A Novel Method for Water Inrush Evaluation Considering Spatial Variability of Strata Permeability: Principles, Optimization, and a Case Study

Abstract The water permeability of the deep limestone strata in North China-type coalfields exhibits significant spatial variability, which presents challenges in accurately assessing the risk of water inrush under mining disturbances using traditional vulnerability index methods. These conventional approaches often rely on averaging and planarizing the water permeability characteristics of aquifers, overlooking their strong spatial variability. As a result, this can lead to error propagation and amplification. To address these issues, this paper introduces a vulnerability index tomography method that explicitly accounts for the spatial variability of water permeability. The method involves subdividing the strata into multiple layers and independently calculating vulnerability indices for each layer, followed by precisely locating the layers with high water inrush risks. This layered approach enhances the accuracy of the evaluation by more effectively capturing vertical index variations. Furthermore, the CPO-Adaboost-RF intelligent algorithm resolves several challenges such as incomplete geological data, extensive computations, and human errors in analysis. A case study conducted at the Dengcao coal mine revealed significant differences in vulnerability among different layers, highlighting that the strong spatial variability cannot be ignored. The combination of the sensitivity analysis results and the results from geophysical prospecting further validated the effectiveness and accuracy of the proposed method. This paper provides a theoretical foundation and practical reference for the assessment of water inrush risk of rock strata with strong vertical water permeability variability, such as the deep limestone in North China-type coalfields, and offers insights for formulating more rational water prevention measures.