Reaching new heights

The built environment consists of structures built by humans in or on the earth, rearrangements of the surface of the earth, bridges over the surface of the earth, and holes dug into or through the earth. Buildings, roads, tunnels, pipes, dams, mines, airports and harbours are examples of things in the built environment. â©

CityGML 2.0 is an OGC standard for the exchange of linked semantic and geometric models of the built environment. CityGML, either directly or with the built-in application domain extension (ADE) feature, can also represent the solid earth, hydrosphere, cryosphere and atmosphere in which the built environment is embedded. This article describes the abstract model that underlies the CityGML encoding.â©

The CityGML encoding standard defines an encoding of an underlying conceptual model of the built environment. This conceptual model is described in the OGC CityGML 2.0 standard but is not separately defined by the standard. It has been informally called the ‘City Model’ or ‘CityModel’ by some of the original developers of what is now called CityGML, and that convention is followed in the rest of this article. â©

Separating the encoding from the concepts has several benefits. The first is simply to isolate the conceptual model from the details of physical storage. Every CityModel application has an internal representation but it is neither necessary nor useful to know how the various elements are stored, only that they conform to the requirements of entities, properties and relationships that must be represented.â©

The CityModel extends traditional GIS models of the built environment in two ways:â©

• It accurately models the full 3D geometry, material properties and appearance of objects in the real world. â©
• It adds semantic content that makes it possible for an application to understand modelled objects from the point of view of people. â©

The CityModel extends geographic representations from the space of 2D maps of large areas, where the height dimension does not play an important role, to the 3D space of detailed building models, where intricate details of single buildings or architectural designs dominate. Most importantly, the CityModel addresses applications where the intrinsic 3D nature of the built environment comes to the fore.â©

A natural fitâ©

There are some compelling arguments for creating 3D representations of buildings and infrastructure:â©

Entities in the built environment are inherently 3D and if they are not directly represented this way, the true spatial relationships between those entities cannot be recovered.â©

3D representation gives a direct connection to people’s cognitive processing capabilities, making it possible to grasp complex relationships at a glance.â©

Because a 3D representation models things ‘the way they are’, developing a new application on top of such a representation does not require special data modelling information, such as layers and relationships between coordinate systems – everything is in one integrated whole.â©

Errors and inconsistencies in geometry are much easier to find, display and correct in an integrated 3D model – again, because all the information is in one place and the relationships are the same 3D relationships as in the physical world.â©

Even with these fundamental arguments for the full and direct representation of 3D geometry in spatial models, it helps to consider some of the obvious shortcomings of lower-dimensional representations. Simply adding a height coordinate to a two-dimensional representation, making a ‘2.5D’ model, only changes the shape of the surface in which the modelled elements are embedded. There is no straightforward way to model vertical relationships between objects located in approximately the same horizontal position. Bridges and tunnels alter the topology of the model and the presence of the resulting handles or wormholes alter spatial relationships in ways that are fundamental for human navigation. Built structures are, in many cases, essentially artificial caves, where rooms and other concavities are the key components of the structure. And many parts of the natural surface of the earth, such as vertical cliffs, overhangs and caves, do not lend themselves to the simple description of location by reference to a horizontal position combined with a single vertical height. It is only when people’s 3D perception and experience is modelled directly that all the important spatial relationships can be easily derived from a spatial model.â©

Enabling navigation, analysis and simulationâ©

Implementations of the CityModel, such as OGC CityGML 2.0, have the expressiveness to model physical form, behaviour, affordances and categories, as well as properties that determine a model component’s compliance with the laws of physics. This rich representation enables them to be used as substitutes for real or realistic built environments for each of the application categories. For example, a blast-effects simulation can be relied on to give accurate predictions of damage from the controlled detonation of an unexploded bomb buried in an old building because most of the important properties are represented.â©

Having modelled three geometric dimensions provides the advantage of direct representation of spatial relationships and geometric details. But that is not enough to make a model useful for any purpose other than visualisation because semantics are required to understand the nature and purpose of built objects. A complete model represents a collection of natural and built things with the following characteristics:â©

• Physical form – geometry and material properties. â©
• Physics – universal rules for all objects based only on physical form.â©
• Behaviour – rules for specific object instances and interactions with other objects.â©
• Affordances – rules that describe how an object may be engaged by another object in a purposeful activity.â©
• Categories – groups of things that have similar behaviour or affordances.â©

All of these aspects must be represented to fully support navigation, analysis and simulation applications. CityModel groups entities in semantic categories based on behaviour and affordances as well as capturing their physical form. â©

For example, a CityModel distinguishes the roof, door, or window of a building as a special building part with specific behaviour and affordances. The creator of a model includes this semantic information to enable applications to treat the roof, door, or window in a special way. For a roof, this might be analysis of roof building parts for their solar heating or electric energy generating potential. For a door or window, it might be in support of navigation. â©

Semantic information cannot be determined from physical form alone – or in most cases at all – and must be added to indicate how people in a particular culture and setting will treat an object with specific semantic properties. The data that describe the material and physical forms of objects are mostly independent of information about intended use, particularly affordances and behaviour. â©

In short, the CityModel addresses the complexities of the built environment in a direct and powerful way that supports a wide range of applications.â©

The CityModel addresses the complexities of the built environment in a direct and powerful wayâ©

Carl Stephen Smyth is co-chair of the OGC CityGML standards working group