Abstract: Karst depressions and gully convergence zones offer favorable topographic conditions for the construction of pumped storage power stations, reducing excavation volumes and construction costs. However, intense karstification in these regions introduces complex engineering geological risks. Accurate evaluation of rock mass integrity and karstification degree is therefore critical for safe and efficient project development.(Purpose/Significance) To resolve challenges in the assessment process—such as difficulty meeting index measurement requirements, tedious repetitive work, and insufficient differentiation precision—this study takes a pumped storage project in the Zigui karst area as a case. It uses multi-source exploration data (including drilled rock cores, borehole sound wave, and borehole TV images) for mutual complementarity. This reveals the patterns of rock mass integrity and dissolution degree across different strata in the study area, identifies the geological characteristics of each stratum, and assigns stratum labels accordingly. With stratum labels and rock mass acoustic wave velocity data as inputs, the Weighted Random Forests (WRF) method is applied to propose a multi-source information fusion assessment method for rock mass integrity and dissolution degree in karst areas. (Method) The results show that the various types of information at the same location are interrelated, and the differences in the presentation forms of the information only arise from the different exploration methods and information sources; there are significant differences in the geological characteristics of rock masses in different strata of the study area, and the corresponding characteristic manifestations of information from various exploration methods also vary, which affects the evaluation of rock mass integrity and dissolution degree using multi-source exploration information. The proposed method was trained with 1073 samples and tested with 118 samples. The consistency rates for integrity assessment in the training set/test set reached 95.67%/94.92%, and the consistency rates for dissolution degree assessment in the training set/test set reached 98.02%/97.46%. Moreover, the results are in good agreement with the field targeted re-exploration and verification results. (Result) Compared with traditional methods, the proposed approach achieves higher accuracy, finer resolution, and improved automation, while effectively accounting for stratigraphic geological features. It provides a practical and efficient tool for evaluating engineering rock mass quality in karst-area pumped storage projects.(Conclusion)