Automated geo-computations driven by knowledge graphs: illustrated by geomorphological and soil cases

Abstract

Geo-computation is a crucial approach in geographic information science for detecting, predicting, and simulating geographic entities, events, and phenomena, which is achieved through a series of geographic models applied to geographic data. Geo-computation has become a key technical approach for quantitative analysis, simulation, and prediction in modern geography, as well as a vital means by which geography contributes to socio-economic development. However, current geo-computations require users to select appropriate geo-models for geo-computational tasks and to collect and preprocess the geo-data needed by these models. This poses challenges for users, particularly novice users, as they are required to invest substantial time and effort in understanding the technical details (e.g., installation and configuration) of geo-models, and the associated data preparation procedures (e.g., collection and preprocessing). Moreover, with the advancement of research on global change and sustainable development, geo-computational tasks are becoming increasingly complex. This complexity necessitates more sophisticated models and a greater amount of data, which further intensifies the challenges for users. To address these challenges, automated geo-computation—automatically selecting the suitable geo-models and matching the corresponding geo-data resources for geo-computational tasks—has become a promising solution and research hotspot in academia.

Therefore, we propose a systematic, descriptive, and procedural method driven by knowledge graphs to capture, represent, organize, and process the essential features of components, relationships, and dynamic computational procedures in geo-computations, aiming to reduce manual involvement and assist in the automation of model selection and data matching. In this approach, the first part involves designing a geo-computation ontology and constructing knowledge graphs that integrate knowledge from geo-models, geo-data, and domain knowledge. Subsequently, key methods are designed to achieve automated task analysis, model selection and linking, and data matching using the constructed knowledge graphs. Finally, an application prototype system is developed to perform automated geo-computations driven by knowledge graphs. Three application cases, namely, geomorphological classification, soil erosion, and soil potential productivity, are computed to illustrate the applicability and effectiveness of automated geo-computations supported by the proposed method. As demonstrated by the cases studied, the proposed knowledge graph-driven method improves the efficiency of model selection and configuration, enhances the utility of open data, generates reliable results, and promotes deeper integration between data and models for automated geo-computation.

Date of Award	18 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Ping Fu (Supervisor), Yunqiang Zhu (Supervisor), Stuart Marsh (Supervisor) & Amin Farjudian (Supervisor)