Abstract: Water-retaining mining of ultra-thick coal seams in ecologically fragile mining areas is a major scientific and engineering challenge in the development of coal resources in northern Shaanxi, with the evaluation of water-retaining mining effectiveness being a critical link. Addressing the water-retaining mining effectiveness assessment for the 2⁻² ultra-thick coal seam (average thickness of 11.18 m) at Caojiatan Coal Mine, this study systematically tested and analyzed 20 samples from the Zhiluo Formation aquifer, the Salawusu Formation aquifer, the Quaternary aeolian sand aquifer, surface water, and mine water. In terms of hydrochemistry, a water source identification model based on Fisher discriminant analysis was established, achieving a conventional hydrochemical backtracking accuracy of 64.71% and an isotopic hydrochemical backtracking accuracy of 88.23%. Neither method alone could fully meet the requirements for water inrush source tracing, but isotopic results revealed extremely low tritium values in the Zhiluo Formation aquifer, indicating ancient confined water, with mine water tritium values matching those of the Zhiluo Formation. For microbiology, high-throughput sequencing technology was introduced for the first time in water-retaining mining effectiveness evaluation. Through multidimensional analysis including α-diversity analysis, species annotation and composition analysis, and β-diversity analysis, it was found that mine water (Group E) and Zhiluo Formation water (Group A) exhibited highly overlapping distributions in nine diversity metrics, such as the Chao1 index (medians of 300 and 240–280), Shannon index (medians of 3.0 and 3.2), and Simpson index (medians of 0.82 and 0.92). Additionally, the γ-Proteobacteria class showed nearly identical proportions in both groups (47% in Group A and 46% in Group E), with key OTUs (e.g., OTU_134 and OTU_27) serving as “signature species” unique to each group. The joint quantitative discrimination model based on hydrochemical and microbial indicators demonstrated a coupling model discrimination accuracy of 100%. The Zhiluo Formation accounted for 78.67% of the mine inflow water, surface water contributed 20.13%, the Salawusu Formation represented only 1.2%, and the Quaternary aquifer constituted 0. Water level monitoring data further confirmed minimal fluctuations in the water levels of the Salawusu Formation and the Quaternary aquifer. The study yielded the following main conclusions: The primary water inflow source for mining in the ultra-thick coal seam of Caojiatan Coal Mine is the Zhiluo Formation aquifer, accounting for 78.67% of the total inflow. The Baode red soil (with an average thickness of approximately 70 m) effectively blocks the recharge from the Salawusu Formation and Quaternary aquifers into the mine. Surface water enters the mine through conductive fracture zones at a local rate of about 20.13%, primarily distributed in areas with significantly reduced impermeable soil thickness, such as the Yiji River Reservoir and Gaogaotu Reservoir. Microbial high-throughput sequencing technology effectively compensates for the insufficient differentiation capability of conventional hydrochemical methods in water source identification, achieving a 100% accuracy rate, and provides a new technical approach for evaluating water conservation effects in ecologically fragile regions. The Salawusu Formation aquifer, as the primary target aquifer for water-preserving coal mining, has not yet been significantly impacted by coal mining activities, demonstrating effective water-preserving mining outcomes.The research results provide theoretical basis and technical support for water conservation coal mining in ecologically fragile mining areas with thick coal seams, and have reference significance for evaluating the effectiveness of water conservation mining under similar conditions.