed, and the method of laying floors across joists is universal. The uncertainty of the reconstructed flooring can therefore be viewed as being relatively low despite the lack of extant remains.

The final addition to the main structural elements was the roof structure. No original roofing framework has been identified in the archaeology. However, remains of a roofing thatch attributed to the main roof were excavated (Bowsher & Miller, 2009). This material evidence coupled with the visual documentary evidence of the Rose strongly suggest a thatch roof covered the building structure. This is further supported by common building practices of the day (Moir & Letts, 1999). The actual reconstruction of the roofing frame is effectively informed conjecture (Figure 4.5). Techniques of thatching and the construction of the underlying supporting framing have changed very little in 600 years (Moir & Letts, 1999). We can, therefore, be reasonably certain of the type, and style, of frame required to support a thatched roof. The roof frame and pitch was determined from living examples of thatched roofs of the era coupled with documented historical methods of thatch roof construction (Moir & Letts, 1999).

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 101

Figure 4.5: Derived building framework showing roofing framework, flooring and thatch roofing material 4.2.2.4 Entrance and rear door - Placement and Size It is known that the 'means by which the audience entered, and left was an important part of the playhouse structure, for it was part of the overall experience of theatre-going' (Bowsher & Miller, 2009, p. 125). The entrance to the Rose Theatre cannot be definitively placed. No remains were uncovered in the outer circuit of the theatre foundation excavations to definitively indicate an entrance. The decision- making for locating and constructing the entrance was based on several factors.

• Documentary evidence

• Other examples

• Utility and physical access

• Standard theatre practise of the time

• Academic postulation

Numerous graphical depictions in the form of sketches and maps of the time show the entrance to the Rose Theatre to be on the southern side of the building opening onto what is now known as Maiden Lane (see Norden's Panorama) (Figure 4.6) and Claes Van Visscher’s 1616 ‘View of London’ (Figure 4.7). Bowsher and Miller postulate that there was 'was little likelihood of entrance from the north-east' and point out, the main entrance or 'Front Door' was usually placed opposite the stage (Bowsher & Miller, 2009, p. 125). Bowsher and Miller 2009 also note the entrance was usually sized 'so as to collect the entrance fees, which may have accounted for its narrowness (Adams 1961,33)' (Bowsher & Miller, 2009, p.155).

102 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

Figure 4.6: The Rose (with circular roof, lower right area), ca 1600 From Norden's Panorama “Civitas Londini” (Source: http://www.luminarium.org/encyclopedia/nordenbankside.jpg)

Figure 4.7: View of London, Claes Visscher 1616 (Source: https://en.wikipedia.org/wiki/File:London_panorama,_1616b.jpg) It is therefore probable that in accordance with practices of the time, the entrance was in fact narrow (Figure 4.8) This is perhaps the most difficult feature of the reconstruction to quantify as the width of the entrance can only be postulated. The early playhouses of the era appear to have utilised this system of narrow entrance opposite the stage and it is therefore 'also most plausible for the Rose' (Bowsher & Miller, 2009, p.125).

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 103

Figure 4.8: Proposed building entrance showing the narrow egress

It is very likely that 'back' door was a stage or service door at the rear of the building (Bowsher & Miller, 2009). The rear door of the playhouses of the time was often also used by 'lords or gentlemen to gain access to their private rooms and perhaps by those provisioning victuals to the playhouse' (Bowsher & Miller, 2009, p. 124). The final positioning of the entrance has been based on postulations by Bowsher and Miller (Figure 4.9). This point is also the closest to Maiden Lane and the most logical position for the walkway across the ditch between the building and Maiden Lane.

Figure 4.9 Postulated building entrance position (Bowsher & Miller, 2009, p. 125). 4.2.2.5 Walls and Windows A considerable amount of Lath and plaster fragments were uncovered in dumps within the yard of the Rose Theatre (Bowsher & Miller, 2009). Henslowe's records of the 1592 Rose renovations specifically note 'infill panels of lath and plaster' for the walls which was consistent with building techniques of the time (Henslow, 1904). It

104 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

is highly likely that the walls were therefore made of Lath infill (Figure 4.10) and whitewashed as is consistent with the building architecture of the time.

Figure 4.10: Exterior of the Rose Theatre showing Lath infill walls Fragments of window glass that appear 'largely to be pale green' were uncovered in the archaeology (Bowsher & Miller, 2009, p. 139). No archaeological evidence exists to suggest the placement and size of windows or indeed that the windows contained glass. Documentary evidence (Figure 4.6) suggest a single row of windows around the playhouses located at the middle tier but this cannot be used to determine if glass was used. The size, shape and placement of the windows is unknown. The documentary evidence is limited and inconclusive. The placement of the windows in the reconstructed model (Figure 4.10) are determined by the underlying building framework and are placed in the most logical position between supporting posts. Only the second tier contains windows as per the depictions mentioned above.

4.2.2.6 Stage and Stage Canopy Extensive remains of the stage foundations have been unearthed (Bowsher & Miller, 2009). The foundations leave little doubt as to the spatial extent and shape of the stage as can be seen in Figure 4.11. The extant remains do not indicate the height of the stage or the material it was constructed of.

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 105

Stage Foundations

Building Foundations

Figure 4.11: View of the Rose Theatre excavations in 1989. (Source Bowsher & Miller, 2009) It can be reasonably assumed that the stage was of wood construction resting on known brickwork foundations. The reconstructed stage (Figure 4.12) assumes an underlying support framework clad in wood planking. The stage framework joined to the main building framework in a logical manner and did not present any difficulties. The spacing between joists and main bearers are based on built examples evident in existing structures.

Figure 4.12: Reconstructed Stage and underlying framework For the purposes of this reconstruction the interpretations by Bowsher and Miller (2009) have been used to set the height of the stage relative to the excavation datum.

106 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

Documentary evidence, (Figure 4.6), shows the Rose with a canopy extending over the stage. Square depressions in the soil within the boundary of the stage have been interpreted as possible foundations for pillars.(Bowsher & Miller, 2009) These foundations suggest that two supporting pillars existed to support the stage canopy. Additionally, wood roof tiles were discovered during the excavations and it is possible these tiles were used on the stage canopy (Bowsher & Miller, 2009). A drip line skirting the stage was also unearthed during the excavation (Bowsher & Miller, 2009). This drip line borders the stage area and suggests a canopy overhanging the actual stage. This dripline has been used as an indicator of the possible size of the canopy eaves.

The reconstruction of the stage canopy was a process of iteration. The presumed known extents of canopy are derived from the drip line discovered during excavations (Bowsher & Miller, 2009). It stands to reason that if a drip line developed around the stage there had to have been some form of a canopy in place long enough to allow time for a sizeable drip line to form. Additionally, a wooden drain was unearthed intersecting the stage that possibly drained water from the stage area. The base of the canopy and the top of the canopy are inferred from the reconstructed roof height and pitch. The orientation of the stage relative to the building presented significant challenges. The excavated stage foundation show that the stage was not ‘square’ to the building but instead had a slight rotational offset (Figure 4.13).

Figure 4.13: Top view showing the stage Figure 4.14: Position of Columns relative to offset the stage shape The supporting column pads uncovered in the archaeology are also offset and follow the line of the front of the stage foundations further supporting the position of the stage columns (Figure 4.14). The joining of the stage canopy frame to the building frame required a process of matching the base structure of the canopy to the main

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 107

building frame in a way that allowed logical ‘tie-points’ to be utilised. These ‘tie- points’ are on intersections where supporting posts and bearers meet. The resulting reconstruction (Figure 4.12) was a highly iterative process steered by the physical extents of the main building frame.

Figure 4.15: The stage canopy roof showing the interpreted covering and underlying framework As can be seen in Figure 4.15 the underlying canopy framework is complex. This configuration was arrived at through the iterative process established in the workflow pipeline. The stage canopy was almost certainly present although the current reconstruction cannot be verified and can only be considered a knowledge representation.

4.2.2.7 Tiring house Records from Henslowe's Diary suggest that a Tiring house (the area behind the stage for actors to don their ‘attire’) was erected in 1592. The tiring house area was used as 'changing rooms and management offices, as well as providing storage space for company valuables, such as costumes, play texts and props.' (Bowsher & Miller, 2009, p. 119)

The excavated remains of the stage wall show a '…distinctly heavier build than the only other small area of new inner walling.' (Bowsher & Miller, 2009, p.119). The wall behind the stage is wider and appears designed to carry heavier load. This would support the assumption that a Tiring house structure and stage roof existed. This section of the inner wall would have been the intersection between the main building and the stage canopy, it would have been required to carry more load than other sections of the inner wall. Other evidence of a more complex structure to the rear of

108 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

the stage is the great '…amount of differing deposits of debris associated with demolition’ suggesting a large amount of structure needing dismantling (Bowsher & Miller, 2009, p. 119).

It is almost certain that if the Tiring house existed it would have been in the space directly behind the stage. This would allow actors and props access to the stage during performance. In the Rose reconstruction, the tiring house has been placed behind the stage and occupies the three ‘bays’ adjoining the stage (Figure 4.16). Although there is no direct evidence to confirm the use of the three bays for the Tiring House it is highly likely it was the case. The rationale used to wall in the bays to effectively create the Tiring house is based primarily on the fact that it would have been impractical to use only the central bay. Collectively the three bays occupy approximately 43.2 square meters of floor space. The central bay occupies only 17 square meters of floor space. This would not have been adequate area to fulfil the roll of the Tiring House. Although the actual extents of the Tiring House are unknown, the archaeology supports the hypothesis that it was present. This is further supported by the fact that it was common practice in the theatres of the time to have a Tiring House (Greenfield & Gurr, 2003). All three bays are directly bordered by the stage. It is logical to assume that the entire space behind the stage would be dedicated to the Tiring House. Finally, the digitally reconstructed building frame supports the inclusion of the walls that define the space (Figure 4.17).

Figure 4.16: The Tiring House bays Figure 4.17: The Tiring House interiors showing stage access and the side walls. 4.2.2.8 Stairways No physical evidence of internal or external stairs and stairways have been unearthed in the archaeological excavation. The lack of any external stairway

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 109

foundations may suggest that internal stairs were employed. This cannot be confirmed, and it must be remembered that a significant portion of the western side of the site is covered by a modern tower block. It is quite possible that the foundations to external stairways were removed as part of the final demolition or only existed on the western side of the building though no documentary evidence exists to support this. Certainly, the structure was multi-storied and therefore required some form of stair system. It has been suggested that internal stairs were used in the Rose and that they would have been ' …limited in number because they would potentially have reduced the seating capacity.' (Bowsher & Miller, 2009, p. 115).

It has been postulated that the '…obvious' place for the stairs is' near to the entrances (Bowsher & Miller, 2009, p. 115). In addition, the location of the 'ingressus' may be associated with access to the upper galleries. Documentary evidence pertaining to the monetary collection practices of the time suggest that there is a '…clear association between the ingressus and access to higher levels' (Bowsher & Miller, 2009, p. 114). Typically these ingressus were narrow to control and monitor the flow of paying patrons. It has been postulated that there would '…have been further private stairs at the back of the building to allow access up to any stage balconies and elsewhere in the tiring playhouse.' (Bowsher & Miller, 2009, p. 115).

Three significant factors determined the final placement of the internal stairs in this reconstruction.

I. Conclusions cited by Bowsher and Miller 2009

II. The location of the 'ingressus' (which would have assisted in separating patrons who wanted to pay additional fees for the upper galleries)

III. The best possible placement that allowed a reasonable stair-rise while minimising the impact on seating, standing room and sightlines.

The stairs are a crucial aspect of the building. Not only did they allow access to the upper tiers but were pivotal to the separation of paying customers and a means to transport theatrical props. As no archaeological evidence exists, the design and placement of the stairs can only be viewed as a knowledge representation informed by utility and the constraints suggested by Bowsher and Miller. Numerous locations were tested for the inclusion of stairs throughout the building and the final positions chosen (Figure 4.18) appear the most logical in terms of use of space, flow of patrons and

110 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

utilisation of the building framework. The location of the stairs would also have allowed more effective policing of patron traffic as tolls could be collected at the base of the stairs starting from the pit entrance. The locations chosen enabled the stairs to join with the underlying building framework and are the only locations that allowed an acceptable rise in the stairs without significantly impacting floorspace.

Figure 4.18: Proposed rear 'Tiring house' stairs and internal general access stairways relative to stage and yard The reconstruction of the stairs is effectively based on postulation, but it is not conceivable that stairs did not exist. 4.2.2.9 Railing Balusters and Doors It is highly likely that each tier contained railings of some form. No evidence remains of railings except for a baluster (Figure 4.19) which was unearthed during the archaeological excavation (Bowsher & Miller, 2009). The unearthed baluster was relatively complete and allowed the height of the railings to be determined with a high degree of confidence. Figure 4.19 shows the reconstructed railing with balusters.

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 111

Figure 4.19: Reconstructed railing and balusters. Unearthed baluster from the Rose excavation (Bowsher & Miller 2009). The spacing of the balusters cannot be determined. It is unknown if a bottom rail existed for the railings. For this reconstruction, it was decided that the spacing of the balusters would be based on what ‘appeared’ to be correct and that the bottom rail would not be included. Overall it can be assumed that railings did exist. The height of the railings has been inferred but could be assumed to be reasonably correct based on the segment of baluster unearthed.

No evidence of doors or door furniture has been found. It is highly likely that doors did exist to separate the tiring house and the ‘Lords’ from the rest of the building. Doors have been placed where it is most logical for them to be placed and were supported by the building framework. The style of door has been left deliberately simple and does not attempt to recreate a period door nor does it attempt to confirm door sizes (Figure 4.20). The overall configuration of the doors is based on living examples of period architecture of the time. This included aspects such as the exposed heavy beams comprising the door frames.

112 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

Figure 4.20: Typical Doorways used in the reconstruction 4.2.3 Analysis of the Rose Theatre Reconstruction The workflow pipeline is most effective when applied to individual building components. It is designed to build upon higher certainty data to arrive a solution. Each component of the building is based on variable data types. In some cases, the process of reconstruction is driven entirely by quantitative data such as archaeological records and is linear whereas in other components multiple qualitative data inputs are utilised and require an iterative approach. The workflow pipeline facilitated both linear and iterative decision-making processes. Phase one of the pipeline allows the sorting of data into classes and the utilisation of quantitative data to form the ‘foundations’ of the subsequent reconstruction. Phase one is a linear pathway and does not incorporate iterative cycles. Phase 2 is the highly iterative stage of the workflow pipeline. Phase 2 builds on the outputs of Phase 1. It is in this stage that extensive ‘testing’ takes place. Qualitative data inputs are tested against the known elements of the reconstruction. The combination of the two phases of the pipeline allows the user to arrive at logical solutions based on the organisation and comparison of available data inputs.

In the case of the Rose Theatre this can be effectively illustrated with the process of creating foundations and then erecting the building framework on the foundations. This is the linear process that would have been followed by the actual builders of the structure. Using the first phase of the workflow pipeline the foundations were created using known archaeological evidence. This was a straight forward process as the inner

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 113

and outer rings of the building foundations were clearly evident. Sufficient archaeological remains of the foundations existed to determine the overall building footprint. It is a logical assumption that the foundations would be the same height over the entirety of the building footprint to provide a level building surface. The archaeological evidence also supports this assumption as the base of the foundations clearly followed the microtopography of the excavated site. This process did not require iteration as the outcomes were evident from the outset. The first phase of the workflow pipeline is not aimed at using iteration as a method of testing and experimentation to arrive at a solution. In general, the resulting 3D model generated from the first phase will have a high degree of certainty in its construction and final form.

The reconstruction of the building framework was an iterative process utilising a variety of input data. The data included existing examples of half-timbered building from the era and sketches of the general south bank area from the time of the Rose. The building framework was created based on existing examples. The reconstructed foundations provided the platform on which to test the building technique. The entire reconstruction followed this process and was very effective in arriving at the most likely solutions to determining building components especially where very little physical evidence was available. Two notable examples from the Rose reconstruction are the reconstruction of the stage canopy and implementation of the internal stairways.

The stage canopy was a particularly challenging component that involved numerous iterations to arrive at an acceptable solution. Archaeological evidence suggested a canopy existed. Two foundation depressions were uncovered that sat within the bounds of the stage and suggested the location of the canopy supporting columns. This evidence was supported by the presence of a ‘drip-line’ around the boundary of the stage suggesting an overhanging roof. To reconstruct the stage canopy, the two ‘pillars’ were placed in the positions identified by the archaeological evidence. The pillar positions aligned very well with the support columns of the main building and allowed the canopy to be ‘attached’ to the main building at logical tie-points.

The three ‘knowns’ were that the supporting posts of the main building aligned well with the extents of the stage, the positions of the stage canopy pillars are relatively certain, and, the extent of the stage canopy could be inferred by the drip-line. These

114 4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case

three factors determined the main framework of the stage canopy. The entire process required several iterations to arrive at a logical solution. The three ‘knowns’ made the construction of the canopy support framework a straight forward process. Effectively the main for the canopy base could only attach to the main building at the intersection of the building pillars. The placement of the canopy joists and battens was problematic due to the irregular shape of the stage and required several iterations to arrive at a logical solution. This process showed that, at a bare minimum, the workflow pipeline, assisted by classified inputs, produced a solution which is technically possible, and, that the solution accounts for input data quality.

The constructing and final placement of the internal stairways was also a highly iterative process. The iterative nature of the workflow pipeline allowed the testing of many possible solutions to arrive at a ‘most likely’ scenario based on assumptions of scholars and the practicalities of monitoring paying patron movement through the space. Without the initial building framework established to a relatively high degree of certainty it would not have been possible to determine a likely location for the stairways. One over-riding factor determined the placement of the stairways. If the stairs were indeed internal, they had to begin and end on the underlying support frame of the building. The main beams and cross-bearers’ positions, to a great degree, influenced the location of the stairs. It was, and still is, common practice for stairways to be anchored on supporting beams or bearers. This restriction coupled with issues of access and space requirements were instrumental in establishing a guide as to where the stairways could be placed. The final solution represents the best ‘fit’ given the frame of the building, the lack of any evidence of external stairs and known theatre practices of the time.

The workflow pipeline has proved to be effective in understanding the relationships between multiple data types and their contribution to a reconstruction. The pipeline provides a logical structure for archaeological reconstruction as it builds on the highest certainty data as a starting point. The iterative nature of the second phase is effective in that it allows ideas based on qualitative inputs to be rapidly tested. The workflow pipeline is applied in subsequent case studies.

4 Results - Application of the Classification System and Workflow Pipeline – The Rose Theatre Test Case 115

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

In this chapter the revised workflow pipeline is applied to the Komediehuset, Bergen Theatre reconstruction

5.1 THE (KOMEDIEHUSET) BERGEN THEATRE RECONSTRUCTION 5.1.1 Bergen Theatre description The Centre for Ibsen Studies at the University of Oslo commissioned the digital reconstruction of the Komediehuset, Bergen, as part of an international research project entitled ‘Visualising Lost Theatres’. (Holledge et al 2018) The project utilises digital reconstructions of lost theatrical venues to study the relationship between space and performance in a historical content. Many of these venues have been ‘lost,’ due to destruction or significant remodelling or rebuilding. The digital reconstructions of these venues are based on archaeology, photographic records, sketches, descriptions, and other documentary evidence that provide evidence of structure, appearance, and spatial dynamics.

The original Bergen Theatre located in the city of Bergen in Norway was a wooden building built by an amateur dramatic society in 1800. It was destroyed in 1944 by English bombers targeting the German-held harbour fortifications. The theatre is significant as it was the home of first Norwegian professional theatre company and major Norwegian playwrights and actors such as Henrik Ibsen and Bjørnstjerne Bjørnson did their theatrical apprenticeship with the company. The theatre had several interior alterations over the course of its life. The original architect is not known, and the architectural plans do not exist. It is known that there were some alterations to the seating in the gallery in 1856 to accommodate the visit of some minor royalty (Holledge et al 2018). There was the major renovation in 1870 conducted by the architect Peter Blix (Holledge et al 2018). Original architectural drawings of the 1870 alteration exist in the official Bergen Theatre at the University of Bergen Department of Linguistic, Literary and Aesthetic Studies. In addition to Blix’s

116 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

drawings the Bergen Theatre Archive contains original survey drawings of the Bergen Theatre produced in 1938, six years prior to its destruction. The 1938 survey drawings contribute significantly to this reconstruction of the Bergen Theatre.

The 1856 version of Bergen Theatre was reconstructed and classified for this study. A variety of data sources were used to achieve the final reconstruction. No physical evidence or archaeology of the original building remain. The reconstruction was reliant on documentary inputs. These inputs took the form of architectural drawings, exterior and interior photography taken in the late 1800’s and analogous example of other Norwegian and European theatres of the same era.

5.1.2 Workflow pipeline This section describes the data inputs used in the reconstruction and how they were used in the workflow pipeline as described in Chapter 4.

5.1.2.1 Data Inputs The primary data inputs for the reconstruction of the original Bergen theatre prior to the major 1870 renovation were exclusively documentary.

• 1870 renovation drawings produced by the architect Peter Blix

• 1938 official survey drawings of the building

• Original seating plans 1850

• Scale model of the 1915 Bergen Theatre

• External photography from the late 1800’s

• Internal photography and film footage after the 1870 renovation

• Existing analogous theatre example in both Norway and the United Kingdom

The original renovation drawings by Peter Blix show the renovations carried out in 1870 (Figure 5.1). These drawings show measurements consistent with the official 1938 survey drawings. The drawing contains numerous measurements and notations although scale accuracy across all drawings is questionable. In addition. Blix provided several differing designs for the new 1870 gallery renovation. The survey drawings, in contrast, are very detailed (figure 5.2) and consistently accurate. The drawings included plans and elevations detailing both the interior and the exterior of the building.

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 117

Figure 5.1: Original elevation (left) and plan (right) Blix drawings of 1870 renovation in the Bergen Theatre. Courtesy of Holledge et.al.

Internal fixtures and mouldings are detailed. It is unknown if the mouldings and details recorded in the survey drawings existed from the time the theatre was built or added later. It is likely that the décor as detailed in Figure 5.2 was added at the time of the gallery renovation and in keeping with the style of the 1870 renovation.

Figure 5.2: Example original 1938 survey of the Bergen Theatre showing elevation and decorative details. Courtesy of Holledge et.al.

118 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

Detail in the Blix drawings suggests that the stage front area was also renovated at the same time as the galleries. It is known that the exterior and main internal structural framework of the building, including the stage, were not altered throughout its life. The 1870 renovation involved the alteration of the gallery and gallery wings. The 1938 survey drawings therefore provide an accurate base for the construction of the building. The shape and decor of the gallery’s elements are primarily in question.

Elevation and plan drawings by Blix (Figure A-7) contains hand-drawn linework which has been interpreted as the outline of the original pre-1870 gallery. The Blix drawings exhibit a reasonable level of accuracy and therefor of medium uncertainty. In addition to the Blix drawings an 1850 seating plan (Figure A-1) shows the gallery outline with slightly curved corners which aligned with the Blix drawings. The seating plan was instrumental in verifying the gallery shape pre-1870 and is known to be a reasonably high accuracy as the seating is accurately laid out within the auditorium and gallery space.

These drawings in conjunction with the 1938 survey drawings formed the basis of the pre-1870 reconstruction. Photography of the exterior and interior of the building (Figure A-2) dating from the early 1900’s was used to supplement the architectural drawings and determine interior decor.

Supplementary data inputs such as the scale model of the Bergen theatre (Figure A-3) and surviving examples of a similar theatre located in Norway were valuable in verifying elements of décor and stage layout.

5.1.2.2 Workflow 5.1.2.2.1 Reconstruction of the Main Building Structure The first stage of the workflow pipeline requires the use of data inputs for the construction of any extant archaeological remains. In the case of the Bergen no archaeology exists. However, detailed architectural plans exist which are known to be accurate. In place of the reconstruction of extant remains the entire main structure of the building could be created with a high degree of accuracy. The same process as detailed with the Rose reconstruction was followed in the construction of the Bergen theatre. Mapping planes representing plan and elevations were constructed in the 3D modelling software. The 1938 survey drawings were mapped to the relevant plane and were scaled to real-world size. The exterior and interior mass structures where then created as can be seen in figure 5.3.

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 119

Figure 5.3: Interior and exterior walls reconstructed based on 1938 plan survey drawing.

The supplied survey drawings allowed the accurate sizing and positioning of window and doorway recesses. The survey drawings provided detailed information on the style of the doors and windows. It is highly likely that the exterior doors and windows remained the same throughout the life of the building. Exterior doors and windows were modelled based on the designs described in the 1938 survey drawings. It is not known if the interior doors remained the same throughout the life of the building or were changed during the interior renovation of 1870. The survey drawings clearly show the exterior of the building clad in weatherboard. This is verified by the exterior photography (Figure A-2) and was reconstructed accordingly (Figure 5.6).

With the completion of the main structures, the floors and stage were reconstructed. The 1938 survey drawings show that the main auditorium floor, the stage and the rear gallery were all raked as is common in purpose-built theatres. The angle of rake for each floor was calculated from the survey drawings. The survey drawings do not detail the type or style of flooring. No internal photography available shows the floors. The 1915 scale model suggests exposed timber flooring laid longitudinally along the main axis of the building. This is supported by the type and style of flooring still in existence in the Fredrikshalds Theater (Figure A-4) located in Halden, Norway.

The theatre was built in 1838 and is Norway’s oldest surviving baroque theatre and shares many common design and décor features with the Bergen Theatre. The spatial extents of the gallery floors were based on the Blix drawings and seating plan

120 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

mentioned previously. The Blix drawings did not provide accurate measurements of the original gallery extents and were only hand-drawn. The seating plan clearly shows the gallery extents although no scale existed in the drawing to suggest placement. A common feature shared between the Blix drawings and the 1938 survey data was the position of the supporting columns in the auditorium. The column positions did not change over time and the 1870 gallery renovation was structured around these existing key supporting columns. It can be safely assumed that the extents of the original gallery were determined by the positions of the underlying supporting columns. Figure 5.4 shows the inclusion of the floors and the supporting columns for the gallery.

Figure 5.4: Auditorium floor and gallery with supporting columns 5.1.2.2.2 External Windows, Doors and Roofing External windows, doors and Roofing were all reconstructed based on the 1938 survey drawings. The survey drawings contained elevation detail regarding roof trusses thereby allowing the construction of the roof to a high degree of accuracy. Figure 5.5 shows an example of the completed roof trusses on which the exterior roof skin was added.

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 121

Figure 5.5: Detail of roof truss system.

Windows, doors and external weatherboard cladding were added following the completion of the roof system (figure 5.6). As previously stated, it is assumed external features did not change over the life of the building. However, as this cannot be definitively ascertained, an element of uncertainty exists with relation to the accuracy of the pre-1870 external building features.

Figure 5.6: Completed external features of the Bergen Theatre: 5.1.2.2.3 Internal Stairs and Doors Internal stairways and doors were added after the completion of the major structural features of the building. The stairways were all reconstructed utilising the 1938 survey drawings. It is highly unlikely that the location and positioning of the stairways had changed from the time the theatre was constructed. Doing so would

122 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

require major structural renovation and no evidence has been found to support such an assertion. The doors, on the other hand, present interesting issues. Comparison of the 1938 survey drawings and Blix’s 1870 drawing show a disparity in the design of the doors (Figure 5.7).

Figure 5.7: Example comparison of door designs between 1938 official survey drawing (left) and Blix's renovation drawings of 1870 (right). Courtesy of Holledge et.al.

Comparison between drawings show it is not possible to arrive at a definitive design. It is possible that the designs in the Blix drawings are simply indications of doors and not representative of the actual design in use prior to 1870. The designs evident in the 1938 survey drawings are quite detailed and drawn to scale, it is equally possible that the designs described were added at the 1870 renovation. As no definitive answer can be arrived at the door designs evident in the 1938 survey drawings were used as a guide (figure 5.8).

Figure 5.8: View of the auditorium showing internal doors.

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 123

Comparison of the 1850 seating plan to the 1938 survey drawing clearly shows fewer entrances to the gallery pre-1870. The 1850 seating plan shows only 3 entrances to the gallery wings compared to four in the 1938 survey drawings. Examination of the Blix renovation drawings aligns to some degree with the 1938 survey drawings (figure 5.9). The Blix drawings show the same number of doorways as in the 1938 survey drawings although the positions do not align. It is highly likely that additional entrances to the wings were included in the 1870 renovation to accommodate the boxes established in the wings of the gallery.

Figure 5.9: Comparison of doorways, the pre-1870 seating plan (Top), a Peter Blix renovation drawing (Middle) and the official 1938 Survey drawing (Bottom). Doorways outlined in red. Courtesy of Holledge et.al.

This aspect of the reconstruction required several iterations in the modelling phase to ascertain if the doorways in the 1850 seating plan were supported by the structural elements of the building.

The positioning of the stairways was a very straight forward process. The 1938 survey drawings clearly show the positions and style of the stairways. Stairways were reconstructed with a high degree of certainty (Figure 5.10). As with the doorway positions, what is unknown is if the associated styling of the bannisters and railings were original features or changed over the life of the building. It is possible that during the 1870 renovation the stairs were refurbished as would be in keeping with an internal

124 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

renovation. A level of uncertainty exists for this section of the reconstruction and is detailed in the classification section.

Figure 5.10: Internal stair detail. 5.1.2.2.4 Orchestra Pit and Stage The reconstruction of the orchestra pit and stage required the completion of the main building framework. No records exist to suggest that the stage or orchestra pit underwent any renovation of the life of the building. It is highly likely that the spatial extents of the stage and the orchestra pit did not change over time. This would have required extensive modification of the building structure. The 1938 survey drawings were used to reconstruct the stage and orchestra pit (Figure 5.11).

Figure 5.11: View showing reconstructed Orchestra Pit and Stage

It is possible that both the orchestra pit and proscenium arch were renovated during the 1870 Blix renovation. Comparison of the Blix drawings with the 1938

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 125

survey drawings (Figure A-5) show similarities of stage columns and proscenium arch. The 1870 gallery intersected the front of the stage and overhung the orchestra pit, it follows that the orchestra pit would have been renovated at the same time as the gallery. It is highly likely that the 1938 survey drawings are not indicative of the original stage front and orchestra pit. The style and design of these pre-1870 elements are somewhat uncertain as no definitive information exists to support a conclusion. The addition of a prompter’s box to the stage front represents an area of very high uncertainty. It is highly likely that a prompter’s box existed. This was standard practice in a purpose-built theatre. The 1938 survey drawing show that ample room below the stage existed to accommodate it and the stage machinery. The design of the prompter’s box, a shell, (Figure A-6) was quite common at the time the Bergen theatre was in operation. It is unlikely a prompter’s box did not exist, but the design and exact positioning is uncertain.

5.1.2.2.5 Gallery The reconstruction of the pre-1870 gallery presented a challenge as no definitive architectural drawings of the original theatre exist. The 1870 renovation was concentrated primarily on the gallery area. It is known that prior to 1870 there were no private boxes in the gallery wings. This is supported by the 1850 seating plan (Figure A-1). The Blix drawings clearly show a hand drawn outline of what has been interpreted as the original gallery outline and the supporting columns (Figure A-7). The columns marked on the Blix drawings match very well with the 1938 survey drawings. This strongly suggests that the original supporting columns for the gallery were not altered and that the renovated gallery and wings used the original columns for support. Conversely this evidence also supports the possibility that the original gallery was straight sided and rested on the columns. The shape and extents of the original gallery can be assumed to be of a high certainty. Both the documentary evidence and the physical restraints of the building support this conclusion.

The original survey drawings show that the 1870 renovation of the gallery area resulted in a horseshoe gallery with private boxes in the wings and angled partition. The gallery is heavily decorated with presumably plaster fixtures (Figure 5.2). The Blix drawings also show similar designs to the 1938 survey drawings further reinforcing the assertion that these features were added in 1870. It is highly likely that the pre-1870 gallery was much more utilitarian in design and lacked the ornate features

126 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

of later renovations. There is no physical or documentary evidence to suggest the exact design of the original galleries. Other examples of similar theatres of the time such as the Georgian Theatre Royal located in Richmond, North Yorkshire suggest a much simpler design with flat gallery fronts and distinctly square design (Figure A-8). The main difference in design between the Georgian example and the Bergen theatre is that both the Blix drawings and the 1850 seating plan clearly show rounded corners at the intersection of the wings and gallery back. The resultant reconstruction draws on the Blix drawings and seating plan to determine the spatial layout of the gallery and wings whilst examples such as the Georgian Theatre are used to determine style and décor. The reconstructed gallery and wings (Figure 5.12) has a high certainty with regards the spatial extents but relatively low uncertainty with regards the exact décor.

Figure 5.12: The reconstructed Gallery and Wings.

The final aspect of the galleries which remains uncertain is the position of the entrances. The 1850 seating plan (Figure A-1) clearly shows only one entrance per wing and three entrances to the area containing the ‘Prince’s Box’ at the rear of the Gallery. This is very different to the 1938 survey drawings which show multiple access-ways in the wings and only two entrances at the rear of the gallery. The location of the doorways in the 1856 reconstruction are primarily informed by the seating plan supported by the underlying building framework.

5.1.2.2.6 Ceiling and Internal Linings

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 127

The 1938 survey drawings provide no indication of the design of the auditorium ceiling other than the position and dimensions of the recess for the chandelier or lighting element (Figure 5.13).

Figure 5.13: Side Elevation showing the chandelier recess in the auditorium ceiling. Courtesy of Holledge et.al.

The Blix drawings present a detailed drawing of the auditorium ceiling. It is unknown if the drawing (Figure A-9) is the original ceiling, a proposed design or a renovation of the existing ceiling. The design of the ceiling (figure 5.14) is therefore highly uncertain. The ceiling design was based on the basic layout in the Blix drawings and deliberately avoids the inclusion of detail. This aspect of the reconstruction is highly uncertain with only the chandelier recess being of low uncertainty.

Figure 5.14: Reconstructed ceiling showing chandelier recess and low detail ceiling relief.

The 1938 survey drawings (Figure 5.2) explicitly detail wall linings for the stage area, the orchestra pit and the auditorium. In all these cases the linings are clearly either Vertical Joined (VJ) wood planks as in the Orchestra Pit and Stage or horizontally laid

128 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

boards as in the auditorium. The survey drawings allow for precise measurement of the board width. As has been argued, it is unlikely that elements such as the orchestra pit, the stage and the auditorium underwent changes over the life of the building. It is reasonable to conclude that wall linings for the stage and auditorium were the original ones and remained till the buildings demise. The Orchestra Pit may have undergone cosmetic renovations in 1870 with regards the wings although this cannot be determined based on the available evidence. The reconstruction of the orchestra pit (Figure 5.15) is based on the 1938 survey drawings and represents a level of uncertainty with regards the linings and décor. The spatial extents of the pit are highly certain. It is also likely that the actual lining of the orchestra pit with the exclusion of the wings would have remained unchanged.

Figure 5.15: Detail of the Orchestra Pit Reconstruction 5.1.2.2.7 Fixtures The fixtures were the final elements of the reconstruction to be added. These included, chandelier, wall lighting and curtain rods. The 1938 survey drawings do not detail the interior décor. Numerous photographs of the interior of the theatre taken in the early 1900’s clearly show a gas lamp chandelier (Figure A-10). Comparison of early interior photography and the Blix drawings (Figure A-11) show similarities in style of lighting fixtures and use of curtains. The Blix drawings give no indication if the fittings included in the drawings are original items or new additions. It is certainly possible that the central chandelier remained unchanged over time although the theatre was converted to gas lighting in the early 1850’s. Although it is highly likely that the chandelier existed from the inception of the building it cannot be verified. The same is

5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case 129

true for the wall fittings and curtains. This element of the reconstruction carries a high degree of uncertainty. It is certain that some form of lighting existed, it is uncertain what that lighting looked like. It is not certain that the doorways had curtains as indicated in later iterations. The final reconstruction and placement of fittings (Figure 5.16) is based on a combination of varying documentary data sources.

Figure 5.16: Render showing internal fittings.

5.1.3 Discussion Application of the workflow pipeline has demonstrated that it provides a structured approach to the whole reconstruction process. The first stage of the pipeline facilitated the organisation of the data inputs. Data inputs were assigned into classes of uncertainty. The data with the highest levels of certainty, in this case the 1938 survey drawings, were used as the ‘base’ for the entire reconstruction. In this reconstruction there were no archaeological remains, the 1938 survey effectively became the archaeological survey. This demonstrated that the workflow pipeline is applicable to cases where the absence of archaeological remains can be substituted for reliable documentary evidence in the form of the 1938 survey drawings. The workflow pipeline relies on iteration to arrive at a solution. The creation of the gallery was a case in point. The pre-1870 gallery required multiple input sources. Corroborating the evidence in the varying inputs required testing and iteration to arrive at a solution. This reconstruction has demonstrated the effectiveness of the workflow pipeline.

130 5 Results - Application of the Classification System and Workflow Pipeline – The (Komediehuset) Bergen Theatre Test Case

6 Results - Application of the Visual Classification - Case Studies

In this chapter visual classification system is applied to the two case studies.

I. The Rose Theatre II. The Bergen Theatre The revised visual classification system is applied to the Rose and Bergen theatre digital reconstructions and discussed. The work-flow pipeline is briefly discussed in the context of the described reconstructions. The visual classification system is applied to both studies and discussed.

6.1 ROSE THEATRE – APPLICATION OF THE VISUAL CLASSIFICATION 6.1.1 Summary of the visual classification procedure The final iteration of the decision tree resulted in a colour classification system as previously detailed. Colours were chosen to represent each of the three data input types (branches) of the decision tree (Table 6.1). The premise of classification system is that a hierarchy of input data type exists. That is, physical extant remains will always be considered first when ordering and classifying the input data. Physical extant (where available) remain the basis of any reconstruction. Documentary and Analogous data inputs will always be preceded by physical evidence.

6 Results - Application of the Visual Classification - Case Studies 131

Class (Data Description Colours type) Extensive extant remains allowing an accurate and authentic reconstruction Extant Indicative extant remains allowing a partial Archaeological reconstruction and strong basis for inference Remains Sparse extant remains allowing very little reconstruction and requiring much inference Extensive and accurate documentary evidence. Includes Archaeological records, Architectural plans, Remotely Sensed data etc. Documentary Descriptive and historical texts. Includes text and Evidence images directly related to the subject matter. Stories, narratives, myths related to the subject matter or specific practises associated with the subject matter. Onsite analogous remains. Structures and buildings like the subject matter and sharing the same building techniques and of the same era. Off-site analogous remains. Structures and buildings Analogous like the subject matter and sharing the same building Evidence techniques and of the same era. Remains with similarities. Not necessarily of the same era or location but exhibit similarities of construction and materials. Table 6.1: Summary of the colour classification system

Each colour is graded into three shades. The colour shades represent ranges of uncertainty in the data for that class. The darker the shade the higher the uncertainty of the data inputs. The three colours chosen for each class are somewhat arbitrary and primarily based on the fact that Red, Green and Blue are primary colours that form the additive colour model (Silva, Sousa Santos, & Madeira, 2011). The logic behind the choice of colours is based on the type of evidence class represented. In the case of archaeological evidence this is considered the most solid type of evidence and was assigned the colour Green. Documentary evidence has been assigned the colour Blue as this class generally carries moderate to high level of certainty. Both Green and Blue can be considered as having ‘positive links in the natural realm (e.g., blue sky and water, green foliage and vegetation) and both have been shown to be associated with positive content ’ (Elliot & Maier, 2014, p. 108). Both these classes represent the higher levels of certainty inherent in these data inputs. Conversely the Analogous evidence class carries a wider range of certainties. Although a structure may be

132 6 Results - Application of the Visual Classification - Case Studies

analogous it is not the same structure as that in question. Analogous evidence generally leads to inference and conjecture. The Analogous evidence class has been assigned the colour Red. The colour Red is acknowledged as having greater visibility than the other colours and hypothesized to represent ‘…negative meaning (failure) and aversive implications’ (Elliot & Maier, 2014, p. 99). Red is typically associated with warnings or to draw attention. There are many examples of reconstructions based primarily on analogous evidence such as the Gladiator School reconstruction (W. Neubauer et al., 2014).

6.1.2 Reconstruction Colour Classification Table A-4 details the application of the colour classification system to the Rose Theatre reconstruction. Each major component of the reconstruction has been assigned a colour based on the outputs of the decision tree. The classification system is not limited to building components but can also be applied to individual construction elements as will be detailed in the next section. In some cases, it is necessary to classify individual building elements to accurately depict the overall uncertainty inherent in a building component. This is exemplified in the Rose Theatre reconstruction where the stage and stage canopy are comprised of varying levels of data input uncertainty and yet the overall component can be considered moderately certain. An interactive visualisation of the Rose Theatre classification is presented at https://ortelia.com/Downloads/Rose_PHD/.

6.1.3 Applying the colour classification to the Rose Theatre Three main building components of the Rose Theatre reconstruction colour classification will be the focus of this section. These are:

• The Rose Exterior and Entrance

• The Stage and Canopy

• The Stair system

6.1.3.1.1 The Rose Theatre Exterior and Entrance Figure 6.1 shows the Rose Theatre reconstruction as both a completed textured model and with the major building components classified according to the decision support tree.

6 Results - Application of the Visual Classification - Case Studies 133

Figure 6.1: The exterior of the reconstructed Rose Theatre textured (top) and classified (bottom)

As can be seen above most of the exterior of the structure is classified based on available archaeological inputs. Although the amount of remains uncovered were limited it is highly likely that the walls were constructed of lath and the roof of thatch. The supporting framework for the building is classified in medium Red as no remains were unearthed and are based on analogous half-timbered building examples still in existence. The windows, surrounding fences, and bridges have been based solely on documentary evidence and represent a high uncertainty in terms of input data. The type, placement and size of windows cannot be determined. The extant fragments of what may have been window glass and documentary evidence support the presence of

134 6 Results - Application of the Visual Classification - Case Studies

windows. The classified render is effective in allowing the user to quickly discern which parts of the reconstructing are based on archaeology and what are based on other less certain inputs.

The entrance to the Rose Theatre (Figure 6.2), as discussed, has been a source of debate amongst scholars (Gurr, 1992). .

Figure 6.2: The exterior entrance of the reconstructed Rose Theatre textured (top) and classified (bottom)

The exact size and position of the entrance cannot be determined from the archaeology. Documentary evidence suggests the entrance was located on the South side of the building. Theatre practices of the time suggest a narrow entrance to the theatre and normally located opposite the stage (Bowsher & Miller, 2009). The

6 Results - Application of the Visual Classification - Case Studies 135

entrance in this reconstruction has been based on documentary evidence As can be seen from Figure 6.2 the entire entrance component is classified as originating from documentary inputs alone and represents a high degree of uncertainty.

6.1.3.1.2 The Rose Theatre Stage and Stage Canopy Figure 6.3 shows both the completed textured stage with components classified according to the decision support tree. The majority of the stage is classified as being sourced from documentary evidence.

Figure 6.3: The Rose Theatre Stage textured (Top) and classified (Bottom)

Although the stage outline itself has been derived from extant foundations unearthed during the archaeological survey it is not visible in the reconstruction as it is covered by the hypothesised wood cladding. Figure 6.4 shows the stage foundations

136 6 Results - Application of the Visual Classification - Case Studies

and framework exposed. As can been seen the foundations of the stage are classified as highly certain archaeological inputs.

Figure 6.4: Rose Theatre classified Stage frame and Foundations exposed

The stage framing, in contrast, is based on documentary evidence such as existing examples of timber joinery of the era. The bases of the pillars are coloured dark Green considering the archaeological evidence. These foundations suggest that two pillars existed to support the stage canopy. The pillar bases are classified thus to reflect the archaeology. The pillars however have been classed based on documentary evidence. Documentary evidence suggests a covered stage as this was a standard feature (Figure A-12) of playhouses of the time.

The stage itself has been classified as having a high certainty based on the documentary inputs used. The central ‘trapdoor’ is classified as having a high uncertainty. Theatre historians analysis of Shakespearian theatre suggest that stage trapdoors were a feature of playhouses of the era (Gurr, 1997). The size and positioning of the trapdoor is based on the most central location achievable within the physical framework of the stage.

6.1.3.1.3 The Rose Theatre stair system The positions of the internal stairs (Figure 6.5) have been discussed in earlier sections.

6 Results - Application of the Visual Classification - Case Studies 137

Figure 6.5: The Rose Theatre Stair system textured (Top) and classified (Bottom)

The placement and design of the stairs throughout the structure represents a ‘most likely’ scenario based on the available evidence. It is almost certain that the Rose playhouse was a multi-tiered building and would therefore require stairs for access to the upper levels. External staircases were not uncommon as was evidenced in the 1599 Globe Theatre. However, no evidence of external stair foundations was identified in the archaeological survey although a significant portion of the foundations were unearthed. In the absence of physical evidence, it is reasonable to assume internal stairs. No evidence of external stairs can be seen in the available documentary evidence. The stair system has been classified (Figure 6.5) as being derived from inference of documentary evidence and represents a high uncertainty input.

138 6 Results - Application of the Visual Classification - Case Studies

The application of the visual classification to the Rose Theatre reconstruction provides a means by which a user can make a rapid assessment as to the veracity of the reconstruction based on the types of data inputs. The use of standardised colour scheme allows the identification of the source data types and the degree to which they contribute to individual elements of the reconstruction. Application of the classification system shows that it can be tailored to either individual reconstruction components such as the stage elements or broader reconstructed elements such as the building framework.

6.2 BERGEN THEATRE – APPLICATION OF THE VISUAL CLASSIFICATION The application of the visual classification system to the Bergen Theatre reconstruction followed the same methodology as that used in the Rose Theatre. Individual reconstruction elements were classified according to the assigned certainty of the data inputs used. In some cases, multiple sources of data with varying levels of certainty were used (eg. Doorways into gallery and auditorium). In these situations, an average level of certainty was used as the final classification colour. The classified components of the Bergen Theatre reconstruction are detailed in Table A-5.

6.2.1 Applying the colour classification to the Bergen Theatre Building Elements 6.2.1.1.1 Main Building mass structure As has been stated the main building mass structure is based on the 1938 official survey drawings of the Bergen theatre. There are no records to suggest that the main building structure and layout was altered over the life of the building. In addition to the survey drawings numerous external photographs captured after 1870 confirm the details recorded in the survey drawings. It is reasonable to assume that the high accuracy of the survey drawings coupled with the external photography provide high certainty to the exterior reconstruction.

6 Results - Application of the Visual Classification - Case Studies 139

Figure 6.6: Bergen Theatre external reconstruction showing the unclassified (top) and classified (bottom) views.

Figure 6.6 above shows the external building reconstruction as unclassified and classified. The external reconstruction has been assigned a high level of certainty. The individual reconstruction elements such as the Walls, Doors, Windows, Roofing were clearly detailed in the survey drawings and supported by external historical photography.

6.2.1.1.2 Internal Stairs and Doors The internal stair system is documented in the 1938 survey drawings (Figure A 13). The plan and elevation drawings allow for the accurate reconstruction of the spatial extents of the individual stair systems. The width, rise, tread depth and height can all be accurately measured and verified across drawings. The Blix 1870 renovation

140 6 Results - Application of the Visual Classification - Case Studies

drawings corroborate the 1938 official survey drawings. The reconstructed stair systems (Figure 6.7) is arrived at using high certainty data.

Figure 6.7: Bergen Theatre internal stairway reconstruction showing the unclassified (top) and classified (bottom) views

The stair bannisters are not clearly detailed in the official survey drawings (Figure A-13The reconstructed bannisters are an approximation extracted from the survey drawings. The height and position of the bannisters can be assumed to correct as they are consistent across all the survey drawing. It is possible that the bannisters may have been changed as part of the overall gallery renovation of. It can be safely assumed that the stairways had bannisters however their design is less certain.

6.2.1.1.3 Orchestra Pit and Stage The orchestra pit (Figure 6.8) has been arrived at using the 1938 official survey drawings. The orchestra pit is an integral component of the main building structure and is a purpose-built component.

6 Results - Application of the Visual Classification - Case Studies 141

Figure 6.8: Bergen Theatre orchestra pit reconstruction showing the unclassified (top) and classified (bottom) views

It is highly unlikely that the main orchestra pit was altered throughout the life of the theatre. To do so would have involved major structural changes to the stage area. The lines of the gallery extents changed between 1856 and 1870. The new gallery extents intersected the orchestra pit and an angle. This would suggest that, at a minimum, minor structural changes were made to the orchestra pit wings to accommodate the new gallery. It can be assumed that the wings of the gallery may well have been renovated to be in keeping with the new gallery. The certainty of the orchestra wings representing the 1856 iteration of the building is less than that of the main orchestra pit as reflected in the classified image.

The stage was reconstructed (Figure 6.9) using the 1938 survey drawings as the primary data input. In addition to the survey drawings, the Blix renovation drawings (Figure A-14) and internal photography (Figure A-15) were referenced. The stage, as

142 6 Results - Application of the Visual Classification - Case Studies

with the orchestra pit, is integral to the main building structure. Not only are the visible parts of the stage integral components of the structural framework but also the underlying space which housed the stage machinery and floor supports. There is no evidence to suggest that the stage underwent any significant structural changes. It is possible that the stage proscenium arch and pillars underwent renovation in 1870 in keeping with the gallery changes.

Figure 6.9: Bergen Theatre stage reconstruction showing the unclassified (top) and classified (bottom) views

The main structural elements of the stage are of a high certainty, however, there are elements which cannot be verified through the available inputs. The proscenium arch and pillars may have undergone cosmetic renovation in 1870. It is unknown if the proscenium arch and pillars are the original components of the 1856 version of the theatre. As the design of this aspect of the stage reconstruction cannot be verified it has a higher level of uncertainty to the main stage structure. It is known that the Bergen

6 Results - Application of the Visual Classification - Case Studies 143

theatre employed the use of under-stage machinery to manipulate sets on stage. (Holledge et. al 2018). The floor contained slots which allowed for the movement and changing of side scenery flats. These floor slots were a regular feature of Baroque theatres of the era (Figure A-16) of which there are many surviving examples. It is known that stage scenery was used extensively in the Bergen theatre from its inception. Extensive records of plays performed at the Bergen theatre remain providing an insight into the type of scenery used (Holledge et al 2018). The official 1938 survey drawings do not contain any information confirming the presence of the stage slots and the associated stage machinery. In addition to the floor slots, there is no indication in the 1938 survey drawings of the Prompter’s box located at the front of the stage. Prompter’s boxes (Figure A-6) were, and are, a common feature of many theatres. It is highly likely the stage contained a Prompter’s box. The inclusion of the Prompters’ box is effectively supposition based on theatre practise of the time it is assigned a high uncertainty.

6.2.1.1.4 Gallery The reconstruction of the 1856 gallery (Figure 10) contained the highest degree of uncertainty in terms of individual components.

144 6 Results - Application of the Visual Classification - Case Studies

\ Figure 6.10: Bergen Theatre gallery reconstruction showing the unclassified (top) and classified (bottom) views

The final spatial extents of the 1856 gallery were determined from the 1856 seating plan (Figure A-1) and comparison with the Peter Blix renovation drawings. The 1856 seating plan is a well-drawn to-scale drawing of the gallery and auditorium seating. Features such as doorways are consistent on both sides of the gallery as are door widths. The spatial extents of the gallery can be considered to have high certainty, and this is reflected in the classified image (Figure 10) with the main gallery floor being classified low uncertainty (Light blue).

The position of the entrance doors to the gallery are less certain. The 1856 seating plan clearly shows two entrances per side to the wings and three entrances at the rear of the gallery (Figure A-1). As discussed in previous sections comparison of the gallery entrances show three entrances per wing in 1856 and six entrances per wing after 1870. It is highly likely that additional entrances to the wings were included in

6 Results - Application of the Visual Classification - Case Studies 145

the 1870 renovation to accommodate the boxes established in the gallery wings. As the exact position of the 1856 gallery entrances cannot be verified with any other evidence they have been classified as highly uncertain. This is primarily in reference to the exact location of the entrances. The physical appearance and dimensions of the Princes’ box can only be inferred from the surrounding gallery structure. The exact physical appearance of the structure cannot be verified.

The rear of the gallery appears to have had a raked floor. This was, and still is, common practice in theatres to accommodate sight-lines for views at the rear of the gallery. Examination of the Peter Blix renovation drawings clearly shows the pre-1870 gallery floor pencilled in and following the same angle of rake as the auditorium floor. It is assumed that the rake for the gallery floor would have been identical to that of the auditorium floor for ease of build. The floor rake is of only a moderate uncertainty as no official drawings or plans exist.

The only element of the gallery reconstruction which carries a high degree of uncertainty is the bannisters. Neither the Blix drawings nor the 1856 seating plan contain any information to suggest the design and physical dimensions of the bannisters. The reconstruction of the bannisters is based on analogous evidence in the form of existing Baroque theatres of a similar era (Figure A-8). However, no evidence exists to verify that the physical dimensions and the decor of the bannisters in the reconstruction. Although we can be highly certain that bannisters did exist, and in the positions indicated, they remain a highly uncertain aspect of the gallery reconstruction.

6.2.1.1.5 Auditorium, Ceiling and Seating The auditorium (Figure 6.11) represents an aspect of the reconstruction of the Bergen Theatre that contains a high degree of certainty. No evidence exists to suggest the auditorium underwent any renovation. The auditorium is an element of the building that is firmly tied to the underlying structural framework. As with the exterior of the building, any major change to the auditorium would have required structural alteration. Similarly, the rake of the auditorium floor is unlikely to have changed.

It is possible that the auditorium entrance linings and doors may have changed over time. The 1938 survey drawings provide details of door appearance and décor. It is unknown if the doors and entrance linings are representative of the 1856 version of the space.

146 6 Results - Application of the Visual Classification - Case Studies

Figure 6.11: Bergen Theatre auditorium reconstruction showing the unclassified (top) and classified (bottom) views

The position of the auditorium entrances represents high certainty as they correspond with the key structural supports in the underlying framework and are clearly detailed in the 1938 survey drawings. The floor of the auditorium is based on existing analogous theatres of the same era (Figure A-8). Typically, floors were constructed of exposed timber boards laid laterally along the main axis of the space. The 1938 survey drawings do not provide any detail regarding the flooring.

A variety of sources were employed in the final construction of the ceiling and associated plasterwork detail. The 1938 official survey drawings (Figure A-17) were used for the main construction of the internal ceiling. The drawings clearly and accurately depict the overlying roof framework and the position of the chandelier recess. As with other major structural elements of the building, there is no evidence to suggest the roofline or internal ceiling of the building underwent significant structural

6 Results - Application of the Visual Classification - Case Studies 147

change. The height and position of the reconstructed ceiling (Figure 6.12) is of high certainty as it the position and size of the central chandelier recess.

Figure 6.12: Bergen Theatre internal ceiling reconstruction showing the unclassified (top) and classified (bottom) views

The reconstructed design of the ceiling plasterwork is less certain. The 1938 official survey drawings do not detail the ceiling decor. It is highly likely that the ceiling plasterwork was altered during the 1870 gallery renovation. The Peter Blix renovation drawings present a study of the ceiling décor (Figure A-9) and it would be in keeping with the gallery renovation to include the ceiling décor. In addition to the Blix drawings a scale model of the post-1870 renovation (Figure A-3), housed at the university of Oslo, depicts a design of the ceiling décor which similarities to the Blix proposal. The reconstructed ceiling detail was modelled from the scale model mentioned. This aspect of the reconstruction represents a high uncertainty as the original ceiling detail cannot be ascertained.

148 6 Results - Application of the Visual Classification - Case Studies

The reconstruction of the seating for the 1856 version of the Bergen theatre (Figure 6.13) used both documentary and analogous evidence to arrive at a solution. The 1856 seating plan (Figure A-1) clearly details the number of seats available in both the Auditorium and galleries. The seating plan shows the seating type to be benches in both the auditorium and the galleries. The 1856 seating plan is a data source of high certainty. There is no indication in the seating plan as to the type or style of benches used.

Figure 6.13: Bergen Theatre seating reconstruction showing the unclassified (top) and classified (bottom) views

It was common for Baroque theatres of the era (Figure A-7) to utilise very simple utilitarian benches as seating. Analogous examples such as the Georgian Theatre Royal located in North Yorkshire were used as a guide to the style and dimensions of the reconstructed seating. The position and distribution of the seating can be considered

6 Results - Application of the Visual Classification - Case Studies 149

highly certain. The design and style of seating on the other hand is much less certain and is classified accordingly.

6.2.1.1.6 Fixtures The internal fixtures such as lights (Figure 6.14) have been reconstructed directly from internal photography taken after 1870. It is highly unlikely that these elements represent the original elements installed when the building was constructed. It is known that the theatre began life using candles for lighting. In the mid 1850’s the theatre was converted to gas lighting. It is possible that the lighting elements detailed in the photography are the original gas lighting installed. It is also possible however, that the lighting fixtures were upgraded in the 1870 renovation. Internal photography post-1870 clearly show the chandelier and wall lighting elements. The lighting fixtures are determined to be of a high uncertainty as it cannot be verified that they existed in the 1856 version of the theatre.

Figure 6.14: Bergen Theatre internal fixtures reconstruction showing the unclassified (top) and classified (bottom) views

150 6 Results - Application of the Visual Classification - Case Studies

The application of the classification system to the Bergen Theatre reconstruction reinforces the outcomes of the Rose Theatre experience, that is, colour classification of reconstruction elements allows a rapid assessment of the authenticity of the reconstruction based on the types of data inputs. The reconstruction of the Bergen theatre offered a unique opportunity to recreate a lost structure to a high degree of accuracy. On the face of the evidence used in the reconstruction, it can be considered to have high authenticity. However, even though accurate drawings were available for the major parts of the reconstruction, several key areas required inference from other data sources. Typically, these ‘inferences’ are not explicit and obvious in a digital reconstruction. The application of the colour classification allows a user to quickly identify, and therefore question, elements whose originating data sources are of low certainty. It is clear when reviewing the classified reconstruction that most of the reconstruction has been arrived at with high certainty data. Highlighting the elements of lower certainty inputs provides transparency and serves to give the final reconstruction greater degree of validity.

6 Results - Application of the Visual Classification - Case Studies 151

7 Results - Application of the visual classification system to external case studies

In the following two test case studies I seek to apply the classification system developed during this study to digital 3D archaeological reconstructions undertaken by researchers unrelated to the current study.

I. School of Gladiators, Carnuntum, Austria II. Temple of Apollo, Kos, Greece In both cases a digital reconstruction has been undertaken based on a variety of input sources and inferences related to those sources. It is acknowledged that both case studies have been developed independently of the classification system described here and the lack of a workflow methodology leading to the final reconstruction imposes limits on the classification. Overall results of the application of the classification system show it remains effective despite the lack of a workflow pipeline description. The presence of a workflow pipeline greatly enhances the application of the classification as individual reconstruction components are considered from the outset. Application of the classification system to a digital reconstruction that does not include a workflow pipeline results in some elements of a reconstruction not being visually classified. These limits are discussed.

7.1 SCHOOL OF GLADIATORS, CARNUNTUM, AUSTRIA 7.1.1 Background The School of Gladiators is located within a greater archaeological complex known as the Roman Carnuntum. The site is approximately 40km south-east of Vienna and located on the southern bank of the River Danub. (Wolfgang Neubauer et al., 2014). The site ‘constitutes the largest archaeological landscape in Austria’ and has seen continuous archaeological exploration for more than 100 years (Neubauer et al., 2014 p.174). Despite this long-term archaeological research only a small portion of the entire site has been subjected to archaeological excavation. The School of Gladiators site has not been part of these archaeological excavations (Wolfgang Neubauer et al., 2014). The presence of intensive farming practises, infrastructure development and

152 7 Results - Application of the visual classification system to external case studies

historical looting by treasure hunters has resulted in a ‘dramatic increase in the irreversible destruction of the archaeological deposits’ (Neubauer et al., 2014, p. 176).

7.1.1.1 Study Summary In 2000 a series of initial Magnetic and Ground Penetrating Radar (GPR) surveys were conducted in association with capture. These remotely sensed surveys revealed traces of the foundation walls of a large building complex and associated infrastructure (Wolfgang Neubauer et al., 2014). The GPR data provided detailed 3D information about the ‘approximate depth, shape and location of archaeological structures’ (Wolfgang Neubauer et al., 2014, p. 177). These initial remotely sensed data provided the justification for a more detailed remote sensing survey of the site. In 2011 the researchers employed a variety of remote sensing techniques to provide a highly detailed 3D survey of the site known as the School of Gladiators. The remote sensing tools employed were:

• High-definition Ground Penetrating Radar (GPR) survey

• High-resolution Electromagnetic Induction (EMI)

• Airborne imaging spectroscopy (AIS) survey in 105 spectral bands

• Airborne Laser Scanning (ALS) for high-resolution topographical data

The data acquired was imported and registered within a Geographic Information System (GIS) (Figure A-18). The GIS formed the basis for all further interpretation. It must be stressed at this point that at no time did the researchers undertake a physical archaeological investigation of the site. That is, no soil was turned to expose the foundations, no physical evidence was unearthed and, most importantly, no ground- truthing of the remotely sensed data in the form of excavation was undertaken.

The archaeological interpretation of the remotely sensed data identified a rectangular walled area of approximately 11000 square meters with no adjacent buildings or significant infrastructure (Wolfgang Neubauer et al., 2014). Within the walled area the researchers identified footprints of buildings and structures arranged around a central inner courtyard. The total building areas is estimated at 1700 square meters (Wolfgang Neubauer et al., 2014). The archaeological data imported into the GIS was used to generate an initial visualisation of the still unearthed foundations (Figure A-18). The visualisation of the extant remains closely mirrors the proposed workflow pipeline which requires the inclusion and visualisation of the extant remains

7 Results - Application of the visual classification system to external case studies 153

as a starting point. The researchers, however, do not elaborate on this step, nor do they refer to it as an important step in the reconstruction. This is an important point to note and all subsequent reconstruction and associated interpretation of the site is derived from this initial step of reconstructing the extant archaeology.

Numerous structures within the bounds of the site were identified through the archaeology (Table A 7). The site has been interpreted by the researchers as a Gladiator School (ludus) (Wolfgang Neubauer et al., 2014, p. 183). This interpretation was based on several factors. The main reasoning is the ‘close topographical relationship to the civilian amphitheatre.’ (Wolfgang Neubauer et al., 2014, p. 183). The authors do not specify why this is important although it is presumed that the ‘Civilian Amphitheatre’ would have been the site of Gladiatorial games. The second main reasoning is the internal layout of the site itself which has analogies with other sites of similar use unearthed at other locations which ‘confirm the proposed interpretation ‘ (Neubauer et al., 2014, p. 183). The site is interpreted as having a single ‘easily controlled’ entrance to the compound with a ‘driveway’ and ‘separate path’ forming the paved access to the compound (Wolfgang Neubauer et al., 2014, p. 183). The walls lining the pathway have been interpreted as the ‘base of a portico or canopy’. The path ends at a formal entranceway to the compound (Wolfgang Neubauer et al., 2014, p. 183). The formal entranceway has been interpreted to have had a Triumphal Arch based on foundation size. The interior of the compound has been interpreted as housing two large rectangular buildings (Figure 7.1). These buildings have been interpreted to house a number of spaces which included the Administration block, the Gladiator living quarters, Bath Houses and preparation area for training.

154 7 Results - Application of the visual classification system to external case studies

Figure 7.1: Perspective view of the reconstructed Gladiator school (ludus) showing the interpreted layout. Source: (W. Neubauer et al., 2014, p. 2)

The most prominent feature within the compound, and that which weights the interpretation towards a Gladiatorial school, is the 19 metre diameter circular structure which has been interpreted as the ‘training area’ for the Gladiators (Wolfgang Neubauer et al., 2014, p. 184). The interpretation of the remotely sensed data by the researchers is that the training area was surrounded by ‘wooden spectator stands set on stone foundations’ (Wolfgang Neubauer et al., 2014, p. 184). In addition to the stone foundations the data shows a central feature in the circle interpreted as the foundation of a ‘palus’, a wooden pole set in a stone foundation used by Gladiators for sword and body blow training (Wolfgang Neubauer et al., 2014, p. 184). It is worth noting that the researchers consider this feature one of the ‘strongest arguments’ for interpreting the site as a Gladiatorial school (Wolfgang Neubauer et al., 2014, p. 184).

7.1.2 Application of the Classification The decision support tree was employed to classify the individual components of the reconstruction based on the described data inputs. These are detailed in Table A-7. For the purposes of this classification the foundation footprints identified through remote sensing are considered to be extant archaeological remains even though they have not been physically unearthed. Modern remote sensing techniques such as GPR and Magnetic survey are highly accurate. The Ground Penetrating Radar (GPR) used for the survey in this study has an accuracy of between 4 and 8 centimetres with significant depth penetrations and the Airborne Imaging spectroscopy has a ground sampling distance of 40cm (Wolfgang Neubauer et al., 2014). The point being stressed

7 Results - Application of the visual classification system to external case studies 155

here is that there is very little doubt that the extant remains exist and are buried beneath an overburden of topsoil. However, the exact composition of the foundation material cannot be definitively identified without archaeological excavation. All subsequent interpretation based on the site footprint is inference of the documentary evidence or examples from other non-geographically related sites (Analogous evidence). The logic for treating interpretations of the archaeology as documentary evidence is that the researchers draw on Gladiatorial practices documented by other researchers to show that the layout of the site fits with these descriptions. In addition, the researchers discuss the proximity of one structure to another to justify its proposed purpose. The training area and the presence of the ‘palus’ is a prime example of these conclusions.

The original 3D files of the reconstructed Gladiator School of Carnuntum were not available for this case study, therefore the only option was to overlay the classification on the final rendered images presented by the researchers (Figure 7.2). The application of the classification in this manner means that identified structures and features not visible in the original image cannot be classified. A significant premise of the methodology and classification proposed is that each individual building component can be identified during the reconstruction phase and assigned a data input class. This allows all aspects of the reconstruction to be rendered as a classified image. It also allows for the investigation of any part of the reconstruction. In this case the only elements that can be classified in a visual form are those visible in the available image.

Figure 7.2: Figure 2: Application of the classification to the Gladiator School image as presented. Original image source: (W. Neubauer et al., 2014, p. 2)

156 7 Results - Application of the visual classification system to external case studies

7.1.2.1.1 Overall Building complex Footprint The overall building complex footprint is classified as low uncertainty extant (light green). The accuracy and fidelity of the remote sensing equipment employed ensures the presence of building and wall foundations is indisputable. The resulting classified image shows the foundations of the entire complex outlined in light green to signify the presence of known foundations. In contrast, the entrance walkways and pathways of the colonnaded Gladiator living quarters are classified as extant remains with a higher uncertainty (Medium Green) compared to the footprint of the complex. The materials employed, and physical appearance of these features cannot be determined without excavation therefore representing a higher level of uncertainty.

7.1.2.1.2 Eastern Entrance and Gateway Arch The remotely sensed data clearly shows the position of the eastern (and only) entrance to the compound. The only details provided for the entrance is a width of 2.4 metres. (Wolfgang Neubauer et al., 2014) The eastern entrance has been reconstructed with a triumphal style arch (Figure 7.1). It is possible that an arch was present although no evidence has been provided to suggest this. It is highly probable that the authors drew on the numerous analogous examples built all over the , although not specifically from the Carnuntum site. However, as no evidence is available to suggest that an arch was indeed present it must be assumed it is based on analogous evidence and possibly a degree of ‘artistic license’. The arch had been classified as a high uncertainty analogous input.

7.1.2.1.3 Driveway and Paved Path Eastern Entrance The presence of a path and driveway leading from the Eastern compound entrance to the main entrance has been identified in the remotely sensed data (Wolfgang Neubauer et al., 2014). As with the pathways of the colonnaded Gladiator living quarters, the existence of these components is indisputable. However, the materials and physical appearance of these components cannot be confirmed without physical excavation. This represents a medium level of uncertainty and has been classified accordingly. The driveway from the eastern entrance to the main entrance has been rendered as a covered structure supported by columns. The authors have interpreted the extant footprint as the ‘base of a portico or canopy’ (Wolfgang Neubauer et al., 2014, p. 183). The authors do not describe how they have arrived at this interpretation. The authors refer to analogous examples although not from the

7 Results - Application of the visual classification system to external case studies 157

Carnuntum site. No evidence of columns was identified in the remote sensing survey to indicate that indeed a canopy and supporting columns existed. Accordingly, the columns have been assigned a high degree of uncertainty based on assumed analogous inputs whilst the overlying canopy has been assigned a medium level of uncertainty. The canopy has been assigned a medium level of uncertainty throughout the classification as there are numerous existing examples of the style of canopy rendered in the final reconstruction. The canopy structure comprised of a tiled roof has more certainty than the columns as the canopy could have been supported by a wooden framework.

7.1.2.1.4 Formal Entranceway and Triumphal Arch At the end of the driveway the remote sensing surveys identified a formal entranceway enclosed by two rectangular buildings to the right and left. The building footprints are clearly in the survey data and they formed the ‘frontage of the entire building complex’ (Wolfgang Neubauer et al., 2014, p. 184). Between the two buildings, large foundations measuring 4.4m x 3.8m wide have been identified. The depth of the foundations is not specified. The researchers have interpreted this component as a structure ‘like’ a triumphal arch (W. Neubauer et al., 2014, p. 184). The size of the foundations indicate that they were designed to carry significant load such as a triumphal arch. However, without detailed knowledge of the depth of the foundations and the type of materials employed it is difficult to ascertain their true purpose. It is highly likely that an arch was present as the main entrance to what was clearly a large and well-financed compound. In the absence of additional evidence such as the arch remnants a reasonable degree of uncertainty surrounds the actual physical appearance of the entranceway. It is assumed the researchers have drawn on the numerous examples of triumphal arches throughout the Roman Empire as a guide to the final reconstruction. The arch has been assigned a medium level of uncertainty as it is highly likely an arch did at one time exist despite the lack of physical evidence to support this assertion.

7.1.2.1.5 Administrative Block and Owners Living Quarters The building complex enclosing the main entrance identified in the archaeological surveys have been interpreted by the researchers as the Administrative block to the North and the Owners’ living quarters (lanista) to the South (Wolfgang Neubauer et al., 2014). The researchers draw on scholarly articles as the basis of their

158 7 Results - Application of the visual classification system to external case studies

interpretation. It makes sense that both the administrative building and owners’ quarters would be near the main entrance. The exact appearance of the buildings in terms of window placement, render, colour etc cannot be determined without physical evidence. The ground level components of both structures have been derived from primarily documentary evidence and represent a medium level of uncertainty. The owners’ living quarters have been interpreted to be a multi-level building. The researchers arrive at this conclusion due to the presence of what has been interpreted from the survey data to be stairway corridors and stairway foundations. It is quite possible the building was a multi-storied structure. The absence of physical ground- truthing, lack of extant remains beyond foundations adds an additional layer of uncertainty. The researchers have inferred the second story from the presence of building components that have not been verified. The authors draw on scholarly articles related to Gladiatorial practise. The upper level of the owner accommodation has been assigned the highest class of uncertainty based on extant inputs. The paving of the portico to the owner’s accommodation has been identified in the survey data although the exact material and physical appearance has not been determined. The paved area has been assigned a medium uncertainty class of extant data input. The colonnades in contrast, are a highly uncertain input. No evidence has been proposed to suggest the presence of columns. It is assumed the columns have been drawn from analogous off-site example and have therefore been assigned the high uncertainty analogous class.

7.1.2.1.6 North-East Main Building and Bath Complex The northern wing of the compound consists of three individual parts connected by a 30m x 3.9m corridor (Wolfgang Neubauer et al., 2014). The north-east main building forms the most prominent part of the northern wing. This protruding structure has been interpreted as a heated hall. This interpretation has been arrived at due to the presence of a collapsed pavement structure identifiable in the GPR data. The authors conclude that the degree of collapse ‘is only imaginable by assuming a corresponding hollow space underneath the pavement’ and that from their ‘experience, this clearly indicates the presence of a hypocaust system for under-floor heating’ (Wolfgang Neubauer et al., 2014, p. 185). This interpretation is supported by the presence of what has been interpreted as a large furnace located close to the south-east corner of the structure. Significant extant remains exist although they have not been excavated. The

7 Results - Application of the visual classification system to external case studies 159

final reconstruction of this area of the compound has been assigned a medium level of certainty based on the extant data used.

West of the furnace room are located ‘six paved rooms (100m2 in area) arranged in a row’ (Wolfgang Neubauer et al., 2014, p. 185). The footprint of the rooms and the paving is evident in the GPR survey data. Although the foundations have not been excavated the extant remains are certain. What is not known is the morphology or appearance of the paving. In addition to the foundations and paving, evidence indicating the presence of a pool and supporting pillars for a heated floor have been identified. The amount of evidence identified in the survey data lends weight to the interpretation proposed by the researchers. Evidence of a sophisticated sewer and heating system running beneath this section of the compound have been identified. This data coupled with evidence of water basins and a direct supply of fresh water have led the researchers to interpret this part of the compound as a bath complex.

On the north-western corner of the compound the survey a 200m2 building was identified. The building is separated into two large spaces by internal walls with evidence of some paving on the north-eastern edge (Wolfgang Neubauer et al., 2014). The researchers conclude that this area represented a storage facility as these rooms were accessible without entering the main training arena. A significant amount of extant evidence exists to determine the layout of this part of the compound. Elements such as paving, and tiles, can be identified through the magnetic and GPR surveys. However, the actual physical appearance of these elements cannot be determined. Similarly, the actual appearance of the external walls rendered in the report cannot be determined nor can the position of windows etc be determined. The bath Complex and storage area have been assigned a medium level of certainty based on the volume of extant data and its influence on the final reconstruction

7.1.2.1.7 Memorials A row of rectangular foundations spaced four metres apart in front of the barracks were identified in the GPR survey data (Wolfgang Neubauer et al., 2014). The researchers have interpreted these features as foundation blocks of memorials. No evidence is provided for this interpretation. It is assumed that the position of the memorials in relation to the barracks and the documentary evidence of such practices in Gladiatorial life are the prime reasons for the interpretation. The interpretation of these elements is determined to be highly uncertain based on the extant inputs

160 7 Results - Application of the visual classification system to external case studies

available. The actual appearance of the structures that resided on the foundations cannot be definitively ascertained. The foundations of the ‘memorials’ are outlined as low uncertainty extant inputs. The inferred memorials have been assigned a high degree of uncertainty.

7.1.2.1.8 Gladiator Barracks, Ground Level and First Floor The foundations identified on the southern side of the compound have been interpreted by the researchers as the Gladiator barracks (Wolfgang Neubauer et al., 2014). The foundations indicate the building included a portico facing the training yard. The interpretation of the foundations as barracks is supported by the presence of ‘small cells 3–7m2 in area arranged within an elongated cell block (total size 130m2)‘ (Wolfgang Neubauer et al., 2014, p. 185). A corridor is identified linking the cells from the north and south. The researchers note that the same design layout was also found at the ludus magnus, a gladiatorial school in Rome. Magnetic survey data identified features consistent with small tiles covering the floors of the cells, further supporting the interpretation as barrack cells (Wolfgang Neubauer et al., 2014). Features identified as foundations for staircases and stairway corridors suggest that the barracks consisted of two stories. It is highly probable that the barracks consisted of two stories although this cannot be definitively ascertained without further evidence. The barracks have been outlined in light green (low uncertainty extant input) as it is highly certain that the foundations exist. The portico (medium Green) has been assigned a medium level of uncertainty based on the extant evidence. As with other aspects of this reconstruction, the physical appearance of the portico paving cannot be determined although it is known that paving exists. The portico columns in contrast have been assigned a high degree of uncertainty as they appear to be based on analogous examples elsewhere in the Roman world. The second story of the building has been assigned a high degree of uncertainty (dark green) based on extant inputs.

7.1.2.1.9 Gladiator Preparation Hall The north-western side of the compound is bound by a hall interpreted as a preparation hall for the Gladiators (Wolfgang Neubauer et al., 2014). The building has a single entrance into the compound courtyard. The entrance to the hall is comprised of a 1.9m wide entrance indicating a ‘massive door frame ‘ (Wolfgang Neubauer et al., 2014, p. 185). The researchers consider the size and position of the entrance important evidence in their interpretation of the space. It is quite probable that the

7 Results - Application of the visual classification system to external case studies 161

building was used as a preparation area, however the exact form and final appearance of the building cannot be definitively determined. The interpretation arrived at by the researchers is primarily based on the extant evidence available and analogous structures. The foundations can be determined to be of high certainty. The building itself is an interpretation of the extant inputs and has been classed as medium certainty extant.

7.1.2.1.10 Main Gladiatorial Training Arena The Gladiator training arena, as identified in the GPR survey, is the main feature within the courtyard of the compound (Figure 7.1) It is a 19-metre diameter free- standing circular structure. The structure features a raised ring of stone. The researchers postulate that the arena ‘was surrounded by wooden spectator stands set on stone foundations’ (Wolfgang Neubauer et al., 2014, p. 184). The stone foundations of the arena are clearly identifiable in the GPR survey data. In addition to the foundations the GPR survey data revealed the presence of stone feature in the centre of the arena interpreted to be a posthole (Wolfgang Neubauer et al., 2014). The authors have drawn on scholarly articles to provide a strong argument that this feature is the foundation of a palus, ‘a wooden pole used for exercising blows with the sword and body slams with the shield’ (Wolfgang Neubauer et al., 2014, p. 184). In contrast, no evidence is presented for the wooden spectator stands although it is conceivable that they would have existed. It appears that the researchers draw on both documentary and physical evidence to determine that wooden seating existed. The researchers draw on numerous scholarly articles detailing Gladiatorial life and practise to describe the arena. In addition, the physical attributes of the extant remains are used to arrive at the conclusion that wooden stands existed. The foundations of the ‘spectator stands’ have a width of 2.4 metres and the researchers estimate that assuming ‘0.6m for each row of seats, up to four levels can be assumed’ (Wolfgang Neubauer et al., 2014, p. 185). The authors do not elaborate on this estimation or what sources were employed to do so. The presence of larger foundations in the north-east section of the arena are interpreted as stair access to the tiers (Wolfgang Neubauer et al., 2014). The presence of the arena foundations and what possibly is stair foundations are indisputable. The arena components can be accurately measured. However, in the absence of any physical evidence regarding the wooden stands it cannot be definitively stated that the stands existed in the form rendered by the authors if indeed at all. The foundations

162 7 Results - Application of the visual classification system to external case studies

have been classed as low uncertainty extant as has the inner ring and palus foundation. The base has been assigned a medium uncertainty classification and the wooden stands a high uncertainty. The style, size and layout of the rendered woodwork cannot be determined. The authors point out that the customary shape of a Roman amphitheatre is elliptical and not circular as evident with the Carnuntum site. However, examples of ‘circular buildings that were probably used for gladiatorial games, both in Italy and in the provinces’ (Wolfgang Neubauer et al., 2014, p. 185). This is a clear example where the authors have drawn on other documentary evidence in the form of scholarly studies to arrive at the interpretation.

7.1.2.1.11 Minor Training Area In the north-east section of the main courtyard another circular structure was identified in the GPR survey data. This area has been interpreted as additional training area devoid of spectator stands. This feature has been assigned an extant class of low uncertainty as it is clearly evident in the survey data.

7.2 TEMPLE OF APOLLO (TEMPLE C), KOS, GREECE

7.2.1 Background Temple ‘C’ is a Roman temple erected in the 3rd century BC and dedicated to the worship of Apollo. The temple is located within the Asklepieion archaeological site on the Greek island of Kos, in the Dodecanese group. The site lies on the eastern slopes of Mount Dikeous (De Mattia, 2012). The Asklepieion dates to the 4th century BC and was dedicated to the god of healing, Asclepius. The Asklepieion site was first excavated by the German archaeologists Rudolf Herzog and Paul Schazmann in 1902. Herzog and Schazmann undertook excavations from 1902 to 1904. Excavations were subsequently continued by Italian archaeologist Luciano Laurenzi of the Italian Archaeological Service in 1928 (De Mattia, 2012). It was during this phase of the excavations that a major study was devoted to Temple C.

Temple C, typified by Corinthian order peripterals1, was erected on what is known as the ‘second terrace’ (Figure A 19) and is one of the ‘best preserved monuments’ of the Asklepieion (Herzog, 1932, p. 42). The temple is oriented in a

1 A single row of columns or pillars along a side typical of classical Greek and Roman architecture.

7 Results - Application of the visual classification system to external case studies 163

north-east direction and occupies an area of 15.47 x 10.32 meters. The initial survey by Herzog noted that the main base (the Stylobate3) of the temple was largely intact (Figure A-23) although no other elements or fragments of the temple were ‘left in-situ’ (Herzog, 1932, p. 42). Fragments of pediment, columns and supporting beams uncovered on the peripheries of the site were assumed to belong to the temple by virtue of their proximity. These fragments were recorded by Herzog (Figure A-25).

7.2.1.1 Case Study Background Between 1938 and 1942 a partial anastylosis2 of Temple C (based on an architectural study by Herzog) was undertaken by the Italian architects Mario Paolini and Luigi Morricone (Figure A-20) (De Mattia, 2012). The anastylosis was intended to incorporate fragments of column and pediment present at the site of the temple (Figure A-20). In 2011 the researcher Daniela De Mattia undertook a review of the 1938 anastylosis2 of Temple C. The review examined the evidence recorded by both Herzog and Paolini. This evidence included archaeological reports, photographic evidence and in-situ research. In addition to the review a ‘complete catalogue of architectural fragments belonging to the Roman temple of the Asklepieion’ was compiled (De Mattia, 2012, p. 61). The review of the anastylosis in conjunction with the catalogue of fragments and the Herzog report were used to test the anastylosis proposed by Paolini and Morricone. De Mattia digitally reconstructed the proposed Paolini anastylosis of Temple C. The resulting reconstruction was reviewed.

De Mattia identified inaccuracies with the proposed Paolini reconstruction (Figure A-21). In particular, ‘lack of precision in defining the blocks to be reused and the poor correctness of their repositioning’ was identified (De Mattia, 2012, p. 73). Examination of photography taken at the time of Herzog excavations (Figure A-22) document the position of architrave-frieze and additional extant fragments in-situ. It is likely that position of the photographed fragments remained ‘unchanged from the moment of collapse’ of the building (De Mattia, 2012, p. 73). Comparison of the original position of the blocks and the Paolini proposed reconstruction led the researcher to conclude that they were not considered correctly in the proposed anastylosis2. This is exemplified in the use of extant fragments uncovered in the Southwest corner of the temple that were used for the reconstruction of the front

2 Reconstruction of extant remains aimed at achieving the original architectural structure.

164 7 Results - Application of the visual classification system to external case studies

northwest side of the building. Additionally, Paolini used a stylobate3 uncovered on the south-west corner of the building when examination of excavation photography showed the complete absence of the stylobate on the northwest of the building (Figure A-23). The Northwest stylobate3 was not uncovered until after the proposed Paolini reconstruction (De Mattia, 2012).

The actual anastylosis2 of the building undertaken by Paolini and Morricone did not realise the complete restoration originally envisaged. The north-east corner of the building was a compromise of fragments located during excavations whilst the columns where constructed of a mixture of concrete and extant fragments (Figure A- 20). The Corinthian style capitals are reproductions based the reconstruction drawings of Herzog (Figure A-24). There no references to surviving capitals of Temple C in either the Herzog or Paolini excavations. De Mattia proposed a revised reconstruction of sections of Temple C based on the new catalogue of fragments, the positions of the original extant fragments and, the extensive archaeological survey reports of Herzog and Paolini.

The digital reconstruction undertaken by De Mattia is largely based on the original position of extant fragments as identified by Herzog between 1902 and 1905. The premise that the position of the fragments was largely unchanged since the time of collapse of the building is significant given that large volumes of stonework and masonry from the site in general, have been removed or destroyed. In short, it is not certain that the fragments recorded by Herzog have not been moved from the time of the building collapse.

The position of the extant fragments as recorded by Herzog were used by De Mattia to extrapolate their original positions (pre-collapse) in 3D space. The results are depicted in Figure 7.3.

3 A continuous base supporting a row of columns typical of classical Greek and Roman architecture.

7 Results - Application of the visual classification system to external case studies 165

Figure 7.3: Proposed position of extant architectural elements of Temple C as interpreted by De Mattia 2009. Source: De Mattia, 2009, P. 79

The extant fragments can be separated into four components, Pediment, Entablature4, Columns and Stylobate. The pediment components comprise the south- east corner of the temple and the entablature fragments the north-east corner. Fragments of column were located in both the south-east and north east corners. One column fragment (south-east) exhibited fluting whilst the north-east column exhibited both fluting in the upper section and smooth surface on the lower (Figure A 24, No.18). This has been interpreted as all the original columns were characterised by a smooth lower section with fluting above (Figure A-21). The position of the column plinths is clearly visible on the extant stylobate and therefore the number of columns per side was known. With the components in place the researchers modelled the missing components to partially complete the reconstruction of the corners (Figure 7.4). The positions of the columns were used to determine the correct distance between the extant fragments.

4 Upper part of the temple resting on columns and supporting the pediment. Comprised of architrave, cornice and frieze.

166 7 Results - Application of the visual classification system to external case studies

Figure 7.4: Partial reconstruction and fill in pieces (dark grey) of the south-east and north- east corners of Temple C as proposed by De Mattia. Source: De Mattia, 2009, P. 79

The final step in the reconstruction of the corners of the temple combined the pediment fragments (including fill-in pieces) of the south-east corner with the entablature fragments of the north-east corner. The resulting reconstruction (Figure 7.5) was considered indicative of the pediment and entablature components of the entire temple. The outer columns of the temple were also included in the reconstruction. It should be noted that the precise height of the upper architectural elements cannot be definitively determined as only fragments of columns remain.

7 Results - Application of the visual classification system to external case studies 167

Figure 7.5: Reconstruction of the north-east corner of Temple C showing position of columns, entablature and pediment components. Source: De Mattia, 2009, P. 77

The reconstruction of the entablature and pediment components allowed the testing of the remaining reconstruction within the constraints of the temple footprints. Based on the reconstructed north-east corner of Temple C and the extant stylobate, De Mattia proposed a final reconstruction (Figure 7.6).

Figure 7.6: Final reconstruction of Temple C as proposed by De Mattia 2009. Source: De Mattia 2009, p. 74

168 7 Results - Application of the visual classification system to external case studies

7.2.2 Application of the Classification The decision support tree was employed to classify the individual components of the Temple C reconstruction. The data inputs, as described by De Mattia, were applied to the Decision support tree and the class outputs detailed in Table A-6. Five components were classified, these are:

• Stylobate

• Columns

• Entablature

• Pediment

• Capitals

The inner chamber (cella) of the temple was not detailed in the current case study and is not included in the classification of data inputs. As with the School of Gladiators, only the final published renders of the reconstruction could be used for the visual classification (Figure 7.7)

7 Results - Application of the visual classification system to external case studies 169

Figure 7.7: Application of the classification to Temple C images as presented. Original image source: (D. De Mattia, 2009, p. 74, P. 79) 7.2.2.1.1 Stylobate The archaeological excavations by Herzog detail the extant stylobate in the south-east section of the temple. Subsequent excavations by Paolini indicate that the stylobate in the north-east corner was not present. This section of was replaced during the Paolini anastylosis. The Stylobate has been classified low-uncertainty extant (Figure 7.7 top) with the reconstructed section classified as low-uncertainty analogous (Figure 7.7 bottom). The reconstructed section has been classified analogous as it is based on on-site extant remains. The reconstructed section of stylobate maintains the same appearance and dimensions as the extant remains and it is highly likely to be indicative of the original missing section.

7.2.2.1.2 Columns

170 7 Results - Application of the visual classification system to external case studies

Two sections of column and a column plinth were identified in the archaeological excavation by Herzog (Figure A-25 No. 14 & 15). Neither fragment represented a complete column. One fragment of column (number 16) appears to be from the top section of a column where the capital rested whilst the other fragment (number 14) appears to be from the lower section of the column. The lower fragment of column is characterised by a smooth lower section abutted by fluting. This indicates that the columns had a smooth base (of unknown length) with fluting for the upper section (Figure A-21). The overall height of the columns is not known. The extant columns have been classified as low-uncertainty. The reconstructed columns have been classified as medium-uncertainty analogous as they are based on the on-site extant remains although uncertainty remains with regards their actual height. The plinths of the reconstructed columns have been classified low-uncertainty analogous as they are based on on-site extant remains. It is unlikely that the plinths would differ across the extent of the peripterals and it is logical to assume the surviving plinth is indicative of the type and style used in the reconstruction.

7.2.2.1.3 Entablature Several extant fragments of the temple entablature were uncovered in the archaeology. These extant fragments (Figure 7.3) provide detail of each component of the entablature. Additionally, the fragments of entablature are collectively from two sides of the temple providing a reliable indication of the thickness and associated decor. The interpretation of the extant fragments allows a reliable reconstruction of the entablature. This component of the reconstruction has been assigned a low level of uncertainty as it is an inference from analogous on-site extant remains.

7.2.2.1.4 Pediment Four significant extant fragments of pediment have been recorded. These fragments have been interpreted as originating from two abutting sides of the pediment (Figure 7.3). The extant fragments when repositioned into 3D space (Figure 7.3) provide indicative detail of the height of the pediment at its apex. Only one fragment exists from the tympanum5 which is not enough to determine if it was decorated or plain as indicated by De Mattia (Figure 7.6). The Lintel and arch of the pediment have

5 A triangular or semi-circular or wall surface bounded by a lintel or arch. The wall surface is usually decorative or adorned with friezes.

7 Results - Application of the visual classification system to external case studies 171

been classified as low-uncertainty analogous as they are based on on-site extant remains. The tympanum is less certain as only one fragment has been identified and this is located on the edge of this component. The tympanum has been classified as medium-uncertainty analogous as it is based on limited on-site extant remains and a degree of uncertainty exists as to the facia décor.

7.2.2.1.5 Capitals No extant remains exist of the original column capitals. Herzog assessed and classified the temple as belonging to the ‘genus of the Corinthian Peripteral temple’. This assessment and classification was based on the ‘art forms’ of the extant fragments although no further description is provided (Herzog, 1932, p. 42). Herzog does draw on analogous off-site examples of Etruscan ‘models’ with similar widely distributed columns (Herzog, 1932, p. 42). Herzog states that ‘even in Asia Minor there are temples from the Antonian period, which could be used for comparison, eg. in Pamphylia and Pisidia.’(Herzog, 1932, p. 42). Although Herzog draws on analogous examples, the complete lack of capitals from the site leads to a high degree of uncertainty with regards their actual appearance. Neither Herzog nor Paolini provide any evidence or reasoning for the decorative style of the capitals. Scholars such as Jones (1989) have shown that many differing capital designs within Corinthian order existed and that comparable Corinthian capitals from different buildings ‘often had differing proportions’ (Jones, 1989, p. 35). The reconstruction of the capitals by De Mattia has clearly adopted the Herzog, and subsequent Paolini interpretations. The capitals have been assigned to the high-uncertainty as they have been based on documentary evidence pertaining to the Corinthian order.

7.3 DISCUSSION The application of the colour classification system to the two case studies has shown that it is flexible and can be applied to sub-elements of a reconstructed component. This is exemplified in the Gladiator school of Carnuntum Gladiatorial arena which was broken down into components based on extant and documentary inputs. However, the application of the classification system to the case studies has highlighted deficiencies and need for additional refinement. This is discussed in detail in the concluding discussion. The classification system has been designed to be an integral part of the reconstruction process whereby input data is classified and ordered (using the Decision Support Tree) prior to the reconstruction. Applying the

172 7 Results - Application of the visual classification system to external case studies

classification to a reconstruction that is outside of this process has uncovered limitations relating to the assessment of the data inputs and, in particular, inference in a reconstruction project.

The primary issues identified with the classification system when applied to external reconstruction projects can be summarised as follows.

I. It is possible a clear ‘Inference’ class may be needed II. A more robust method of assigning classes is required III. The use of colour and display mediums requires refinement It is clear from the case studies that a high degree of inference is usually required in a reconstruction and this is directly proportional to the amount low-uncertainty input data available. Currently the decision support tree does not cater for an ‘inference’ class. Inference can be a direct result of a variety of data input types. These data input types range in their levels of uncertainty and origin. More work is required in the development of the decision support tree to accommodate inference more clearly. A detailed analysis and review are undertaken in concluding discussion.

The application of the classification system has highlighted the need for a more robust method of assigning data classes whilst using the data decision tree. Currently input data is assigned to final classes based on a fuzzy approach. That is, data is assigned to a class by the user based on an assessment of the quality of the data. This data assigning process lacks any quantitative assessment. It may be of use to retain an aspect of the numerical calculation system originally proposed to assign the final colour classes. This may prove to be a more objective method and assigning final classes and is discussed in the concluding discussions.

It is clear from the enormous body of research surrounding visual classification systems that colour plays an important role in the communication of information. In this study a simplistic colour scheme was used to depict three categories of input data used in an archaeological reconstruction. The method of display is effectively a colour overlay on the underlying reconstruction. It is questionable if both these approaches are appropriate in terms of effectively communicating the required information regarding the origins of the reconstruction being viewed. The primary purpose of the classification is to effectively inform the viewer of the veracity of the reconstruction being viewed. The use of colour schemes and methods of visual delivery are discussed in the conclusion.

7 Results - Application of the visual classification system to external case studies 173

[Extra page inserted to ensure correct even-page footer for this section

174 7 Results - Application of the visual classification system to external case studies

8 Discussion and Conclusions

The purpose of this study has been to investigate and develop a method of visually conveying the authenticity of an archaeological reconstruction. Essentially the question proposed was, ‘Can a methodology be developed that facilitates the observer of a digital reconstruction in making assessments of the authenticity or veracity of the viewed reconstruction?’ To answer this question the approach was to firstly develop a robust framework for the digital reconstruction process. This took the form of the workflow pipeline. Secondly, a classification system was developed to classify the data inputs used in a digital reconstruction as eventual authenticity indicators of the final reconstruction. The data input classes should be in a visual form that a user can interpret to arrive at their own assessment of the authenticity of a digital reconstruction. It is not the purpose of this study to question the work of the archaeologists or the reconstructions in the case studies analysed but rather to interrogate and assess the inputs used to arrive at the reconstruction.

8.1 THE WORKFLOW PIPELINE 8.1.1 Workflow Pipeline Summary A workflow pipeline was presented that aimed to standardise the process of digital reconstruction of archaeological structures. The focus of this study was in the context of lost buildings. The workflow pipeline was applied to two test case digital reconstructions, the 1596 Rose Theatre in London and the 1856 Bergen Theatre in Bergen, Norway. The workflow pipeline expands on previous work by Hermon which investigated the modelling stages of a 3D archaeological reconstruction. Hermon identified the data inputs at each stage with the aim of ‘turning data into information and into knowledge’ (Sorin. Hermon, 2008, p. 36). The presented workflow pipeline draws on the same principals as Hermon in that a critical part of this process is the identification of data inputs from the outset. These inputs can be used not only in the reconstruction, but also in the assessment of the final outputs. Hermon proposed a somewhat linear approach to the digital reconstruction where data inputs are ‘fed’ into a single linear pipeline arriving at a final reconstruction. In contrast, the presented workflow pipeline draws on agile software development methodologies which encourages iteration in discrete phases (Coleman, 2016). The three-phase iterative

8 Discussion and Conclusions 175

approach facilitated the identification of issues within the reconstruction that resulted from uncertain data inputs and provided a setting for addressing issues through experimentation. The use of a Decision Tree to identify and classify data inputs into parent classes of type and sub-classes of certainty facilitated the iterative approach of the workflow pipeline. All data inputs could be considered and used in the reconstruction process but, most importantly, their influence could be identified in the final output.

8.1.2 Workflow Pipeline Discussion

The workflow pipeline is characterised by two major phases. The first phase of the reconstruction process involves data sorting and classification with the use of the highest certainty data to begin the reconstruction. High certainty data such as extant remains, archaeological reports, architectural plans etc. are utilised in this stage whilst lower certainty data is ‘transferred’ to the second phase for incorporation. The second phase incorporates the outputs of phase one and the lower certainty data to arrive at a final reconstruction. Phase 2 is highly iterative.

Application of the workflow pipeline to the Rose and Bergen theatres showed that it provides a structured approach to the digital reconstruction process. Using the workflow in the same manner as a structure would be erected is very effective as it can be applied to individual building components. This allows for the high certainty data inputs such as extant remains and architectural plans to be used to reconstruct elements of the structure with a high degree of confidence. These elements that represent high certainty are ‘built’ upon in later stages of the reconstruction. In the case of the Rose Theatre the presence of extant remains and an accurate archaeological survey allowed the reconstruction of the building foundations and establishment of spatial layout to a high degree of certainty. The availability of official survey drawings for the Bergen Theatre allowed the reconstruction of the building framework and mass-structure with a high degree of confidence. In both cases the outputs of Phase One of the workflow pipeline were crucial inputs for the second phase. Phase One of the pipeline is a very linear process. The high certainty data inputs are used to create components of the structure of which there is little or no uncertainty. The decision-making process is driven purely by the hard data with little to no inference. These inputs are significant drivers for the highly iterative Phase Two. Testing in both the Rose and Bergen

176 8 Discussion and Conclusions

Theatres led to minor revisions of Phase one to more fully reflect the data sorting in this stage.

Both test cases highlighted that each element or component of a building can also be based on data inputs of variable certainty requiring an iterative approach to the reconstruction. Iteration in Phase Two allows experimentation. Higher uncertainty inputs can be tested against the low uncertainty outputs of Phase One. In short, testing can take place against the known elements of the building or structure. The combination of the two phases of the pipeline allows the user to arrive at logical solutions based on the organisation and comparison of available data inputs. After testing, the workflow pipeline has been demonstrated to be flexible in terms of the data inputs used. In the case of the Rose Theatre the primary low uncertainty data inputs consisted of the archaeology of the site and subsequent survey. The Bergen Theatre in contrast, was destroyed with no archaeological record remaining. The available 1938 official survey record became, in effect, the archaeological survey. In both cases the high certainty data inputs provided a foundation on which the remainder of the building elements could be reconstructed using the additional higher uncertainty data inputs. This demonstrated that the workflow pipeline was flexible in accommodating multiple data types whilst retaining a structure and clear direction in terms of output.

The workflow pipeline is presented as a structured approach to the process of reconstructing a lost structure. It allows the ordering and classification of data types into levels of certainty. It utilises high certainty data to build upon and, accommodates both linear and iterative approaches to reconstruction. The iterative nature of Phase two allows for the testing of areas of limited knowledge and considers all aspects of the structure under study. The workflow pipeline has been shown to be adaptable to the type and range of data available for a digital reconstruction and data inputs can be clearly identified at any stage of the pipeline. The two case studies did not highlight any significant issues with the workflow pipeline. It is acknowledged that pipeline was only applied to the two case studies and that application to a range of other case studies may highlight the need for refinement or adaptation. Applying the agile software development approach of iteration in discrete phases to the reconstruction process allows for rapid testing of more uncertain inputs. Both test cases demonstrated that this approach facilitated the rapid and effective testing of decisions related to data inputs.

8 Discussion and Conclusions 177

8.2 THE CLASSIFICATION SYSTEM 8.2.1 Classification System Summary A data classification system was presented in this study. The principal purpose of the classification system was to facilitate assessment of the authenticity of a digitally reconstructed archaeological structure. The foundation of the classification system draws on fuzzy sets and allows the assigning of data inputs used in a digital reconstruction into parent classes of origin and sub-classes of certainty. The premise of the classification system is that the level of certainty of the data inputs is directly related to the authenticity, or conversely, veracity of the reconstruction. That is, an archaeological reconstruction which is primarily reliant on high certainty data inputs (such as extensive archaeological remains), will likely have a higher authenticity than one that has primarily used high uncertainty data such as anecdotal evidence. The classification system presented was applied to the Rose Theatre reconstruction and underwent modification. Initially the classification system was presented as a numerical system with the intention to create a quantitative measure of authenticity. Numerous deficiencies were identified, and the system was modified to a qualitative assessment of authenticity based on a visual output. The classification system is discussed below.

8.2.2 Findings of the Classification System The classification system, of which the Decision Tree is the core, was primarily developed to work in conjunction with the Workflow pipeline. This allows for the all data inputs to be classified at the beginning of a reconstruction and their influence on authenticity to be tracked and identified. The Decision Tree facilitated the organisation of data inputs and informed the reconstruction. Initially the numerical classification system was applied to the Rose Theatre reconstruction. The results of the application raised significant questions with regards the rigour of the approach and the validity of the resulting authenticity assessment. The Decision Tree was modified to address the questions raised in the Rose reconstruction. A simpler, more effective decision tree was developed and applied to the Rose and Bergen theatre reconstructions. In both test cases the classification system was an integral part of the reconstruction from the beginning. The Decision tree was used to classify the data inputs and reconstruction decisions were based on the assigned classes. The results of the reconstruction test cases were used to finalise the Decision Tree structure. The finalised Decision tree was then applied to an additional two test cases where the reconstruction had been

178 8 Discussion and Conclusions

undertaken by third parties. This exercise was undertaken to test if the classification system could be applied independently of the workflow pipeline.

8.2.2.1 Numerical Classification Initially a decision tree based on a numerical structure was proposed (Figure A 26). The decision tree used to separate input data in parent classes of physical evidence and non-physical evidence. Each parent class was then separated into three subclasses and each subclass into three ‘fuzzy’ uncertainty classes with numeric values. The uncertainty classes were:

• Low Uncertainty – (Numeric value – 1)

• Medium Uncertainty - (Numeric value – 3)

• High Uncertainty - (Numeric value – 5)

The uncertainty class values were based on classification criteria detailed in a numerical tabular matrix (Table A-8). The matrix was used to arrive at an average uncertainty score for each data input used in the reconstruction. The classification system was applied to the Rose Theatre as an initial test case.

The application of the decision tree to the Rose Theatre reconstruction demonstrated that it provides a robust method of organising data inputs and is valuable in informing the reconstruction process. The rigour of the approach and resulting outputs, was questionable. The calculation of an average numerical score was an attempt to quantify the various data inputs. This was demonstrated to be overly simplistic. The use of an average numerical score for each data input to calculate overall authenticity assumed that all inputs would have the same weight in terms of impact on the reconstruction when this clearly was not the case. Small elements of the reconstruction had the same impact on the final authenticity score as major elements. In the case of the Rose Theatre the combined impact of individual minor components on the overall reconstruction authenticity assessment was disproportionate. The final calculation of a single numerical ‘score’ of authenticity for the reconstruction based on an average of all the data inputs did not reflect possible extremes that may be assigned to individual reconstruction elements and provided no way of showing the effects of individual components in the final reconstruction.

A further aspect of the numerical approach which was of concern was the assigning of data inputs to classes of uncertainty. Although the decision tree made the

8 Discussion and Conclusions 179

process of separating data into parent and sub-classes simple and straight forward, the final assigning of a value of uncertainty was effectively subjective. It became clear that although the aim of the numerical decision tree was to quantify the level of uncertainty for each data input, the varied nature of the data inputs themselves prevented a purely quantitative assessment. The process of choosing an uncertainty class value based on a series of class descriptions was, by nature, subjective. These subjective decisions impact the final computed value of authenticity which was intended to be a quantitative indicator of authenticity. An additional significant outcome of the test case was length of time required to complete the classification procedure. The procedure was lengthy and onerous. Each individual component that is considered in the reconstruction must be sorted into parent classes and then sub- classes of type. The sub-classes were then classified into levels of certainty using the tabular matrix. In the case of the Rose Theatre there were 15 major building components and the related data inputs considered. It became clear that the more data inputs utilised, and the more components identified, the more cumbersome the numerical classification system would rapidly become.

The resulting single numerical value as an indicator of reconstruction authenticity is questionable for the reasons detailed above. Additionally, this single score or value, does not afford the viewer any opportunity to evaluate the evidence used to achieve the reconstruction. A numerical score gives no indication of which elements of the reconstruction are more certain and which are not. A greater danger lies in the fact that the use of a numerical score is a definitive statement of authenticity of a reconstruction. As has been demonstrated, a few minor components can adversely affect the score of what is otherwise an authentic reconstruction. This can possibly lead the viewer to believe that the reconstruction is of low authenticity when in fact the opposite is true.

8.2.2.2 Visual classification Application of the initial decision support tree and classification system necessitated modifications to address the issues discussed above. A refined decision tree was developed (Figure A-27). The decision tree was simplified by reducing the subclass types from four to three. The subclasses ‘Construction Techniques’ and ‘Analogous Structures’ were merged into a single sub-class. The ‘Construction Techniques’ class proved to be superfluous as any analogous structure would

180 8 Discussion and Conclusions

incorporate the construction techniques of the time. Initial parent classes were renamed ‘Related Evidence’ and “Associated Evidence’ to more clearly differentiate between what is essentially quantitative and qualitative data inputs.

The modification of the decision tree facilitated a much clearer distinction and simpler method of rapidly assigning input data to a class of uncertainty. Additionally, this modification affords a rapid and effective method of distinguishing and sorting data based on evidence type. Data inputs have been categorised into three basic types, Archaeological evidence, Documentary evidence, and Analogous structures. The initial system of numerical classes was replaced with colour elements as classes of uncertainty. This modification fulfilled the role of the decision tree as a data sorting and classification tool. In the previous iteration the classification resulted in a sum-of- the-parts measure of authenticity without regard for how major or minor a reconstructed component was. The modified decision tree allowed the individual components of a reconstruction to be classified based on their origin and uncertainty. This provides more transparency for the entire reconstruction by affording a visual assessment of the individual reconstructed components. The addition of colour facilitated a method of visualising reconstructed elements based on their inputs. The modified classification system was applied to the both reconstruction case studies, the Rose and Bergen Theatres. Additionally, the modified classification system was applied to two case studies where the reconstruction had been performed by third parties. This was done to test the application of the classification system on a reconstruction that was carried out independently of the proposed workflow pipeline.

8.2.2.2.1 Rose and Bergen Theatres Application of the modified classification system in conjunction with the workflow pipeline to the Rose and Bergen theatres demonstrated that the process of sorting data was more efficient. The reduction in the sub-classes from four to three greatly simplified the data classification stage. It was found that across both case studies the three sub-classes were adequate to cater for the variety of data input types. The assigning of data inputs to classes of uncertainty based on colour obviated the need for a decision matrix and acknowledges that subjective decision making in an archaeological reconstruction is effectively, unavoidable. The removal of the quantitative approach significantly improved the time taken to sort and classify the

8 Discussion and Conclusions 181

data inputs. The most significant outcome of the modified classification system was the ability to visually display the classification outputs in a meaningful manner.

Individual building components can be displayed based on the data inputs used to construct them. This also resulted in a fundamental shift in how this study approached the process of authenticity assessment. Initially the aim was to be able to produce an authenticity score based on a quantitative assessment of the data inputs used. This approach provided little transparency of process and prevented the viewer of reconstruction from making an assessment. It effectively negated the reasons for this study by, once again, asking the viewer to rely on an authenticity assessment that provided no information other than a score and no information as to how the assessment was arrived at. The use of a single colour to represent the sub-classes of information type aids interpretation, as it is possible to quickly distinguish the data origins of a reconstructed component. The use of shades of a class colour to differentiate between levels of certainty also appears to have been effective. Light shading represents low uncertainty whilst dark shading represents high uncertainty. The existence of only three levels of uncertainty per class appears to facilitate easy differentiation between those classes.

Application to the Rose and Bergen reconstruction case studies demonstrated that the classification system is clearly effective when used as the starting point of a reconstruction. Doing so makes the data sorting and classification process integral to the workflow pipeline. It allows each data input to be tracked through the reconstruction process and, more importantly, allows it to be visualised in the final reconstruction. As each component is created it can be assigned the class attributes derived from the decision tree. It is acknowledged that cases may arise where multiple data types are used to arrive at a final solution for an individual reconstruction component. This was evidenced by the Gladiator School of Carnuntum reconstruction. This was not the experience with either the Rose or Bergen theatres where all the data inputs were assigned to one of the three parent classes without difficulty.

Display of the final classified reconstruction in conjunction with the unclassified reconstruction appears to be effective. An interactive ‘proof of concept’ visualisation is presented at https://ortelia.com/Downloads/Rose_PHD/. Within the 3D modelling environment used to create the reconstruction it is possible to switch between the classified and unclassified reconstruction. A slider bar provides the ability to switch

182 8 Discussion and Conclusions

between the textured components and flat-shaded colour representing input data type and assigned certainty level. Users can interrogate the reconstructed stage in this proof- of-concept by clicking on the front of the stage. Clicking on the front of the stage invokes a pop-up interactive panel containing the evidence utilised to reconstruct this element. This interactivity appears to facilitate interpretation of the reconstruction and the assessment of a level of authenticity. It is acknowledged that no user interpretation testing has been undertaken to test the efficacy of the colour classification and method of display. This is discussed later.

8.2.2.2.2 Gladiator School and Temple of Apollo Application of the visual classification system to third party reconstructions demonstrated that the classification does have applicability which is afforded by the ability to identify and classify individual reconstruction components. However, the two test case studies (Gladiator School and Temple ’C’) highlighted significant issues, reinforcing the notion that the classification system is most effective when it is an integral part of the workflow pipeline and, by extension, the reconstruction. Two third party reconstructions were used in the application of the classification system. The archaeological reconstruction case studies were very different in terms of data inputs used. In the case of the Temple of Apollo the virtual reconstruction was based exclusively on traditional data sources such as obvious onsite extant remains and extensive archaeological surveys. The extant remains could be physically examined and measured. The Gladiator School of Carnuntum virtual reconstruction was, in contrast, derived exclusively from remotely sensed, analogous and documentary data. No extant remains were unearthed that could be physically examined and measured. The actual composition of the foundation building materials, for example, could not be determined. In contrast, the type and size of materials comprising the foundations of the Temple of Apollo are precisely known.

The Temple of Apollo reconstruction was clearly most suited to the application of the classification system. The data inputs used (extant remains and survey data) in the reconstruction were clearly detailed. Elements such as the temple foundations and fragments column and pediment were enough in extent to be able to reliably determine major components of the structure. The reconstruction also draws on documentary evidence of varying degrees of uncertainty where no extant evidence exists. The degree of inference in the reconstruction is low. The parent and sub-classes of the

8 Discussion and Conclusions 183

decision tree were sufficient to incorporate all the described data inputs used in the reconstruction. Although the reconstruction was not undertaken using the workflow pipeline proposed it did follow the central premise of beginning with known low- uncertainty inputs. The original extant remains were created on which the remaining reconstruction was based. The visual classification allowed each component of the resulting reconstruction to be visualised based on the data inputs used. A critical factor in the successful application of the classification system to the Temple of Apollo case study was the availability of detailed information regarding the original data inputs used in the reconstruction.

The Gladiator School reconstruction, in contrast, highlighted deficiencies in the decision tree. Unlike the previous test cases, which had in common high certainty extant inputs, the Gladiator school ‘extant’ inputs were derived from inference. The remotely sensed data provides a definite footprint of the complex. Within the main site footprint individual building footprints and the like can be identified. The initial issue identified was in where to assign the data inputs. There is little doubt that the foundations of the complex exist, however, they are not extant in terms of being exposed for viewing and measurement. This data is highly certain in terms of the accuracy of the remote sensing and mapped layout, however the question of which type class to assign these inputs was raised. The data is, strictly speaking, documentary but there is little doubt the remains exist. The remotely sensed data detailing the foundations of the complex was assigned as extant, low uncertainty although it is acknowledged it could have equally been assigned to low-uncertainty documentary. This highlighted that the classification system may either require additional classes or a set of additional criteria which clearly assign inputs to classes.

The described reconstruction of the School of Gladiators clearly shows that the authors drew on sources such as scholarly articles and analogous off-site evidence to arrive at a reconstruction solution for each component of the complex. In effect, a high degree of inference was undertaken based on the available evidence. In many cases multiple varying data inputs were used to arrive at a reconstruction for a single component. This was clearly illustrated with the arena which was a combination of ‘hard’ data (the footprint), analogous examples elsewhere in the Roman world, scholarly articles, and inference based on foundation sizes. Although individual inputs in this example can each be assigned a class and certainty level, the individual

184 8 Discussion and Conclusions

reconstructed component could not be visualised effectively to reflect these inputs. In the case of the arena, the solution was to decide which body of evidence (in this case documentary) contributed the most to the reconstruction. It is acknowledged that this is a subjective decision which is influenced by the available classes that data inputs can be assigned to.

The high degree of inference applied by the author in this reconstruction highlighted that the Decision Tree has limitations and may not be applicable in all cases. The classification of data inputs was based on a description presented in an article. Not all the input data used was completely described. A simple example is the final presentation of the walls and roofing of the compound. There are many examples of rendered walls and tiled roofs from the Roman world, however, the authors do not describe how the final presentation of these features were arrived at. This is further compounded by the fact that no extant evidence of these features exists on-site. The point here is that to apply the classification system effectively, prior knowledge of data input sources is essential. The final classification of the Gladiator school reconstruction itself was determined using inference. It is likely that the reconstructed Gladiator School has a high level of authenticity and there is a distinct danger that a classification that may not faithfully represent the veracity of the final reconstruction through the presentation of inaccurate classes. The Gladiator school presented the greatest challenges in terms of assigning inputs to classes. This reflects the limitations in the decision tree and not of the data inputs used by the authors to reconstruct the school. These limitations suggest that the decision tree may not be suitable in all contexts. Future work in improving the decision tree to work in such contexts may be fruitful.

8.2.3 Discussion of the Methodology Application of the workflow pipeline and classification as a complete reconstruction solution has been shown to be effective. The application of the system to the test cases demonstrated that when applied as a start-to-finish solution authenticity can be assessed as a measure of the data inputs. A quantitative measure of authenticity is difficult to achieve when the various inputs are both quantitative and qualitative in origin. A visual method of highlighting the types of inputs used in a reconstruction permits the viewer to form an assessment of authenticity. The proposed workflow pipeline and associated classification system has been shown to facilitate

8 Discussion and Conclusions 185

viewer assessment and overall reconstruction transparency. Information knowledge is essential to the successful application of the proposed system. Application of the classification system to a third-party reconstruction such as the Gladiator School of Carnuntum clearly demonstrated the deficiencies of the classification system with regards lack of input data information. The use of multiple varying data inputs to infer the appearance of a building component highlight that the classification system would require modification to address these types of reconstructions.

Modification of the decision tree in the initial iteration of this study highlighted its flexibility. The modified decision tree maintained the same structure and methodology after the modification continuing to be a data sorting a classification tool. The decision tree is a critical part of the workflow pipeline and greatly facilitates the reconstruction process. The decision tree can adapt to class changes as was demonstrated. This flexibility allows the decision tree to conceivably be adapted to any reconstruction project. It may be possible to replace classes of uncertainty with classes of inference to address the issues highlighted by the Gladiator School reconstruction and is a possible avenue of future research. The application of the classification system to the third-party reconstructions highlighted the need for a more robust method of assigning data inputs to classes of uncertainty. The initial numeric classification system attempted to address this issue using a tabular matrix of criteria. During this study numerous examples can be identified where data inputs could have been assigned to more than one class. With the current classification system this is addressed by making a subjective assessment as to the most appropriate class. A more robust method of assigning ‘problem’ inputs may be warranted to minimise this subjective influence. A possible solution is to incorporate aspects of the original numerical classification system, such as the matrix to quantitatively assign a data input to a colour class. It is acknowledged that this approach would still entail qualitative assessments of selected inputs, but a more formalised framework may provide a set of clearly defined criteria for assessing data. It is acknowledged that the addition of an additional step in the data sorting and classification may increase the complexity of the process and warrants further research.

From a display and interpretation standpoint the modified classification system appears to be effective although no user testing has been undertaken to assess its efficacy. The classification system, when used as part of the entire reconstruction

186 8 Discussion and Conclusions

process allows each component of a reconstruction to be shaded according to data type and class of certainty. This allows the viewer to assess the entire reconstruction as whole and identify the various data inputs used to arrive at the presented work. Effectively, this visual communication of classes empowers the viewer to form an assessment of authenticity. The colours used for the classes were arbitrarily chosen for each category. There has been an extensive body of work dedicated to the use of colour in visual classification systems and, as discussed, near universal agreement that colour plays an important role in the communication of information. It is quite possible that the colours chosen for this study are not the most appropriate in terms of representing the types of data inputs. Additionally, the shades of colour used to show levels of uncertainty may also not be the most effective means of communication. As has been detailed earlier in this thesis there are a variety of methods of uncertainty visualisation such as the use of animation and sonification that may be more appropriate. The scope of this study did not allow for user interpretation testing. This is an area that warrants further research in the context of this type of study.

In addition to the possible issues surrounding the best and most effective use of colour, is the method of delivery. In this study the two reconstruction case studies allowed the application of the colour classes directly onto the 3D modelled components. Conversely the two third-party reconstructions were limited to overlaying class colours on the available published images. In both cases the final method of delivery is a 2D depiction in the form of an image. In the context of a 3D reconstruction this is very limiting. Users are forced to compare images side-by-side to gain context and arrive at authenticity assessment. The viewer has only the available imagery at their disposal with no detailed information on the inputs used beyond data type and level of authenticity. It is suggested that the use of 3D interactive environments would greatly enhance the efficacy of the classification system and the ability of viewers to make informed assessments. An interactive virtual environment of a reconstruction would allow a user to freely navigate the reconstruction thereby forming a better understanding of the context of the structure. Additionally, interactive environments facilitate features such as the ability to embed meta data in the form of precise textual information, photography, video and the like. The ability to turn the visual classification on and off at will allows users to explore the classification more directly with reference the reconstruction. Combining these types of features to create

8 Discussion and Conclusions 187

an information rich visualisation would greatly enhance the viewer experience and understanding. This area warrants additional research.

8.3 CONTRIBUTIONS AND APPLICATION The past decade has seen significant advancement in computing hardware, and by extension, virtual reality and 3D modelling tools. These tools have become widely accessible to archaeologists and researchers. The increasing power and general availability of these tools has resulted in a proliferation of virtual reconstructions. Along with this proliferation has been a growing recognition of the dangers inherent in any reconstruction project utilising highly variable data inputs. Researchers such as Denard have recognised that there is a lack of account of how data input sources have been evaluated and interpreted leading to the inability of the viewer to effectively assess a reconstruction. (Denard, 2011)

The presented workflow pipeline and classification system attempts to directly address the concerns expressed by many researchers regarding lack of transparency in reconstructions. The proposed system has been inspired by the London charter for the Computer-based Visualisation of Cultural Heritage. The main aim of the classification system is to provide greater transparency with regards the data inputs used to arrive at a reconstruction. It directly addresses the lack of account of how data input sources have been evaluated and interpreted. The workflow pipeline is a highly iterative process which provides the ability to track inputs throughout the reconstruction process. This process validates the reconstruction and adds methodological rigour to the process.

The presented reconstruction methodology contributes to the domain of archaeological virtual reconstructions by providing a formalised and structured approach to digital reconstruction. The organisation of input data coupled with a clearly defined and repeatable process provides a methodological framework for the digital reconstruction process.

8.4 RECOMMENDATIONS The workflow pipeline, as a digital reconstruction methodology has been shown to be robust and repeatable. The organisational process of sorting and classifying data inputs is essential to the digital reconstruction process. The workflow pipeline has

188 8 Discussion and Conclusions

immediate application in archaeological reconstruction where multiple input data types are the norm. However, several areas would benefit from additional research. Further research and user testing are required to establish the efficacy of the colour classification. Colour interpretation in the context of a classified reconstruction has not been addressed in this study. It is possible that the selected colour scheme, and method of delivery, is inappropriate in the context of the information represented. The testing of a variety of colour schemes based on human colour cognition would be a valuable addition to the current classification system. Other methods of visualising uncertainty, such as those explored by Zuk et al. (2005), Dudek and Blaise (2004), and Bonde et al. (2009) may be equally effective and deserve attention. The use of animation techniques such as time and colour sliders may provide an effective delivery method whereby viewers can easily switch between the classified and final render views. Additionally, the early work conducted by Dudek and Blaise (2004) using graphical properties such as colour and translucency may have application in the context of the classification. The combination of animation and colour properties deserves further investigation.

A digital reconstruction of an archaeological structure will, by default, produce the core foundations for a digital environment. That is, the 3D reconstructed model provides the focus of the digital environment. Modern 3D and VR interactive environments offer a setting which allows users not only free movement and immersion, but also the ability to interact with the synthetic environment. Interactive environments offer the unique opportunity to embed and present metadata (data inputs) associated with a reconstruction. Synthetic environments such as an interactive archaeological reconstruction allow any part of the environment to become an information portal. That is, a building element can be interrogated through interaction with a mouse, wand or any other interaction device. Pop-up panels or interfaces that contain information relating to the element selected can be displayed. This information can be audio, text, video or a mix of all. The proof of concept visualisation (https://ortelia.com/Downloads/Rose_PHD) presented illustrates these techniques. The use of such techniques effectively transforms a digital reconstruction into an information rich environment. The testing of user experiences within such environments would be invaluable. Further research in this area would yield valuable knowledge in terms of how users access information, what cues they best respond to,

8 Discussion and Conclusions 189

and, how this information influences their assessment of authenticity. Additionally, research in this area would shed light on design of interactive, information rich, reconstruction environments.

The results of the classification system application to the Gladiator School reconstruction highlighted the need for more work in how to classify inference. Multiple inputs of varying degrees of certainty can be employed to arrive at the reconstruction of a single building element. Although this is common practise, visualising inference has proven to be difficult as assigning classes to data inputs is highly subjective. Modification of the decision tree to account for inference would be a valuable addition to this study and digital archaeological reconstruction in general.

8.5 CONCLUSION This study began with the intention of developing a methodological approach to digital archaeological reconstruction, and a measure of authenticity for the reconstruction. The aim was to classify the data inputs used in a digital archaeological reconstruction with the intention of using the classified values as a measure of authenticity. As has been extensively discussed, an archaeological reconstruction utilises data inputs of varying type and veracity. Data inputs are both qualitative and quantitative. Initial attempts to quantify all data inputs regardless of origin proved to be problematic. Application of the initial numerical classification system to quantify levels of data input uncertainty were overly simplistic. It was clear from the initial case studies that the nature of the data types used in a reconstruction precludes it from strict quantitative assessment. Certainly, some data inputs such as extant remains and detailed surveys may be quantified to some degree in terms of levels of certainty. However, even an archaeological survey will be influenced by an archaeologist’s interpretation as was shown with the Temple of Apollo case study. Overall most data used in an archaeological reconstruction is qualitative. This is not to say it is inaccurate or misleading, but it cannot be quantified in any strict form when assigning levels of certainty.

Subjective assessments will always be made when reconstructing an archaeological structure where, in general, information is limited and varied. Although this is effectively unavoidable, the presented workflow pipeline and classification system provides a formalised structure in which to make subjective decisions. The iterative nature of the workflow pipeline encourages testing and reflection.

190 8 Discussion and Conclusions

Assumptions based on uncertain data inputs can be tested against known high certainty components of the reconstruction. The workflow pipeline encourages exploration of possibilities within the bounds of the known hard data. Classification of data inputs prior to the application of the workflow pipeline ensures that the influence of any data used in a reconstruction can be identified.

Although this study began as an attempt to determine the authenticity of a reconstruction with a numerical value, the outcome was somewhat different. The initial goal was to present the viewer with a score that could be trusted as a measure of reconstruction authenticity. In doing so it became apparent that this approach lacked transparency and did not achieve the intended outcomes. The viewer was not reliably informed when presented with an authenticity score and was given no opportunity toi make an informed assessment. The visual classification system, in contrast, allows the viewer to make informed decision as to the authenticity of a reconstruction through transparency of data inputs. The workflow pipeline and classification system have been demonstrated to work effectively when used as a beginning-to-end solution. Deficiencies have been identified that merit further investigation. The presented system, although only a first step, contributes significantly to making digital archaeological reconstructions more transparent and accessible to viewers by allowing them to form their own assessment of authenticity or veracity. Provision of a structured framework for reconstructing archaeological structures brings methodological rigour to the reconstruction process and maximises the use of visualisation as a means of faithfully depicting lost structures.

8 Discussion and Conclusions 191

Bibliography

Addison, A. C. (2000). Emerging Trends in Virtual Heritage. IEEE Multimedia, 22– 25.

Almagro, A. (2007). Preserving the Architectural Heritage of al-Andalus. From Restoration to Virtual Reconstruction. Al-Masaq: Islam and the Medieval Mediterranean, 19(2), 155–175. https://doi.org/10.1080/09503110701581985

Almeida, V. M. De, & Barceló, J. A. (2012). Computer Simulation of Multidimensional Archaeological Artefacts. Virtual Archaeology Review, 3(7), 77–81.

Andrés, A. N., Pozuelo, F. B., Marimón, J. R., & Gisbert, A. de M. (2012). Generation of virtual models of cultural heritage. Journal of Cultural Heritage, 13(1), 103–106. https://doi.org/10.1016/j.culher.2011.06.004

Barceló, J. A. (2000). VISUALIZING WHAT MIGHT BE . An Introduction to Virtual Reality in Archaeology Visualizing archaeological data. In D. Barcelo, J.A., Forte, M., Sanders (Ed.), Virtual Reality in Archaeology (pp. 9–36). ArcheoPress.

Beacham, R. C. (2011). Defining our Terms in Heritage Visualization. In A. Bentkowska-Kafel, H. Denard, & D. Baker (Eds.), Paradata and Transparency in Virtual Heritage (1st ed., Vol. 12, pp. 8–11). Farnham, UK: Ashgate Publishing Limited.

Beacham, R., Niccolucci, F., Denard, H., Hermon, S., & Bentkowska-Kafel, A. (2009). THE LONDON CHARTER FOR THE COMPUTER-BASED VISUALISATION OF CULTURAL HERITAGE. Retrieved from http://www.londoncharter.org/

Bonde, S., Maines, C., Mylonas, E., & Flanders, J. (2009). The Virtual Monastery: Re‐Presenting Time, Human Movement, and Uncertainty at Saint‐Jean‐des‐ Vignes, Soissons. Visual Resources, 25(4), 363–377. https://doi.org/10.1080/01973760903331742

Bowen Loftin, R., Chen, J.X., Rosenblum, L., & Loftin, B. (2005). Visualization

Bibliography 193

Using Virtual Reality. In C. R. Hansen, Charles. D., Johnson (Ed.), The Visualization Handbook (pp. 497–489). Elsevier-Butterworth Heinemann.

Bowsher, J., & Miller, P. (2009). The Rose and the Globe playhouses of Shakespeare’s Bankside, Southwark: Excavations 1988-91. of London Archaeology.

Candy, L., & Edmonds, E. (2018). Practice-Based Research in the Creative Arts: Foundations and Futures from the Front Line. Leonardo, 51(1), 63–69. https://doi.org/10.1162/LEON_a_01471

Cargill, R. R. (2009). An Argument for Archaeological Reconstruction in Virtual Reality. , 72(1), 28–41.

Carrillo Gea, J. M., Toval, A., Fernández A., J. L., & Flores, N. J. M. (2015). The London Charter and the Seville Principles as sources of requirements for e- archaeology systems development purposes. Virtual Archaeology Review, 4(9), 205. https://doi.org/10.4995/var.2013.4275

Cerato, I., & Pescarin, S. (2013). Reconstructing Past Landscapes for Virtual . In V. F. Corsi C., Slapšak B. (Ed.), Good Practice in Archaeological Diagnostics. Natural Science in Archaeology. (pp. 285–296). https://doi.org/10.1007/978-3-319-01784-6_1

Clarke, A. c. (1962). Hazards of Prophecy: The Failure of Imagination. In Profiles of the Future (pp. 19–27). Holt, Rinehart, and Winston.

Coleman, G. (2016). Agile Software Development: Software Quality Professional, 19(1), 23–29.

Collingwood, R. G. (1926). Some Perplexities about Time : With an Attempted Solution. Proceedings of the Aristotelian Society, New Series, 26, 135–150. Retrieved from https://www.jstor.org/stable/4544098

Daniels, R. (1997). The need for the solid modelling of structure in the archaeology of buildings. Retrieved from http://dx.doi.org/10.11141/ia.2.1

Das, V. M., & Garg, Y. K. (2011). Digital reconstruction of pavilions described in an ancient Indian architectural treatise. ACM Journal on Computing and Cultural Heritage, 4(1), 1–16. https://doi.org/10.1145/2001416.2001417

194 Bibliography

Davis, A., Belton, D., Helmholz, P., Bourke, P., & McDonald, J. (2017). Pilbara rock art: Laser scanning, and 3D photographic reconstruction as heritage management tools. , 5(1), 1–16. https://doi.org/10.1186/s40494-017-0140-7

De Mattia, D. (2012). Il tempio romano dell’Asklepieion di Kos: nuovi dati per la sua anastilosi. Thiasos, 1, 61–68. Retrieved from http://www.thiasos.eu/wp- content/uploads/2012/05/10-De-Mattia-Tempio-C-Kos.pdf

Demetrescu, E. (2015). Archaeological stratigraphy as a formal language for virtual reconstruction. Theory and practice. Journal of , 57, 42– 55. https://doi.org/10.1016/j.jas.2015.02.004

Denard, H. (2011). A New Introduction to The London Charter. In Farnham, A. Bentkowska-Kafel, & H. Denard (Eds.), Paradata and Transparency in Virtual Heritage (1st ed., Vol. 2, pp. 57–71). Farnham, UK: Ashgate Publishing Limited.

Dudek, I., & Blaise, J. (2004). Graphic Variables for Dynamic 2D / 3D Documentation Visualisation in the Context of Historical Architecture. In ICHIM 04 - Digital Culture & Heritage (pp. 1–21). Berlin: French Ministry of Culture and Communication. Retrieved from http://www.map.archi.fr

Dunn, S., & Woolford, K. (2012). Reconfiguring using 3D Reconstruction. In Electronic Workshops in Computing (eWiC). London.

Elliot, A. J., & Maier, M. A. (2014). Color : Effects of Perceiving Color on Psychological Functioning in Humans. Annual Review of Psychology, 65(1), 95–120. https://doi.org/10.1146/annurev-psych-010213-115035

Frischer, B. (2009). The Rome Reborn Project. How Technology is helping us to study history. Retrieved from http://www.londoncharter.org/

Frisher Consulting. (2013). Rome Reborn. Retrieved March 6, 2014, from http://romereborn.frischerconsulting.com/about.php

Gasparini, G., Redondo, E., & Dávila, M. (2012). Restoration and Disemmination Through Digital Techniques of Missing Venezuelen Cultural Heritage . Virtual Restoration of the San Jacinto Church in Caracas. A Case Study. EGA : Revista de Expresión Gráfica Arquitectónica, (20), p214-225.

Bibliography 195

Gershon, N. (1998). Visualization of an Imperfect World. Computer Graphics and Applications, 18(4), 43–45.

Greenfield, J., & Gurr, A. (2003). The Rose Theatre , London : the state of knowledge and what we still need to know, (September), 330–340.

Grigoryan, G., & Rheingans, P. (2002). Probabilistic surfaces: point based primitives to show surface uncertainty. In IEEE Visualization, 2002. VIS 2002. (pp. 147– 153). Ieee. https://doi.org/10.1109/VISUAL.2002.1183769

Guidi, G., Frischer, B., De Simone, M., Cioci, A., Spinetti, A., Carosso, L., … Grasso, T. (2005). Virtualizing Ancient Rome: 3D Acquisition and Modeling of a Large Plaster-of-Paris Model of Imperial Rome. In J.-A. Beraldin, S. F. El- Hakim, A. Gruen, & J. S. Walton (Eds.), Videometrics VIII (Vol. 5665, pp. 119– 133). San Jose. https://doi.org/10.1117/12.587355

Gurr, A. (1992). The Shakespearean stage, 1574-1642 (third edti). Cambridge: Cambridge University Press.

Gurr, A. (1997). Traps and Discoveries at the Globe. Proceedings of the British Academy, 94(April 1996), 85–101.

Henslow, J. (1904). Henslows Diary. (W. Walter & M. A. Greg, Eds.). A. H. Bullen.

Henslowe, P. (2002). Henslowe’s Diary. (R. A. Foakes, Ed.) (2nd ed.). Cambridge University Press.

Hermon, S. (2008). Beyond Illustration: 2D and 3D DigItal Technologies as Tools for Discovery In Archaeology. (B. Frischer & A. Dakouri-hild, Eds.). Archaeopress Oxford.

Hermon, S. (2008). Reasoning in 3D: a Critical Appraisal of the role of 3D modelling and virtual reconstructions in archaeology. In B. Frischer & A. Dakouri-Hild (Eds.), Beyond Illustration: 2D and 3D Digital Technologies as Tools for Discovery in Archaeology (pp. 36–45). Oxford : Archaeopress. Retrieved from https://trove.nla.gov.au/version/41836228

Hermon, S. (2011). Scientific Method , chaine operatoire and Visualization : 3D Modelling as a Research Tool in Archaeology. In A. Bentkowska-Kafel, H. Denard, & D. Baker (Eds.), Paradata and Transparency in Virtual Heritage (pp. 13–22). Farnham, UK: Ashgate Publishing Limited.

196 Bibliography

Hermon, S., & Kalisperis, L. (2011). Between the Real and the Virtual : 3D visualization in the Cultural Heritage domain - expectations and prospects. Virtual Archaeology Review, 2(4), 59–63.

Hermon, S., & Kalisperis, L. (2011). Between the Real and the Virtual : 3D visualization in the Cultural Heritage domain - expectations and prospects. Virtual Archaeology Review, 2(4), 59–63.

Hermon, S., & Nikodem, J. (2007). 3D Modelling as a Scientific Research Tool in Archaeology. Proceedings of the 35th International Conference on Computer Applications and Quantitative Methods in Archaeology (CAA), 1–6. Retrieved from http://proceedings.caaconference.org/files/2007/38_Hermon_Nikodem_CAA20 07.pdf

Herzog, R. (1932). Asklepieion: building description and building history (Volume 1). (P. Schazmann, Ed.). Berlin, Heidelberg: Kellar. Retrieved from http://digi.ub.uni-heidelberg.de/diglit/kos1932bd1

Hynst, S., Gervautz, M., Grabner, M., & Schindler, K. (2001). A work-flow and data model for reconstruction , management , and visualization of archaeological sites. In Proceedings of the 2001 conference on Virtual reality, archeology, and cultural heritage (VAST ’01) (pp. 43–52). ACM. https://doi.org/10.1145/584993.585000

Issini, G. (2012). Knowledge Visualization of Large-size Architectural Heritage. A Research Experience on Yanqing Section of Chinese Great Wall. 2012 16th International Conference on Information Visualisation, 523–527. https://doi.org/10.1109/IV.2012.89

Jablonka, P., Kirchner, S., Serangeli, J., Tübingen, E., & Frühgeschichte, U.-. (2002). TroiaVR : A Virtual Reality Model of and the Troad. In M. Doerr & A. Sarris (Eds.), Computer Applications in Archaeology (CAA). Heraklion, Crete.

Johanson, C. (2009). Visualizing History: Modeling in the Eternal City Three- Dimensional Reconstruction: Rome. Visual Resources, 25(4), 403–418. https://doi.org/10.1080/01973760903331924

Johnson, C. R., & Sanderson, A. R. (2003). A Next Step : Visualizing Errors and

Bibliography 197

Uncertainty. IEEE, (October).

Jones, M. W. (1989). Designing the Roman Corinthian order. Journal of Roman Archaeology, 2, 35–69. https://doi.org/10.1017/S1047759400010291

Kastanis, L. (2015). Rebuilding the Rose Theatre: Experiences, Assumptions and Outcomes. Retrieved from https://lazphdlog.tumblr.com/post/97204874734/rebuilding-the-rose-theatre- experiences

Kastanis, L., Pack, D., & Tompkins, J. (2007). Recreation of The Rose Theatre, London. Retrieved from http://ortelia.com/new/portfolio/rosetheatre/

Kelly, M. (2013). The Archaeology of Tomb Raider. Retrieved March 6, 2014, from http://archaeologyoftombraider.com/2013/10/07/arte-factual-minoan-dolphin- fresco/

M. I. Zainal Abidina, A. Bridges, A. R. (2008). The Digital Archaeological Reconstruction of the a Famosa Fortress, Malaysia. In VSMM2008, The 14th International Conference on Virtual Systems and Multimedia: Digital Heritage 2.0 (pp. 209–214). Limnasol: VSMM.

MacEachren, A. M., Robinson, A., Hopper, S., Gardner, S., Murray, R., Gahegan, M., & Hetzler, E. (2005). Visualizing Geospatial Information Uncertainty: What We Know and What We Need to Know. Cartography and Geographic , 32(3), 139–160. https://doi.org/10.1559/1523040054738936

Moir, J., & Letts, J. B. (1999). Thatch: thatching in England 1790-1940. English Heritage Research Transactions: Research and case Studies in Architectural Conservation (Vol. 5). James and James.

Neubauer, W., Gugl, C., Scholz, M., Verhoeven, G., Trinks, I., Doneus, M., … Meirvenne, M. V. (2014). The discovery of the school of gladiators at Carnuntum , Austria, Supplement. Antiquity, 88(339), 1–3. Retrieved from http://antiquity.ac.uk/ant/088/ant0880173.htm

Neubauer, W., Gugll, C., Scholz, M., Trinks, I., Locker, K., Doneus, M., & Meirvenne, M. Van. (2014). The discovery of the school of gladiators. Antiquity, 88(339), 173–190. Retrieved from

198 Bibliography

http://antiquity.ac.uk/ant/088/ant0880173.htm

Niccolucci, F., & Hermon, S. (2004). A fuzzy logic approach to reliability in archaeological virtual reconstruction. In CAA2004: Beyond the – Digital Interpretation of the Past (pp. 26–33). Budapest: public- repository.epoch-net.org.

Olsen, S., Brickman, A., & Cai, Y. (2004). Discovery by Reconstruction : Exploring Digital Archeology. In SIGCHI Workshop (Ambient Intelligence for Scientific Discovery (AISD)).

Palombini, A., Arnoldus-huyzendveld, A., Ioia, M. Di, Gioia, P., Persiani, C., & Pescarin, S. (2012). The everyday-life in neanderthal times : a full-immersive Pleistocene reconstruction for the Casal De ’ Pazzi Museum ( Rome ). Virtual Archaeology Review, 3(7), 73–76.

Pang, A. T., Wittenbrink, C. M., & Lodha, S. K. (1997). Approaches to uncertainty visualization. The Visual Computer, 13(8), 370–390. https://doi.org/10.1007/s003710050111

Petriaggi, B. D., Petriaggi, R., Bruno, F., Lagudi, A., Peluso, R., & Passaro, S. (2018). A digital reconstruction of the sunken “villa con ingresso a protiro” in the underwater archaeological site of Baiae. IOP Conference Series: Materials Science and Engineering, 364(1). https://doi.org/10.1088/1757- 899X/364/1/012013

Petrova, Y. A., Tsimbal, I. V., Laska, T. V., & Golubkov, S. V. (2011). Practice of Using Virtual Reconstruction in the Restoration of Monumental Painting of the Church of the Transfiguration of Our Saviour on Nereditsa Hill. 2011 15th International Conference on Information Visualisation, 389–394. https://doi.org/10.1109/IV.2011.33

Pierdicca, R., Frontoni, E., Malinverni, E. S., Colosi, F., & Orazi, R. (2016). Virtual reconstruction of archaeological heritage using a combination of photogrammetric techniques: Huaca Arco Iris, Chan Chan, Peru. Digital Applications in Archaeology and Cultural Heritage, 3(3), 80–90. https://doi.org/10.1016/j.daach.2016.06.002

Pletinckx, D., & Tartessos, P. (2011). Virtual Archaeology as an Integrated

Bibliography 199

Preservation Method. Virtual Archaeology Review, 2(4), 33–37.

Pollefeys, M., Proesmans, M., Koch, R., Vergauwen, M., & Van Gool, L. (1998). Acquisition of Detailed Models for Virtual Reality. BAR INTERNATIONAL SERIES, (843), 71–78.

Pollini, J., Swartz Dodd, L., Kensek, K., & Cipolla, N. (2005). Problematics of making ambiguity explicit in virtual reconstructions: a case study of the Mausoleum of Augustus. In Theory and Practice, Proceedings of the 21st Annual Conference of CHArt: Computers and the History of Art. British Academy, London. Retrieved from http://www.chart.ac.uk/chart2005/papers/pollini.html

Rajapakse, R. P. C. J., Tokuyama, Y., & Somadeva, R. Virtual Reconstruction and Visualization of Pre and Proto Historic Landscapes in Srilanka, 2011 International Conference on Biometrics and Kansei Engineering § (2011). IEEE. https://doi.org/10.1109/ICBAKE.2011.69

Rashid, M., Rahaman, H., Rashid, M., & Rahman, M. (2010). Heritage Interpretation Collective Reconstruction of Sompur Mahavihara, Bangladesh. In M. Ioannidis, A. Addison, A. Georgopoulos, & L. Kallisperis (Eds.), Virtual Systems and Multimedia (VSMM), 2010 (pp. 30–36). https://doi.org/10.1109/VSMM.2010.5665990

Renfrew, C., & Bahn, P. (2012). Archaeology: Theories, Methods and Practice (6th ed.). London, UK: Thames and Hudson.

Roach, C. (2016). Rehanging Reynolds at the British Institution : Methods for Reconstructing Ephemeral Displays. https://doi.org/10.17658

Schofield, J. (1991). The Construction of Medieval and Tudor Houses in London. Construction History, 7, 3–28.

Silberman, N. . A. (2008). (2008) ICOMOS Charter for the Interpretation and Presentation of Cultural Heritage Sites. International Journal of , 15(4), 377–383. https://doi.org/10.1017/S0940739108080417

Silva, S., Sousa Santos, B., & Madeira, J. (2011). Using color in visualization: A survey. Computers and Graphics (), 35(2), 320–333. https://doi.org/10.1016/j.cag.2010.11.015

200 Bibliography

Skeels, M., Lee, B., Smith, G., & Robertson, G. G. (2010). Revealing uncertainty for information visualization. Information Visualization, 9(1), 70–81.

Smith, H., & Dean, R. T. (2009). Practice-led Research, Research-led Practice in the Creative Arts. (H. Smith & R. T. Dean, Eds.). Edinburgh University Press. Retrieved from http://ebookcentral.proquest.com/lib/qut/detail.action?docID=475756.

Stanley-price, N. (2009). The Reconstruction of Ruins : Principles and Practice The Reconstruction of Ruins : Principles and Practice. In A. Bracker & A. Richmond (Eds.), Conservation: Principles, Dilemmas and Uncomfortable Truths. Elsevier.

Thomson, J., Hetzler, E., Maceachren, A., & Gahegan, M. (2005). A Typology for Visualizing Uncertainty. In R. F. Erbacher, J. C. Roberts, M. T. Gröhn, & K. Börner (Eds.), Visualization and Data Analysis (Vol. 5669, pp. 146–157). Retrieved from http://proceedings.spiedigitallibrary.org/

Thuswaldner, B., Flöry, S., Kalasek, R., Hofer, M., Huang, Q.-X., & Thür, H. (2009). Digital anastylosis of the Octagon in Ephesos. Journal on Computing and Cultural Heritage, 2(1), 1–27. https://doi.org/10.1145/1551676.1551677

Tompkins, J., & Kastanis, L. (2017). Staging supernatural creatures in a computer- based visualisation of London’s sixteenth-century Rose Theatre. International Journal of Performance Arts and , 0(0), 1–17. https://doi.org/10.1080/14794713.2017.1280306

UNESCO. (2013). Operational Guidelines for the Implementation of the World Heritage Convention. UNESCO World Heritage Centre. Retrieved from http://whc.unesco.org/archive/opguide13-en.pdf

Walker, W. E., Harremoës, P., Rotmans, J., van der Sluijs, J. P., van Asselt, M. B. A., Janssen, P., & Krayer von Krauss, M. P. (2003). Defining Uncertainty: A Conceptual Basis for Uncertainty Management in Model-Based Decision Support. Integrated Assessment. https://doi.org/10.1076/iaij.4.1.5.16466

Wells, S., Frischer, B., Ross, D., & Keller, C. (2009). Rome Reborn in Google Earth. In Making History Interactive. 37th Proceedings of the CAA Conference (pp. 373–379). Williamsburg: Archaeopress: Oxford. Retrieved from

Bibliography 201

http://romereborn.frischerconsulting.com/rome_reborn_2_documents/papers/W ells2_Frischer_Rome_Reborn.pdf

Wittenbrink, C. M., Pang, A. T., & Lodha, S. K. (1996). Glyphs for Visualizing Uncertainty in Vector Fields. IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, 2(3), 266–279.

Zadeh, L. (1965). Fuzzy Sets. Information and Control, 8(3), 338–353. https://doi.org/10.1016/S0019-9958(65)90241-X

Zarmakoupi, M. (2010). The virtual reality digital model of the Villa of the Papyri project. In M. Zarmakoupi (Ed.), Sozomena : Studies in the Recovery of Ancient Texts, Volume 1 : Villa of the Papyri at Herculaneum : Archaeology, Reception, and Digital Reconstruction: The virtual reality digital model of the Villa of the Papyri project (pp. 181–194). Walter de Gruyter.

Zuk, T., Carpendale, S., & Glanzman, W. D. (2005). Visualizing temporal uncertainty in 3D virtual reconstructions. In VAST 2005: the 6th International Symposium on Virtual Reality, Archaeology and Intelligent Cultural Heritage, incorporating 3rd Eurographics Workshop on Graphics and Cultural Heritage: ISTI-CNR Pisa, Italy, November 8-11, 2005 (pp. 99–106). Eurographics Association. https://doi.org/10.2312/VAST/VAST05/099-106

202 Bibliography

Appendices

Data Source acknowledgement

The input data sourced for the reconstruction of the Komediehuset and referenced in this Chapter were provided by the academics and from Norway who are attached to the Lost Theatres international research team: Professor Julie Holledge (University of Oslo), Dr Ellen Karoline Gjervan (Dronning Mauds Minne Høgskole), Tove Jensen Holmås (Theatre Archive, University of Bergen), Anders Nilsen (Bergen Byakiv). The 1938 survey drawings came from the Bergen City Council architectural archives. The architectural drawing by Peter Andreas Blix, and the seating plans of the theatre are held in the Theatre Archive at the University of Bergen. I would like to thank both Professor Holledge and Dr Gjervan for the interpretative work that they have shared with me concerning these historical documents.

Appendices 203

Appendices

Appendix A

Tables

Category Subcategories Examples

Accuracy/error • Collection Accuracy • Documents that are translated into English may • Processing errors contain translation errors. • Deception • A report may note that 50 tanks were observed although the tanks may in fact be dummy placements. Precision • Precision of collection • resolution of • Capability Completeness • Composite completeness • images of a site may not be available on a • Information completeness particular day because of adverse weather • Incomplete sequence conditions. • an intercepted conversation may have words that were not clear • the lack of confirming information might signal incompleteness Consistency • Conflict among info • multiple sources or data types may actually • Model/observation conflict • Consistency • models of events may differ from observations Lineage • Translation • Machine translation is more uncertain than • Transformation human linguist translation Interpretation • Measurements or signals may have been transformed • Information that comes directly from the collection capabilities has a different lineage than an • interpretive report produced by an analyst Currency/timing • Temporal gaps • Images that show new objects do not show • Versioning when the object first appeared

204 Appendices

• Time between when events occurred, when they were reported, and when the information is available to analysts • Reports may have multiple versions, sometimes with major changes. Credibility • Reliability • Possibility of deliberate disinformation • Proximity • Source may not have expertise on this subject • Appropriateness • Information may be second hand • Motivation (of the source) Subjectivity • Analytic judgment • Amount of interpretation added rather than pure facts Interrelatedness • Source independence • Likelihood that the source derives from other • reported information (such as repeated news stories)

Table A-1: An Analytic Uncertainty Typology as proposed by Thomson et al. (2005) (Thomson et al. 2005, pp. 152-153)

Appendices 205

Domain Ecology Computational Biology Medicine Goal Identifying species diversity in Modelling viral evolution Diagnosis of disease different regions of the world Data Types List (and numbers) of species and Evolutionary tree and RNA Variety of formats such as maps sequences written/verbal descriptions, numerical values and images Inference Predicting future populations Models of evolution are Signs and Symptoms are used based on past events or current imperfect so inference about to infer diagnosis when often states is imperfect past viral sequences based on there is not a clear line evolutionary models is between one set of signs and uncertain symptoms and one diagnosis Completeness Lack of data for some parts of the Do not have sequences for all Decisions must often be world leaves an incomplete data strains of the virus, which would made with incomplete set be required to have a complete medical records or without data set running all possible tests Measurement Identification of a species within Errors can occur in lab Many diagnostic procedures precision a specific geographic region processes for identifying (for example laboratory tests) could be wrong because location sequences are not perfectly precise data is imprecise Credibility Data gathered from some sources Viral sequences from one data Some expert opinions (for may be trusted over others based source might be trusted over example radiology) may be on source's qualifications another trusted over others Two Disagreement Different sources may disagree Different evolutionary models Two different tests might about the presence or absence of render different results contradict one another or two a species in a region doctors might disagree on a diagnosis

Table A-2: Examples of uncertainty from three different domains Domain, from (Skeels et al. 2010, p. 78)

206 Appendices

Reconstructed Evidence Uncertainty Value Elements (Derived from the Uncertainty Class) Supporting Existing extant remains (incomplete). Sufficient 1.25 Foundations evidence to determine extent and size of structure but not final height. Supporting Sill Beams No extant remains. Known material types based on 2.5 existing buildings and known building materials/methods. The shape the foundations only support one type of supporting beam style Joists No extant remains. Known material types based on 2.5 existing buildings and known building materials/methods. Joists had to exist but spacing and size is unknown Supporting Posts No extant remains. Known material types based on 2.0 existing buildings and known building materials/methods Tiers No extant remains. Known material types and 3.5 dimensions based on existing and known buildings. Extensive documentary evidence to support multi-tier. Tier heights unknown Flooring No extant remains. Known material types based on 3.0 existing buildings and known building materials/methods Roof Structure Some extant remains of Thatch roofing. Existing 1.5 documentary evidence (images) show a pitched thatch roof. Known building techniques for thatch roofing. Documentary evidence supports the reconstruction Walls Some extant remains of lath wall. Known material 1.5 types based on existing buildings and known building materials/methods. Documentary evidence supports the reconstruction Windows Very limited extant remains. Dimensions and 5.0 placement of windows unknown. No documentary or anecdotal evidence to determine placement, size or spacing. Limited documentary evidence supports the inclusion of windows Building Entrance No extant remains. Existing documentary evidence 4.5 (images) show an entrance on Maiden Lane. No visual detail of entrance size available. Analogous written documentary evidence suggests a narrow entrance. Entrance size is a debated issue Stairways and No extant remains. The lack of any external stairway 5.0 Stairwells foundations may suggest that internal stairs were employed. Stage Extensive extant remains of the stage foundations. 1.25 Known position, size and shape of the stage. No indication of the height of the stage and the material it was constructed of. Stage canopy Archaeological evidence to support the presence of 3 canopy support columns. Evidence of a ‘drip-line’ around the stage suggesting a canopy. Documentary evidence exists that stage canopies were in use in playhouses of the era. Images of the Rose suggest a possible canopy.

Appendices 207

Reconstructed Evidence Uncertainty Value Elements (Derived from the Uncertainty Class) Tiring House Documentary records from Henslowe's Diary suggest a 2.5 Tiring house. The archaeological remains of the stage wall show a distinctly heavier build and are designed to carry heavier load. The extent of the Tiring House, as well as access points, are driven by the constraints of the structure. Railings and Balusters It is highly plausible that railings did exist, as evidence 2.0 of balusters have been unearthed. The height of the railings can be determined with a high confidence and the reconstructed space is physically suited to the purpose. Table A-3: Uncertainty values assigned to reconstructed elements of the Rose Theatre

208 Appendices

Reconstructed Evidence Uncertainty Class Elements Colour Supporting Existing extant remains (incomplete). Sufficient evidence to Foundations determine extent and size of structure but not final height. Supporting No extant remains. Known material types based on existing Sill Beams buildings and known building materials/methods. The shape the foundations only support one type of supporting beam style Joists No extant remains. Known material types based on existing buildings and known building materials/methods. Joists had to exist but spacing and size is unknown Supporting No extant remains. Known material types based on existing Posts buildings and known building materials/methods Tiers No extant remains. Known material types and dimensions based on existing and known buildings. Extensive documentary evidence to support multi-tier. Tier heights unknown Flooring No extant remains. Known material types based on existing buildings and known building materials/methods Roof Some extant remains of Thatch roofing. Existing documentary Structure evidence (images) show a pitched thatch roof. Known building techniques for thatch roofing. Documentary evidence supports the reconstruction Walls Some extant remains of lath wall. Known material types based on existing buildings and known building materials/methods. Documentary evidence supports the reconstruction Windows Very limited extant remains. Dimensions and placement of windows unknown. No documentary or anecdotal evidence to determine placement, size or spacing. Limited documentary evidence supports the inclusion of windows Building No extant remains. Existing documentary evidence (images) Entrance show an entrance on Maiden Lane. No visual detail of entrance size available. Analogous written documentary evidence suggests a narrow entrance. Entrance size is a debated issue Stairways No extant remains. The lack of any external stairway and foundations may suggest that internal stairs were employed. Stairwells Stage Extensive extant remains of the stage foundations. Known position, size and shape of the stage. No indication of the height of the stage and the material it was constructed of. Stage canopy Archaeological evidence to support the presence of canopy support columns. Evidence of a ‘drip-line’ around the stage suggesting a canopy. Documentary evidence exists that stage canopies were in use in playhouses of the era. Images of the Rose suggest a possible canopy. Tiring House Documentary records from Henslowe's Diary suggest a Tiring house. The archaeological remains of the stage wall show a distinctly heavier build and are designed to carry heavier load. The extent of the Tiring House, as well as access points, are driven by the constraints of the structure.

Appendices 209

Reconstructed Evidence Uncertainty Class Elements Colour Railings and It is highly plausible that railings did exist, as evidence of Balusters balusters have been unearthed. The height of the railings can be determined with a high confidence and the reconstructed space is physically suited to the purpose. Table A-4: Rose Theatre reconstruction - classified components.

210 Appendices

Reconstructed Evidence Uncertainty Class Elements Colour Main Building Documentary evidence - 1938 official survey drawings, post- mass 1870 exterior photography structure External Walls Documentary evidence - post-1870 exterior photography External Doors Documentary evidence - 1938 official survey drawings, post- 1870 exterior photography Windows Documentary evidence - 1938 official survey drawings, post- 1870 exterior photography Roof Documentary evidence - 1938 official survey drawings, post- 1870 exterior photography Internal Stairs Documentary evidence - 1938 official survey drawings, Peter Blix renovation drawings Internal Stair Documentary evidence - 1938 official survey drawings Railings Internal Documentary evidence - 1938 official survey drawings Doorways Internal Doors Documentary evidence - 1938 official survey drawings Orchestra Pit Documentary evidence - 1938 official survey drawings Orchestra Pit Documentary evidence - 1938 official survey drawings Wings Stage Documentary evidence - 1938 official survey drawings Stage Floor Documentary evidence - 1938 official survey drawings Stage Floor Slots Analogous structures - Indicative analogous surviving examples Stage Front Documentary evidence - 1938 official survey drawings Auditorium Documentary evidence - 1938 official survey drawings, post- Columns 1870 interior photography Gallery Layout Documentary evidence - 1938 official survey drawings, Peter Blix renovation drawings, 1856 Seating plan Gallery décor Analogous structures - Indicative analogous surviving Design examples Seating Analogous structures - 1856 Seating plan, Indicative analogous surviving examples Ceiling Design Documentary evidence - Peter Blix renovation drawings Ceiling 1938 official survey drawings Chandelier Recess Internal 1938 official survey drawings Auditorium Lining Seating Documentary evidence - 1856 Seating plan Analogous Examples – Existing theatres of the same period Internal Fixtures 1938 official survey drawings, post-1870 interior photography Table A-5: Bergen Theatre reconstruction classified building elements.

Appendices 211

Reconstructed Evidence Uncertainty Elements Class Colour Overall Building Extant Remains – identified through remotely sensed data complex Footprint Eastern Entrance Analogous Structures – Off-site similar examples (Gateway Arch) Driveway and Paved Documentary Evidence - Inference from combined survey Path Eastern results Entrance Formal Entranceway Analogous Structures – Off-site similar examples Triumphal Arch Administrative Block Documentary Evidence - Inference from combined survey results and documented practices Owners Living Extant Evidence - Inference from combined survey results Quarters Documentary Evidence - Inference from combined survey results and documented practices North-East Main Documentary Evidence – Inference from combined survey Building and Bath results Complex Extant remains – identified through remotely sensed data Memorials Documentary Evidence - Inference from combined survey results

Analogous Structures – Off-site similar examples Gladiator Barracks Documentary Evidence - Inference from combined survey Ground Level results

Gladiator Barracks Analogous Structures – Off-site similar examples Colonnades Ground Level Gladiator Barracks First Extant Evidence - Inference from combined survey results Floor

Gladiator Preparation Documentary Evidence - Inference from combined survey Hall results Main Gladiatorial Extant Remains – identified through remotely sensed data - Training Arena Main Training Area Documentary Evidence - Inference from combined survey Foundations results Main Training Area Documentary Evidence - Inference from combined survey Spectator Stand results Minor Training Area Extant Remains – identified through remotely sensed data Table A-6: Temple C reconstruction - classified components.

212 Appendices

Reconstructed Evidence Uncertainty Elements Class Colour Stylobate Extant Extant remains in-situ Stylobate Inferred North-east corner inferred from analogous onsite remains Columns Extant Limited extant remains in-situ Columns Inferred Columns inferred from analogous onsite remains Entablature Extant Limited extant remains in-situ for the north-east corner Entablature Inferred Remaining entablature inferred from analogous onsite remains Pediment Extant Limited extant remains in-situ for the south-east corner Pediment Inferred Remaining pediment inferred from analogous onsite remains Peripterals Inferred from original Herzog interpretation of the Temple C Table A-7: Gladiator School reconstruction - classified components.

Appendices 213

Uncertainty Class Description

A very low level of uncertainty associated with: • Complete or extensive archaeological evidence – I • Very detailed documented evidence, literary or cultural evidence – II 1 • Many similar examples of analogous structures or methods of construction available on site or geographically close – III The digital reconstruction is supported by known construction methods/materials of the time- period and the laws of physics. – IV •

A low level of uncertainty associated with: • Substantial significant archaeological evidence – I • Detailed documentary, anecdotal, literary or cultural evidence – II 2 • Similar examples of analogous structures or methods of construction exist – III • The digital reconstruction is supported by known construction methods/materials of the time-period and the laws of physics. – IV

A moderate level of uncertainty associated with: • Significant archaeological evidence – I • Documentary, anecdotal, literary or cultural evidence exists although not detailed – II 3 • Somewhat similar analogous structures or methods of construction – III • The reconstruction is supported by the laws of physics and known construction methods/materials of the time period – IV

A high level of uncertainty associated with: • Limited archaeological evidence – I • Limited documentary, anecdotal, literary or cultural evidence – II 4 • Few analogous structures or exemplified methods of construction – III • The components of the reconstruction support the laws of physics and some known construction methods/materials of the time period – IV

A very high level of uncertainty associated with: • Very limited archaeological evidence – I • Very limited documentary, anecdotal, literary or cultural evidence – II 5 • Lack of analogous structures or exemplified methods of construction – III • The reconstruction is not supported by the laws of physics and known construction methods/materials of the time-period. – IV

Table A-8: The tabular matrix for assigning classes of uncertainty

214 Appendices

Appendices

Appendix B

Figures

Figure A-1: 1850 seating plan showing underlying gallery outline (Red) and position of doorways (Blue). Image courtesy of Holledge et.al.

Appendices 215

Figure A-2: Exterior photograph of the Bergen Theatre showing the Right Wing and Interior detail photograph of the post 1870 gallery. Courtesy of Holledge et.al.

Figure A-3: Scale model of the post 1870 renovation. Courtesy of Holledge et.al.

Figure A-4: The Fredrikshalds Thater (Source: https://ostfoldmuseene.no/wp- content/uploads/revslider/fredrikshalds-teater/hovedbilde.jpg)

216 Appendices

Figure A-5: Comparison of Blix Stage columns and arch (Left) with the 1938 survey drawings (Right) showing style similarities. Courtesy of Holledge et.al.

Figure A-6: Example of a stage prompter's box (Image source: https://goo.gl/images/xXjEAv)

Appendices 217

Original pre-1870 gallery outline

Supporting columns

Figure A-7: A section of one of the Blix plan drawings showing the line of the original gallery and supporting columns. Courtesy of Holledge et.al.

218 Appendices

Figure A-8: The Georgian Theatre Royal showing the square utilitarian gallery. (Image source: https://goo.gl/images/2UX8hS)

Appendices 219

Figure A-9: Ceiling detail from the 1870 Blix renovation proposals. Courtesy of Holledge et.al.

220 Appendices

Figure A-10: Interior of the Bergen Theatre, early 1900's showing the gas chandelier. Courtesy of Holledge et.al.

Figure A-11: Side elevation Blix drawing showing internal decor and lighting fixtures. Courtesy of Holledge et.al.

Appendices 221

Figure A-12: Aernout van Buchel's copy of Johannes de Witt's drawing of the Swan playhouse, Circa 1596 Copyright Utrecht, University Library, Ms 842. I

222 Appendices

Figure A-13: Example plan and elevation drawings of the internal Bergen Theatre stair system from the 1938 official survey drawings. Courtesy of Holledge et.al.

Figure A-14: Peter Blix 1870 renovation drawing depicting the stage. Courtesy of Holledge et.al.

Appendices 223

Figure A-15: Photograph showing the Bergen theatre stage post 1870. Unknown date. Courtesy of Holledge et.al.

Figure A-16: An example Baroque theatre stage (Český Krumlov Castle Theater) showing floor slots and scenery flats. (Source: https://fotosdenatalia.files.wordpress.com/2012/06/dsc_0138edit.jpg)

224 Appendices

Figure A-17: Elevation drawing for the official 1938 survey showing the roof framework and chandelier recess. Courtesy of Holledge et.al.

Figure A-18: Remotely sensed data (from GPR and Magnetic survey) incorporated in the GIS system. Inset 3D visualisation (Top right) shows the visualised archaeological data. Source:(Neubauer et al., 2014, p. 182)

Appendices 225

Figure A-19: Plan view of the Asklepieion archaeological site. Temple C outlined in red. (Image Source: M.S. Kiapokas, Hippocrates of Cos and the Hippocratic Oath, 1999, p. 164.)

226 Appendices

Figure A-20: Temple C, Asklepieion of Kos, showing partial anastylosis undertaken by Mario Paolini and Luigi Morricone in 1938 (Image: Lazaros Kastanis 2017)

Appendices 227

Figure A-21: North-east elevation drawing of proposed anastylosis by M. Paolini 1938. Source: De Mattia 2012 P. 70

228 Appendices

Figure A-22: Temple C photography of excavation site with extant architectural fragments identified in-situ. Source: R. Herzog 1902, De Mattia P.78

Appendices 229

Figure A-23: Original survey drawings of Temple C (Top) and interpreted reconstruction (bottom) by Herzog. Source: Herzog, 1932, Plate 23, P. 139

230 Appendices

Figure A-24:Reconstruction of a section of Temple C based on extant fragments as proposed by Herzog. Note the lack of Capitals. Source: Herzog R., 1932, Plate 25, p. 141

Appendices 231

Figure A-25: Example of architectural fragments from Temple C and nearby Temple A as described by Herzog. Source: Herzog, 1932, Plate 26, P. 142

232 Appendices

Figure A-26: Initial Numerical Decision tree detailing the process of assigning data a class of uncertainty.

Appendices 233

Figure A-27: The modified Decision Tree

234 Appendices

Appendices 235