Environmental impact assessment (EIA) is a required process for projects in many countries, which results in the preparation and publication of a detailed report. Such EIA reports describe impacts on, e.g., humans and the environment. In this paper, we describe our eforts in modeling the metadata of EIA reports and their description of environmental impacts and mitigations with Wikidata and Wikibase. We show that it is possible to record the bibliographic metadata of Danish EIA reports and in doing so link it to the rest of the Wikidata knowledge graph, allowing multilingual search. With our dedicated instance of a Wikibase, we show how EIA reports and their associated projects along with activities, impacts, recipients, and mitigations can be represented and how we can show aggregated views of the data based on SPARQL templates so that users can explore the data and make eficient use of it.
environmental data, and within the DREAMS project, the EIA reports have now been stored as PDF ifles and made available.
Part of the DREAMS project is to establish a system capturing the specific environmental efects and mitigations that are described in the EIA reports — an efect assessment tool we refer to as CAUSA. The relations for environmental efects and mitigations are often not just a simple triple relation, but may be of higher order and form a small chain or graph. The project also explores how the Sustainable Development Goals (SDG) can be integrated into EIA processes [3] and sets out to establish links between impacts identified through the EIA process and the SDGs. We have annotated a set of EIA reports with the cloud service tagtog5 which can handle PDF files and link relations between annotated text pieces [4]. The annotation can be exported as JSON and we have further converted it to a spreadsheet format.
Wikidata [5] is a knowledge base closely linked to Wikipedia. With the Wikicite initiative, see, e.g., [6], metadata about publications have been added to Wikidata so that a considerable and growing part of Wikidata is now publications. At Wikicite.org, the estimated fraction in Wikidata of items about publications grew to 42% in 2022, corresponding to almost 42 million items.6 As of 2023, over 38 million Wikidata items are tagged as scholarly articles.7
Wikidata is continuously converted to a Semantic Web representation that is downloadable as well as query-able through the Wikidata Query Service (WDQS) using the SPARQL query language.8 The conversion to a Semantic Web representation also entails a complex structure with qualifiers and references [7].
Various tools have taken advantage of the data in Wikidata available through WDQS. One such tool is Scholia which was originally designed to display the metadata of scientific publications [ 8]. It has now been extended to display information about, e.g., taxa, biological pathways, and chemicals, see, e.g., [9]. Currently, Scholia does not support other languages than English, other Wikibases with other SPARQL endpoints than the WDQS, and other specialized data that would not be appropriate to enter into Wikidata. In our case, the issues are obstacles as we would like to support Danish and specialized data from EIA reports.
Wikibase9 is the software that runs Wikidata. Knowledge workers may download the software and run it locally or use a cloud service where Wikibase is set up. One cloud Wikibase service provider is Wikimedia Deutschland which provides its service from https://wikibase.cloud. Wikibase instances may be used by organizations to manage their data — either public or private.
The Wikibase set up at Wikibase.cloud is bundled together with a triple store, a Blazegraph instance, corresponding to the Wikidata Query Service for Wikidata. Data from the Wikibase are continually piped into the triple store so the data in a Wikibase.cloud Wikibase instance is available via SPARQL queries to the endpoint.
Compared to Wikidata the curation in (another) Wikibase can be tailored to the data of the organization with specialized properties and items that would not usually appear in Wikidata. Furthermore, editing may be restricted to those with authorization.
In this paper, we describe our eforts to model EIA reports within Wikidata and Wikibase as well as supporting ontologies and modeling the environmental efects with what we call an efect pattern . We also present a new method for displaying the information from the wiki via SPARQL queries, the Wikibase pages, and a simple webpage with Javascript. Our eforts are available in the DREAMS Wiki, which is a Wikibase.cloud instance at https://dreams.wikibase.cloud/.
Garrido and Requena described an ontology for environmental impact assessment [10]. The ontology features classes such as Impact, VisualImpact, NoiseImpact, TraficChanges, and Amphibian. An OWL ifle is available 10 and the authors also described a web application to display the ontology.
The SuperPattern Ontology has been suggested for the representation of individual scientific claims [11]. This ontology is a scheme with five slots: classes for context, subject, qualifier, relation, and object. The ontology has been used to express a scientific claim such as “strong static magnetic field generally afects cell cortex in the context of dejellied fertilizable stage VI Xenopus laevis oocyte” [12].
Wikidata can represent cause and efect between entities with the has efect (P1542) and has cause (P828) properties. This data and the textual data in Wikipedia has been used to build a knowledge graph of events and consequences [13]. Articles describe using Wikibase, e.g., for a food composition knowledge base [14], as an enslavement resource [15], and for biological collection information [16].
We established several ontologies on the DREAMS Wiki. The items of these ontologies are typically linked hierarchically with a subclass of property. The ontologies we consider are project types, report types, significance types, taxa, efects, and SDGs.
An ontology of project types was created from the itemized hierarchy of the Danish environment assessment law. For instance, the project type “elektricitetstransportprojekt (bilag 2, 3c)” refers to the project description of the electricity transport in appendix 2, item 3.c of the Danish law. The corresponding item in Directive 2011/92/EU is part of annex II, item 3.b. Each report links to one or more project types. The project types are linked together with subclass of with the project item as the top entity. Some entities do not correspond to entities in the law, e.g., (the generic) electricity transport project aggregates two items from diferent appendixes in the law.
Our taxa ontology records species and other taxa, but also some aggregated entities, such as “annex IV species or species group”, which is an aggregated group of protected species and species groups. A small ontology for report types was created with just four entities: report, environment assessment report, environmental report, and combined environmental and environmental assessment report as the only entities. Another small ontology records the significance types , where instances are, e.g., “not significant and negative impact” or “significant and positive impact”. We create an ontology of possible efects . Items in this ontology are, for example, “wind turbine noise”, subclass of “noise”, subclass of “physical/environmental determinant’,’ subclass of “human health impact”, finally subclass of “impact”. An ontology for SDGs is set up for the 17 SDG goals and a selection of the corresponding targets (as a subclass to an SDG goal). Efects of an activity may link to one or more targets. Identifying relevant SDG targets was based on the SDG targets provided in [17].
Our ontologies in the wiki were set up from information from experts recorded in tabular format in documents and gradually expanded and modified based on annotated EIA reports. We record possible aliases for ontology entities in the DREAMS Wiki, e.g., noise, which in Danish is “støj”, would have aliases such as “støjpåvirkning”, “støjen”, “Støj” and “støjbidrag” reflecting inflections and compounds. We use these aliases for semi-automated entity linking. 10https://github.com/julian-garrido/EAonto/blob/master/eia.owl class title main topic country of origin language publication date number of pages project type Miljørapporter ID Wikidata ID citations P2 P1 — — P11 P6 P27 P16 P40 P3 — P31 P1476 P921 P495 P407 P577 P1104
— P10930
— P2860 environmental impact assessment report “Energiparken ved Vennerslund Energi- og Naturpark” Vennerslund, wind farm, PV system Denmark Danish “2020” “441” photovoltaic system project, project with wind turbine 377355e9-9db6-4163-aad0-cf0e5a0004e5 Q112506586
Dansk pattedyrs atlas, Behavior of Scandinavian Bats during . . .
We represent metadata about an EIA report on a page in the DREAMS Wiki, see Table 1. The properties we consider are, e.g., instance of for the class of the report, title, author, language (always Danish), publication date, project type, and an identifier for the report in Danmarks Miljøportal. Some of these property values are set up from information obtained for entry in Danmarks Miljøportal while others have been manually entered — optionally and not systematically. We have currently recorded 632 reports in the DREAMS Wiki.11
We also represent the EIA report in Wikidata with properties such as instance of, title, main subject, country of origin (always Denmark), language (always Danish), date of publication, number of pages, and Miljørapporter File ID that links to the report at Danmarks Miljøportal. The DREAMS Wiki has a link to the report entity in Wikidata. Some EIA reports contain references — often a highly diverse set of references, where the target of the reference may include databases, various reports, scholarly articles, and books. We have added some references to Wikidata.
Geographic entities described in a report might not yet exist, e.g., “Veddum Kaer” (Veddum marsh) is not (yet) found in Wikidata. Other resources, such as OpenStreetMap or Geonames also return no results. Reports may describe projects that are pointlike (e.g., small plants), lines (e.g., railway tracks), or areas (e.g., wind turbine park). Projects may also detail multiple alternative solutions for alternative routes, e.g., alternative routes for a new railway line.
Document identifiers are not commonly used in Danish EIA reports. When the EIA reports are represented in Danmarks Miljøportal they are each assigned a unique identifier and we can use that. Both Wikidata and the DREAMS Wiki each as a property for that identifier.
Inspired by the SuperPattern Ontology, we developed a scheme for the DREAMS Wiki to represent the individual environmental efect relations described in an EIA report. Our initial experiments defined a scheme with subject class, object class, context class, and relation type properties closely following 11SPARQL query to the DREAMS Wiki Query Service: https://tinyurl.com/2jygv2o5 instance of described with subject text subject class object text object class recipient text recipient class degree of impact text degree of impact mitigation text mitigation class project type project phase quote page pattern number
Description item class report entity page number in PDF number used during tagtog input the SuperPattern Ontology scheme. However, to incorporate recipient and mitigation in a chain of environmental efects, we extended the scheme with recipient and mitigation classes and omitted the context class. Furthermore, we omit the relation type as the relationship between subject, object, recipient and mitigation are implicitly given, e.g., the relation from the subject would in most cases correspond to “afects” in the SuperPattern Ontology scheme. We call our scheme an efect pattern . This efect pattern also contains an identifier (the Q-identifier of the Wikibase page) a link to the entity of the EIA report, a quote from the PDF, page, an indication of the significance of the efect, and link to the entity of the project phase class (e.g., planning, construction, operation and/or decommissioning). The four essential parts of the environmental efect chain are subject (usually a physical construction or an activity), object (also called impact), recipient, and mitigation. These four parts may be represented with both a string and an entity link: From the manual annotation extracted from tagtog, we get the string. Using the ontology of efects we manually or semi-automatically—with the aliases presented in the wiki—make the entity linking for the string to the entities.
Mitigation in one efect pattern can be a subject in another efect pattern, e.g., a road (subject) causes noise (object/impact) which afects neighbors (recipient) and can be mitigated by a noise wall (mitigation). The noise wall (subject in another efect pattern), in turn, causes a visual impact (object) that afects neighbors (recipient).
Table 2 displays an example of an efect pattern from an EIA report about a specific photovoltaic system and wind turbine project. The efect pattern represents that the report describes that the specific planned photovoltaic system creates a light impact (reflections) that afects the neighbors with a “not significant and negative impact”. Any efect is mitigated with a vegetation zone and the efect is present in the operational phase of the project. We have currently recorded 4303 efect patterns in the DREAMS Wiki.12
Inspired by Scholia, we created a system for displaying aggregated information from the DREAMS Wiki via SPARQL queries to the endpoint associated with our Wikibase. Instead of a Python/Flask-based approach like Scholia, we employ a serverless single-page Web application (SPA). We mirror Scholia’s URL patterns to distinguish diferent aspects of the data. However, as an SPA we use the URI fragment to control which aspect and entity to show. For instance, the URI fragment #report/Q153 will display entity Q153 (the report Vindmøller ved Bjørnstrup) as a report.
Whereas Scholia stores template SPARQL queries in separate Jinja2 files read during application startup, our present system uses pages on the wiki itself to store the templates with SPARQL queries. For instance, the wikipage Dreams:report defines a template for a report corresponding to a URI fragment regular expression pattern of #report/Q\d+, while Dreams:report-index corresponds to an index page for reports where the URI fragment pattern is #report. Likewise, the Dreams:projecttype wikipage defines the template for a project type. Other aspects we have defined on the wiki besides report and projecttype are pattern, efect, recipient, sdg, location, taxon, and author. The Web application can display faceted aspects. For instance, #projecttype/Q264/report will show reports for wind turbine projects. The DREAMS Wiki Query Service can perform federated SPARQL queries against the Wikidata Query service. We have used that to display “Related works from co-citations” from the citation information present in Wikidata.
When a user visits a new “page” on the SPA, the application reads the URI fragment, determines which template should be downloaded from the wiki, and then builds the HTML page from the 12SPARQL query to the DREAMS Wiki Query Service: https://tinyurl.com/2kczebk3 template. Currently, the SPA recognizes wiki section headers and converts them to HTML headings and SPARQL queries embedded in a MediaWiki template. The SPARQL queries define a target entity that is interpolated with the entity identifier extracted from the URI fragment. The interpolated SPARQL queries are then sent of to the SPARQL endpoint of the wiki. The response from the endpoint is converted by Javascript to an HTML table like the method in Scholia. Like Scholia, we use jQuery and SpryMedia’s DataTables library13 to display the tables. The DataTables library can be configured, so we have switched the language to Danish. The SPARQL query service of the Wikibase.cloud has currently relatively slow response times when multiple simultaneous queries are issued, hence we have limited the number of tables shown on each page — usually to just one or two.
Figure 1 shows a screenshot from a page on the Web application displaying efect patterns for photovoltaic system projects. Each blue text fragment is linked to another page of the Web application, displaying other aspects of the data. The “Citat” (quote) column displays an extracted paragraph, where the efect pattern description originates. The last column in the table is a link to the PDF. With the page number for the quote, we are able to make a URL that will direct the browser to the page with the quote. With property path SPARQL queries we can aggregate information across the hierarchy of our ontologies, e.g., aggregation of all efect patterns from energy industry projects.
Our manual annotation from tagtog did not correspond exactly to the scheme we settle on for representing the environmental efects in the wiki. Hence manual and time-consuming editing in the wiki has been necessary, particularly around information about recipients. Manual extraction of efect patterns from EIA reports does not scale. Machine learning and large language models could help with relationship extraction and entity linking. However, the integration of machine learning or language models with the Wikibase.cloud system is not straightforward.
The use of on-wiki SPARQL templates makes the development of the Web application for displaying the wiki content more agile. Adding a new aspect requires only the definition of a new wikipage with SPARQL queries, whereas a Flask/Jinja2 approach requires software developer involvement.
Although developed for Danish EIA reports, the approach should also work for environmental impact assessment in other countries. Wikibase is multilingual and some entities in the DREAMS Wiki have been translated from Danish to English. Future development could explore how to improve the response time of the display of data in the web application as well as ensuring that search and navigation within the Web application are eficient for the user.
This research is part of the DREAMS project, Digitally Supporting Environmental Assessment for the Sustainable Development Goals. This research is made possible by the Innovation Fund of Denmark (Grant agreement number, 0177-00021B DREAMS).