The EPISECC Ontology model: spatio-temporal ontology for disaster management

Tracking #: 1758-2970

Martina Baucic
Snjezana Knezic
Georg Neubauer

Responsible editor: 
Mark Gahegan

Submission type: 
Ontology Description
The EPISECC project’s investigation revealed that 70% of tools used by first responders during disaster management reach physical and syntactical interoperability and only 30% are covering semantic interoperability. To improve information sharing, the EPISECC project developed a concept of Common Information Space (CIS) including three semantic structures: taxonomy, semantic repository and ontology model. As there is no reference ontology for disaster or emergency management, the EPISECC project team developed a spatio-temporal ontology for disaster management, modelling common operational picture for the first responders. A two steps approach is applied: the first step models the application knowledge and the second step links concepts with reference ontologies. The ontology backbone is a subset of the EPISECC Taxonomy representing the application knowledge. GeoSPARQL, W3C Time and DOLCE-Lite ontologies are referenced. Representing dynamic features ask for a new concept with properties varying in both space and time. Thus, the concept of “spatio-temporal part” is introduced in the ontology. The conceptual model is formalised in the Ontology Web Language (OWL) schema and validated by the competency questions based on the EPISECC scenario. Future research should tackle a new paradigm in disaster management: including ad-hoc participants and trustworthiness of their data.
Full PDF Version: 

Major Revision

Solicited Reviews:
Click to Expand/Collapse
Review #1
Anonymous submitted on 19/Apr/2018
Minor Revision
Review Comment:

This is a very interesting paper on ontology to support emergency respondents. The ontology is developed taking into consideration many different developments on related topics. This ontology is a good example of a ontology built on top of carefully selected and motivated existing ontologies. The literature review on ontologies is very complete. The explanation of the introduced classed is clear and logical. The comments I have are only related to the introduced classes/sub-classes to describe the disasters.
1. Apparently the EPISECC taxonomy created within the mentioned project has been utilised. However the guidelines/use cases//documents that have been followed are not mentioned in the paper. UNSDR as well as IRDR have been working for year to establish peril classification and hazard glossary. The authors may want to refer to some literature. IRDR classiciation is availble here:
2. Motivations for the sub-classes of Resources, Process and Organisation is not specified as well. It is unclear how generic the proposed classification is and whether it can serve all kind of case studies. In this respect is good to elaborate on the level of generalization which aimed.

The whole validation part should be extended and provide more details on the results. The current version of the paper refers to previous publications and reports, which is difficult to find. The final 8 questions in Table 1 look like a composition of several questions. Why such compositions? Can they be asked one by one? Who wants to have these questions?Are there other options for improvement.
Please elaborate on these questions and the entire validation. It is a very important part of the paper.

Review #2
By Sven Schade submitted on 28/May/2018
Major Revision
Review Comment:

The authors present an ontology that resulted from an EU-funded research project. Although the topic area is important and it would be valuable to publish an ontology in the field of disaster management, the current description misses details and explanations and thereby does not justify acceptance for the journal. If the authors are willing to invest the extra effort and re-submit the article, the following recommendations might be considered.

Main comment: Several of the modelling decisions are unclear and need further explanation/clarification. These include:
- Elements of the description of major classes do not appear in the ontology. Just to give one example: the description of a process mentions a set of “actions”, “executing organization”, “response”, “critical event”.
- “Resource: Assets an organization has available…” definition is unclear.
- Is a “Common Operational Picture” a “Resource”?
- The distinctions that apply in is-a relationships are rarely specified, e.g. what exactly determines the different kinds of physical responses?
- Concepts such as “Epicenter”, “Magnitude” and others are mentioned in the text but the related classes are not explained in detail and are not included in the figures.
- “Resource” appears in figures 2 and 4, but with different relationships. This needs an explanation.
- It also needs an explanation why a “Disaster” is “Situational Data” in Fig. 4, whereas in Fig 5 “Disaster” is a “State”.
- Also in Fig 5 there are many relationships that are not intuitive, e.g. a “Resource” is modelled as a subclass of “Human” of “Physical” and of “Dynamic Data”. The text explaining Fig 5 should be a major part of the article and is too short at the moment.
- The needs for a new class (“SpatioTemporalPart”) that covers spatial temporal relationships (in addition to e.g. GeoSPARQL) needs more explanation, incl. its relationships to the existing OGC and W3C.

Additional major comments:
- It remains unclear if the developed solution is integrated with already existing community tools. It should be discussed if previously existing systems can be still used, if adaptations are required, or if interested users have to change their tools (which might be difficult).
- Whereas the overall knowledge engineering approach builds on a well-known and solid methodology, the mix of re-used ontologies needs more explanation. Which classes are really re-used and how do different ontological commitments relate to each other? Also the seven tasks may be described in more detail and the competency questions might be listed earlier (not only in table 1). It might be worth to also discuss any possibly required evolution of the approach (which was basically developed a decade ago).
- Many of the central references for this publication are project deliverables. While this could be fine in principle, they should be provided for future use. Project web pages tend to be shut down a few years after the project finished. A third-party repository (ideally one also assigning DOIs) might be used in order to ensure long term access to these central documents. Zenodo might be one option.
- Given that the example covers multiple countries, any discussion about languages would add value to the document.
- It would be equally valuable to learn more about the way in which the already existing communities were engaged in the ontology development so that their possible different views are well represented.

Minor comments:
- The introduction section mentions many different concepts “taxonomy”, “concepts”, “classes”, “hierarchies”, “facets”, “semantic annotation”, etc. It might be worth to introduce how these are understood by the authors.
- The details about existing models remains vague and could be extended, e.g. what exactly exists from the W3C and why does this not fit the use cases.
- The acronym FP7 needs an explanation.
- In table 1, the text on the right for item 7 and 8 repeats that of item 4. I would replace this with something like “None, as already addressed by the extension related to question 3”. Also, I would talk about “extensions” not about “improvements” of the ontology.

Review #3
By Yingjie Hu submitted on 27/Jun/2018
Major Revision
Review Comment:

This paper presents a spatio-temporal ontology for modeling the common concepts, operations, and datasets involved in disaster management. The authors modeled concepts, such as "disaster", "process", and "common operational picture", and linked the concepts to existing ontologies, such as GeoSPARQL and W3C Time. Overall, this paper addresses an important and meaningful topic. However, it also suffers from some major issues that should be resolved.

First, as also pointed out by the authors, the lack of ontologies specifically focusing on disaster management may be due to the complexity of this domain and the composed multiple sub components. Indeed, the sub components, such as events and organizations, can be modeled by other existing ontologies. In the later text, the authors mentioned SWEET and SSN ontologies, which is good. However, two other ontologies, the simple event model [1] and the semantic trajectory ontology [2], are also worth mentioning. The former can help annotate the actors, places, and time involved in an event, while the latter can help annotate the trajectories of rescue teams.
[1] Van Hage, W.R., Malaisé, V., Segers, R., Hollink, L. and Schreiber, G., 2011. Design and use of the Simple Event Model (SEM). Web Semantics: Science, Services and Agents on the World Wide Web, 9(2), pp.128-136.
[2] Hu, Y., Janowicz, K., Carral, D., Scheider, S., Kuhn, W., Berg-Cross, G., Hitzler, P., Dean, M. and Kolas, D., 2013, September. A geo-ontology design pattern for semantic trajectories. In International Conference on Spatial Information Theory (pp. 438-456). Springer, Cham.

Second, since one can combine a number of existing ontologies for organizing disaster management data, the authors may need to enhance the justifications of developing a new and high-level disaster management ontology. In other words, what functionalities cannot be done without your ontology? The authors may consider adding a separate section (a new section 2) before the ontology development section, and specifically focus on discussing the competence questions that cannot be answered by using existing ontologies and why answering such new questions are important.

Third, the current section 5 "Implementation and validation" is thin (also section 4 and section 5 can be combined). The authors may consider adding a case study and using your ontology to annotate some real disaster response data. A number of SPARQL queries, especially GeoSPARQL queries, can then be added to demonstrate the new spatio-temporal questions that can be answered by your ontology. This would make this paper much stronger and can emphasize the value of the developed ontology.

Finally, the authors mentioned that further developments should tackle new paradigms of disaster management. It is worth mentioning the recent remote and online participation of volunteers in disaster response, organized by the Humanitarian OpenStreetMap Team and used in multiple recent disasters, such as the Nepal Earthquake [3]. This is a new paradigm especially useful in international disaster responses.
[3] Hu, Y., Janowicz, K. and Couclelis, H., 2017. Prioritizing disaster mapping tasks for online volunteers based on information value theory. Geographical Analysis, 49(2), pp.175-198.

In sum, this paper proposes a spatio-temporal ontology for disaster management. This paper can benefit from having enhanced justifications on the development of such a new ontology and clarifications on the values of this ontology. This paper can also be improved by adding a more concrete case study with real data to demonstrate the new spatio-temporal queries enabled by the developed ontology.

Review #4
By Carsten Keßler submitted on 02/Jul/2018
Major Revision
Review Comment:

This paper describes the EPISECC ontology model that addresses disaster management and has been developed in an EU-funded project under the same name. As an ontology description, the paper should be judged based on two criteria:

(1) Quality and relevance of the described ontology (convincing evidence must be provided):

This is clearly a relevant field where heterogeneous information is being exchanged between different organizations, so a well-defined ontology could potentially improve the messy situation in the field substantially. Unfortunately, the ontology itself does not seem to be accessible, so that the quality can only be judged based on the description (more on that below). Moreover, there is no indication of how/where the ontology is actually in use, which is also a review criterion for this kind of submission.

(2) Illustration, clarity and readability of the describing paper, which shall convey to the reader the key aspects of the described ontology.

The ontology development has followed a well-established methodology and the paper describes and illustrates the ontology in a (somewhat) clear way, giving the reader an idea of the key aspects. Having said that, there are some concerns about the nature of the competency questions, which seem quite high-level to me. My own experience from the development of the Humanitarian eXchange Language (HXL) ontology [1] tells me that the actual challenges in this area are actually the much more fine-grained concepts (e.g. to record the number and demographics of affected population and their needs, and to which extent they have been met) which are ambiguous and can benefit from a clear definition. This makes me wonder whether there is a clear benefit in knowing that e.g. Earthquake is a subclass of Disaster. I.e., where is the semantic ambiguity the makes this a relevant problem in practice?

Based on these two points, I think the paper needs major revisions to convince the reader that (a) the ontology is of practical use, and (b) why the included concepts are relevant for the problem space.

Some more detailed comments:

- The last sentence about future work should be removed from the abstract.
- Introduction: Your definition of "response" includes " or more situations which straightforwardly lead to a disaster". So the response also leads to a disaster? That doesn't seem to make sense.
- p.2, right column: How likely is it that semantic descriptions are actually entered into the Semantic Repository prior to a disaster? Doesn't it depend largely on the type of disaster what descriptions will be required, and is it realistic to assume that any such situation can be foreseen and described?
- Some column: The difference between operational and situational data should be explained for readers not familiar with the domain.
- The end of the introduction talks about the "EPISECC use case" without ever clearly describing what that use case is.
- Section 3.1: what is the rationale to omit static data from any further analysis?
- Section 3.2: THE EPISECC taxonomy needs some introduction: What does it include and how was it defined?
- Figure 3 does not match the description on the previous page. In the figure, "search for people" and "decontamination" are subclasses of "physical response", which is a subclass of "process". The text describes them as subclasses of "resource".
- The modeling of the spate-temporal dynamics is also ambiguous. How exactly are you modeling e.g. a contaminated area that changes over time? Will the object always have the same identifier and its spatial and temporal properties change in snapshots? And if so, how is this implemented ([2] discusses different ways to solve this problem).
- In the conclusions, I would have liked to hear about any applications producing and/or consuming data annotated with this ontology, but instead it talks about including social media data and trustworthiness, which does not seem directly relevant for the work presented here.

[1] Keßler, C., and Hendrix, C. "The humanitarian exchange language: coordinating disaster response with semantic web technologies." Semantic Web 6.1 (2015): 5-21.
[2] Trame, J., Keßler, C., & Kuhn, W. (2013, September). Linked data and time–modeling researcher life lines by events. In International Conference on Spatial Information Theory (pp. 205-223). Springer, Cham.