Knowledge graph-enhanced carbon impact assessment of construction and demolition waste with large language model-based decision support

Abstract

Construction and demolition waste sector generates substantial carbon emissions across sorting, transportation, and disposal phases. However, existing assessment approaches often generate static numerical results that offer limited traceability, regulatory interpretation, and decision-support capability. This study proposes an integrated framework that combines a knowledge graph with a large language model for construction and demolition waste carbon emission management. A domain-specific construction and demolition waste Carbon Knowledge Graph is constructed in Neo4j to integrate waste categories, treatment methods, facility information, source-referenced emission factors, and regulatory constraints. Carbon emissions from sorting, transportation, and disposal are quantified using an emission factor-based approach and embedded into the graph for traceable accounting. A retrieval-augmented generation mechanism connects the knowledge graph with a large language model, enabling natural language querying and evidence-based explanations. A residential demolition case study demonstrates the applicability of the framework. The results indicate that transportation is the dominant emission phase, accounting for 63.6% of total emissions, and show that the recycling-based pathway for concrete waste reduces emissions by 36.46% compared with the landfill-only baseline. This study contributes to bridging knowledge management and carbon modeling, providing an explainable, traceable, and regulation-aware decision-support approach for low-carbon construction and demolition waste management.