Foundation Models for Geospatial Reasoning: Assessing Capabilities of Large Language Models in Understanding Geometries and Topological Spatial Relations

TL;DR

This study probes the ability of large language models to reason about geospatial geometries and topological spatial relations by encoding vector data as Well-Known Text (WKT) and evaluating three tasks: topological relation qualification, spatial query processing, and vernacular-to-formal predicate conversion. It compares embedding-based and prompting approaches across GPT-3.5-turbo, GPT-4, and DeepSeek-R1-14B, with GPT-4 achieving the highest performance in few-shot settings ($>0.66$) and embedding-based methods exceeding $0.60$ accuracy on average. The results show that LLMs can preserve geometry types in WKT embeddings and generate plausible, relation-preserving synthetic geometries, though strict topological discrimination remains challenging and context often influences outcomes. The work offers insights for developing geo-foundation models, highlights the value of tailored prompts and context, and points toward neurosymbolic and retrieval-augmented approaches to improve geospatial reasoning in practice.

Abstract

Applying AI foundation models directly to geospatial datasets remains challenging due to their limited ability to represent and reason with geographical entities, specifically vector-based geometries and natural language descriptions of complex spatial relations. To address these issues, we investigate the extent to which a well-known-text (WKT) representation of geometries and their spatial relations (e.g., topological predicates) are preserved during spatial reasoning when the geospatial vector data are passed to large language models (LLMs) including GPT-3.5-turbo, GPT-4, and DeepSeek-R1-14B. Our workflow employs three distinct approaches to complete the spatial reasoning tasks for comparison, i.e., geometry embedding-based, prompt engineering-based, and everyday language-based evaluation. Our experiment results demonstrate that both the embedding-based and prompt engineering-based approaches to geospatial question-answering tasks with GPT models can achieve an accuracy of over 0.6 on average for the identification of topological spatial relations between two geometries. Among the evaluated models, GPT-4 with few-shot prompting achieved the highest performance with over 0.66 accuracy on topological spatial relation inference. Additionally, GPT-based reasoner is capable of properly comprehending inverse topological spatial relations and including an LLM-generated geometry can enhance the effectiveness for geographic entity retrieval. GPT-4 also exhibits the ability to translate certain vernacular descriptions about places into formal topological relations, and adding the geometry-type or place-type context in prompts may improve inference accuracy, but it varies by instance. The performance of these spatial reasoning tasks offers valuable insights for the refinement of LLMs with geographical knowledge towards the development of geo-foundation models capable of geospatial reasoning.

Figures (13)

• Figure 1: The conceptual neighborhood of topological relations in RCC-8 randell1992spatial on the left (redrawn for comparison) and 9-IM egenhoferReasoningGradualChanges1992 on the right.
• Figure 2: Overview of the workflow in this research.
• Figure 3: The decision tree for the topological spatial relations.
• Figure 4: The workflow for the topological spatial relation qualification task.
• Figure 5: Topological spatial relation qualification example inputs and outputs with different prompt techniques.