Enhancing interoperability between geospatial data: NPL semantic similarity alignment metrics with AI between land cover and land use data

Abstract

The evolution of geospatial data sources and their varied classification systems presents data integration and interoperability challenges. This research addresses these challenges by introducing an AI-driven methodology using Natural Language Processing (NLP) to measure semantic similarity between land use, vegetation classification systems and the national topographic database. Leveraging NLP techniques, such as those in ChatGPT-4.0, this approach automates the semantic alignment process, reducing manual work. The study aimed to align the Brazilian ET-EDGV topographic mapping with broader national (IBGE Vegetation and Land Use Manuals) and international (Dynamic World, Global Forest Resources Assessments (FRA)) classification systems. By applying semantic similarity coefficients, the research sought to create a harmonized framework for integrating geospatial data. The methodology combined AI-based semantic similarity measures, ensuring consistent data alignment. Results showed strong alignments for classes like “Cultivated Vegetation” and “Crops” and identified challenges for unique Brazilian ecosystems such as “Campinarana”. The “Mangrove” class highlighted the need for context-specific definitions. The study concludes that NLP can contribute to automated semantic alignment, enhancing geospatial data integration and interoperability. Although focused on Brazilian data, this methodology is adaptable globally, supporting better landscape representation and decision-making. Future research should integrate advanced AI models and broader ecosystems to refine the process.