ARCHITECTURE OF AN ANALYTICAL SYSTEM FOR FORECASTING AND ASSESSING EMERGENCY RISKS USING ARTIFICIAL INTELLIGENCE AND GIS

Abstract

The article presents a scientifically grounded architecture of an analytical system for forecasting and assessing emergency risks developed within project IRN AR 275064/0224. The study addresses the need to move from fragmented departmental records and predominantly manual analysis toward a unified spatiotemporal framework integrating historical events, operational observations, geospatial data, and information on response resources. A normalized integrated risk model R = H × E × V / C is proposed, combining hazard, exposure, vulnerability, and response capacity. A multilayer architecture is developed that includes data sources, an integration layer, specialized repositories, an analytical layer, GIS-based modelling, and decision-support interfaces. A hybrid approach to artificial intelligence is substantiated: physically interpretable and expert models provide the basic scenario, while machine-learning methods refine parameters using regional historical data. Automated outputs are accompanied by quality indicators and uncertainty estimates. A flood-threat scenario demonstrates the complete analytical sequence from data-quality control and water-level forecasting to inundation mapping, impact assessment, and optimization of resource allocation. Project acceptance criteria and a staged implementation scheme are proposed. The system is designed to support, rather than replace, authorized decision-making, combining analytical efficiency, explainability, information security, and meaningful human control.