KnowWhereGraph is one of the largest fully publicly available spatially enabled knowledge graphs. It includes data on natural hazards (e.g., hurricanes, wildfires), climate variables (e.g., air temperature, precipitation), soil properties, crop and land-cover types, demographics, and human health, among other themes. These have been leveraged through the graph by a variety of applications to address challenges in food security and agricultural supply chains; sustainability related to soil conservation practices and farm labor; and delivery of emergency humanitarian aid following a disaster. This paper showcases the KnowWhereGraph ontology, which acts as the schema for the KnowWhereGraph. We discuss how it enables the powerful spatial and semantic integration across these datasets, our validation paradigm, and the applications it supports.
KnowWhereGraph1 (KWG) is one of the largest, publicly available geospatial knowledge graphs in the world. The KWG supports applications in the food, agriculture, humanitarian relief, and energy sectors and their attendant supply chains, generally; environmental policy issues relative to interactions among agricultural sustainability, soil conservation practice, and farm labor; and delivery of emergency humanitarian aid, within the US and internationally. It brings together over 30 datasets related to observations of natural hazards (e.g., hurricanes, wildfires, and smoke plumes), spatial characteristics related to climate (e.g., temperature, precipitation, and air quality), soil properties, crop and land-cover types, demographics, and human health, among others, resulting in a knowledge graph with over 16 billion triples.
We present the KnowWhereGraph ontology, resulting from significant knowledge and ontology
engineering and reuse, which integrates these datasets, to address meaningful use cases, and
provides answers to questions such as “what is here”, “what happened here before”, “how does
this region compare to…” at a high spatial resolution across the entire globe [
There are a few ontologies that deal with geospatial information, but not to the extent or breadth
that the KWG ontology provides. In some cases, we re-use related ontologies (e.g., SOSA/SSN [
Some standardized and structured vocabularies for describing environmental concepts exist but have limitations that impacted their usability and efectiveness within the context of the KWG. Moreover, these vocabularies are either large and complex, with many concepts and relationships, or too simple, which makes them challenging to use efectively.
The Environmental Ontology (ENVO)
ENVO covers a wide range of environmental concepts
[
cogan.shimizu@wright.edu (C. Shimizu); shirlystephen@ucsb.edu (S. Stephen); rui.zhu@bristol.ac.uk (R. Zhu); https://coganshimizu.com/ (C. Shimizu); https://people.cs.ksu.edu/~hitzler/ (P. Hitzler) 0000-0003-4283-8701 (C. Shimizu)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR characteristics of the natural world. Another example is where ‘part of‘ and ‘has part‘ are used interchangeably to describe hierarchical relationships between concepts.
Google Data Commons & Schema.org The Google Data Commons,2 powered by annotations from Schema.org3 provides a mechanism via its “Map Explorer” tool to visualize data that has a geospatial component. However, by its nature, Schema.org is a semantically shallow representation, which generally indicates the type of data something is. This was insuficient for the heavy semantic harmonization needs of the KWG ontology.
The KWG ontology satisfies several requirements: enabling geospatial integration, facilitating data integration, providing rich inferencing, and expediting maintainability. We describe these below.
Enable Geospatial Integration The primary purpose of KWG is to provide a convenient method for integrating data along a geospatial dimension. This is integral to the mission of the project and, subsequently, a core requirement for the graph and its schema.
Facilitate Data Integration KnowWhereGraph must be capable of providing an overarching framework for the semantic harmonization of key terms and concepts.
Provide Rich Inferencing Beyond a 1:1 representation of the (integrated) datasets, KWG’s schema must be expressive enough to infer latent relationships between datasets, such as causality of events or the inheritance of spatial characteristics.
Highly Maintainable To be maximally useful, KWG must be easily maintained by the
community. This includes both the degree of facilitation of data integration, but how amenable the
schema – and thus the graph – are to modification: either through the incorporation of new or
evolving use cases, rectifying conceptual errors in the graph, or adapting to changes.
To facilitate satisfying these requirements we utilize the Modular Ontology Methodology (MOMo;
[
KWG represents a myriad of domains – and is capable of integrating more. Any domain that is capable of being represented along a geospatial axis can be incorporated. As of now, the graph generally supports the hazard (and related) domains. That is, it supports given physical phenomena that can negatively impact places, people, or the economy, including the specifics of who is impacted, what is impacted, and how the impacts can be mitigated. In addition to the four general requirements above, the KWG ontology supports several pilot use cases across diferent domains.
Humanitarian Relief Disasters are complex and dynamic situations requiring humanitarian organizations to evaluate and respond rapidly to many diferent issues simultaneously. Often what 2https://www.datacommons.org/ 3https://schema.org
Thematic Dataset
Soil Properties
Wildfires
Earthquakes
Climate Hazards Experts (Covid-19 Mobility)
Expert (General) Cropland Types
Air Quality Smoke Plumes Forecasts
Climate Disaster Declarations Smoke Plume Extents
BlueSky Forecasts
Highway Networks Public Health Observations Public Health Infrastructure
Social Vulnerability Hurricane Tracks
Source Agency
USDA USGS, USDA, USFS, NIFC
USGS
NOAA
Direct Relief KWG, UC System, Direct Relief,
Semantic Scholar
USDA
EPA NOAA NOAA FEMA NOAA BlueSky
DoT CDC, USCB, University of
Wisconsin
HIFLD CDC, ATSDR
NOAA
Example Attributes soil type, farmland class wildfire type, num acres burned
magnitude casulaties, property damage name, afiliation, expertise name, afiliation, expertise crop types (raster data)
air quality index daily smoke plume forecast temperature, precipitation area, amount approved smoke plume extent
PM10, PM5 road type, road length, signage poverty, diabetes, obesity
pharmacies, hospitals social vulnerability index max wind speed, min pressure is most needed to improve efective response is quick access to the right experts at the right time. To assist in identifying people with expertise in humanitarian aid and relief, with a particular focus on health and the health care impacts of disasters, we are working with Direct Relief to showcase how our knowledge graph can give them rapid access to area briefings, including previous events and physical properties of, for example, climate and transportation infrastructure in the afected regions.
Food Supply Chain Resilience Moreover, understanding and improving the robustness and adaptability of the food supply chain is of critical importance to make it more resilient to disturbances in food supply and demand networks. Network fracturing and delayed recovery during extreme weather events is always an inherent risk when it comes to wildfires, floods, and other natural hazards. In the face of uncertain natural hazards, which are increasing in frequency and severity, it is vital that the implications of these disruptions are evaluated for the source nodes of our supply chains, such that resiliency in the whole supply chain can be promoted. To solve this challenge, we are partnering with the Food Industry Association (FMI), which has identified food safety and food quality issues rising from environmental disasters or disturbances as high-priority industry concerns.
Tables 1 and 2 show condensed views of the datasets that the KWG Ontology integrates. These datasets are widely sourced, originating from non-governmental organizations (NGOs), US governmental agencies, open source data, and the commercial sector (with attribution).
Place-Centric Dataset
S2 Cells Global Administrative Regions
US Federal Judicial Districts
National Weather Zones
FIPS Codes Designated Market Areas
ZIP Codes
Climate Divisions Census Metropolitan Area
Drought Zone
GNIS
Defining Authority
Google GADM.org DoJ, ESRI
NOAA USCB Nielsen USPS NOAA USCB NDMC USGS
Spatial Coverage Lvl 9 (Global), Lvl 13 (US)
Global
US US US US US US US US US
KWG, and thus the KWG Ontology, are used in the pilots described above, as well as in several tools
(e.g., Knowledge Explorer [
We have developed several standalone ontologies and resources, and reuse a number of wellknown, standardized, or W3C-recommended vocabularies, taxonomies, and ontologies. This helps to support greater interoperability with other knowledge graphs, to maintain consistency in our data model, and to leverage existing tools that support these vocabularies.
Ontology Design Patterns During the development of the KWG ontology, we both create new
and adapt ontology design patterns (ODP; [
GeoSPARQL We used GeoSPARQL [15, 16], an Open Geospatial Consortium (OGC) standard, to represent our geospatial data in RDF. It, in turn, reuses the OGC Simple Features (SF) standard, which defines a set of geometric primitives (e.g., points or polygons) and their spatial relationships. Within the KWG ontology, we represent any discrete geographic feature type (Hazard, Region, and their subclasses) that has a spatial extent as a subclass of GeoSPARQL’s geo:SpatialObject class. We also use the spatial relationships from GeoSPARQL (based on DE-9IM spatial relations) to establish pre-computed spatial relationships between any two spatial features. While several graph databases support GeoSPARQL, we found a number of features of GraphDB including its support of GeoSPARQL [17] made it the best candidate for KWG.
SOSA/SSN The Sensors, Observations, Sampling, and Actuator Ontology [18], coupled with the
Semantic Sensor Network ontology (SOSA/SSN; [19]) are used to model observations made by
sensors that detect, measure, or observe properties of features [
OWL-Time The Time ontology is (re)used for all representations of time within the KWG ontology. The super-property for most time-related conceptualizations is kwg-ont:hasTemporalScope, which efectively has a range of any temporal entity from OWL-Time. For serializations, we reuse the XML schema datatypes.
Metadata and Provenance To easily maintain metadata and provenance, we reuse several vocabularies to describe the KWG ontology. For example, we use annotation properties from Dublin Core Metadata Initiative (DCMI) Metadata Terms [20] to describe the title, description, rights, license, date created, and creator. We use Friend of a Friend (FOAF [21]) to describe the development team and their roles. The Simple Knowledge Organization System (SKOS) is used to annotate definitions, examples, and the taxonomic structure between domain concepts. Finally, we use PROV-O [22] to describe the provenance of resources, e.g., to track the provenance and lineage of a dataset.
QUDT The QUDT (Quantities, Units, Dimensions and Data Types) ontology [23] is used for representing climate measurement data (such as temperature, Palmer drought severity index, cooling degree days) and their corresponding units of measure. Specific climate quantity types (such as mean or value) are denoted using the kwg-ont:Quantity class, a subclass of kwg-ont:QuantityValue. Measured values and corresponding data properties are then captured using data properties qudt-unit:unit and qudt-unit:numericValue.
The Expertise Ontology KWG contains information on agents who are experts on topics related to specific disaster types, disaster management activities, named disasters, and public health. The Expertise Ontology4 (EO) was developed to represent all varied expertise-related information consistently. At a high level, EO consists of a core set of classes and properties to 1) model experts (individuals or groups), topics, and their relations, 2) represent hierarchical relations between topics of diferent levels of granularity, and 3) connect topics with relevant content in a knowledge graph. EO facilitates representing not only research- and theory-based expertise, but also experience-based expertise by modeling the activities that an expert may have engaged in or their role and afiliation within an organization, and scopes these spatially and temporally.
The Disaster Management Domain Ontology KWG contains at least 11 hazard datasets and
at least one hazard (Fire) from four diferent sources (see Table 1). To model their semantics and
enable integration using any existing ontologies, we are developing the Disaster Management
Domain Ontology (DMDO), which will provide a framework to align diverse hazard types,
formats of data, and domain vocabularies consistently within KWG, but also for better situational
awareness through clarification of the spatiotemporal interactions of similar events. The ontology
disambiguates hazards from disasters and their impacts, but also distinguishes spatiotemporal
events from their observations. DMDO also formalizes the UNDRR hazard classification [ 24]
using the taxonomy alignment ODP [
The KnowWhereGraph Ontology has been evaluated through its ability to meet use-case requirements, as outlined in Section 3. We do this through interviewing domain experts and analyzing 4https://github.com/KnowWhereGraph/expertise-ontology competency questions and their corresponding SPARQL queries with results. In formulating aspects of the ontology, and especially in understanding specific thematic datasets, it is necessary to draw in the expertise of knowledgeable practitioners and subject matter experts. By iterating through multiple versions of the ontology with multiple diferent experts, we are able to converge on a common conceptualization. To evaluate the materialization, as well as the efectiveness of querying against the graph, we develop suites of competency questions. These allow for the connection between natural language, expected usage of the graph, and the KWG ontology (via the formulation of a SPARQL query). The analysis of the actual results, as opposed to expected results, allows us to evaluate if our conceptualization mirrors domain expertise and meets our use-case needs.
In a secondary manner, the KWG ontology is evaluated through its usability. That is, how well it meets the needs of developers creating applications against the entire knowledge graph. To this end, we realized that the materialization (aka shortcuts) would be necessary to link more efectively places and their identifiers (and subsequently simplify queries). For example, regions from diferent place-centric datasets could previously only be obtained through a complex query that drilled down to a cell-based representation, and then abstracted back upwards to the region in question. An entirely new version of the ontology was rolled out to accommodate this identified need. Finally, we provide a set of shapes (defined in the SHApes Constraint Language; SHACL; [25]) to validate the materialization of the ontology, which can be found online5 and in [26].
The KnowWhereGraph is a complex project with multiple evolving use cases, a large team, and
an ambitious goal. We have presented the KnowWhereGraph Ontology, which integrates over 30
datasets to power several motivating use cases. It was developed using a pattern-based method
(i.e., modular ontology modeling [
Availability We provide multiple types of documentation for the KnowWhereGraph Ontology: living documentation (generated using [27]) coupled with schema diagrams (generated manually) can be found at [28], alongside a static, technical report (generated using [29]). Finally, the ontology itself can be found in [30] and is released under the CC BY 4.0 license. The KnowWhereGraph is maintained by the KnowWhereGraph team; more details can be found in [31].
This work was funded by the National Science Foundation under Grant 2033521 A1: KnowWhereGraph: Enriching and Linking Cross-Domain Knowledge Graphs using Spatially-Explicit AI Technologies. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. 5https://github.com/KnowWhereGraph/KWG-SHACL October 24, 2022, volume 3352 of CEUR Workshop Proceedings, CEUR-WS.org, 2022. URL: https://ceur-ws.org/Vol-3352/pattern3.pdf. [13] C. Shimizu, P. Hitzler, C. F. V. II, A pattern for modeling computational observations, in: V. Svátek, V. A. Carriero, M. Poveda-Villalón, C. Kindermann, L. Zhou (Eds.), Proceedings of the 13th Workshop on Ontology Design and Patterns (WOP 2022) co-located with the 21th International Semantic Web Conference (ISWC 2022), Online, October 24, 2022, volume 3352 of CEUR Workshop Proceedings, CEUR-WS.org, 2022. URL: https://ceur-ws.org/Vol-3352/ pattern5.pdf. [14] C. Shimizu, Q. Hirt, P. Hitzler, MODL: A modular ontology design library, in: K. Janowicz, A. A. Krisnadhi, M. P. Villalón, K. Hammar, C. Shimizu (Eds.), Proceedings of the 10th Workshop on Ontology Design and Patterns (WOP 2019) co-located with 18th International Semantic Web Conference (ISWC 2019), Auckland, New Zealand, October 27, 2019, volume 2459 of CEUR Workshop Proceedings, CEUR-WS.org, 2019, pp. 47–58. [15] F. Knibbe, J. Herring, J. Abhayaratna, M. Bonduel, M. Perry, N. J. Car, S. J. D. Cox, T. Homburg, GeoSPARQL Ontology, OGC Standard, OGC, 2021. Https://opengeospatial.github.io/ogcgeosparql/geosparql11/index.html. [16] R. Battle, D. Kolas, GeoSPARQL: enabling a geospatial semantic web, Semantic Web Journal 3 (2011). [17] graphdb, GraphDB, http://graphdb.ontotext.com/, 2023. [18] K. Janowicz, A. Haller, S. J. Cox, D. Le Phuoc, M. Lefrançois, Sosa: A lightweight ontology for sensors, observations, samples, and actuators, Journal of Web Semantics 56 (2019) 1–10. [19] A. Haller, K. Janowicz, S. J. Cox, M. Lefrançois, K. Taylor, D. Le Phuoc, J. Lieberman, R. GarcíaCastro, R. Atkinson, C. Stadler, The modular ssn ontology: A joint w3c and ogc standard specifying the semantics of sensors, observations, sampling, and actuation, Semantic Web 10 (2019) 9–32. [20] H. Wagner, S. Weibel, The dublin core metadata registry: Requirements, implementation, and experience, J. Digit. Inf. 6 (2005). URL: https://journals.tdl.org/jodi/index.php/jodi/article/ view/70. [21] D. Brickley, L. Miller, Foaf vocabulary specification 0.91, 2007. [22] S. Sahoo, D. McGuinness, T. Lebo, PROV-O: The PROV Ontology, W3C Recommendation,
W3C, 2013. Http://www.w3.org/TR/2013/REC-prov-o-20130430/. [23] H. Rijgersberg, M. van Assem, J. L. Top, Ontology of units of measure and related concepts, Semantic Web 4 (2013) 3–13. URL: https://doi.org/10.3233/SW-2012-0069. doi:10.3233/ SW- 2012- 0069. [24] UNDRR Hazard definition and classification review (Technical Report)., https://www.undrr.
org/publication/hazard-definition-and-classification-review, 2021. [25] D. Kontokostas, H. Knublauch, Shapes Constraint Language (SHACL), W3C Recommendation, W3C, 2017. Https://www.w3.org/TR/2017/REC-shacl-20170720/. [26] R. Zhu, C. Shimizu, S. Stephen, L. Zhou, L. Cai, G. Mai, K. Janowicz, M. Schildhauer, P. Hitzler, SOSA-SHACL: shapes constraint for the sensor, observation, sample, and actuator ontology, in: IJCKG’21: The 10th International Joint Conference on Knowledge Graphs, Virtual Event, Thailand, December 6 - 8, 2021, ACM, 2021, pp. 99–107. URL: https://doi.org/10.1145/3502223. 3502235. doi:10.1145/3502223.3502235. [27] D. Garijo, WIDOCO: A wizard for documenting ontologies, in: C. d’Amato, M. Fernández, V. A. M. Tamma, F. Lécué, P. Cudré-Mauroux, J. F. Sequeda, C. Lange, J. Heflin (Eds.), The Semantic Web - ISWC 2017 - 16th International Semantic Web Conference, Vienna, Austria, October 21-25, 2017, Proceedings, Part II, volume 10588 of Lecture Notes in Computer Science, Springer, 2017, pp. 94–102. URL: https://doi.org/10.1007/978-3-319-68204-4_9. doi:10.1007/ 978- 3- 319- 68204- 4\_9. [28] The KnowWhereGraph Schema, The KnowWhereGraph schema, https://stkokwg.geog.ucsb.edu/lod/ontology, 2023. [29] C. Shimizu, P. Hitzler, Automatically generating human readable documentation for ontology design patterns, in: O. Seneviratne, C. Pesquita, J. Sequeda, L. Etcheverry (Eds.), Proceedings of the ISWC 2021 Posters, Demos and Industry Tracks: From Novel Ideas to Industrial Practice co-located with 20th International Semantic Web Conference (ISWC 2021), Virtual Conference, October 24-28, 2021, volume 2980 of CEUR Workshop Proceedings, CEUR-WS.org, 2021. URL: http://ceur-ws.org/Vol-2980/paper305.pdf. [30] The KnowWhereGraph Ontology, The KnowWhereGraph ontology, https://stkokwg.geog.ucsb.edu/lod/ontology.ttl, 2023. [31] kwg-site, Knowwheregraph, https://knowwheregraph.org/, ????