Page Text: Project Story
Improving environmental impact assessments using knowledge graphs
Environmental, Social and Health Impact Assessments (ESHIAs) are required by investors and regulatory authorities for large scale engineering and infrastructure projects. The early identification of the impact of such projects on habitats and species can avoid significant financial penalties and costs in later stages. This project demonstrated the use of “knowledge graphs” in linking together all the relevant information needed to produce an ESHIA, thereby increasing the efficiency of producing ESHIAs and increasing knowledge sharing between projects within the client’s organisation and with their customers.
Process
Through a series of consultation meetings with the client, we gained an understanding of the ESHIA process, the datasets that are involved and the challenges faced by environmental consultants in drawing all the relevant information together. The knowledge graph - and a custom-built user interface, - were built in an iterative fashion, gaining feedback from the client at each stage.
Solution
We developed an online software tool, consisting of a graph database, query tool and a web-based graphical user interface for searching and visualising the data. The knowledge graph approach is a highly flexible means of storing data, representing information as a set of objects that are linked together by relationships, forming a rich network (or graph) of information. For example, information about a species can be linked to information about its vulnerabilities, habitats and protected status, which can in turn be linked to other information, such as scientific publications and biodiversity surveys.
This approach allows the graph database to grow organically with time and permits powerful and rapid searching and querying. Using reference data provided by the client, and accessing a number of third party biodiversity and habitat datasets, a trial database was created storing 897 species and 4125 protected or key biodiversity areas. We also automatically extracted information from a set of publications and reports and linked these into the knowledge graph.
Impact
A key result of the project was to show that complex queries across multiple data sources can be made much easier. For example, the analyst may wish to find all species that are vulnerable to oil and are also on threatened species lists. Answering this question involves cross-referencing multiple data sources (both in-house and third-party) and can be time-consuming with traditional tools, but is made significantly more efficient using a knowledge graph approach.
In addition to the creation of a knowledge base and query tool, it was demonstrated how an open satellite data - such as optical imagery from the Sentinel 2 satellites - could potentially be used to detect changes in habitats (for example, that have occurred since other reference sources have been issued). This could allow analysts to stay informed on relevant habitat changes at minimal cost.
Tell us about your data challenge
Get in touch
The University of Reading's Data Protection policy .
The icons in How we do it are from Icons8 .