Impact Statement
This research introduces a novel standards-based methodology for validating integrated 3D semantic city model snapshots that represent concurrent and successive urban evolution scenarios. By leveraging Semantic Web Ontologies and logical rules, the study enables the temporal validation of these urban evolution scenarios. A reproducible proof-of-concept test suite and example queries demonstrate practical applications, including constructing a 3D web application to visualize and analyse urban changes. This innovative approach empowers urban planners, policymakers, and researchers with robust tools to model and understand complex urban transformations, facilitating data-driven decision-making. The methodology’s emphasis on interoperability and real-world applicability positions it as a significant contribution to advancing urban evolution analysis.
1. Introduction
In recent years, 3D semantic city models have become more frequently adopted as tools to understand long-term changes in urban landscapes. Here, temporal information can be used to represent the historical evolution of a city as well as its imagined future. This information could be related to changes in the structure, geometry, or semantic information used to describe city snapshots across different instances of time. In addition, these changes could be analysed at different scales, from the changes of individual urban objects over time to the development of entire districts and cities (Chaturvedi and Kolbe, Reference Chaturvedi and Kolbe2019).
These 4D (3D + Time) city models may also be extended to represent concurrent or competing representations of city evolution. For instance, historians may discuss different hypotheses of how cities evolved in their studies to understand city urbanization (Kutzner et al., Reference Kutzner, Chaturvedi and Kolbe2020; Samuel et al., Reference Samuel, Servigne and Gesquière2020). Similarly, urban planners may propose new buildings and districts in existing zones or territories when designing new construction projects (Chaturvedi and Kolbe, Reference Chaturvedi and Kolbe2019; Nguyen and Kolbe, Reference Nguyen and Kolbe2021). To this end, the study of urban evolution and the associated hypotheses related to changes in urban objects or features requires the complex integration of data from multiple sources. This process of combining data from different sources requires data validation to ensure the data can be holistically exploited to provide users with more complete views of city evolution.
The major contribution of this article is a proposed novel standards-based methodology for validating integrated 4D urban data as concurrent scenarios of urban evolution (Section 3). Additionally, we present the following contributions created thanks to this methodology:
-
• Formalization of rules for a standardized 4D urban data conceptual model for representing concurrent viewpoints of urban changes (Section 3.3).
-
• Logical validation of urban scenarios of evolution (Section 3.4) using a reproducible proof-of-concept test suite.
-
• Formulation of queries and data views for retrieving concurrent points of view of evolving cities using the proposed data model and generated data (Section 4).
The article is presented as follows: Section 2 details related works in modeling, integrating and logically validating 4D urban data; Section 3 details the aforementioned methodology for reusing standards in 4D urban data integration and eventual data validation; Section 4 presents the results of integrating and validating standardized open urban data snapshots of the Gratte-Ciel neighborhood of Lyon, France with this methodology; Finally, Section 5 offers our perspective future works and concludes the article.
2. Related works
The evolving 3D urban data used to create 4D city models is frequently complex. In this context, a process of combining data from multiple sources (i.e., data integration Tran et al. (Reference Tran, Plumejeaud, Bouju and Bretagnolle2016)) can be used to construct these 4D city models and provide users with more complete views of city evolution. This section will discuss existing approaches for integrating these data at the data instance level, existing approaches for representing city evolution (both real and imagined), and previous works in logical 4D data validation. This section also positions the contributions of this article among these works.
2.1. Representing urban data evolution
Many approaches have been proposed for portraying different digital representations of the same real-world city object (over time). In geospatial data, distinctions between these representations are often identity-based, i.e., where identity is the characteristic that distinguishes one object from another over time, and where a change to the object may or may not change its identity (Hornsby and Egenhofer, Reference Hornsby and Egenhofer2000; Stefani et al., Reference Stefani, De Luca, Véron and Florenzano2008; Van Ruymbeke et al., Reference Van Ruymbeke, Carré, Delfosse, Pfeiffer and Billen2015).
In this context, city object changes are often either semantic and/or geometric. Semantic changes are often straightforward to identify, but may be challenging to categorize depending on the user’s perspective, application, or use-case (Nguyen and Kolbe, Reference Nguyen and Kolbe2021). Certain geometric or spatial changes may also complicate the definition of an object’s identity, such as the splitting of buildings into two or more buildings or the fusion of multiple distinct buildings (Stefani et al., Reference Stefani, De Luca, Véron and Florenzano2008). Furthermore, one major requirement for identity-based changes is to ensure the use of unique identifiers for a given city object. This is not guaranteed as city object identifiers may be different between independent data providers, may be created through the use of different data capture techniques (e.g., by using different LIDAR, arial photography, and/or photogrammmery technologies). Even the same data provider may use different identifiers for the same urban feature over time due to various administrative or technical reasons (Nguyen and Kolbe, Reference Nguyen and Kolbe2021).
Several works have been proposed to extend these representations to facilitate the comparison of different historical or future evolutions of urban spaces (Pfeiffer et al., Reference Pfeiffer, Carré, Delfosse, Hallot and Billen2013; Chaturvedi et al., Reference Chaturvedi, Smyth, Gesquière, Kutzner, Kolbe and Abdul-Rahman2017). Notably, the versioning module in CityGML 3.0 (Kutzner et al., Reference Kutzner, Chaturvedi and Kolbe2020) has been proposed to provide a standard for representing multiple versions of a city object and, at a larger scale, multiple versions of a city or district. Extensions to these standards have been proposed that permit the comparison of alternate or concurrent scenarios of urban evolution, such as the Workspace model proposed in Samuel et al. (Reference Samuel, Servigne and Gesquière2020); Vinasco-Alvarez et al. (Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024f). The specific concepts used in this article are discussed in detail in Section 3.1
2.2. 4D urban data integration
In the domain of 4D data integration, approaches may propose applying feature matching or entity resolution techniques that compare the spatial and/or semantic characteristics of city objects to determine when two digital representations are modeling the same real-world object (Sehgal et al., Reference Sehgal, Getoor and Viechnicki2006; Pédrinis et al., Reference Pédrinis, Morel and Gesquière2015; Bernard, Reference Bernard2019; Balsebre et al., Reference Balsebre, Yao, Cong and Hai2022). These characteristics may include a representation’s geospatial coordinates (in 2D or 3D) and/or non-spatial attributes such as unstructured text, identifiers, administrative information, and so forth By incorporating the temporal dimension into their proposed 3D city models, these approaches can provide a continuous view of urban evolution instead of isolated, interrupted snapshots.
These techniques can be applied to stitch together static digital city models from different points in time to provide digital views of how their real-world counterparts have changed over time (Nguyen et al., Reference Nguyen, Yao and Kolbe2017; Jaillot et al., Reference Jaillot, Servigne and Gesquière2020). These approaches can also be used to integrate data based on frequently used 3D urban data standards such as Industry Foundation Classes (IFC) and CityGML (buildingSMART, 2024; Kutzner et al., Reference Kutzner, Chaturvedi and Kolbe2020; Diakite and Zlatanova, Reference Diakite and Zlatanova2020; Liu et al., Reference Liu, Wang, Wright, Cheng, Li and Liu2017; Liu et al., Reference Liu, Ellul and Swiderska2021; Hijazi et al., Reference Hijazi, Krauth, Donaubauer and Kolbe2020).
2.3. Semantic Web technologies for urban data integration
New research works have turned to Semantic Web data standards such as the World Wide Web Consortium’s (W3C) Semantic Web technology suite (Lassila et al., Reference Lassila, Hendler and Berners-Lee2001; W3C, 2025) have been adopted to facilitate urban data modeling and integration on the web. Here, expressive and formal modeling languages are proposed, such as Ontology Web Language (OWL) family of languages that can be used to create models that permit the description of complex data models while being machine-readable through flexible graph data formats such as Resource Description Framework (RDF) (Malinverni et al., Reference Malinverni, Naticchia, Lerma Garcia, Gorreja, Lopez Uriarte and Di Stefano2020). RDF permits the definition of knowledge graphs through triple statements that are composed of a subject, predicate, and object (Figure 1). OWL builds on this format to permit the creation of ontological knowledge graphs that are sufficiently formal to be considered logical theories and can be used to logically infer new information (Uschold and Gruninger, Reference Uschold and Gruninger2004). These ontologies can store the concepts, relationships, and constraints of a data model (i.e., the collection of terminological knowledge or TBox of the ontology) as well as information regarding the data instances of these concepts (i.e., the collection of assertions or ABox of the ontology) (Figure 1). Several data transformation approaches have been proposed to integrate 2D, 3D, and 4D urban data using these data formats (Kyzirakos et al., Reference Kyzirakos, Savva, Vlachopoulos, Vasileiou, Karalis, Koubarakis and Manegold2018; Tran et al., Reference Tran, Aussenac-Gilles, Comparot and Trojahn2020; Chadzynski et al., Reference Chadzynski, Krdzavac, Farazi, Lim, Li, Grisiute, Herthogs, von Richthofen, Cairns and Kraft2021; Hor and Sohn, Reference Hor and Sohn2021; Usmani et al., Reference Usmani, Jadidi and Sohn2021; Vinasco-Alvarez et al., Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2021).
Figure 1. An illustration of the RDF triple structure (top left). An example of RDF triples in the Turtle syntax (bottom left). An illustration of the RDF triple example (right).
Dynamic knowledge graphs (Pujara and Getoor, Reference Pujara and Getoor2014; Trivedi et al., Reference Trivedi, Dai, Wang, Song, Precup and Teh2017; Aparicio et al., Reference Aparicio, Arsenio, Santos and Henriques2024) have been explored to understand the evolution of knowledge graphs over time and to integrate data from different heterogeneous sources with different levels of data accuracy and completeness. In case of urban evolution studies, urban digital twins (Grieves, Reference Grieves2014; Batty, Reference Batty2018) have been explored and implemented as a method for integrating and maintaining spatio-temporal urban data. These approaches propose maintaining a digital representation (or “twin”) of an evolving urban environment up to date through a “twinning” processes. This process often involves applying changes to the digital twin—either periodically or continuously—with real-world data from the physical twin. Furthermore, dynamic knowledge graphs such as The World Avatar (TWA) have been implemented as approaches to integrate cross-domain digital twins by levraging Semantic Web technologies for interoperability (Eibeck et al., Reference Eibeck, Lim and Kraft2019; Akroyd et al., Reference Akroyd, Mosbach, Bhave and Kraft2021; Chadzynski et al., Reference Chadzynski, Krdzavac, Farazi, Lim, Li, Grisiute, Herthogs, von Richthofen, Cairns and Kraft2021). In this domain, the digital twin may be refered to as the ‘base world’ and the twinning process is effectuated through a set of software agents that can update, retrieve information from, and restructure the “base world.” Some dynamic knowledge graphs also permit the modeling of hypothetical scenarios, such as the Parallel World Framework (Eibeck et al., Reference Eibeck, Chadzynski, Lim, Aditya, Ong, Devanand, Karmakar, Mosbach, Lau and Karimi2020), which proposes storing secondary “parallel base-world for simulating changes to the primary base-world.”
While these approaches have been proven to be effective for integrating heterogeneous urban data, the validation of this integration remains necessary and complex for ensuring a high-quality representation of the urban environment. Here, the formal nature of languages such as OWL lends themselves to validation approaches using logical rules.
2.4. Logical consistency validation of 4D urban data
Many validation approaches exist for urban data, each implemented to identify different aspects of data quality in 4D urban data applications. For example, 4D geospatial data applications using Semantic Web technologies may use data model constraints and rules formalized in languages such as OWL and SWRL (Semantic Web Rule Language) to verify that their integrated data is logically consistent (referred to as consistency checking) (Tran et al., Reference Tran, Plumejeaud, Bouju and Bretagnolle2016; Batsakis et al., Reference Batsakis, Petrakis, Tachmazidis and Antoniou2017; Samuel et al., Reference Samuel, Servigne and Gesquière2020). This verification step is especially important in applications that propose the automatic and logical inference of new information from existing data using reasoners (Tran et al., Reference Tran, Aussenac-Gilles, Comparot and Trojahn2020) or spatio-temporal analysis of territorial events (Batsakis and Petrakis, Reference Batsakis, Petrakis, Bassiliades, Governatori and Paschke2011; Harbelot, Reference Harbelot2015; Tran et al., Reference Tran, Plumejeaud, Bouju and Bretagnolle2016; Bernard, Reference Bernard2019).
In the case of SWRL, its language is based on several sublanguages from OWL and Rule Markup Language that permit Horn-like rules to be written in OWL ontologies (Horrocks et al., Reference Horrocks, Patel-Schneider, Boley, Tabet, Grosof and Dean2004). These rules are composed of an antecedent (or the rule body) and a consequent (or the rule head). In SWRL, the antecedent and the consequent are sets of atoms such that if the conjunction of all the atoms of the antecedent holds true, then the conjunction of all the atoms of the consequent must hold true.
$$ antecedent\to consequent $$
For example, the rule if
$ a $
is a building, it must have a height greater than 0, can be written as:
building(?a) -> height(?a,? h), greaterThan(?h, 0).
Inferencing can be performed by using reasoners such as Pellet, HermiT, or Fact++ (Tsarkov and Horrocks, Reference Tsarkov, Horrocks, Furbach and Shankar2006; Sirin et al., Reference Sirin, Parsia, Grau, Kalyanpur and Katz2007; Glimm et al., Reference Glimm, Horrocks, Motik, Stoilos and Wang2014), through queries using the SPARQL query language (Buil Aranda et al., Reference Buil Aranda, Olivier, Das, Feigenbaum, Gearon, Glimm, Harris, Herman, Humfrey, Michaelis, Ogbuji, Perry, Passant, Polleres, Prud’hommeaux, Seaborne and Williams2013; Tran et al., Reference Tran, Plumejeaud, Bouju and Bretagnolle2016), or with tools that can parse SWRL rules with SQWRL, such as the SWRLTab plugin in the Protégé ontology editor or OwlReady2 (O’Connor and Das, Reference O’Connor and Das2010; Musen, Reference Musen2015; Lamy, Reference Lamy2017). In the context of integrating open data conformant to international standards, approaches that rely on purely syntactic transformations—i.e., conversion of data format as opposed to semantics (Kutzner, Reference Kutzner2016)—may implement validation steps to ensure that the integrated data rests conformant to the initial standard and thus internationally interoperable.
For example, Chadzynski et al. (Reference Chadzynski, Krdzavac, Farazi, Lim, Li, Grisiute, Herthogs, von Richthofen, Cairns and Kraft2021) proposes a similar validation to preparing the CityGML 2.0 ontology proposed by the University of Geneva in Zalamea et al. (Reference Zalamea, Orshoven and Thérèse2016) for integration in TWA that illustrates the importance of this validation. This ontology was created using XSLT transformations of the CityGML 2.0 XML Schema and evaluated according to several Semantic Web metrics proposed in Hlomani and Stacey (Reference Hlomani and Stacey2014). It was determined that the ontology was inconsistent (the ontology contained logical contradictions), insufficiently concise (the ontology contains irrelevant or redundant semantic elements concerning a domain of knowledge), and incomplete (the ontology is missing appropriate semantic elements to sufficiently answer questions in a domain of knowledge) in the context of their use-case. The ontology was thus manually corrected to enable reasoning over it, simplified by removing unnecessary classes and properties leftover from the XML Schema in OWL, and extended with the help of domain experts as necessary to cover missing concepts required by the use-case.
2.5. Summary
Taking these approaches into consideration illustrates how complex the 4D urban data integration process can be for providing users with more complete views of the urban landscape and its evolution. It also underlines the importance of data validation during these processes. However, while standards such as CityGML 3.0 do provide general constraints and abstract normative tests (Kolbe et al., Reference Kolbe, Kutzner, Smyth, Nagel, Roensdorf and Heazel2021, Annex A), they leave the interpretation and implementation of these constraints and tests to the users of the standard. Additionally, as standards themselves evolve over time, the proposed rules for testing the conformance of data to the standard may also require periodic updates to work with the latest versions of the standard. Therefore, manual intervention is required to translate abstract tests, constraints, and rules to more machine-readable languages and formats.
Furthermore, research in the domain of concurrent scenarios of urban evolution is relatively new. As far as we know, there is currently a lack of existing works in the domain of validation of standardized data used to create these urban scenarios of evolution. Building on these proposed works, the next section presents our methodology for validating integrated heterogeneous data in applications that require temporal and analysis of urban phenomena.
3. Methodology
The methodology of our approach is decomposed into 3 main activities (Figure 2) as detailed in this section: formalize an extension for OWL-Time, formalize rules for validating urban evolution, and use the results of these activities to validate urban evolution scenarios. To validate whether a proposed urban evolution scenario is temporally logical, we first propose an extension of the OWL-Time model, presented in Section 3.2 for inferring temporal relations between temporal entities (such as instants and intervals). This extension is based on the temporal relations originally defined in Harbelot (Reference Harbelot2015) for temporal intervals. Subsequently, we adapt the rules proposed in Samuel et al. (Reference Samuel, Servigne and Gesquière2020) to infer when scenarios of urban evolution and their versioned entities are temporally valid. These rules, originally formalized in a description logic (DL) language, are translated to SWRL (presented in Section 3.3) following the practices described in Vinasco-Alvarez (Reference Vinasco-Alvarez2023, Chapter 5). Additionally, as the CityGML standard has evolved since the original proposal of these DL-rules, they are updated to conform to the newer CityGML 3.0 Conceptual model.
Figure 2. UML activity diagram of proposed methodology. External artifacts or data sources are located on the bottom of the diagram while artifacts created from an activity are placed in the center.
Finally, these contributions are used to validate existing scenarios of urban evolution with a reproducible proof-of-concept test suite presented in Section 3.4. Unlike the first two activities, which manually produce artifacts in Semantic Web formats, this final activity takes advantage of this fact to fully automate the data validation process.
3.1. Preliminaries for representing scenarios of urban evolution
As discussed in Section 2.1, many works define concepts for representing urban evolution. This section presents the concepts used in this article for this purpose, as defined by CityGML 3.0 and the Workspace model. Feature with a lifespan (or a “versioned” feature): a representation of a real-world phenomenon with a specific existence time. Version: a set of versioned features such that the existence time of the version occurs during the existence time of all of its versioned feature members (i.e., the interval of time when the feature members are stable, i.e., it does not undergo any modification). Version Transition: a set of Transactions or individual changes between the corresponding versioned features of two consecutive Versions (e.g., whether a feature is added, modified, or removed from the Version); the existence time of a Version Transition occurs between the existence time of its corresponding Versions. In addition, a string of sequential alternating versions and version transitions can be composed into a Scenario for representing a specific evolution of the city, hypothetical or otherwise. These scenarios exist either within either a Consensus Space or Proposition Space—where concensus space contains the (official or generally accepted) scenario (and all the versions and transitions) that represent the agreed upon real-world evolution of a city (or district), and proposition space contains all the scenarios that include hypothetical or imagined evolutions of the city. Finally, a Workspace is composed of a single proposition and consensus space and can be used to store the relevant representations of city evolution for a specific application or community of users. These concepts are exemplified in Figure 3.
Figure 3. Illustration of a scenario of evolution containing two city objects (
$ O1 $
,
$ O2 $
) and their properties (
$ {P}_1 $
–
$ {P}_4 $
) that change between versions (
$ {V}_0 $
–
$ {V}_7 $
) or “snapshots” of an urban area (shown above). These scenarios are also composed of Transitions (
$ {T}_0 $
–
$ {T}_7 $
) that contain the changes between each successive Version. These scenarios can intersect one of 2 abstract spaces (shown below): a consensus space containing an agreed upon real-world representation of urban evolution, and a proposition space containing 0 or more hypothetical or unverified scenarios of evolution (Samuel et al., Reference Samuel, Servigne and Gesquière2020).
Furthermore, two major temporal intervals or instants are often used for representing the evolution of cities and city objects:
-
• Existence time (or validity time) for representing when urban data and urban objects exist or are relevant in the real world
-
• Transaction time for recording when the data is stored, updated, or deleted from a data source (e.g., a file or database)
These are modeled with start and end timestamps, indicating the beginning and completion of said interval or instant (where in a temporal instant, the beginning and end are concurrent). Approaches, such as those that follow the European INSPIRE model (INSPIRE Drafting Team “Data Specifications”, 2014), support both of these representations, including the CityGML standard.
The rest of this article uses the namespace prefixes available in Table A1 of the Appendix section when referring to these concepts in the context of existing or proposed RDF/OWL identifiers. E.g., time:Instant references a temporal instant as defined in the OWL-Time ontology.
3.2. Extending OWL-Time for validating urban evolution scenarios
As a part of the first activity of our methodology (2), we propose an alternative interpretation of an existing extension of OWL-Time for modeling relationships between temporal intervals and instants (Harbelot, Reference Harbelot2015) (Figure 4). We use OWL-Time, since it is a recommended W3C standard for representing temporal relations and is a widely accepted temporal model for representing the basic temporal relations proposed in Allen (Reference Allen1984). As detailed in (Vinasco-Alvarez, Reference Vinasco-Alvarez2023, chapter 5), our proposal consists of 3 definitions of temporal instant-interval relations (time_ext:in, time_ext:during, time_ext:finishes) and 1 relation between instants (time_ext:equals). In addition to the temporal relations, we propose two object properties as subproperties of time:hasTime for distinguishing between the existence time (time_ext:hasExistenceTime) and transaction time (time_ext:hasTransactionTime) of versionned urban entities.
Figure 4. A synthesis from (Vinasco-Alvarez, Reference Vinasco-Alvarez2023) of the temporal relations as defined in OWL-Time according to Allen (Reference Allen1984) with our alternative interpretation of the Instant-Interval relations proposed in Harbelot (Reference Harbelot2015) (highlighted in orange).
To simplify the existing definition of these relations, we use a left-closed, right half-open interpretation of temporal intervals, whereas the original extension proposes a purely closed interpretation. For example, given two temporal intervals
$ {T}_1 $
and
$ {T}_2 $
, where
$ {T}_1 $
defined as
$ {T}_1=\left[{t}_{b1},{t}_{e1}\right),{T}_2=\left[{t}_{b2},{t}_{e2}\right) $
where
$ {t}_{b1} $
,
$ {t}_{e1} $
,
$ {t}_{b2} $
, and
$ {t}_{e2} $
are temporal instants, the temporal rule
$ meets $
can be defined as the following DL statement:
$$ {\displaystyle \begin{array}{r} Interval\left({T}_1\right)\hskip0.3em \wedge \hskip0.3em Interval\left({T}_2\right)\hskip0.3em \wedge \\ {} has End\left({T}_1,{t}_{e1}\right)\wedge \hskip0.3em has\ Beginning\left({T}_2,{t}_{b2}\right)\hskip0.3em \wedge \\ {} equal\left({t}_{e1},{t}_{b2}\right)\\ {}\to meets\left({T}_1,{T}_2\right)\end{array}} $$
This reduces the number of extended temporal relations from 4 to 3 and affects when instants and intervals meet and also permits a continuous representation of time. Under the previous interpretation, a discrete interpretation of time is required where
$ meets\left({T}_1,{T}_2\right) $
is true if
$ {t}_{e1}+1={t}_{b2} $
as defined in Samuel et al. (Reference Samuel, Servigne and Gesquière2020). Also, unlike Harbelot (Reference Harbelot2015) we use the vocabulary from OWL-Time to refer to temporal instants instead of points.
These relations are first formalized as object properties in OWL with constraints on their domain and range. For example, the instant-interval relation, time_ext:in, is defined as follows in the RDF Turtle syntax:
time_ext:in a owl:ObjectProperty;
rdfs:domain time:Instant;
rdfs:range time:ProperInterval;
rdfs:comment “If an instant T1 is in a proper interval …
The next section illustrates how we translate the DL representation of these rules to SWRL.
3.3. Rules for the temporal validation of urban evolution scenarios
Within the second methodology activity (2), we propose a translation of the DL rules from Samuel et al. (Reference Samuel, Servigne and Gesquière2020) to SWRL for representing concurrent scenarios of urban evolution. These rules can be divided into two subsets. The first ruleset is a set of basic rules for validating temporal entities as temporally consistent, largely based on the rules described in Allen (Reference Allen1984); Harbelot (Reference Harbelot2015) and is closely interoperable with OWL-Time. The second ruleset can be used specifically for validating urban evolution, as it is an extension of the CityGML 2.0 model rules for structuring versioning and Workspace ontologies. Figure 5 illustrates the proposed rule translation process in three steps.
Figure 5. 3-step DL to SWRL rule translation process. In our context, the process is applied to the DL rules proposed in Samuel et al. (Reference Samuel, Servigne and Gesquière2020) and the target ontology is composed of the “CityOWL” ontology network Vinasco-Alvarez et al. (Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024f) and the proposed OWL-Time extension (Section 3.2).
First, the class and property definitions proposed in Samuel et al. (Reference Samuel, Servigne and Gesquière2020) are updated to reflect the recent changes to CityGML between versions 2.0 and 3.0, such as the adoption of the Versioning extension (Chaturvedi and Kolbe, Reference Chaturvedi and Kolbe2019) as an official CityGML module. Additionally, Vinasco-Alvarez et al. (Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024f) proposes the “CityOWL” ontology network with OWL definitions for CityGML 3.0 and Workspace class hierarchies, and class and role restrictions. By reusing these ontologies, these declarations can be ignored in the translation of the original DL ruleset. For example, the Turtle definition of a Version is as follows (Vinasco-Alvarez et al., Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2022):
vers:Version a owl:Class;
rdfs:label “Version”@en;
rdfs:subClassOf.
[a owl:Restriction;
owl:allValuesFrom vers:ADEOfVersion;
owl:onProperty vers:Version.adeOfVersion],
[a owl:Restriction;
owl:allValuesFrom core:AbstractFeatureWithLifespan;
owl:onProperty vers:Version.versionMember],
[a owl:Restriction;
owl:allValuesFrom xsd:string;
owl:onProperty vers:Version.tag],
core:AbstractVersion;
skos:definition “Version represents a defined state of a city model consisting of the dedicated versions of all city object instances that belong to the respective city model version. Versions can have names, a description and can be labeled with an arbitrary number of user defined tags.”@en .
In the second step, the DL rules are rewritten to align with a concrete SWRL syntax that can be interpreted by a parser. This work uses the OwlReady2 SWRL syntax since it closely resembles the abstract SWRL syntax (Horrocks et al., Reference Horrocks, Patel-Schneider, Boley, Tabet, Grosof and Dean2004), and OwlReady2 has support for parsing and inferencing, and both SWRL and OWL (Lamy, Reference Lamy2017). This translation requires replacing conjunctions (
$ \wedge $
) with commas and removing the universal closure (
$ \forall $
) from the antecedent (as it is assumed in OwlReady2). For example, following the first two steps, the DL rule 33 from Samuel et al. (Reference Samuel, Servigne and Gesquière2020) states that “the time limits of [a] version transition must be greater than or equal to the end time of the previous version and the start time of the following version”:
$$ {\displaystyle \begin{array}{l}\forall vt\Big( hasPreviousVersion\left( vt,{vt}_p\right)\wedge \\ {}\hskip2em has\ New\ Version\left( vt,{vt}_n\right)\wedge \\ {}\hskip2em has\ Existence\ Time\left({vt}_p,{t}_1\right)\wedge \\ {}\hskip2em has\ Existence\ Time\left({vt}_n,{t}_2\right)\wedge \\ {}\hskip2em has\ Existence\ Time\left( vt,t\right)\Big)\\ {}\to meets\left({t}_1,t\right)\wedge meets\left(t,{t}_2\right)\end{array}} $$
This rule is rewritten as the following SWRL rule, taking into consideration “CityOWL” and the OWL-Time extension as defined in 3.2:
vers:VersionTransition.from(?vt, ?v1),
vers:VersionTransition.to(?vt, ?v2),
time_ext:hasExistenceTime(?v1, ?i1),
time_ext:hasExistenceTime(?vt, ?i2),
time_ext:hasExistenceTime(?v2, ?i3),
time:hasEnd(?i1, ?e1), time:inXSDDateTimeStamp(?e1, ?t1),
time:hasBeginning(?i2, ?bt), time:inXSDDateTimeStamp(?bt, ?t2),
time:hasEnd(?i2, ?et), time:inXSDDateTimeStamp(?et, ?t3),
time:hasBeginning(?i3, ?b2), time:inXSDDateTimeStamp(?b2, ?t3),
-> equal(?t1, ?t2), equal(?t3, ?t4).
Here we see that, instead of inferring that two historically successive versions must
$ meet $
, we instead infer that if their respective beginning and end timestamps must be equal.
Finally, rewriting these rules as “safe” or “DL-safe” is required as the expressivity of OWL 2 and SWRL permits the definition of computationally expensive, undecidable logical statements for logical reasoning (Horrocks et al., Reference Horrocks, Patel-Schneider, Boley, Tabet, Grosof and Dean2004; Hitzler and Parsia, Reference Hitzler and Parsia2009). To ensure this, first SWRL rules with conjunctions in the consequent must be split into multiple rules using the Lloyd-Topor transfomation (Lloyd, Reference Lloyd2012) such that only one atom remains in the consequent. Additionally, no variables may occur in the consequent that are not defined in the antecedent. For example, the SWRL rule provided above violates the former of these principles and should be rewritten as:
vers:VersionTransition.from(?vt, ?v1),
time_ext:hasExistenceTime(?v1, ?i1),
time_ext:hasExistenceTime(?vt, ?i2),
time:hasEnd(?i1, ?e1), time:inXSDDateTimeStamp(?e1, ?t1),
time:hasBeginning(?i2, ?bt), time:inXSDDateTimeStamp(?bt, ?t2),
-> equal(?t1, ?t2).
vers:VersionTransition.to(?vt, ?v2),
time_ext:hasExistenceTime(?vt, ?i2),
time_ext:hasExistenceTime(?v2, ?i3),
time:hasEnd(?i2, ?et), time:inXSDDateTimeStamp(?et, ?t3),
time:hasBeginning(?i3, ?b2), time:inXSDDateTimeStamp(?b2, ?t3),
-> equal(?t3, ?t4).
Additionally, we can rewrite these rules as their contrapositives we would like to infer an inconsistency. In this case, if the time limits of a version transition are not greater than or equal to the end time of the previous version, that version transition is inconsistent:
vers:VersionTransition(?vt),
vers:VersionTransition.from(?vt, ?v1),
time_ext:hasExistenceTime(?v1, ?i1), time_ext:hasExistenceTime(?vt, ?i2),
time:hasEnd(?i1, ?e), time:inXSDDateTimeStamp(?e, ?t1),
time:hasBeginning(?i2, ?b), time:inXSDDateTimeStamp(?b, ?t2),
notEqual(?t1, ?t2),
-> owl:Nothing(?vt).
In SWRL an empty consequent could also be provided to induce an inconsistency; however, this manner allows the inconsistent instance to be directly tagged as inconsistent, facilitating the identification of errors.
3.4. Validating concurrent scenarios of urban evolution
We developed a proof of concept SWRL test suite (Table 1) to validate integrated datasets as the final activity of our methodology (2). Through reasoning, the suite can then infer new RDF triples based on our proposed ruleset and verify if the given ontology is logically consistent. Figure 6 shows an activity diagram of how the test suite is utilized. In order to reuse existing ontological models such as “CityOWL,” the input ontology is divided into its TBox (the ontological data model) and its Abox (datasets conforming to the model). In addition to the ontology and the ruleset, a configuration for these tests can be provided for orchestrating which TBoxes, ABoxes, and rulesets to use for each test.
Table 1. The URLs and SWHID (Software Heritage identifiers) of the SWRL rules, test suite, and web application in this article. SWHIDs begin with “swh” and can be used at https://archive.softwareheritage.org/ to view the archived resources
Figure 6. A UML activity diagram of the input and output artifacts used and produced by the test suite. These elements are contextualized within the M1 and M0 metamodeling layers proposed in Atkinson and Kuhne (Reference Atkinson and Kuhne2003).
These elements are passed to the test suite, which uses RDFLib and OwlReady2 (Lamy, Reference Lamy2017) to read OWL ontologies and SWRL rules, reason upon them, and output newly inferred triples. These triples can be added back to the input ontology. In the case inconsistencies are detected for any test, these entities are logged in an output file.
4. Results
To illustrate how the proposed methodology can be used in a real-world use case, we integrated concurrent and successive open data city snapshots of a zone of interest and validated them using the methodology. Additionally, we created a web application for visualizing the integrated and validated data. This application was created using the UD-Viz urban data visualization framework (Samuel et al., Reference Samuel, Jaillot, Colin, Vinasco-Alvarez, Boix, Servigne and Gesquière2023) for displaying the 3D environment, a Blazegraph SPARQL endpoint (SYSTAP, LLC, 2013) that stores and serves the validated ontology (TBox and Abox), and a fileserver for serving 3DTiles (Cozzi and Lilley, Reference Cozzi and Lilley2022) containing the 3D geometry of the urban features (Figure 7). Links to these reproducible components are available in Table 1.
Figure 7. UML component diagram of the services used for the proposed web application for visualizing concurrent scenarios of urban evolution.
The use case arises from an urban planning project of the Gratte-Ciel neighborhood of Villeurbanne, France, from 2009 to 2018, at different levels of detail (Figure 8). These skyscrapers, developed in the 1930s and some of the first of their class at the time, are today considered built cultural heritage. The datasets used to represent proposed scenarios of evolution of this zone of interest come from the Lyon, France, metropole open urban data repository (Métropole de Lyon, 2021a, 2021b, 2021c, 2022) and are integrated to conform to CityOWL and the Workspace model using the approach proposed in Vinasco-Alvarez (Reference Vinasco-Alvarez2023).
Figure 8. An aerial photo of the Gratte-Ciel neighborhood from 1936 Commons (Reference Commons2020) (left) and from 2018 as a 3D city model (Vinasco-Alvarez, Reference Vinasco-Alvarez2023) (right).
Once validated, we use SPARQL queries to provide views of the workspace data and the evolving urban features contained within. For example, the following query permits to construction a knowledge graph of all the versions and version transitions of a workspace and their immediate neighbors. Figure 9 provides an illustration of the results of this query in an urban data web application.
Figure 9. A screenshot of the 3D urban data application visualizing the state of the Gratte Ciel neighborhood in 2018 (right) and the versions and version transitions of the workspace (left).
CONSTRUCT {
?subject ?predicate ?object.
}
WHERE {
?subject a vers:VersionTransition;
(vers:VersionTransition.to|vers:VersionTransition.from) ?object; ?predicate ?object .
}
While the previous query is useful for presenting an overview of the evolution of the zone of interest over time, queries can also be proposed at a more granular level to retrieve the individual changes features may have undergone between two specific versions. The query below can be used to construct a graph of all the transactions between version 2015 and 2018 of the neighborhood in the 1st proposed scenario of evolution. Figure 10 provides an illustration of the results of this query in an urban data web application.
Figure 10. A screenshot of the application visualizing the state of the Gratte Ciel neighborhood between 2015 and 2018 (right) and the transactions (or changes) between each building from each of these versions. The type of transaction is also used to color the buildings in the right view using the 3D tiles extension proposed in Jaillot et al. (Reference Jaillot, Rigolle, Servigne, Samuel and Gesquière2021).
CONSTRUCT {
?subject ?predicate ?object.
}
WHERE {
{
?subject a wksp:Scenario;
?predicate ?object .
FILTER(?subject = ws:scenario_1).
} UNION {.
?subject a vers:VersionTransition;
?predicate ?object .
FILTER(?subject = vt2015_2018:versionTransition_2015_2018).
} UNION {.
?subject a vers:Version;
?predicate ?object .
FILTER(?subject = v2015:version_2015 | |
?subject = v2018:version_2018).
} UNION {
vt2015_2018:versionTransition_2015_2018.
vers:VersionTransition.transaction? subject .
?subject a vers:Transaction;
?predicate? object .
}
}
Taking into account the work of Samuel et al. (Reference Samuel, Servigne and Gesquière2020), we focused on generalizing the Workspace model to work for any urban data. The original workspace model focused on rules for the documentation of historical evolution, which were not translated. Therefore, we did not propose rules for the existing or imagined version types and the influenced version transition type. We determined that these two aspects of the Workspace model were conceived for a specific use case in cultural heritage documentation and fall out of the scope of general urban data evolution. However, these rules could be translated using the methodology proposed in our work for documentary use cases that require them. Additionally, our translated rules do not include the subsumption and equivalence DL rules proposed in Samuel et al. (Reference Samuel, Servigne and Gesquière2020) as they are already defined by rdfs:subClassOf and owl:equivalentClass axioms in the CityOWL ontologies.
In our context, we focused on official datasets from the Lyon Metropole, which are updated at a low frequency (e.g., every 3 years). However, the proposed ruleset can be used to validate datasets that are updated at a higher frequency (e.g., daily, weekly, monthly, etc.) as long as the data is structured according to the CityOWL and Workspace models. The datasets used in this work were validated for conformance to the CityOWL and Workspace models, and did not contain any inconsistencies; hence, we do not encounter any data quality issues. However, we note that the proposed ruleset is not exhaustive and does not cover all the possible inconsistencies that may arise in urban data. Furthermore, we do not take into account the probabilistic nature of dynamic knowledge graphs Pujara and Getoor (Reference Pujara and Getoor2014), as this work focuses on deterministic scenarios of urban evolution. But we have not yet tested this in practice as well as the performance of the ruleset on larger datasets. An exhaustive evaluation of the performance of the ruleset on dynamic and larger datasets is left for future work.
To promote the reproducibility of this work, links to all artifacts, applications, and tools discussed here are available in Table 1. As is denoted in Figure 7, containerization technologies (i.e., Docker (2025)) are used to allow the components of the web application and the proof of concept test suite to be easily installed and used through portable ‘containers’ independent of hardware architecture or installed programs.
5. Conclusion
This article puts forth a novel methodology for validating integrated 3D city models for illustrating concurrent and successive views of urban evolution. This methodology also capitalizes on new standards and works for representing concurrent scenarios of urban evolution. Thanks to this methodology, we also propose a formalization of machine-readable SWRL rules for validating these evolution scenarios proof of concept tests suite. An integrated dataset based on open urban data was validated for conformance and temporal consistency using these rules. To demonstrate how these validated data can be exploited, they were visualized through a proof-of-concept 4D urban data web application with an extension to the UD-Viz urban data visualization framework. Using SPARQL queries, we also demonstrated how such validated datasets could be used in a 3D virtual environment for visualizing these scenarios of urban evolution. Additionally, the technical contributions of this work are reproducible, including the proposed test suite and web application. Our methodology is well-documented, modular, and makes use of open-source components and libraries. Thanks to these modular components for validating, storing, and visualizing ontological graph data of urban evolution, these components can be replaced as necessary.
When comparing this work to other Semantic Web approaches for integrating 4D urban data, there are several complementary aspects to note. In particular, approaches such as the Parallel World Framework with dynamic knowledge graphs are also capable of comparing concurrent or successive—real and imagined—urban evolution scenarios. One advantage of a versioning approach is that it can also be used to compare a real multiple imagined successive and concurrent states of the city, whereas the Parallel World Framework only focuses on one real and one imagined world state. These agent-based approaches also lend themselves well to implementing high-frequency urban digital twins, such as those that integrate streaming sensor data, which has not been done with our approach. Future work could explore using the proposed SWRL rule test suite as a validating agent for permitting more expressive data constraints. In addition, implementing a snapshot-based versioning approach within a dynamic knowledge graph can be a strategy for efficiently storing a history of states within Parallel Framework-based approaches. This could permit existing dynamic urban knowledge graph approaches to more easily compare concurrent AND successive urban states of the city. While within the context of this work, the proposed rule test suite performs adequately for our use-case with the low frequency updates (Section 4), applications that require validation in near real-time pose more demanding data validation needs.
Additionally, we are currently exploring how to easily translate rules from standardized abstract test suites such as the CityGML 3.0 Conceptual Model test suite, which defines rules through unstructured text. To make inference on the proposed rules more optimal at scale, we are working on assuring all translated rules are DL-safe according to the proposed approach. We note that there may be applications to using unsafe rules on small-scale datasets; however, this was not examined in this work. Future works will include more quantitative studies, exploring the limitations of our approach at different urban scales and levels of detail (e.g., from buildings to large-scale territories) with respect to related works. Finally, we would like to complete the proposed ruleset for Workspaces, including rules for validating proposed changes with additional heterogeneous data (e.g., multimedia documents, cultural heritage thesauri).
Data availability statement
Research data and code supporting this publication (Vinasco-Alvarez et al., Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024a, Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024b, Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024c, Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024d, Reference Vinasco-Alvarez, Samuel, Servigne and Gesquière2024e) are available on Software Heritage (2025) using the permanent identifiers listed below (Table 1). This includes the proposed SWRL ruleset, the test suite for validating urban evolution scenarios, and the web application for visualizing these scenarios. Additionally, the concurrent scenario data input data referenced in Section 4 are originally issued from the Métropole de Lyon (2021a, 2021b, 2021c, 2022) (accessed on July 26, 2024) before transformation to RDF/OWL.
Acknowledgements
The authors would like to thank the University Lumière Lyon 2 and INSA Lyon for funding respectively the Ph.D. Thesis and the post-doctoral position of Diego Vinasco-Alvarez. This work has been done within the LIRIS laboratory as a part of the VCity project (LIRIS, 2023).
Author contribution
The following indivduals contributed to this article: Diego Vinasco-Alvarez: Conceptualization, Methodology, Software, Validation, Investigation, Data Curation, Writing - Original Draft, Visualization. John Samuel: Conceptualization, Methodology, Validation, Investigation, Writing - Original Draft, Project administration. Sylvie Servigne: Conceptualization, Methodology, Writing - Review & Editing, Project administration. Gilles Gesquière: Conceptualization, Methodology, Resources, Writing - Review & Editing, Project administration, Funding acquisition. All authors approved the final submitted draft.
Funding statement
Diego Vinasco-Alvarez and Sylvie Servigne report a relationship with National Institute of Applied Sciences of Lyon that includes: employment. Diego Vinasco-Alvarez and Gilles Gesquière report a relationship with University Lumière Lyon 2 that includes: employment. John Samuel reports a relationship with the Ecole d’Ingénieurs en Chimie et Sciences du Numérique de Lyon that includes: employment.
Competing interests
The authors declare no competing interests exist.
Ethical standard
The research meets all ethical guidelines, including adherence to the legal requirements of the study country.
Appendix: URI prefixes
Table A1. The prefixes used in this article with their respective URIs (Vinasco-Alvarez, Reference Vinasco-Alvarez2023)
Open access
Comments
No Comments have been published for this article.