2025 , Volume 30, № 4, p.26-40

Purpose. The article addresses analyzing of the introduced new approaches for solving applied problems in geoecology and nonlinear geomechanics using large language models — a rapidly developing branch of artificial intelligence. The authors seek to demonstrate how LLM can address limitations of traditional computational methods, especially when working with large, heterogeneous,and multimodal datasets.

Methodology. The authors review existing computational approaches (e.g., boundary integral equation and finite element methods) and highlight their constraints in handling unstructured textual and multimodal data. They propose the integration of LLMs, trained on extensive textual corpora, to extract, synthesize, and generate domain-specific knowledge. The article discusses several practical applications, such as modelling nonlinear soil behavior, predicting environmental impacts, optimizing mining operations, and implementing transfer learning. The integration of LLMs with geographic information systems and real-time monitoring systems is also explored.

Findings. LLMs can significantly expand the range of solvable problems in geoecology and geomechanics. They enhance predictive accuracy, support knowledge transfer, and facilitate the creation of educational resources. However, challenges include high computational demands, data quality and availability, model interpretability, and ethical risks such as misinformation and privacy concerns.

Originality/value. The article is among the first to systematically evaluate the potential and challenges of LLMs in geoecological and geomechanical applications. It emphasizes the importance of prompt engineering and domain adaptation for effective LLM use, and highlights the transformative potential of these models for research, industry, and education in the field.