Automated Construction of Geospatial Analysis Workflows: Recursive Solving of Elementary Modeling Questions Adaptive to Workflow Construction Contexts

Abstract

ABSTRACT Geospatial analysis workflows are widely used to address complex geographic problems, but their construction is challenging due to dynamic contextual factors like the characteristics of the study area, geo‐models, and datasets during construction process. Current automated approaches typically focus on the workflow executability, neglecting such dynamic influences—defined in this paper as “workflow construction contexts.” This ignorance undermines the rationality and accuracy of workflow modeling. To address this limitation, we propose a new automated approach adaptive to workflow construction contexts, which recursively decomposes the workflow construction process into elementary modeling questions (EQs). This allows the dynamic contexts to be characterized based on pre‐formalized modeling knowledge, enabling the automated construction of appropriate workflows. Implemented in a prototype system with a knowledge graph, our approach is validated through a digital soil mapping case study. The results demonstrate that the proposed approach can successfully construct distinct, context‐specific workflow structures that adaptively match different contexts, thus significantly enhancing the rationality and feasibility of geospatial analysis workflows.