The style is closely connected to the visual representation. It specifies how the visualization is to be applied to the object. In general, the style direction can be seen as a group of visualization rules. The only difference is that the rules and values are usually already linked to a form of data interpretation. One extreme is photorealism. The use of a photorealistic style (fig. 5) can denote high certainty since the actual geometry and appearance are based on actual measured data. Those renderings usually have a high level of detail and are expected to be seen by the public. Consequently, a lot of discussion has emerged about its use and ability to deliver uncertainty (Wittur 2013, 48). The most common belief is that this method is unsuitable for the academic environment (Olivito and Taccola 2004, 181). The first years of virtual archaeology were characterized by a trend of photorealistic renderings (Strothotte 1999, 37; Sifniotis et al. 2006, 1). However, following increasingly intense discussion, the use of this style peaked off again (Olivito and Taccola 2014, 182). According to Reimersdahl, photorealistic renderings should always be the aim, as long as they are not based upon assumptions (Reimersdahl et al. 2008, 147). Hence, a photorealistic style can be used in a mixed context with other style directions (Bakker et al. 2003, 161; Olivito and Taccola 2014, 182). However, a purely photorealistic rendering does not distinguish between real and interpreted structures (Kantner 2000, 47), as is required in many charters. Therefore, this style usually implies high certainty of the displayed object and is not suitable to display uncertainty (Kensek et
al. 2004, 183; Strothotte 1999, 36). Bakker et al. summarize it as follows: “Truth and credibility as double ambition: reconstruction of the built past experience and dilemmas.”
However, its use of visualization is more artistic than scientific (Bakker et al. 2003, 5ff). To sum up, photorealism is avoided for interpreted data in most recent scientific reconstructions.
Figure 5: Example of a photorealistic rendering of an ancient building structure. The illustration refers to the investigations of the Agora of Segesta, for which several virtualization methods were tested (Olivito and Tavvola 2014, 176 and 181).
32 Simple (fig. 6) or abstract styles are more used for uncertain structures (Reimersdahl et al. 2008, 147). The objects are usually default to simple geometrical forms and textures. Furthermore, the amount of details shown is decreased. Since uncertain parts are not modeled anymore, the certainty of the whole object increases. On the other hand, the accuracy will decrease (Alusik and Sovarova 2015, 438). Similar approaches are known to be followed by architects in which only the rough outline is shown first (Strothotte 1999, 36). Moreover, fewer details also mean fewer polygons, which enables a better use in real time rendering applications7 and the connection of data and model (Fanini and Ferdani 2012, 112). As a result, schematic renderings can emphasize uncertain areas (Murgatroyd 2008). Thus, they indicate the speculative nature of the objects (Zuk et al. 2005, 3). Furthermore, they are easy to apply and are therefore often used. Frischer and Stinson use lighter and less saturated colors to indicate uncertainty in their reconstruction of a villa. Delicate details are neglected completely to emphasize the uncertainty (Frischer and Stinson 2007, 66). Strothotte completely avoids details in uncertain areas, which facilitates easy access to other methods to enrich the result (Strothotte et al. 1999, 36f).
Figure 6: Example of a schematic rendering of an ancient building structure. Hereby, the render format emphasizes the point of interest due to higher details. The picture itself is of a reconstruction of the Palace of Margaret (Brusaporci 2017, 130).
Mixing both of the previous approaches (fig. 7) is defined as overlay (Schwerin et al. 2016, 212). The value of this style lies in the ability to distinguish real and interpreted data quickly and easily (Schwerin et al. 2016, 212; Olivito and Taccola 2014, 182). For example, Schwerin uses laser scan data that is incorporated into the schematic reconstruction of the temples. Furthermore, both datasets can be easily enriched by further information, such as different degrees of uncertainty (Schwerin et al. 2016, 211f). However, in parts
7 Real time rendering applications are usually the end product of game engines. They enable one
to view and manipulate a rendered 3D model in real time. This means that this type of application is also strongly tied to computer performance. More well-known uses in archaeology are, for example, integrated models with connected databases or a 3D GIS.
33 with high certainty she also increases the level of detail and segmentation to provide a better scale and organization (Schwerin et al. 2016, 211). Likewise, Olivitio and Taccola mix their styles for their reconstruction of the agora of Segesta (Olivito and Taccola 2014, 179 and 181). It works by placing simple geometrical forms over the reality-based model to indicate possible structures that no longer exist. However, it is not possible to encode floating uncertainty with this method (Olivito and Taccola 2014, 182). It more or less resembles the Boolean uncertainty or crisp sets.
Figure 7: Example of an overlay rendering, which shows a reality-based reconstruction as a photorealistic model, combined with a schematic overlay for the hypothetical parts. The model itself refers to the agora of Segesta (Olivito and Tavvola 2014, 182).
As well as the photorealistic, abstract or mixed approaches, changes in the data itself can also give hints of uncertainty. One way is the organization of the data. This can happen with layers or segments. The exact realization is described below.