AI-based procedure for extracting spatial and temporal information on rainfall-induced landslides from textual sources

Abstract Comprehensive landslide catalogues require accurate and reproducible spatial and temporal information, yet manual extraction from textual sources is time-consuming and affected by inter- and intra-operator variability. We present AIDE4LAND (Artificial Intelligence-based procedure for Data Extraction for LANDslides), a novel structured procedure that combines Large Language Models and prompt engineering to automatically extract temporal and spatial information on rainfall-induced landslides from textual sources. The procedure is organized as a two-step process: (i) temporal extraction, which reconstructs the date and time of occurrence according to explicit hierarchical rules and assigns standardized temporal accuracy levels, and (ii) spatial extraction, which progressively determines geographic coordinates through structured queries and reverse-geocoding validation using OpenStreetMap APIs via Anthropic’s Model Context Protocol. The workflow was developed through iterative prompt-refinement cycles, involving 2162 tests conducted on 725 textual sources (664 written in Italian and 61 in English) published between 2011 and 2025. Validation of AIDE4LAND outputs against a manually compiled expert dataset revealed a high level of agreement, with 87% of cases demonstrating maximum concordance in terms of temporal and spatial attribution. By combining standardized rules, explicit accuracy classification, and automated consistency checks, AIDE4LAND formalizes expert reasoning into transparent and reproducible procedural rules, thereby reducing interpretative variability and ensuring transparent and systematic traceability in the decision process. The procedure significantly accelerates catalogue updates and provides a transferable and reproducible methodological framework for constructing structured rainfall-induced landslide catalogues, supporting hazard assessment, rainfall-threshold definition, and early warning applications.