STAD: An Ontology Design Pattern and Ontology for the Semantic Representation of Aggregate Spatial and Temporal Data

This paper presents STAD, an ontology design pattern and OWL 2 ontology for representing the semantics of spatially and temporally aggregated data. The paper makes a strong conceptual contribution and is a good fit for SWJ.

1. In Listing 1, the following triple appears:
ex:NH_SummerMeanTemp2013-2022_1 a stad:SpatioTemporalAggregateData;
However, in Figure 6, a subclass arrow is used to link those two. This may cause confusion, since the listing shows an instance assertion, while the figure appears to suggest a subclass relationship. Please clarify and ensure consistency between the listing and the figure.

2. Figure 6 could be simplified. It currently exposes details that are further explained in subsequent subgraphs. I would suggest starting from the core entity (ex:NH_SummerMeanTemp2013-2022_1) and keeping at most two hops until reaching a class.
For example:
ex:NH_SummerMeanTemp2013-2022_1 stad:hasAggregationPeriod ex:SummerAggregate2013-2022;
ex:SummerAggregate2013-2022 rdf:type …
This would improve readability and reduce visual complexity.

3. The modeling of spatial support appears less formally constrained and less semantically deep compared to the temporal support modeling.
Based on Section 3.3.3 (“Spatial Support”), both Spatial Coverage and Spatial Resolution are expected components. However, Figure 8 only shows Spatial Coverage while referring to “spatial support,” which may be confusing. It would be helpful to briefly illustrate how Spatial Resolution is modeled in this example, even if it is not directly linked to ex:NH_SummerMeanTemp2013-2022_1. This would better echo the design of stad:hasSpatialResolution shown in Figure 9.
Additionally, Page 20, line 41 also uses the term “spatial support,” which should be checked for consistency and clarity.

4. In Listing 4, NewHampshireGeo does not have a label. To be consistent, Figure 8 should consistently use “NewHampshireGeo” if that is the intended identifier.

5. The motivating examples in Section 1.2 compellingly illustrate real semantic ambiguities in aggregated spatio-temporal data. However, the competency questions presented later appear comparatively weak.
The current CQs primarily demonstrate that the ontology supports structured metadata retrieval (e.g., identifying base datasets, transformation kinds, spatial or temporal support). However, they do not clearly demonstrate capabilities that would not already be achievable using existing ontology modeling approaches.
Given that the motivating examples emphasize ambiguity and incompatibility of aggregate data, the evaluation would be significantly strengthened by demonstrating reasoning-based scenarios that explicitly address those challenges.

6. Minor Issues
• Typo: Page 5, line 50
• Missing citation: Page 8, line 25

This manuscript describes an ontology for representing aggregate spatial and temporal data.
It describes datasets and their individual datapoints (qualitative or quantitative), with the dataset having a spatiotemporal support, resolution, and granuarity. Datapoints may be qualitative or quantitative, observed or computed (incl. predicted datapoints or statistical aggregate datapoints).
It is an extended version of a paper presented at the workshop on Ontology Patterns and Design (WOP) 2022.

Overall, the paper motivates the problem well, provides satisfiable grounding to the state of the art, and reuses well established ontologies (plus the existing algorithm, implementation, and execution ODP). The paper addresses an interesting problem and has sufficient raw material, I would say, for an interesting ontology. However, it has been submitted as 'full paper', and suffers from major flaws that make me recommend rejection. These can be summarized as follows:

1. the proposed STAD ODP resembles more an ontology than an ODP. There are different categories of ODPs, but they are usually small and purpose-specific. It's not clear what's part of the STAD ODP, and what's part of the STAD ontology.
2. the contribution seems to me like incremental with respect to those already presented in WOP 2022
3. the ontology does not meet the expected quality standards, and is not published following the expected quality standards. (ex. most terms don't have a label or a definition ; the Turtle version is not available from the ontology IRI ; the import to https://www.w3.org/2006/time/agg does not resolve)
4. the stad-mls module, essential in the ontology and its validation, is not available online for review
5. the paper contains many small issues (ex. including describing axioms that are not in the ontology (see below), using acronyms that are only later defined, not using acronyms when they have already been defined, having unresolved references to bib resources, ...)

Some minor comments:

The ontology documentation at https://theskailab.github.io/STAD/stad.html contains a git conflict mark `>>>>>>> ca56fc87ca94e8ed55e619c03562f869dedcb828`

Many acronyms are either not defined before they are used. For example SCAN in the abstract, MODIS and NOAA in the examples in Section 1.2, MLS on p.5

Some acronyms are not used when they have already been defined. For example the acronym for "Ontology Design Pattern" is defined once in the abstract and twice in the article body, but it's still not used a total of ten times in the body of the article (2x in 1.2, 3x in 1.3, 1x in 3, redefined in 3.3.2, 2x in 6)

One occurrence of a undefined bibliographic reference for Chrismas [?] p8

The paper claims that the remaining Ontology pitfalls identified by the OOPS! scanner are "not important". Yet it doesn't say which are these pitfalls, and according to who they are not important.

Some sentences should be reviesed (in-build -> built-in, TO -> To, It also is much more detailed -> It is also much more detailed,

p23 CQ2: the lines that start with % should be deleted. This is not valid SPARQL.

This paper introduced a new ontology to conceptually formalize the aggregated data in spatial, temporal, and spatiotemporal dimensions.
(1) Originality: understanding how geospatial data are aggregated is indeed a challenge in data science. Often, it relies on the data creators to put these kinds of details into the metadata files. But still there are no explicit standards of doing so. This paper is original in terms of filling this gap.
(2) Significance of the results: Through a set of competency questions, the authors showed that the designed ontology can be used to answer very complex questions that other data schema won’t help answer. In this sense, the results are significant. However, there is a lack of discussion on how the competency questions are designed and how useful they will be for domain scientists. Page 3-4 have some discussions on it, but they are quite general. Also, the experiments are done in a small toy dataset with some predefined aggregation. I am wondering how useful the ontology is to real-world geospatial datasets, which often do not have that granular details about data aggregation.
(3) Quality of writing: the paper is well written and easy to follow. Particularly, the ontology diagrams are well presented. One suggestion for improvement is on the presentation of listings. They are quite chunky and are displayed across the texts, which distracts the reading of the main texts. I am wondering if there are ways to simplify them or put them into appendix.

Further comments:
-Page 3 to 4 has listed a set of questions and the very high-level context they might be useful. More details discussing how these questions were generated (e.g., by domain scientists? )and in what aspects they would help the domain scientists would be useful for the readers to appreciate the work (e.g., why they need to know the details of the aggregation?). Also, it seems that the competency questions (Page 4) are brought up based on these questions., and what are the links between these two sets of questions? – In section 1.3, it might not be useful to list the contribution on top of workshop paper, as it is not self-contained anymore. Namely, for those who haven’t read your workshop paper, it is quite difficult to understand your listed contributions.
- In page 13, line 10, more explainations are needed to justify “therefore, the temporal resolution is […] described as part of the DatasetDescription of the base dataset […]”
- In page 15, line 14, I am not sure why OM is compared here with QUDT. It is quite out of contexts.
- In page 15, line 27, could you be more specific on “non-trivial consistency”?
- There are quite some typos, format issues. Please carefully proof-read the paper.

Regarding to “Long-term stable URL for resources”, the ontology is presented in a standard, accessible way. However, there needs to be more description on the sampled data and queries in the readme to make the work more replicable.