Potential of Sunlight Simulation to Support Conservation of the Bayon Archaeological Site

Potential of Sunlight Simulation to Support Conservation of the Bayon Archaeological Site

Nobuya Watanabe, Hiroyuki Yoshida*, Yuichiro Usuda**
Research Postgraduate, MAG-SFC
Keio University, Japan
Email: nov@sfc.keio.ac.jp
*Post-Doctoral Research Associate
SFC, Keio University, Japan
**Remote Sensing Technology Centre of Japan, Japan

Abstract
An important determinant for surface temperature of an object on the ground is the quantity of heat brought by sunlight. Heat reaches surfaces, and deterioration may begin from the torrid superficies. This is particularly crucial for an archeological site when its conservation is considered an urgent issue. In this study, solar energy influx to Bayon, a major part of the Angkor Monuments in Cambodia, is examined through simulations. The objective of this attempt is to clarify the way in which sunlight covers the site over a day as well as over seasons. The approach taken in this study is complementary to conventional field surveys in two ways. Firstly, while thermographic instruments and other tools enable accurate but only temporary measurements of changing temperatures and other variables in the field, simulations of geometry of the site in relation to the sun enables analyses of effects of solar energy on the site for a prolonged period. Secondly, while field surveys consists of sampling from the target inevitably, simulations enable spatially synoptic observations of the site having complex compositions. Simulation of sunlight, simulation of changing shadows around the three-dimensionally modeled site in other words, is a comprehensive approach which sets data derived from field surveys in temporal and spatial contexts. The sunlight simulation carried out in this study presents geometrical patterns through which heat taking the form of sunlight is given to the site of Bayon. On the basis of such geometrical patterns, the possibility to shed light on concrete mechanisms in which not just heat but also precipitation, humidity, permeability of building material and structure of the site itself result in damages will be envisaged.

Introduction
Simulation of a phenomenon by using a computer has become common throughout the wide spectra of science and engineering. This trend is reaching the domain of archaeological studies. Among various approaches of computer simulations, production of computer graphics (CG) has a significant possibility for many archaeological researches. Graphics produced through simulations can be interfaces for qualitative observations / analyses and compilations of quantitative facts taking forms of data sets. In many archaeological studies, one of the most important characteristics of CG to be exploited is its ability to visualise scenes that are difficult, or impossible, to see under normal circumstances.

In this study, attempts were made to visualise shadows around the large and complex ruin in Cambodia, the Bayon, over a day and a year to formulate a conservation strategy to be deployed in future. Systematic observation of shadows of, and insolation over, an archaeological site like the Bayon in situ is not an easy task to carry out. While it might be possible to observe a site on one day or for one week in a scientifically sound manner, human and economic resources required to do the same for one year could be too enormous to proceed. The difficulty becomes an impossibility for many archaeological sites in areas where sufficient logistics can be hardly maintained. Moreover, the impossibility cannot be confined in the temporal aspect of fieldwork when a target is spatially large and complex. Field observation consists mainly of sampling of variables at points on and around, or at angles to, the target. The number of samples is limited under any circumstance. If a ruin of interest is not simple enough, the limit on sampling forces the observer to omit certain important features from analyses. In the context explained above, visualisation of shadows of the ruin of Bayon in Cambodia is significant in three ways. Firstly, such an approach enables an observation for a prolonged period. Secondly, computer simulation is not restricted by insufficient logistics. Thirdly, and most substantially, it is almost the only way to comprehend the spatially complex target as a whole. Conventional methods to measure insolation over a target like the Bayon are usages of instruments such as sunshine recorder, temperature sensor and thermography. Sunshine recorders and temperature sensors measure the variables only at points, and their locations and numbers are highly likely to bias consequent analyses and inferences. This problem has to be considered critical when structural complexity of the Bayon is taken into account. A thermography can record a series of frames at an angle to the target, but it is hard to set up a number of the costly instruments in the site for a long-term observation. The simulation of sunlight by using a 3D model of the ruin and its geographical location data set in this study has advantages over such conventional methods. It has to be, however, noted that the production of CGs through the simulation in this study is not a substitution for conventional field observations. Data collection in fields, simulation and ground truthing are all supplementary to each other in many scientific disciplines. IT intensive archaeology such as this study is not an exception.

Bayon in Cambodia
The Bayon in Cambodia is a Buddhist temple set up in the ninth century. Several modifications and expansions have been applied to the ruin in the past, and it is a three-dimensionally complex structure today. The dimension of the site is approximately 150 m by 150 m, and has about fifty towers forming a stone-built maze. Structural collapses and surface deterioration are conspicuous on this site, and efforts have been made by both Cambodian and foreign teams for conservation and restoration. Uneven distribution of heat over the ruin is considered one of the causes for structural collapses (e.g. JSA, 1998, p334). More concretely, heat brought by the sunlight might be resulting in thermal expansion of particular stone blocks and joints, hence, imbalance of structures.

Figure 1. Bayon on a sunny day

3D Model of Bayon
Construction of a 3D model of the ruin of Bayon was carried out by using a 3D modeling software. The basis for the modeling process was the plan made by the French institute, Ecole Francaise D'extreme-Orient. This plan was produced through field surveys in the 1920s and 1960s, and contained outlines and height information. The plan was first scanned, then, lines on it were digitised on screen. The 3D modeling software used for this study extruded towers, terraces and other structures to the heights as indicated by this plan.

The number of primitive 3D shapes produced by the above procedure was more than 1200. Every effort was made to make the 3D model as precise as possible. The following two points, however, have to be realised. Firstly, the plan on which the 3D model was based was about 50 years old. It is not necessarily an accurate illustration of the status quo of the ruin of Bayon. Secondly, details of the ruin such as fine relief engravings were omitted. In the production process of the 3D model, features less than 1m had to be neglected. The "spatial resolution" that the model had in the end seemed approximately same to, or slightly less than, that of standard aerial photographs taken for photogrammetric purposes. These two points, however, do not have to be regarded as defects. The age of the plan suggests that the 3D model produced from it can be taken as a reconstructed model showing the state in 1960s rather than the state in 2000s representing the effects of insolation for many decades. The lack of details economizes the available computing power so that production of many scenes having different viewing angles become possible.

Figure 2. 3D model of Bayon

Positioning 3D Model on 3D Globe
Geometrical relationship among the Sun, object and its location on the Earth produces shadows. The 3D model of the Bayon produced in the previous section had to be placed at its precise location on a model of the Earth. Geo-coordinates to determine the position and orientation were derived from a field survey done in 1998 (JSA, 1999, pp145 - 151). This survey was carried out by using Real Time Kinematic GPS, and the position data set acquired was expected to have a high accuracy. Some of the geo-coordinates used for positioning and orientation of the 3D model in this study were as follows.

**Table 1. Examples of geo-coordinates on the site**
Description	Latitude (DD MM SS)	Longitude (DD MM SS)
The middle point between the main entrance and the inner structure block	13 26 28.44 N	103 51 33.74 E
The north-west corner of the outermost periphery	13 26 30.05 N	103 51 29.71 E
The southern edge of the central tower block	13 26 27.99 N	103 51 31.99 E
The north-east edge of the central tower block	13 26 28.83 N	103 51 32.27 E

Variations in ground elevation among these four points were, according to the data set, within the range of 1m. Taking the "spatial resolution" of the 3D model into account, the ground on which the ruin was situated was approximated to a flatland.

The landscape simulation software chosen for positioning of the 3D model and consequent simulations in this study used the normal sphere of which radius is 6,370,997m as the Earth model. The spherical model allowed positioning of the 3D model by using the latitude / longitude coordinates above. The changing movement of the Sun over a year was preset in the software, and it became possible to reproduce the condition of insolation and shadows over the site of Bayon at any given date and time of a year after the placement of the 3D model on the 3D globe.

Simulations and Visualisations
On the basis of the 3D model set at the position on the 3D globe, several experiments were conducted. They can be divided into two groups. One of those is the experiments for verification of the 3D model and its positioning, and the other is the experiments for inductive and deductive inferences to formulate a future conservation strategy.

Comparison with Aerial Photograph
To verify geometry of the 3D model itself as well as its relationship to the Earth and the Sun, a comparison was made between an aerial photograph taken from a camera installed on a blimp and a computer generated scene having the approximately same viewpoint. The aerial photograph was taken at 9:07 a.m. on March 12th, 1999 (Madhavan, 1999). The altitude of the blimp, hence, the camera was about 110m above the ground. In the landscape simulation software, the same lighting condition was reproduced, and a corresponding viewpoint was identified. As shown in figure 3, shapes and sizes of shadows are identical on the photograph and the computer generated image. This comparison also illustrates the degree of generalisation in the modeling process mentioned earlier.

Figure 3. Aerial photograph compared to simulated image
Note: The white line on the aerial photograph is a captive rope of the blimp.

Comparison with Images Recorded by Thermography
A series of images recorded by a thermography were compared to a series of computer generated scenes having the corresponding field of view. The thermal images were recorded as a part of a field work in 1999 (Keio University, 1999). They indicated that roofs of galleries and terraces had kept having high temperatures. On the computer generated images, the same parts of the ruin were illuminated by the sunlight for prolonged periods. A point to be noticed is that some parts kept relatively high temperatures on the thermal images even when the corresponding simulated images indicated that those parts were already covered by shadows. This inconsistency is suggestive for future study to gather more information on heat capacity held by each component of the ruin.

Comprehension of Seasonal Changes of Isolation
Distribution of insolation over the site for a year was estimated by production of hourly overhead images for four particular days, namely, two equinox days, the Summer Solstice and the Winter Solstice. At the location of the Bayon, the date for those four days were March 21st, September 19th, July 4th and December 19th. The previous experiment implied that the heat held by a particular part of the ruin was roughly proportional to the length of insolation for that part. This was applied to the whole of site to enable a synoptic observation in this experiment. The viewpoint for this experiment was set to 400m above the ground to include every part of the site. Orthorectification was applied to the rendering phase of the image generation to make areas of polygons as accurate as possible. Duration of the daytime in those four days was within the range between 5:00 am and 6:00 pm. One orthorectified overhead image was produced for every hour for the daytime of each of the four days, which resulted in 4 sets of 14 images. The landscape simulation software used in this study output them as 8 bit unsigned images. They were exported to an image processing software for remotely sensed data, and the 8 bit unsigned values from 14 images were summed up in a 16 bit unsigned channel. The 16 bit unsigned pixel depth was re-scaled to 8 bit unsigned for display purposes. In the end, four images indicating the total of insolation on the vernal equinox, the summer solstice, the autumnal equinox and the winter solstice came into existence as shown below.

Figure 5. Insolatin on Vernal Equinox, Summer Solstice,
Autumnal Equinox, Winter Solstice (From left to right)
N.B. Areas of shadows change over seasons.

To characterise insolation in summer and winter, a false colour composite was made by assigning the summer solstice image to red, the vernal equinox to green and the winter solstice to blue. Furthermore, a classification algorithm was applied to the four 16 bit unsigned images. The particular procedure used was the K-means (minimum distance) clustering. The false colour composite and classification image are shown below.

Figure 6. False Colour Composite (see text) and Figure 7. Classification image

The images derived from this experiment denote that the Sun's movement along the east-west line shifts between north and south according to the season. This results in seasonal change of locations of shadow. More concretely, the north side of a wall laying between east and west, for example, is heated for a long period in summer, while the south side of wall in winter. On other types of structures, the balance between insolation and shadows may exhibit a more complex pattern.

Scatter Plot Analysis
A scatter plot was produced by assigning the Summer Solstice image to the X axis and the Winter Solstice image to the Y axis. The two axes were highly correlated to each other. The plot were dominated by two sets of rectangulars. The set of rectangulars close to the bottom left corner represent shadows on the ground, and the other larger set of rectangulars represent shadows within the ruin itself. An examination on domains outside of the latter (as annotated on the figure) revealed the following tendencies: Domain 1) Areas having high pixel values on the Summer Solstice image and various pixel values on the Winter Solstice image. Their distributions were, interestingly, only on northern edges of surfaces at more than a certain height such as tops of towers;

Domain 2) Areas having high pixel values on the Winter Solstice image and various pixel values on the Summer Solstice image. Their distributions were only on southern edges of surfaces at more than a certain height;

Domain 3) Pixels in this domain occupy the west to the northwest edge of the top of the central tower; and Domain 4) Pixels in this domain occupy the west to south edge of the top of the central tower.

Figure 7. Scatter Plot

The observations above can be further linked to analyses specific to particular components of the ruin on perspective images.

Conclusion
This study illustrated the potential that computer visualisation / simulation had for archaeological researches. The simulated images of the Bayon showing insolation and shadows of the Bayon was verified and made a number of useful implications. In fact, they showed what field observations can never show. It is, however, necessary to recognise the necessity of conventional field observations and data gathering practices. Without them, simulation / visualisation cannot be carried out and verified. Field works and simulations / visualisations are supplementary to each other. One inspires the other, and synergy between them will push researches forward. The future research scope from this attempt could include not just insolation but other variables and phenomena such as precipitation, humidity and land subsidence.

References

Ecole Francaise D'extreme-Oreint, 1967. Le Bayon Histoire Architechturale du Temple. Librairie Adrien-Maisonneuve, Paris.
Japanese Government Team for Safeguarding Angkor, 1998. Annual Report on the Technical Survey of Angkor Monument. Japan International Corporation Center, Tokyo.
Japanese Government Team for Safeguarding Angkor, 1999. Annual Report on the Technical Survey of Angkor Monument. Japan International Corporation Center, Tokyo.
Madhavan, B., 1999. Development of Large-scale Digital Orthographs from Balloon-Borne Aerial Photos of Archaeological ruins in Angkor, Cambodia. Internal research material, COE, Keio University, Japan.

Acknowledgements
We would like to acknowledge the JSA Project for supporting us in the data acquisition phase.